Current consensus guidelines from the American Diabetes Association and the American Association of Clinical Endocrinologists recommend the use of insulin‐based treatment protocols for most hospitalized patients with hyperglycemia.1 For noncritically ill patients, it is recommended to target a fasting blood glucose (BG) < 140 mg/dL and a random BG < 180‐200 mg/dL, without excess hypoglycemia. Prior studies recommended using a basal‐bolus insulin protocol that specifies starting doses and parameters for dose adjustment, applied by well‐educated teams of physicians and nurses.27 We have shown that insulin detemir given as a once‐daily basal injection coupled with rapid‐acting insulin aspart with meals is an effective regimen for managing hyperglycemia in hospitalized patients with type 2 diabetes.7 We and others have shown that once‐daily basal insulin mealtime rapid‐acting insulin is significantly more effective than sliding‐scale regular insulin in the hospital setting.6, 8

The majority of patients admitted to general medical units are first evaluated in the emergency department (ED), and significant hyperglycemia is not uncommon in ED patients. However, protocols for the treatment of hyperglycemia in the ED have not been widely implemented. Ginde et al studied 160 ED patients with a history of diabetes and BG > 200 mg/dL and found that although 73% were admitted to the hospital, only 31% were treated with insulin, and only 18% had a diagnosis of diabetes charted.9 A recent survey of 152 residents and attendings in 3 academic EDs found that only 32% would give insulin for a BG > 200 mg/dL, 59% for a BG > 250 mg/dL, and 91% for a BG > 300 mg/dL to ED patients with known diabetes.10 We completed a preliminary study of a novel protocol for the administration of subcutaneous insulin aspart in the ED at Rush University in 2008.11 We found that the mean BG was significantly lowered during an ED stay, from 333 to 158 mg/dL, and that the protocol was easily adopted by ED staff, with a low rate of hypoglycemia. Historically, only 35% of hyperglycemic patients with diabetes received insulin in our ED.11 Reasons for limited ED management of hyperglycemia may include presence of more critical issues, time and resource restriction, unfamiliarity with glycemic targets, and concerns regarding hypoglycemia.9, 10

In the current study we focused on 3 questions:

• 1

  Could we further reduce the risk of hypoglycemia by modifying our original Rush ED insulin protocol? We reduced insulin aspart from 0.1 to 0.05 units/kg for BG 200‐299 mg/dL, from 0.15 to 0.1 units/kg for BG 300‐399 mg/dL, and from 0.2 to 0.15 units/kg for BG 400 mg/dL.

• 2

  Could we couple our ED insulin aspart protocol with prompt initiation of a detemir‐aspart protocol in those patients who were subsequently admitted to general medical units from the ED?

• 3

  Would the hospital length of stay, mean BG, and incidence of hypoglycemia be improved by the use of 2 back‐to‐back subcutaneous insulin protocols in a randomized clinical trial compared with the usual care provided in the ED and general medical inpatient units?

Research Design and Methods

From May 2008 through June 2009, patients presenting to the Rush University Medical Center ED with a history of type 2 diabetes and an initial point‐of‐care BG 200 mg/dL were randomized to an intervention group (INT) or to a usual care group (UC) after giving informed consent. Inclusion criteria for the study were: ages 18‐80 years, history of type 2 diabetes for at least 3 months, and prior therapy with dietary management, oral agents, or insulin. Patients were excluded if subsequently found to have diabetic ketoacidosis, hyperosmolar nonketotic syndrome, or critical illness requiring intensive care unit admission/direct surgical intervention. Other exclusion criteria included a positive pregnancy test or an inability to give informed consent secondary to acute drug or alcohol intoxication or active mental illness. Patients with clinically significant liver disease, with ALT or AST > 3 times the upper limit of normal, or with a history of end‐stage renal disease requiring dialysis were also excluded because of their increased risk of hypoglycemia, as they have required a more conservative insulin regimen. Similarly, we excluded patients with a history of type 1 diabetes because our aspart algorithm for the ED had only been tested in a type 2 diabetes population, and we did not want to disrupt the insulin regimen of type 1 diabetes patients, usually glargine‐based or an insulin pump.

The study consisted of 2 phases. Patients randomized to INT or UC in phase 1 stayed in their respective groups through phase 2. After informed consent was obtained by the study staff, implementation of the protocol was carried out by the ED staff. ED nurses were trained in the study protocol. During phase 1, INT patients received subcutaneous aspart every 2 hours while in the ED if BG was > 200 mg/dL. Aspart dosing per protocol was limited to 2 doses. Uncommonly, when a third dose of aspart was needed, physician input was requested. Aspart dosing was weight‐ and BG level based (0.05 units/kg for BG 200‐299 mg/dL, 0.1 units/kg for BG 300‐399 mg/dL, or 0.15 units/kg for BG 400mg/dL; see Supporting Appendix Fig. 1). Regardless of BG level, the ED aspart protocol was discontinued on patient discharge home or admission to the hospital. UC patients received treatment for hyperglycemia at the discretion of their ED physicians. INT subjects who required hospital admission were transitioned to basal‐bolus insulin therapy with detemir and aspart, receiving their first dose of detemir in the ED. Detemir dosing was weight‐based (0.3 units/kg) if the patient was not on home insulin or based on the patient's home dose of insulin (same dose for detemir, unit‐for‐unit conversion from glargine to detemir, or 80% of total NPH dose). If a patient received basal insulin prior to arrival, the first dose of detemir was held until 12 hours after the last dose of NPH or until 20 hours after the last dose of glargine or detemir. We found from our preliminary study that inadvertent overlaps in long‐acting insulin were one cause of hypoglycemia. We compared differences between groups in the final ED BG level, frequency of hypoglycemia, and ED length of stay (LOS).

Patients subsequently admitted to the hospital entered phase 2. During phase 2, INT patients had detemir and premeal aspart titrated by study staff using a predefined protocol (see Supporting Appendix Fig. 2). Detemir was given once daily, 24 hours after the initial ED dose. UC patients had their diabetes managed by medical house staff teams. House staff members have been educated on the Rush inpatient insulin protocol on which the INT protocol was based. The Rush inpatient diabetes protocol is implemented via a single computerized order set in which all patients should receive mealtime insulin using aspart and a basal insulin (glargine or detemir or NPH). We compared differences between groups in the mean admission BG level, mean daily BG level, mean BG level before each meal, hospital LOS, and frequency of hypoglycemic events. Moderate hypoglycemia was defined as a BG between 50 and 69 mg/dL, and severe hypoglycemia was defined as a BG < 50 mg/dL. We also compared the frequency of BG < 60 mg/dL.

The Rush University Medical Center institutional review board approved the study. Statistical analysis was done using SPSS version 11.0. The Student t test was used to determine any significant difference in BG means between the INT and UC groups. The Fisher's exact test or the chi‐square test was used to determine any difference in proportions of hypoglycemic events between INT and UC patients.

Results

Phase 1: Emergency Department

A total of 176 patients were randomized: 87 to the INT group and 89 to the UC group. Baseline characteristics were similar between groups (Table 1). Mean initial ED BG was similar: 300 70 mg/dL for INT patients and 307 82 mg/dL for UC patients. During phase 1, all INT patients were treated with aspart every 2 hours if BG > 200 mg/dL until discharge home or admission to the hospital. They received an initial mean insulin dose of 0.08 0.04 units/kg of subscutaneous aspart. Twenty‐five percent of INT patients received a second dose of aspart, and 3% received a third dose. For UC patients, only 55% received insulin therapy. Aspart was used for all UC patients who received insulin. Of those in the UC group who received insulin, 11% received a second dose for persistent hyperglycemia; none received a third dose. The mean initial ED BG for UC patients who received insulin was 358 73 mg/dL, and they received an initial mean dose of 0.11 0.05 units/kg. UC patients who did not receive insulin had a lower mean initial ED BG, 241 35 mg/dL. The mean final ED BG was 217 71 mg/dL for the INT group and 257 89 mg/dL for the UC group (P < .01; Fig. 1). The mean ED LOS was 30 minutes longer in the INT group (P = .06; Table 1). Sixty INT patients (69%) and 61 UC patients (69%) were admitted to the hospital. Fifty‐six percent of INT patients received the first dose of detemir based on their home insulin dose, and 44% received a weight‐based dose per protocol.

Figure 1

Table 1.Patient Characteristics and Biochemical Parameters

	Intervention (n = 87)	Usual Care (n = 89)	Significance
Data are in means standard deviation or percentages.
Age (y)	55 13	55 13
Sex (% male)	48	39
BMI (kg/m2)	34 9	33 9
Ethnicity (%)
African American	58	66
Hispanic	24	19
White	15	11
Other	3	4
Duration of diabetes (y)	13 9	12 10
HA1C	10.4 2.2	9.8 2.6
Insulin treatment at home (%)	56	62
Presenting complaint/diagnosis (%)
Cardiac	20	23
Gastrointestinal	30	23
Hyperglycemia	13	18
Infection	9	11
Initial ED blood glucose (mg/dL)	300 70	307 82
Final ED blood glucose (mg/dL)	217 71	256 89	P < .01
ED length of stay (h)	5.4 1.7	4.9 1.9	P = .06
Patients treated with insulin in ED (%)	100	54
Initial dose of SQ aspart (units)	7.9 4.2	9.5 4
ED patients admitted (%)	69	69
Admission blood glucose (mg/dL)	184 70	224 93	P < .01
Treatment of hyperglycemia in hospital (%)
Detemir aspart insulin	100	7
Glargine aspart insulin	0	36
NPH aspart insulin	0	34
Oral agents	0	8
None	0	15
Hospital length of stay (days)	2.7 2.0	3.1 1.9	P = .58

Phase 2: Inpatient Setting

In phase 2, mean admission BG was significantly lower in the INT group (184 70 mg/dL) than in the UC group (223 93 mg/dL), P < .01, as a result of aspart given in the ED. The day 1 mean fasting BG for INT patients was 148 54 mg/dL, significantly lower than the day 1 mean fasting BG for UC patients: 212 81 mg/dL (P < .01). The mean fasting BG for the entire hospitalization was significantly lower for INT patients, 135 48 mg/dL, than for UC patients, 185 72 mg/dL (P < .01). During phase 2, all INT patients had detemir and aspart titrated daily per protocol. Treatment of UC patients was as follows: 78.5% with insulin, 8.2% with oral agents, and 11.5% did not receive medication for hyperglycemia. Of those in the UC group who received insulin, 36.0% were treated with lantus/aspart or detemir/aspart, 34.4% with NPH/aspart, 6.5% with lantus or detemir alone, and 1.6% with aspart alone. Overall, 76.9% of UC received basal insulin, and 70.4% received nutritional insulin. Only 47% of UC patients had insulin adjusted on a daily basis despite persistent hyperglycemia. Significant differences were also seen between INT and UC patients in mean prelunch and predinner BG levels, but not in mean bedtime BG level (Fig. 2). Mean daily BG levels for the initial 5 days of inpatient stay were significantly lower in the INT group (P < .01), except for day 5, when only 21 patients remained in the study (Fig. 3). Patient‐day weighted mean glucose was 163 39 mg/dL for INT patients versus 202 39 mg/dL for UC patients (P < .01). On admission, day 1 mean insulin total daily dose (TDD) was 0.65 0.26 units/kg for INT patients and 0.52 0.29 units/kg for UC patients. The final mean TDD was 0.75 0.35 units/kg for INT patients and 0.61 0.38 units/kg for UC patients. Mean hospital LOS was 9.6 hours shorter for INT patients (2.7 2 days) than for UC patients (3.1 1.9 days), P = .58.

Figure 2

Figure 3

Discussion

This is the first randomized trial comparing the Rush Emergency Department Hyperglycemia Intervention (REDHI) protocol with usual care for the treatment of hyperglycemia in the ED. We believe this may be the first trial to initiate subcutaneous basal insulin therapy in the ED at the time of hospital admission. Initiation of our protocol for type 2 diabetic patients with BG > 200 mg/dL resulted in lower final ED and admission BGs compared with those in the UC group. Although a higher mean initial ED BG of 358 73 mg/dL was required to prompt initiation of insulin therapy for UC patients, 3 experienced severe hypoglycemia. By following the REDHI protocol, ED nurses avoided BG < 50 mg/dL in INT patients. Our first version of the REDHI protocol dosed more insulin than our current version, and we saw excess hypoglycemia.11 With a reduced dosing formula, there was less lowering of BG, but we eliminated all BG < 60 mg/dL. There was a trend toward an increased ED LOS in INT patients compared with UC patients. This may be because of delays in the administration of insulin or the requirement for a final BG check prior to discharge from the ED for INT patients. However, we did not receive feedback from ED nursing that either factor was a significant issue.

During phase 2, we observed improved glycemic control in INT, likely due to two factors: early initiation of basal insulin and protocol driven daily titration of both basal and mealtime insulin. We achieved a mean fasting BG of 148 54 mg/dL in INT the morning after the ED dose of detemir. BG levels in both groups continued to improve each day, but since the admission BG for INT was lower, this group maintained significantly lower BG levels throughout most of the hospitalization. There were also significant differences between groups at different times of day. Basal doses for INT patients were adjusted daily per fasting BG. Scheduled mealtime doses were based on the basal dose; each mealtime aspart dose was a third of the basal detemir dose. Therefore, patients who required larger doses for certain meals may have received less aspart than needed. This may explain why fasting BG control was better than BG control later in the day.

Current Rush guidelines recommend the same insulin doses as those that our intervention used, but patients in the UC group were less likely to have insulin titrated daily. Cook et al found that clinical inertia, or failure of health care providers to initiate or intensify therapy when indicated, is a common problem among medical residents treating inpatients with insulin.15 Reasons for clinical inertia may include unawareness of inpatient glycemic targets, lack of training or confidence in titrating insulin, and concerns regarding hypoglycemia. Our study shows that this is still an operative issue, even after residents have participated in multiple small‐group educational sessions. Details of the Rush inpatient insulin protocol are also on pocket cards distributed to residents. However, fewer than half of UC patients had insulin adjusted appropriately for persistent hyperglycemia. This may be one explanation for the improved control seen in the INT group and underscores the importance of daily dose titrations based on a uniform protocol.

During phase 2, despite improved glycemic control in INT, there was no significant difference in rates of hypoglycemia between the groups. The number of patient‐stays with moderate hypoglycemia was more in the INT group than in the UC group, 12 versus 6, respectively, but not statistically different (P = .20). There was a trend toward fewer patient‐stays with severe hypoglycemia in the INT group than in the UC group, 1 versus 6, respectively (P = .11).

Other studies have described improved inpatient glycemic control without excess hypoglycemia. In the RABBIT 2 trial, institution of a glargine‐glulisine insulin protocol, TTD of 0.4‐5 units/kg, among insulin‐naive inpatients resulted in a mean fasting BG of 147 36 mg/dL and a mean hospital BG of 166 32 mg/dL, with 3% of patient‐stays having a BG < 60 mg/dL.6 In a second trial, detemir‐aspart was compared with NPH‐aspart TTD of 0.4‐5 units/kg.7 Both groups achieved a similar mean fasting BG of 146 mg/dL and a mean hospital BG of 157 mg/dL. However, the rate of hypoglycemia was higher: 29% of patient‐stays overall. In our study, we achieved a mean fasting BG of 135 48 mg/dL and a mean hospital BG of 163 40 mg/dL in the INT group, using a mean initial TTD of 0.65 0.23 units/kg. The frequency of hypoglycemia in this trial, 22% of INT patient‐stays and 20% of UC patient‐stays, was less than that in Umpierrez et al,7 despite a lower mean fasting BG in our current trial. Maynard et al found 16% of patient‐stays and 3% of patient‐days had an episode of BG < 60 mg/dL in a trial of glarginerapid‐acting insulin (0.4‐5 units/kg TTD).12 Schnipper et al found that 6.1% of patient‐days had an episode of BG < 60 mg/dL using either glargine or NPH and a rapid‐acting insulin, TTD 0.6 units/kg.13 Our hypoglycemia rates were higher; however, we defined hypoglycemia as BG < 70 mg/dL, as suggested by the ADA workgroup.14 If we use a cutoff of < 60 mg/dL for hypoglycemia, it occurred in 13.3% of patient‐stays and 4.3% of patient‐days in the INT group, comparable to that in previous studies.

Our study has several limitations. First, this was a single‐center study, and our ED protocol should be tested in other ED settings, both academic and community. Second, although there were trends toward lower rates of severe hypoglycemia in the INT group, the study was underpowered to detect possible significant differences. Third, although ED nurses implemented the study protocol, study staff closely monitored nurses to ensure adherence. Therefore, it is difficult to speculate on protocol adherence under normal circumstances. Successful implementation requires ongoing nursing and medical staff education. A fourth limitation is the absence of patients with type 1 diabetes.

In conclusion we demonstrated that weight‐based subcutaneous aspart insulin therapy begun in the ED, coupled with prompt initiation of a detemir‐aspart insulin protocol, results in rapid correction of hyperglycemia and improved inpatient glycemic control without increasing hypoglycemia. Diabetes is a common comorbidity in patients presenting to the ED that is not uniformly addressed. These patients may present with uncontrolled hyperglycemia or diabetes‐related infections, and prompt, efficacious glucose control is important. The nurse‐driven Rush ED hyperglycemia protocol ensures that hyperglycemia is safely addressed, allowing the ED physician to address more critical issues. By initiating basal insulin in the ED, our protocol allows for a prompt and smooth transition to a basal‐bolus insulin regimen for the inpatient setting.

Current consensus guidelines from the American Diabetes Association and the American Association of Clinical Endocrinologists recommend the use of insulin‐based treatment protocols for most hospitalized patients with hyperglycemia.1 For noncritically ill patients, it is recommended to target a fasting blood glucose (BG) < 140 mg/dL and a random BG < 180‐200 mg/dL, without excess hypoglycemia. Prior studies recommended using a basal‐bolus insulin protocol that specifies starting doses and parameters for dose adjustment, applied by well‐educated teams of physicians and nurses.27 We have shown that insulin detemir given as a once‐daily basal injection coupled with rapid‐acting insulin aspart with meals is an effective regimen for managing hyperglycemia in hospitalized patients with type 2 diabetes.7 We and others have shown that once‐daily basal insulin mealtime rapid‐acting insulin is significantly more effective than sliding‐scale regular insulin in the hospital setting.6, 8

The majority of patients admitted to general medical units are first evaluated in the emergency department (ED), and significant hyperglycemia is not uncommon in ED patients. However, protocols for the treatment of hyperglycemia in the ED have not been widely implemented. Ginde et al studied 160 ED patients with a history of diabetes and BG > 200 mg/dL and found that although 73% were admitted to the hospital, only 31% were treated with insulin, and only 18% had a diagnosis of diabetes charted.9 A recent survey of 152 residents and attendings in 3 academic EDs found that only 32% would give insulin for a BG > 200 mg/dL, 59% for a BG > 250 mg/dL, and 91% for a BG > 300 mg/dL to ED patients with known diabetes.10 We completed a preliminary study of a novel protocol for the administration of subcutaneous insulin aspart in the ED at Rush University in 2008.11 We found that the mean BG was significantly lowered during an ED stay, from 333 to 158 mg/dL, and that the protocol was easily adopted by ED staff, with a low rate of hypoglycemia. Historically, only 35% of hyperglycemic patients with diabetes received insulin in our ED.11 Reasons for limited ED management of hyperglycemia may include presence of more critical issues, time and resource restriction, unfamiliarity with glycemic targets, and concerns regarding hypoglycemia.9, 10

In the current study we focused on 3 questions:

• 1

  Could we further reduce the risk of hypoglycemia by modifying our original Rush ED insulin protocol? We reduced insulin aspart from 0.1 to 0.05 units/kg for BG 200‐299 mg/dL, from 0.15 to 0.1 units/kg for BG 300‐399 mg/dL, and from 0.2 to 0.15 units/kg for BG 400 mg/dL.

• 2

  Could we couple our ED insulin aspart protocol with prompt initiation of a detemir‐aspart protocol in those patients who were subsequently admitted to general medical units from the ED?

• 3

  Would the hospital length of stay, mean BG, and incidence of hypoglycemia be improved by the use of 2 back‐to‐back subcutaneous insulin protocols in a randomized clinical trial compared with the usual care provided in the ED and general medical inpatient units?

Research Design and Methods

From May 2008 through June 2009, patients presenting to the Rush University Medical Center ED with a history of type 2 diabetes and an initial point‐of‐care BG 200 mg/dL were randomized to an intervention group (INT) or to a usual care group (UC) after giving informed consent. Inclusion criteria for the study were: ages 18‐80 years, history of type 2 diabetes for at least 3 months, and prior therapy with dietary management, oral agents, or insulin. Patients were excluded if subsequently found to have diabetic ketoacidosis, hyperosmolar nonketotic syndrome, or critical illness requiring intensive care unit admission/direct surgical intervention. Other exclusion criteria included a positive pregnancy test or an inability to give informed consent secondary to acute drug or alcohol intoxication or active mental illness. Patients with clinically significant liver disease, with ALT or AST > 3 times the upper limit of normal, or with a history of end‐stage renal disease requiring dialysis were also excluded because of their increased risk of hypoglycemia, as they have required a more conservative insulin regimen. Similarly, we excluded patients with a history of type 1 diabetes because our aspart algorithm for the ED had only been tested in a type 2 diabetes population, and we did not want to disrupt the insulin regimen of type 1 diabetes patients, usually glargine‐based or an insulin pump.

The study consisted of 2 phases. Patients randomized to INT or UC in phase 1 stayed in their respective groups through phase 2. After informed consent was obtained by the study staff, implementation of the protocol was carried out by the ED staff. ED nurses were trained in the study protocol. During phase 1, INT patients received subcutaneous aspart every 2 hours while in the ED if BG was > 200 mg/dL. Aspart dosing per protocol was limited to 2 doses. Uncommonly, when a third dose of aspart was needed, physician input was requested. Aspart dosing was weight‐ and BG level based (0.05 units/kg for BG 200‐299 mg/dL, 0.1 units/kg for BG 300‐399 mg/dL, or 0.15 units/kg for BG 400mg/dL; see Supporting Appendix Fig. 1). Regardless of BG level, the ED aspart protocol was discontinued on patient discharge home or admission to the hospital. UC patients received treatment for hyperglycemia at the discretion of their ED physicians. INT subjects who required hospital admission were transitioned to basal‐bolus insulin therapy with detemir and aspart, receiving their first dose of detemir in the ED. Detemir dosing was weight‐based (0.3 units/kg) if the patient was not on home insulin or based on the patient's home dose of insulin (same dose for detemir, unit‐for‐unit conversion from glargine to detemir, or 80% of total NPH dose). If a patient received basal insulin prior to arrival, the first dose of detemir was held until 12 hours after the last dose of NPH or until 20 hours after the last dose of glargine or detemir. We found from our preliminary study that inadvertent overlaps in long‐acting insulin were one cause of hypoglycemia. We compared differences between groups in the final ED BG level, frequency of hypoglycemia, and ED length of stay (LOS).

Patients subsequently admitted to the hospital entered phase 2. During phase 2, INT patients had detemir and premeal aspart titrated by study staff using a predefined protocol (see Supporting Appendix Fig. 2). Detemir was given once daily, 24 hours after the initial ED dose. UC patients had their diabetes managed by medical house staff teams. House staff members have been educated on the Rush inpatient insulin protocol on which the INT protocol was based. The Rush inpatient diabetes protocol is implemented via a single computerized order set in which all patients should receive mealtime insulin using aspart and a basal insulin (glargine or detemir or NPH). We compared differences between groups in the mean admission BG level, mean daily BG level, mean BG level before each meal, hospital LOS, and frequency of hypoglycemic events. Moderate hypoglycemia was defined as a BG between 50 and 69 mg/dL, and severe hypoglycemia was defined as a BG < 50 mg/dL. We also compared the frequency of BG < 60 mg/dL.

The Rush University Medical Center institutional review board approved the study. Statistical analysis was done using SPSS version 11.0. The Student t test was used to determine any significant difference in BG means between the INT and UC groups. The Fisher's exact test or the chi‐square test was used to determine any difference in proportions of hypoglycemic events between INT and UC patients.

Results

Phase 1: Emergency Department

A total of 176 patients were randomized: 87 to the INT group and 89 to the UC group. Baseline characteristics were similar between groups (Table 1). Mean initial ED BG was similar: 300 70 mg/dL for INT patients and 307 82 mg/dL for UC patients. During phase 1, all INT patients were treated with aspart every 2 hours if BG > 200 mg/dL until discharge home or admission to the hospital. They received an initial mean insulin dose of 0.08 0.04 units/kg of subscutaneous aspart. Twenty‐five percent of INT patients received a second dose of aspart, and 3% received a third dose. For UC patients, only 55% received insulin therapy. Aspart was used for all UC patients who received insulin. Of those in the UC group who received insulin, 11% received a second dose for persistent hyperglycemia; none received a third dose. The mean initial ED BG for UC patients who received insulin was 358 73 mg/dL, and they received an initial mean dose of 0.11 0.05 units/kg. UC patients who did not receive insulin had a lower mean initial ED BG, 241 35 mg/dL. The mean final ED BG was 217 71 mg/dL for the INT group and 257 89 mg/dL for the UC group (P < .01; Fig. 1). The mean ED LOS was 30 minutes longer in the INT group (P = .06; Table 1). Sixty INT patients (69%) and 61 UC patients (69%) were admitted to the hospital. Fifty‐six percent of INT patients received the first dose of detemir based on their home insulin dose, and 44% received a weight‐based dose per protocol.

Figure 1

Table 1.Patient Characteristics and Biochemical Parameters

	Intervention (n = 87)	Usual Care (n = 89)	Significance
Data are in means standard deviation or percentages.
Age (y)	55 13	55 13
Sex (% male)	48	39
BMI (kg/m2)	34 9	33 9
Ethnicity (%)
African American	58	66
Hispanic	24	19
White	15	11
Other	3	4
Duration of diabetes (y)	13 9	12 10
HA1C	10.4 2.2	9.8 2.6
Insulin treatment at home (%)	56	62
Presenting complaint/diagnosis (%)
Cardiac	20	23
Gastrointestinal	30	23
Hyperglycemia	13	18
Infection	9	11
Initial ED blood glucose (mg/dL)	300 70	307 82
Final ED blood glucose (mg/dL)	217 71	256 89	P < .01
ED length of stay (h)	5.4 1.7	4.9 1.9	P = .06
Patients treated with insulin in ED (%)	100	54
Initial dose of SQ aspart (units)	7.9 4.2	9.5 4
ED patients admitted (%)	69	69
Admission blood glucose (mg/dL)	184 70	224 93	P < .01
Treatment of hyperglycemia in hospital (%)
Detemir aspart insulin	100	7
Glargine aspart insulin	0	36
NPH aspart insulin	0	34
Oral agents	0	8
None	0	15
Hospital length of stay (days)	2.7 2.0	3.1 1.9	P = .58

Phase 2: Inpatient Setting

In phase 2, mean admission BG was significantly lower in the INT group (184 70 mg/dL) than in the UC group (223 93 mg/dL), P < .01, as a result of aspart given in the ED. The day 1 mean fasting BG for INT patients was 148 54 mg/dL, significantly lower than the day 1 mean fasting BG for UC patients: 212 81 mg/dL (P < .01). The mean fasting BG for the entire hospitalization was significantly lower for INT patients, 135 48 mg/dL, than for UC patients, 185 72 mg/dL (P < .01). During phase 2, all INT patients had detemir and aspart titrated daily per protocol. Treatment of UC patients was as follows: 78.5% with insulin, 8.2% with oral agents, and 11.5% did not receive medication for hyperglycemia. Of those in the UC group who received insulin, 36.0% were treated with lantus/aspart or detemir/aspart, 34.4% with NPH/aspart, 6.5% with lantus or detemir alone, and 1.6% with aspart alone. Overall, 76.9% of UC received basal insulin, and 70.4% received nutritional insulin. Only 47% of UC patients had insulin adjusted on a daily basis despite persistent hyperglycemia. Significant differences were also seen between INT and UC patients in mean prelunch and predinner BG levels, but not in mean bedtime BG level (Fig. 2). Mean daily BG levels for the initial 5 days of inpatient stay were significantly lower in the INT group (P < .01), except for day 5, when only 21 patients remained in the study (Fig. 3). Patient‐day weighted mean glucose was 163 39 mg/dL for INT patients versus 202 39 mg/dL for UC patients (P < .01). On admission, day 1 mean insulin total daily dose (TDD) was 0.65 0.26 units/kg for INT patients and 0.52 0.29 units/kg for UC patients. The final mean TDD was 0.75 0.35 units/kg for INT patients and 0.61 0.38 units/kg for UC patients. Mean hospital LOS was 9.6 hours shorter for INT patients (2.7 2 days) than for UC patients (3.1 1.9 days), P = .58.

Figure 2

Figure 3

Discussion

This is the first randomized trial comparing the Rush Emergency Department Hyperglycemia Intervention (REDHI) protocol with usual care for the treatment of hyperglycemia in the ED. We believe this may be the first trial to initiate subcutaneous basal insulin therapy in the ED at the time of hospital admission. Initiation of our protocol for type 2 diabetic patients with BG > 200 mg/dL resulted in lower final ED and admission BGs compared with those in the UC group. Although a higher mean initial ED BG of 358 73 mg/dL was required to prompt initiation of insulin therapy for UC patients, 3 experienced severe hypoglycemia. By following the REDHI protocol, ED nurses avoided BG < 50 mg/dL in INT patients. Our first version of the REDHI protocol dosed more insulin than our current version, and we saw excess hypoglycemia.11 With a reduced dosing formula, there was less lowering of BG, but we eliminated all BG < 60 mg/dL. There was a trend toward an increased ED LOS in INT patients compared with UC patients. This may be because of delays in the administration of insulin or the requirement for a final BG check prior to discharge from the ED for INT patients. However, we did not receive feedback from ED nursing that either factor was a significant issue.

During phase 2, we observed improved glycemic control in INT, likely due to two factors: early initiation of basal insulin and protocol driven daily titration of both basal and mealtime insulin. We achieved a mean fasting BG of 148 54 mg/dL in INT the morning after the ED dose of detemir. BG levels in both groups continued to improve each day, but since the admission BG for INT was lower, this group maintained significantly lower BG levels throughout most of the hospitalization. There were also significant differences between groups at different times of day. Basal doses for INT patients were adjusted daily per fasting BG. Scheduled mealtime doses were based on the basal dose; each mealtime aspart dose was a third of the basal detemir dose. Therefore, patients who required larger doses for certain meals may have received less aspart than needed. This may explain why fasting BG control was better than BG control later in the day.

Current Rush guidelines recommend the same insulin doses as those that our intervention used, but patients in the UC group were less likely to have insulin titrated daily. Cook et al found that clinical inertia, or failure of health care providers to initiate or intensify therapy when indicated, is a common problem among medical residents treating inpatients with insulin.15 Reasons for clinical inertia may include unawareness of inpatient glycemic targets, lack of training or confidence in titrating insulin, and concerns regarding hypoglycemia. Our study shows that this is still an operative issue, even after residents have participated in multiple small‐group educational sessions. Details of the Rush inpatient insulin protocol are also on pocket cards distributed to residents. However, fewer than half of UC patients had insulin adjusted appropriately for persistent hyperglycemia. This may be one explanation for the improved control seen in the INT group and underscores the importance of daily dose titrations based on a uniform protocol.

During phase 2, despite improved glycemic control in INT, there was no significant difference in rates of hypoglycemia between the groups. The number of patient‐stays with moderate hypoglycemia was more in the INT group than in the UC group, 12 versus 6, respectively, but not statistically different (P = .20). There was a trend toward fewer patient‐stays with severe hypoglycemia in the INT group than in the UC group, 1 versus 6, respectively (P = .11).

Other studies have described improved inpatient glycemic control without excess hypoglycemia. In the RABBIT 2 trial, institution of a glargine‐glulisine insulin protocol, TTD of 0.4‐5 units/kg, among insulin‐naive inpatients resulted in a mean fasting BG of 147 36 mg/dL and a mean hospital BG of 166 32 mg/dL, with 3% of patient‐stays having a BG < 60 mg/dL.6 In a second trial, detemir‐aspart was compared with NPH‐aspart TTD of 0.4‐5 units/kg.7 Both groups achieved a similar mean fasting BG of 146 mg/dL and a mean hospital BG of 157 mg/dL. However, the rate of hypoglycemia was higher: 29% of patient‐stays overall. In our study, we achieved a mean fasting BG of 135 48 mg/dL and a mean hospital BG of 163 40 mg/dL in the INT group, using a mean initial TTD of 0.65 0.23 units/kg. The frequency of hypoglycemia in this trial, 22% of INT patient‐stays and 20% of UC patient‐stays, was less than that in Umpierrez et al,7 despite a lower mean fasting BG in our current trial. Maynard et al found 16% of patient‐stays and 3% of patient‐days had an episode of BG < 60 mg/dL in a trial of glarginerapid‐acting insulin (0.4‐5 units/kg TTD).12 Schnipper et al found that 6.1% of patient‐days had an episode of BG < 60 mg/dL using either glargine or NPH and a rapid‐acting insulin, TTD 0.6 units/kg.13 Our hypoglycemia rates were higher; however, we defined hypoglycemia as BG < 70 mg/dL, as suggested by the ADA workgroup.14 If we use a cutoff of < 60 mg/dL for hypoglycemia, it occurred in 13.3% of patient‐stays and 4.3% of patient‐days in the INT group, comparable to that in previous studies.

Our study has several limitations. First, this was a single‐center study, and our ED protocol should be tested in other ED settings, both academic and community. Second, although there were trends toward lower rates of severe hypoglycemia in the INT group, the study was underpowered to detect possible significant differences. Third, although ED nurses implemented the study protocol, study staff closely monitored nurses to ensure adherence. Therefore, it is difficult to speculate on protocol adherence under normal circumstances. Successful implementation requires ongoing nursing and medical staff education. A fourth limitation is the absence of patients with type 1 diabetes.

In conclusion we demonstrated that weight‐based subcutaneous aspart insulin therapy begun in the ED, coupled with prompt initiation of a detemir‐aspart insulin protocol, results in rapid correction of hyperglycemia and improved inpatient glycemic control without increasing hypoglycemia. Diabetes is a common comorbidity in patients presenting to the ED that is not uniformly addressed. These patients may present with uncontrolled hyperglycemia or diabetes‐related infections, and prompt, efficacious glucose control is important. The nurse‐driven Rush ED hyperglycemia protocol ensures that hyperglycemia is safely addressed, allowing the ED physician to address more critical issues. By initiating basal insulin in the ED, our protocol allows for a prompt and smooth transition to a basal‐bolus insulin regimen for the inpatient setting.

Current consensus guidelines from the American Diabetes Association and the American Association of Clinical Endocrinologists recommend the use of insulin‐based treatment protocols for most hospitalized patients with hyperglycemia.1 For noncritically ill patients, it is recommended to target a fasting blood glucose (BG) < 140 mg/dL and a random BG < 180‐200 mg/dL, without excess hypoglycemia. Prior studies recommended using a basal‐bolus insulin protocol that specifies starting doses and parameters for dose adjustment, applied by well‐educated teams of physicians and nurses.27 We have shown that insulin detemir given as a once‐daily basal injection coupled with rapid‐acting insulin aspart with meals is an effective regimen for managing hyperglycemia in hospitalized patients with type 2 diabetes.7 We and others have shown that once‐daily basal insulin mealtime rapid‐acting insulin is significantly more effective than sliding‐scale regular insulin in the hospital setting.6, 8

The majority of patients admitted to general medical units are first evaluated in the emergency department (ED), and significant hyperglycemia is not uncommon in ED patients. However, protocols for the treatment of hyperglycemia in the ED have not been widely implemented. Ginde et al studied 160 ED patients with a history of diabetes and BG > 200 mg/dL and found that although 73% were admitted to the hospital, only 31% were treated with insulin, and only 18% had a diagnosis of diabetes charted.9 A recent survey of 152 residents and attendings in 3 academic EDs found that only 32% would give insulin for a BG > 200 mg/dL, 59% for a BG > 250 mg/dL, and 91% for a BG > 300 mg/dL to ED patients with known diabetes.10 We completed a preliminary study of a novel protocol for the administration of subcutaneous insulin aspart in the ED at Rush University in 2008.11 We found that the mean BG was significantly lowered during an ED stay, from 333 to 158 mg/dL, and that the protocol was easily adopted by ED staff, with a low rate of hypoglycemia. Historically, only 35% of hyperglycemic patients with diabetes received insulin in our ED.11 Reasons for limited ED management of hyperglycemia may include presence of more critical issues, time and resource restriction, unfamiliarity with glycemic targets, and concerns regarding hypoglycemia.9, 10

In the current study we focused on 3 questions:

• 1

  Could we further reduce the risk of hypoglycemia by modifying our original Rush ED insulin protocol? We reduced insulin aspart from 0.1 to 0.05 units/kg for BG 200‐299 mg/dL, from 0.15 to 0.1 units/kg for BG 300‐399 mg/dL, and from 0.2 to 0.15 units/kg for BG 400 mg/dL.

• 2

  Could we couple our ED insulin aspart protocol with prompt initiation of a detemir‐aspart protocol in those patients who were subsequently admitted to general medical units from the ED?

• 3

  Would the hospital length of stay, mean BG, and incidence of hypoglycemia be improved by the use of 2 back‐to‐back subcutaneous insulin protocols in a randomized clinical trial compared with the usual care provided in the ED and general medical inpatient units?

Research Design and Methods

From May 2008 through June 2009, patients presenting to the Rush University Medical Center ED with a history of type 2 diabetes and an initial point‐of‐care BG 200 mg/dL were randomized to an intervention group (INT) or to a usual care group (UC) after giving informed consent. Inclusion criteria for the study were: ages 18‐80 years, history of type 2 diabetes for at least 3 months, and prior therapy with dietary management, oral agents, or insulin. Patients were excluded if subsequently found to have diabetic ketoacidosis, hyperosmolar nonketotic syndrome, or critical illness requiring intensive care unit admission/direct surgical intervention. Other exclusion criteria included a positive pregnancy test or an inability to give informed consent secondary to acute drug or alcohol intoxication or active mental illness. Patients with clinically significant liver disease, with ALT or AST > 3 times the upper limit of normal, or with a history of end‐stage renal disease requiring dialysis were also excluded because of their increased risk of hypoglycemia, as they have required a more conservative insulin regimen. Similarly, we excluded patients with a history of type 1 diabetes because our aspart algorithm for the ED had only been tested in a type 2 diabetes population, and we did not want to disrupt the insulin regimen of type 1 diabetes patients, usually glargine‐based or an insulin pump.

The study consisted of 2 phases. Patients randomized to INT or UC in phase 1 stayed in their respective groups through phase 2. After informed consent was obtained by the study staff, implementation of the protocol was carried out by the ED staff. ED nurses were trained in the study protocol. During phase 1, INT patients received subcutaneous aspart every 2 hours while in the ED if BG was > 200 mg/dL. Aspart dosing per protocol was limited to 2 doses. Uncommonly, when a third dose of aspart was needed, physician input was requested. Aspart dosing was weight‐ and BG level based (0.05 units/kg for BG 200‐299 mg/dL, 0.1 units/kg for BG 300‐399 mg/dL, or 0.15 units/kg for BG 400mg/dL; see Supporting Appendix Fig. 1). Regardless of BG level, the ED aspart protocol was discontinued on patient discharge home or admission to the hospital. UC patients received treatment for hyperglycemia at the discretion of their ED physicians. INT subjects who required hospital admission were transitioned to basal‐bolus insulin therapy with detemir and aspart, receiving their first dose of detemir in the ED. Detemir dosing was weight‐based (0.3 units/kg) if the patient was not on home insulin or based on the patient's home dose of insulin (same dose for detemir, unit‐for‐unit conversion from glargine to detemir, or 80% of total NPH dose). If a patient received basal insulin prior to arrival, the first dose of detemir was held until 12 hours after the last dose of NPH or until 20 hours after the last dose of glargine or detemir. We found from our preliminary study that inadvertent overlaps in long‐acting insulin were one cause of hypoglycemia. We compared differences between groups in the final ED BG level, frequency of hypoglycemia, and ED length of stay (LOS).

Patients subsequently admitted to the hospital entered phase 2. During phase 2, INT patients had detemir and premeal aspart titrated by study staff using a predefined protocol (see Supporting Appendix Fig. 2). Detemir was given once daily, 24 hours after the initial ED dose. UC patients had their diabetes managed by medical house staff teams. House staff members have been educated on the Rush inpatient insulin protocol on which the INT protocol was based. The Rush inpatient diabetes protocol is implemented via a single computerized order set in which all patients should receive mealtime insulin using aspart and a basal insulin (glargine or detemir or NPH). We compared differences between groups in the mean admission BG level, mean daily BG level, mean BG level before each meal, hospital LOS, and frequency of hypoglycemic events. Moderate hypoglycemia was defined as a BG between 50 and 69 mg/dL, and severe hypoglycemia was defined as a BG < 50 mg/dL. We also compared the frequency of BG < 60 mg/dL.

The Rush University Medical Center institutional review board approved the study. Statistical analysis was done using SPSS version 11.0. The Student t test was used to determine any significant difference in BG means between the INT and UC groups. The Fisher's exact test or the chi‐square test was used to determine any difference in proportions of hypoglycemic events between INT and UC patients.

Results

Phase 1: Emergency Department

A total of 176 patients were randomized: 87 to the INT group and 89 to the UC group. Baseline characteristics were similar between groups (Table 1). Mean initial ED BG was similar: 300 70 mg/dL for INT patients and 307 82 mg/dL for UC patients. During phase 1, all INT patients were treated with aspart every 2 hours if BG > 200 mg/dL until discharge home or admission to the hospital. They received an initial mean insulin dose of 0.08 0.04 units/kg of subscutaneous aspart. Twenty‐five percent of INT patients received a second dose of aspart, and 3% received a third dose. For UC patients, only 55% received insulin therapy. Aspart was used for all UC patients who received insulin. Of those in the UC group who received insulin, 11% received a second dose for persistent hyperglycemia; none received a third dose. The mean initial ED BG for UC patients who received insulin was 358 73 mg/dL, and they received an initial mean dose of 0.11 0.05 units/kg. UC patients who did not receive insulin had a lower mean initial ED BG, 241 35 mg/dL. The mean final ED BG was 217 71 mg/dL for the INT group and 257 89 mg/dL for the UC group (P < .01; Fig. 1). The mean ED LOS was 30 minutes longer in the INT group (P = .06; Table 1). Sixty INT patients (69%) and 61 UC patients (69%) were admitted to the hospital. Fifty‐six percent of INT patients received the first dose of detemir based on their home insulin dose, and 44% received a weight‐based dose per protocol.

Figure 1

Table 1.Patient Characteristics and Biochemical Parameters

	Intervention (n = 87)	Usual Care (n = 89)	Significance
Data are in means standard deviation or percentages.
Age (y)	55 13	55 13
Sex (% male)	48	39
BMI (kg/m2)	34 9	33 9
Ethnicity (%)
African American	58	66
Hispanic	24	19
White	15	11
Other	3	4
Duration of diabetes (y)	13 9	12 10
HA1C	10.4 2.2	9.8 2.6
Insulin treatment at home (%)	56	62
Presenting complaint/diagnosis (%)
Cardiac	20	23
Gastrointestinal	30	23
Hyperglycemia	13	18
Infection	9	11
Initial ED blood glucose (mg/dL)	300 70	307 82
Final ED blood glucose (mg/dL)	217 71	256 89	P < .01
ED length of stay (h)	5.4 1.7	4.9 1.9	P = .06
Patients treated with insulin in ED (%)	100	54
Initial dose of SQ aspart (units)	7.9 4.2	9.5 4
ED patients admitted (%)	69	69
Admission blood glucose (mg/dL)	184 70	224 93	P < .01
Treatment of hyperglycemia in hospital (%)
Detemir aspart insulin	100	7
Glargine aspart insulin	0	36
NPH aspart insulin	0	34
Oral agents	0	8
None	0	15
Hospital length of stay (days)	2.7 2.0	3.1 1.9	P = .58

Phase 2: Inpatient Setting

In phase 2, mean admission BG was significantly lower in the INT group (184 70 mg/dL) than in the UC group (223 93 mg/dL), P < .01, as a result of aspart given in the ED. The day 1 mean fasting BG for INT patients was 148 54 mg/dL, significantly lower than the day 1 mean fasting BG for UC patients: 212 81 mg/dL (P < .01). The mean fasting BG for the entire hospitalization was significantly lower for INT patients, 135 48 mg/dL, than for UC patients, 185 72 mg/dL (P < .01). During phase 2, all INT patients had detemir and aspart titrated daily per protocol. Treatment of UC patients was as follows: 78.5% with insulin, 8.2% with oral agents, and 11.5% did not receive medication for hyperglycemia. Of those in the UC group who received insulin, 36.0% were treated with lantus/aspart or detemir/aspart, 34.4% with NPH/aspart, 6.5% with lantus or detemir alone, and 1.6% with aspart alone. Overall, 76.9% of UC received basal insulin, and 70.4% received nutritional insulin. Only 47% of UC patients had insulin adjusted on a daily basis despite persistent hyperglycemia. Significant differences were also seen between INT and UC patients in mean prelunch and predinner BG levels, but not in mean bedtime BG level (Fig. 2). Mean daily BG levels for the initial 5 days of inpatient stay were significantly lower in the INT group (P < .01), except for day 5, when only 21 patients remained in the study (Fig. 3). Patient‐day weighted mean glucose was 163 39 mg/dL for INT patients versus 202 39 mg/dL for UC patients (P < .01). On admission, day 1 mean insulin total daily dose (TDD) was 0.65 0.26 units/kg for INT patients and 0.52 0.29 units/kg for UC patients. The final mean TDD was 0.75 0.35 units/kg for INT patients and 0.61 0.38 units/kg for UC patients. Mean hospital LOS was 9.6 hours shorter for INT patients (2.7 2 days) than for UC patients (3.1 1.9 days), P = .58.

Figure 2

Figure 3

Discussion

This is the first randomized trial comparing the Rush Emergency Department Hyperglycemia Intervention (REDHI) protocol with usual care for the treatment of hyperglycemia in the ED. We believe this may be the first trial to initiate subcutaneous basal insulin therapy in the ED at the time of hospital admission. Initiation of our protocol for type 2 diabetic patients with BG > 200 mg/dL resulted in lower final ED and admission BGs compared with those in the UC group. Although a higher mean initial ED BG of 358 73 mg/dL was required to prompt initiation of insulin therapy for UC patients, 3 experienced severe hypoglycemia. By following the REDHI protocol, ED nurses avoided BG < 50 mg/dL in INT patients. Our first version of the REDHI protocol dosed more insulin than our current version, and we saw excess hypoglycemia.11 With a reduced dosing formula, there was less lowering of BG, but we eliminated all BG < 60 mg/dL. There was a trend toward an increased ED LOS in INT patients compared with UC patients. This may be because of delays in the administration of insulin or the requirement for a final BG check prior to discharge from the ED for INT patients. However, we did not receive feedback from ED nursing that either factor was a significant issue.

During phase 2, we observed improved glycemic control in INT, likely due to two factors: early initiation of basal insulin and protocol driven daily titration of both basal and mealtime insulin. We achieved a mean fasting BG of 148 54 mg/dL in INT the morning after the ED dose of detemir. BG levels in both groups continued to improve each day, but since the admission BG for INT was lower, this group maintained significantly lower BG levels throughout most of the hospitalization. There were also significant differences between groups at different times of day. Basal doses for INT patients were adjusted daily per fasting BG. Scheduled mealtime doses were based on the basal dose; each mealtime aspart dose was a third of the basal detemir dose. Therefore, patients who required larger doses for certain meals may have received less aspart than needed. This may explain why fasting BG control was better than BG control later in the day.

Current Rush guidelines recommend the same insulin doses as those that our intervention used, but patients in the UC group were less likely to have insulin titrated daily. Cook et al found that clinical inertia, or failure of health care providers to initiate or intensify therapy when indicated, is a common problem among medical residents treating inpatients with insulin.15 Reasons for clinical inertia may include unawareness of inpatient glycemic targets, lack of training or confidence in titrating insulin, and concerns regarding hypoglycemia. Our study shows that this is still an operative issue, even after residents have participated in multiple small‐group educational sessions. Details of the Rush inpatient insulin protocol are also on pocket cards distributed to residents. However, fewer than half of UC patients had insulin adjusted appropriately for persistent hyperglycemia. This may be one explanation for the improved control seen in the INT group and underscores the importance of daily dose titrations based on a uniform protocol.

During phase 2, despite improved glycemic control in INT, there was no significant difference in rates of hypoglycemia between the groups. The number of patient‐stays with moderate hypoglycemia was more in the INT group than in the UC group, 12 versus 6, respectively, but not statistically different (P = .20). There was a trend toward fewer patient‐stays with severe hypoglycemia in the INT group than in the UC group, 1 versus 6, respectively (P = .11).

Other studies have described improved inpatient glycemic control without excess hypoglycemia. In the RABBIT 2 trial, institution of a glargine‐glulisine insulin protocol, TTD of 0.4‐5 units/kg, among insulin‐naive inpatients resulted in a mean fasting BG of 147 36 mg/dL and a mean hospital BG of 166 32 mg/dL, with 3% of patient‐stays having a BG < 60 mg/dL.6 In a second trial, detemir‐aspart was compared with NPH‐aspart TTD of 0.4‐5 units/kg.7 Both groups achieved a similar mean fasting BG of 146 mg/dL and a mean hospital BG of 157 mg/dL. However, the rate of hypoglycemia was higher: 29% of patient‐stays overall. In our study, we achieved a mean fasting BG of 135 48 mg/dL and a mean hospital BG of 163 40 mg/dL in the INT group, using a mean initial TTD of 0.65 0.23 units/kg. The frequency of hypoglycemia in this trial, 22% of INT patient‐stays and 20% of UC patient‐stays, was less than that in Umpierrez et al,7 despite a lower mean fasting BG in our current trial. Maynard et al found 16% of patient‐stays and 3% of patient‐days had an episode of BG < 60 mg/dL in a trial of glarginerapid‐acting insulin (0.4‐5 units/kg TTD).12 Schnipper et al found that 6.1% of patient‐days had an episode of BG < 60 mg/dL using either glargine or NPH and a rapid‐acting insulin, TTD 0.6 units/kg.13 Our hypoglycemia rates were higher; however, we defined hypoglycemia as BG < 70 mg/dL, as suggested by the ADA workgroup.14 If we use a cutoff of < 60 mg/dL for hypoglycemia, it occurred in 13.3% of patient‐stays and 4.3% of patient‐days in the INT group, comparable to that in previous studies.

Our study has several limitations. First, this was a single‐center study, and our ED protocol should be tested in other ED settings, both academic and community. Second, although there were trends toward lower rates of severe hypoglycemia in the INT group, the study was underpowered to detect possible significant differences. Third, although ED nurses implemented the study protocol, study staff closely monitored nurses to ensure adherence. Therefore, it is difficult to speculate on protocol adherence under normal circumstances. Successful implementation requires ongoing nursing and medical staff education. A fourth limitation is the absence of patients with type 1 diabetes.

In conclusion we demonstrated that weight‐based subcutaneous aspart insulin therapy begun in the ED, coupled with prompt initiation of a detemir‐aspart insulin protocol, results in rapid correction of hyperglycemia and improved inpatient glycemic control without increasing hypoglycemia. Diabetes is a common comorbidity in patients presenting to the ED that is not uniformly addressed. These patients may present with uncontrolled hyperglycemia or diabetes‐related infections, and prompt, efficacious glucose control is important. The nurse‐driven Rush ED hyperglycemia protocol ensures that hyperglycemia is safely addressed, allowing the ED physician to address more critical issues. By initiating basal insulin in the ED, our protocol allows for a prompt and smooth transition to a basal‐bolus insulin regimen for the inpatient setting.

The clinical outcome of patients with pulmonary embolism (PE) has a wide spectrum, ranging from a benign course, with early resolution of symptoms, to death. Prognostic stratification of patients with PE is therefore essential to guide therapy. Currently, this stratification is primarily based on blood pressure on admission. Hemodynamic instability or shock carries a bad prognosis, and is associated with a 3‐fold to 7‐fold increase in mortality.1 Thrombolytic therapy is warranted in these cases.2 There is a subset of patients that, despite normal blood pressure on presentation, can subsequently suffer hemodynamic deterioration and death.3

Echocardiography has been proposed as a useful tool to identify a subset of patients at a higher risk. It has been suggested that the finding of right ventricle dysfunction (RVD) in echocardiography predicts a worse prognosis, even in patients with normal blood pressure on presentation.4 Also, computed tomography pulmonary angiography (CTPA) is increasingly used as the main test to diagnose PE. This technique can also assess the degree of pulmonary arterial obstruction and the presence of signs suggestive of RVD. It has been suggested that CTPA findings can predict outcome in PE.5

Electrocardiography (ECG) has a widespread use and is routinely carried out and interpreted in the emergency room. Some ECG findings in patients with PE are suggestive of right ventricular strain and might therefore be used to prioritize cases that require closer monitoring.6

This study was designed with the objectives of assessing the correlation of standard ECG findings with echocardiography and CTPA signs of RVD, with the anatomic severity of the pulmonary vascular obstruction, and with the degree of hypoxemia, in hemodynamically stable patients with PE. Secondary objectives were determining if ECG could accurately predict RVD, and determining the interobserver agreement for the ECG score.

Patients and Methods

Results

A total of 103 patients were included in the study (mean age = 69 15 years). One patient died from fatal PE soon after the diagnosis, before echocardiography could be performed. During the 6‐month follow‐up, one patient died from septic shock and another suffered a thromboembolic recurrence (deep vein thrombosis). Echocardiography‐assessed RVD was found in 25 cases (24.5%) and isolated pulmonary hypertension in 20 (19.6%). Median value of the vascular obstruction index was 43.8% (25‐65). Thirty‐three patients (32%) had signs of RVD in CTPA.

Median value of ECG score was 2.5 (1‐6). The level of interobserver agreement regarding ECG score was substantial ( = 0.80). Table 2 shows correlation between ECG score and echocardiography, arterial blood gases, and CTPA parameters. ECG correlated significantly with all the variables measured. Correlation was higher for the vascular obstruction index and the RV/LV ratio, measured with echocardiography. ECG scores were significantly different for groups with varying degrees of severity of PE. Table 3 shows the ECG scores for patients grouped according to different criteria of severity of the disease.

Figure 1 shows the ROC curve for the ECG score and the presence of echocardiography‐detected RVD. Area under the ROC curve was 0.82 (95% confidence interval [CI]: 0.72‐0.89). A cut‐off point of 6 for the ECG score showed the highest accuracy for detecting RVD; sensitivity = 76.5% (95% CI: 50.1‐93.0), specificity = 88.6% (95% CI: 78.7‐94.9), positive likelihood ratio (+LR) = 6.69, negative likelihood ratio (LR) = 0.37. If the ECG score was to be used as a screening method to exclude RVD, a cut‐off point of 0 showed sensitivity = 94.1% (95% CI: 71.2‐99.0), specificity = 27.1% (95% CI: 17.2‐39.1), +LR = 1.29, LR = 0.22. If the ECG score was considered for confirming RVD, the more useful value was 9; sensitivity = 58.8% (95% CI: 33.0‐81.5), specificity = 92.0% (95% CI: 84.1‐97.6), +LR = 8.23, LR = 0.44. Area under the ROC curve to predict CTPA‐detected RVD was slightly lower than for echocardiography: 0.74 (95% CI: 0.61‐0.84).

Figure 1

Discussion

This study shows that a simple, easy‐to‐use ECG scoring system correlates well with the severity of PE, assessed by different methods, in hemodynamically stable patients. Interobserver agreement regarding the scoring system was substantial. Several previous studies have found that ECG signs of RV strain correlate with the presence of RVD.6, 1316 However, some of these studies are limited by a retrospective design that could lead to selection bias or by the inclusion of relatively few (eg, <100) cases.6, 13, 14 Moreover, only a few of these articles have specifically examined patients with normal blood pressure.6, 15 Although our study is still small, one of its strengths is that it prospectively recruited a population of consecutive, normotensive patients and both echocardiography and CTPA were systematically performed. Classifying cases with normal blood pressure by severity is particularly relevant, because of the uncertainty about best treatment when hemodynamic instability is not present. These patients who are not at high risk for PE can be further stratified according to the presence of RVD and/or myocardial injury into intermediate‐risk and low‐risk PE.17 Early hospital discharge or initial outpatient treatment might be considered in low‐risk patients.17, 18 However, echocardiography can be difficult to perform in the emergency setting. Also, the presence of RVD is not easy to predict on clinical grounds. Our study suggests that a standard 12‐lead ECG can be useful to prioritize patients for more exhaustive monitoring or additional risk stratifying tests. An ECG score of 0 would exclude RVD with enough sensitivity to avoid getting an echocardiogram in these cases. On the other hand, a score 9 would suggest the possibility that RVD was present, and an echocardiography ought to be considered in these patients.

Previous studies regarding correlation between ECG and anatomic extension of the pulmonary vascular obstruction in PE have revealed conflicting data. Iles et al. found that ECG score identified those patients with the greatest percentage of perfusion defect on ventilation/perfusion scan.19 However, Kanbay et al. did not find significant differences in ECG scores for patients with pulmonary vascular obstruction scores <50% or 50%, also assessed with ventilation/perfusion scan.20 Subramaniam et al. used CTPA to assess clot burden score, and they did not see a significant association between ECG score and the anatomic severity of PE.21 This study found a significant association between higher ECG score and a more severe vascular obstruction index. This finding might have potential clinical implications, because some reports suggest that a higher CTPA‐measured clot load score can be associated with poorer outcome in PE.22 However, the utility of CTPA scores of vascular obstruction severity as a marker of clinical evolution is controversial, and several other authors have found that these scores are poor predictors of mortality or adverse outcome events.23, 24 Therefore, the positive correlation between ECG and the vascular obstruction index in our study is interesting mainly in the sense that it independently supports the validity of the ECG score as a marker of a potentially more severe PE.

In this study, we have not systematically excluded patients with comorbidities (eg, chronic obstructive pulmonary disease) that could account for RVD or PH. Our intention was to reflect a real‐life clinical scenario. When a patient attends the emergency room with PE, clinical decisions must usually be taken without complete information on comorbidities that may contribute to the existence of RV overload. Also, these subjects have a reduced cardiorespiratory reserve, and the PE might impose an unbearable strain to the RV, so risk stratification is especially important. Therefore, we felt that the study should include all patients presenting with PE, so that the results can be extrapolated to usual clinical practice.

Our results suggest that, by selecting the appropriate cut‐off value, the ECG score can be used with adequate sensitivity and specificity to exclude and confirm, respectively, RVD in hemodynamically stable PE patients. This finding implies that ECG may be useful to design management strategies in these cases, selecting patients that may benefit from further tests.

There are some limitations for the present study. Echocardiography was not immediately performed after the ECG. Thus, we might have missed some patients who presented with transient RVD, which resolved between tests. It is likely that the correlation between ECG and echocardiography would have improved if this delay had been avoided. Also, we did not use a multidetector scanner for performing CTPAs. Consequently, small emboli might have been missed and the clot score might have been underestimated. It is plausible that this may have reduced to some extent the correlation between ECG and CTPA. Also, this a single‐center study, with relatively few patients studied. Finally, due to the low incidence of adverse events in our patients, our study is underpowered to make conclusions about the independent prognostic value of ECG for predicting adverse events.

In conclusion, an easy‐to‐use ECG score correlates significantly with the severity of PE, as assessed with echocardiography, CTPA, and arterial blood gas analysis, in normotensive patients. It can be used to predict with acceptable sensitivity and specificity the presence of RVD in these cases and, accordingly, it has potential value in risk‐stratification strategies. Larger, multicenter studies, should confirm these results before they can be applied to clinical practice.

The clinical outcome of patients with pulmonary embolism (PE) has a wide spectrum, ranging from a benign course, with early resolution of symptoms, to death. Prognostic stratification of patients with PE is therefore essential to guide therapy. Currently, this stratification is primarily based on blood pressure on admission. Hemodynamic instability or shock carries a bad prognosis, and is associated with a 3‐fold to 7‐fold increase in mortality.1 Thrombolytic therapy is warranted in these cases.2 There is a subset of patients that, despite normal blood pressure on presentation, can subsequently suffer hemodynamic deterioration and death.3

Echocardiography has been proposed as a useful tool to identify a subset of patients at a higher risk. It has been suggested that the finding of right ventricle dysfunction (RVD) in echocardiography predicts a worse prognosis, even in patients with normal blood pressure on presentation.4 Also, computed tomography pulmonary angiography (CTPA) is increasingly used as the main test to diagnose PE. This technique can also assess the degree of pulmonary arterial obstruction and the presence of signs suggestive of RVD. It has been suggested that CTPA findings can predict outcome in PE.5

Electrocardiography (ECG) has a widespread use and is routinely carried out and interpreted in the emergency room. Some ECG findings in patients with PE are suggestive of right ventricular strain and might therefore be used to prioritize cases that require closer monitoring.6

This study was designed with the objectives of assessing the correlation of standard ECG findings with echocardiography and CTPA signs of RVD, with the anatomic severity of the pulmonary vascular obstruction, and with the degree of hypoxemia, in hemodynamically stable patients with PE. Secondary objectives were determining if ECG could accurately predict RVD, and determining the interobserver agreement for the ECG score.

Patients and Methods

Results

A total of 103 patients were included in the study (mean age = 69 15 years). One patient died from fatal PE soon after the diagnosis, before echocardiography could be performed. During the 6‐month follow‐up, one patient died from septic shock and another suffered a thromboembolic recurrence (deep vein thrombosis). Echocardiography‐assessed RVD was found in 25 cases (24.5%) and isolated pulmonary hypertension in 20 (19.6%). Median value of the vascular obstruction index was 43.8% (25‐65). Thirty‐three patients (32%) had signs of RVD in CTPA.

Median value of ECG score was 2.5 (1‐6). The level of interobserver agreement regarding ECG score was substantial ( = 0.80). Table 2 shows correlation between ECG score and echocardiography, arterial blood gases, and CTPA parameters. ECG correlated significantly with all the variables measured. Correlation was higher for the vascular obstruction index and the RV/LV ratio, measured with echocardiography. ECG scores were significantly different for groups with varying degrees of severity of PE. Table 3 shows the ECG scores for patients grouped according to different criteria of severity of the disease.

Figure 1 shows the ROC curve for the ECG score and the presence of echocardiography‐detected RVD. Area under the ROC curve was 0.82 (95% confidence interval [CI]: 0.72‐0.89). A cut‐off point of 6 for the ECG score showed the highest accuracy for detecting RVD; sensitivity = 76.5% (95% CI: 50.1‐93.0), specificity = 88.6% (95% CI: 78.7‐94.9), positive likelihood ratio (+LR) = 6.69, negative likelihood ratio (LR) = 0.37. If the ECG score was to be used as a screening method to exclude RVD, a cut‐off point of 0 showed sensitivity = 94.1% (95% CI: 71.2‐99.0), specificity = 27.1% (95% CI: 17.2‐39.1), +LR = 1.29, LR = 0.22. If the ECG score was considered for confirming RVD, the more useful value was 9; sensitivity = 58.8% (95% CI: 33.0‐81.5), specificity = 92.0% (95% CI: 84.1‐97.6), +LR = 8.23, LR = 0.44. Area under the ROC curve to predict CTPA‐detected RVD was slightly lower than for echocardiography: 0.74 (95% CI: 0.61‐0.84).

Figure 1

Discussion

This study shows that a simple, easy‐to‐use ECG scoring system correlates well with the severity of PE, assessed by different methods, in hemodynamically stable patients. Interobserver agreement regarding the scoring system was substantial. Several previous studies have found that ECG signs of RV strain correlate with the presence of RVD.6, 1316 However, some of these studies are limited by a retrospective design that could lead to selection bias or by the inclusion of relatively few (eg, <100) cases.6, 13, 14 Moreover, only a few of these articles have specifically examined patients with normal blood pressure.6, 15 Although our study is still small, one of its strengths is that it prospectively recruited a population of consecutive, normotensive patients and both echocardiography and CTPA were systematically performed. Classifying cases with normal blood pressure by severity is particularly relevant, because of the uncertainty about best treatment when hemodynamic instability is not present. These patients who are not at high risk for PE can be further stratified according to the presence of RVD and/or myocardial injury into intermediate‐risk and low‐risk PE.17 Early hospital discharge or initial outpatient treatment might be considered in low‐risk patients.17, 18 However, echocardiography can be difficult to perform in the emergency setting. Also, the presence of RVD is not easy to predict on clinical grounds. Our study suggests that a standard 12‐lead ECG can be useful to prioritize patients for more exhaustive monitoring or additional risk stratifying tests. An ECG score of 0 would exclude RVD with enough sensitivity to avoid getting an echocardiogram in these cases. On the other hand, a score 9 would suggest the possibility that RVD was present, and an echocardiography ought to be considered in these patients.

Previous studies regarding correlation between ECG and anatomic extension of the pulmonary vascular obstruction in PE have revealed conflicting data. Iles et al. found that ECG score identified those patients with the greatest percentage of perfusion defect on ventilation/perfusion scan.19 However, Kanbay et al. did not find significant differences in ECG scores for patients with pulmonary vascular obstruction scores <50% or 50%, also assessed with ventilation/perfusion scan.20 Subramaniam et al. used CTPA to assess clot burden score, and they did not see a significant association between ECG score and the anatomic severity of PE.21 This study found a significant association between higher ECG score and a more severe vascular obstruction index. This finding might have potential clinical implications, because some reports suggest that a higher CTPA‐measured clot load score can be associated with poorer outcome in PE.22 However, the utility of CTPA scores of vascular obstruction severity as a marker of clinical evolution is controversial, and several other authors have found that these scores are poor predictors of mortality or adverse outcome events.23, 24 Therefore, the positive correlation between ECG and the vascular obstruction index in our study is interesting mainly in the sense that it independently supports the validity of the ECG score as a marker of a potentially more severe PE.

In this study, we have not systematically excluded patients with comorbidities (eg, chronic obstructive pulmonary disease) that could account for RVD or PH. Our intention was to reflect a real‐life clinical scenario. When a patient attends the emergency room with PE, clinical decisions must usually be taken without complete information on comorbidities that may contribute to the existence of RV overload. Also, these subjects have a reduced cardiorespiratory reserve, and the PE might impose an unbearable strain to the RV, so risk stratification is especially important. Therefore, we felt that the study should include all patients presenting with PE, so that the results can be extrapolated to usual clinical practice.

Our results suggest that, by selecting the appropriate cut‐off value, the ECG score can be used with adequate sensitivity and specificity to exclude and confirm, respectively, RVD in hemodynamically stable PE patients. This finding implies that ECG may be useful to design management strategies in these cases, selecting patients that may benefit from further tests.

There are some limitations for the present study. Echocardiography was not immediately performed after the ECG. Thus, we might have missed some patients who presented with transient RVD, which resolved between tests. It is likely that the correlation between ECG and echocardiography would have improved if this delay had been avoided. Also, we did not use a multidetector scanner for performing CTPAs. Consequently, small emboli might have been missed and the clot score might have been underestimated. It is plausible that this may have reduced to some extent the correlation between ECG and CTPA. Also, this a single‐center study, with relatively few patients studied. Finally, due to the low incidence of adverse events in our patients, our study is underpowered to make conclusions about the independent prognostic value of ECG for predicting adverse events.

In conclusion, an easy‐to‐use ECG score correlates significantly with the severity of PE, as assessed with echocardiography, CTPA, and arterial blood gas analysis, in normotensive patients. It can be used to predict with acceptable sensitivity and specificity the presence of RVD in these cases and, accordingly, it has potential value in risk‐stratification strategies. Larger, multicenter studies, should confirm these results before they can be applied to clinical practice.

The clinical outcome of patients with pulmonary embolism (PE) has a wide spectrum, ranging from a benign course, with early resolution of symptoms, to death. Prognostic stratification of patients with PE is therefore essential to guide therapy. Currently, this stratification is primarily based on blood pressure on admission. Hemodynamic instability or shock carries a bad prognosis, and is associated with a 3‐fold to 7‐fold increase in mortality.1 Thrombolytic therapy is warranted in these cases.2 There is a subset of patients that, despite normal blood pressure on presentation, can subsequently suffer hemodynamic deterioration and death.3

Echocardiography has been proposed as a useful tool to identify a subset of patients at a higher risk. It has been suggested that the finding of right ventricle dysfunction (RVD) in echocardiography predicts a worse prognosis, even in patients with normal blood pressure on presentation.4 Also, computed tomography pulmonary angiography (CTPA) is increasingly used as the main test to diagnose PE. This technique can also assess the degree of pulmonary arterial obstruction and the presence of signs suggestive of RVD. It has been suggested that CTPA findings can predict outcome in PE.5

Electrocardiography (ECG) has a widespread use and is routinely carried out and interpreted in the emergency room. Some ECG findings in patients with PE are suggestive of right ventricular strain and might therefore be used to prioritize cases that require closer monitoring.6

This study was designed with the objectives of assessing the correlation of standard ECG findings with echocardiography and CTPA signs of RVD, with the anatomic severity of the pulmonary vascular obstruction, and with the degree of hypoxemia, in hemodynamically stable patients with PE. Secondary objectives were determining if ECG could accurately predict RVD, and determining the interobserver agreement for the ECG score.

Patients and Methods

Results

A total of 103 patients were included in the study (mean age = 69 15 years). One patient died from fatal PE soon after the diagnosis, before echocardiography could be performed. During the 6‐month follow‐up, one patient died from septic shock and another suffered a thromboembolic recurrence (deep vein thrombosis). Echocardiography‐assessed RVD was found in 25 cases (24.5%) and isolated pulmonary hypertension in 20 (19.6%). Median value of the vascular obstruction index was 43.8% (25‐65). Thirty‐three patients (32%) had signs of RVD in CTPA.

Median value of ECG score was 2.5 (1‐6). The level of interobserver agreement regarding ECG score was substantial ( = 0.80). Table 2 shows correlation between ECG score and echocardiography, arterial blood gases, and CTPA parameters. ECG correlated significantly with all the variables measured. Correlation was higher for the vascular obstruction index and the RV/LV ratio, measured with echocardiography. ECG scores were significantly different for groups with varying degrees of severity of PE. Table 3 shows the ECG scores for patients grouped according to different criteria of severity of the disease.

Figure 1 shows the ROC curve for the ECG score and the presence of echocardiography‐detected RVD. Area under the ROC curve was 0.82 (95% confidence interval [CI]: 0.72‐0.89). A cut‐off point of 6 for the ECG score showed the highest accuracy for detecting RVD; sensitivity = 76.5% (95% CI: 50.1‐93.0), specificity = 88.6% (95% CI: 78.7‐94.9), positive likelihood ratio (+LR) = 6.69, negative likelihood ratio (LR) = 0.37. If the ECG score was to be used as a screening method to exclude RVD, a cut‐off point of 0 showed sensitivity = 94.1% (95% CI: 71.2‐99.0), specificity = 27.1% (95% CI: 17.2‐39.1), +LR = 1.29, LR = 0.22. If the ECG score was considered for confirming RVD, the more useful value was 9; sensitivity = 58.8% (95% CI: 33.0‐81.5), specificity = 92.0% (95% CI: 84.1‐97.6), +LR = 8.23, LR = 0.44. Area under the ROC curve to predict CTPA‐detected RVD was slightly lower than for echocardiography: 0.74 (95% CI: 0.61‐0.84).

Figure 1

Discussion

This study shows that a simple, easy‐to‐use ECG scoring system correlates well with the severity of PE, assessed by different methods, in hemodynamically stable patients. Interobserver agreement regarding the scoring system was substantial. Several previous studies have found that ECG signs of RV strain correlate with the presence of RVD.6, 1316 However, some of these studies are limited by a retrospective design that could lead to selection bias or by the inclusion of relatively few (eg, <100) cases.6, 13, 14 Moreover, only a few of these articles have specifically examined patients with normal blood pressure.6, 15 Although our study is still small, one of its strengths is that it prospectively recruited a population of consecutive, normotensive patients and both echocardiography and CTPA were systematically performed. Classifying cases with normal blood pressure by severity is particularly relevant, because of the uncertainty about best treatment when hemodynamic instability is not present. These patients who are not at high risk for PE can be further stratified according to the presence of RVD and/or myocardial injury into intermediate‐risk and low‐risk PE.17 Early hospital discharge or initial outpatient treatment might be considered in low‐risk patients.17, 18 However, echocardiography can be difficult to perform in the emergency setting. Also, the presence of RVD is not easy to predict on clinical grounds. Our study suggests that a standard 12‐lead ECG can be useful to prioritize patients for more exhaustive monitoring or additional risk stratifying tests. An ECG score of 0 would exclude RVD with enough sensitivity to avoid getting an echocardiogram in these cases. On the other hand, a score 9 would suggest the possibility that RVD was present, and an echocardiography ought to be considered in these patients.

Previous studies regarding correlation between ECG and anatomic extension of the pulmonary vascular obstruction in PE have revealed conflicting data. Iles et al. found that ECG score identified those patients with the greatest percentage of perfusion defect on ventilation/perfusion scan.19 However, Kanbay et al. did not find significant differences in ECG scores for patients with pulmonary vascular obstruction scores <50% or 50%, also assessed with ventilation/perfusion scan.20 Subramaniam et al. used CTPA to assess clot burden score, and they did not see a significant association between ECG score and the anatomic severity of PE.21 This study found a significant association between higher ECG score and a more severe vascular obstruction index. This finding might have potential clinical implications, because some reports suggest that a higher CTPA‐measured clot load score can be associated with poorer outcome in PE.22 However, the utility of CTPA scores of vascular obstruction severity as a marker of clinical evolution is controversial, and several other authors have found that these scores are poor predictors of mortality or adverse outcome events.23, 24 Therefore, the positive correlation between ECG and the vascular obstruction index in our study is interesting mainly in the sense that it independently supports the validity of the ECG score as a marker of a potentially more severe PE.

In this study, we have not systematically excluded patients with comorbidities (eg, chronic obstructive pulmonary disease) that could account for RVD or PH. Our intention was to reflect a real‐life clinical scenario. When a patient attends the emergency room with PE, clinical decisions must usually be taken without complete information on comorbidities that may contribute to the existence of RV overload. Also, these subjects have a reduced cardiorespiratory reserve, and the PE might impose an unbearable strain to the RV, so risk stratification is especially important. Therefore, we felt that the study should include all patients presenting with PE, so that the results can be extrapolated to usual clinical practice.

Our results suggest that, by selecting the appropriate cut‐off value, the ECG score can be used with adequate sensitivity and specificity to exclude and confirm, respectively, RVD in hemodynamically stable PE patients. This finding implies that ECG may be useful to design management strategies in these cases, selecting patients that may benefit from further tests.

There are some limitations for the present study. Echocardiography was not immediately performed after the ECG. Thus, we might have missed some patients who presented with transient RVD, which resolved between tests. It is likely that the correlation between ECG and echocardiography would have improved if this delay had been avoided. Also, we did not use a multidetector scanner for performing CTPAs. Consequently, small emboli might have been missed and the clot score might have been underestimated. It is plausible that this may have reduced to some extent the correlation between ECG and CTPA. Also, this a single‐center study, with relatively few patients studied. Finally, due to the low incidence of adverse events in our patients, our study is underpowered to make conclusions about the independent prognostic value of ECG for predicting adverse events.

In conclusion, an easy‐to‐use ECG score correlates significantly with the severity of PE, as assessed with echocardiography, CTPA, and arterial blood gas analysis, in normotensive patients. It can be used to predict with acceptable sensitivity and specificity the presence of RVD in these cases and, accordingly, it has potential value in risk‐stratification strategies. Larger, multicenter studies, should confirm these results before they can be applied to clinical practice.

If you wish to receive credit for this activity, please refer to the website: www.wileyblackwellcme.com.

Accreditation and Designation Statement

Blackwell Futura Media Services designates this journal‐based CME activity for a maximum of 1 AMA PRA Category 1 Credit.. Physicians should only claim credit commensurate with the extent of their participation in the activity.

Blackwell Futura Media Services is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians.

Educational Objectives

Upon completion of this educational activity, participants will be better able to:

• Illustrate the elements of a systematic approach to successful hospital smoking cessation programs.

• Describe the efficacy of a coordinated real world&#148; hospital smoking cessation program in a U.S. hospital.

• Evaluate the barriers to successful hospital smoking cessation programs.

This manuscript underwent peer review in line with the standards of editorial integrity and publication ethics maintained by Journal of Hospital Medicine. The peer reviewers have no relevant financial relationships. The peer review process for Journal of Hospital Medicine is blinded. As such, the identities of the reviewers are not disclosed in line with the standard accepted practices of medical journal peer review.

Conflicts of interest have been identified and resolved in accordance with Blackwell Futura Media Services's Policy on Activity Disclosure and Conflict of Interest. The primary resolution method used was peer review and review by a non‐conflicted expert.

Instructions on Receiving Credit

For information on applicability and acceptance of CME credit for this activity, please consult your professional licensing board.

This activity is designed to be completed within an hour; physicians should claim only those credits that reflect the time actually spent in the activity. To successfully earn credit, participants must complete the activity during the valid credit period, which is up to two years from initial publication.

Follow these steps to earn credit:

• Log on to www.wileyblackwellcme.com

• Read the target audience, learning objectives, and author disclosures.

• Read the article in print or online format.

• Reflect on the article.

• Access the CME Exam, and choose the best answer to each question.

• Complete the required evaluation component of the activity.

This activity will be available for CME credit for twelve months following its publication date. At that time, it will be reviewed and potentially updated and extended for an additional twelve months.

If you wish to receive credit for this activity, please refer to the website: www.wileyblackwellcme.com.

Accreditation and Designation Statement

Blackwell Futura Media Services designates this journal‐based CME activity for a maximum of 1 AMA PRA Category 1 Credit.. Physicians should only claim credit commensurate with the extent of their participation in the activity.

Blackwell Futura Media Services is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians.

Educational Objectives

Upon completion of this educational activity, participants will be better able to:

• Illustrate the elements of a systematic approach to successful hospital smoking cessation programs.

• Describe the efficacy of a coordinated real world&#148; hospital smoking cessation program in a U.S. hospital.

• Evaluate the barriers to successful hospital smoking cessation programs.

This manuscript underwent peer review in line with the standards of editorial integrity and publication ethics maintained by Journal of Hospital Medicine. The peer reviewers have no relevant financial relationships. The peer review process for Journal of Hospital Medicine is blinded. As such, the identities of the reviewers are not disclosed in line with the standard accepted practices of medical journal peer review.

Conflicts of interest have been identified and resolved in accordance with Blackwell Futura Media Services's Policy on Activity Disclosure and Conflict of Interest. The primary resolution method used was peer review and review by a non‐conflicted expert.

Instructions on Receiving Credit

For information on applicability and acceptance of CME credit for this activity, please consult your professional licensing board.

This activity is designed to be completed within an hour; physicians should claim only those credits that reflect the time actually spent in the activity. To successfully earn credit, participants must complete the activity during the valid credit period, which is up to two years from initial publication.

Follow these steps to earn credit:

• Log on to www.wileyblackwellcme.com

• Read the target audience, learning objectives, and author disclosures.

• Read the article in print or online format.

• Reflect on the article.

• Access the CME Exam, and choose the best answer to each question.

• Complete the required evaluation component of the activity.

This activity will be available for CME credit for twelve months following its publication date. At that time, it will be reviewed and potentially updated and extended for an additional twelve months.

If you wish to receive credit for this activity, please refer to the website: www.wileyblackwellcme.com.

Accreditation and Designation Statement

Blackwell Futura Media Services designates this journal‐based CME activity for a maximum of 1 AMA PRA Category 1 Credit.. Physicians should only claim credit commensurate with the extent of their participation in the activity.

Blackwell Futura Media Services is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians.

Educational Objectives

Upon completion of this educational activity, participants will be better able to:

• Illustrate the elements of a systematic approach to successful hospital smoking cessation programs.

• Describe the efficacy of a coordinated real world&#148; hospital smoking cessation program in a U.S. hospital.

• Evaluate the barriers to successful hospital smoking cessation programs.

This manuscript underwent peer review in line with the standards of editorial integrity and publication ethics maintained by Journal of Hospital Medicine. The peer reviewers have no relevant financial relationships. The peer review process for Journal of Hospital Medicine is blinded. As such, the identities of the reviewers are not disclosed in line with the standard accepted practices of medical journal peer review.

Conflicts of interest have been identified and resolved in accordance with Blackwell Futura Media Services's Policy on Activity Disclosure and Conflict of Interest. The primary resolution method used was peer review and review by a non‐conflicted expert.

Instructions on Receiving Credit

For information on applicability and acceptance of CME credit for this activity, please consult your professional licensing board.

This activity is designed to be completed within an hour; physicians should claim only those credits that reflect the time actually spent in the activity. To successfully earn credit, participants must complete the activity during the valid credit period, which is up to two years from initial publication.

Follow these steps to earn credit:

• Log on to www.wileyblackwellcme.com

• Read the target audience, learning objectives, and author disclosures.

• Read the article in print or online format.

• Reflect on the article.

• Access the CME Exam, and choose the best answer to each question.

• Complete the required evaluation component of the activity.

This activity will be available for CME credit for twelve months following its publication date. At that time, it will be reviewed and potentially updated and extended for an additional twelve months.

Severe, disabling pain, often requiring opioids, is the most common medical presentation for children and adults with sickle cell disease (SCD), an autosomal recessive red blood cell disorder affecting those of African, Mediterranean, and Asian descent.1, 2 A genetically controlled hemoglobin alteration impairs oxygen binding, and enables polymerization of deoxy‐hemoglobin, resulting in, classically, sickle‐shaped erythrocytes3 and a complex cascade of ischemia and vaso‐occlusion in the microcirculation.4, 5

Dramatic gains in the treatment of SCD in childhood have resulted in markedly improved survival through adulthood.68 Thus, the need for adult SCD care is relatively new and rapidly growing. In 2005, approximately 70% of the nearly 80,000 US SCD hospitalizations occurred in adults versus children (Table 1). These hospitalizations occurred in the context of a poorly coordinated American health care system,9 despite the hopes raised by the Patient‐Centered Medical Home10 and the Chronic Care Model.11

Adults with SCD are vulnerable both because they are usually members of racial and ethnic minority groups, and because they have a Food and Drug Administration (FDA)‐defined orphan disease.12 They often do not receive the only FDA‐approved medication for SCD, life‐saving hydroxyurea,13 recommended for adults with homozygous sickle cell anemia (Hb SS) and sickle‐^othalassemia (Hb S^oThal).14 Young adults often fail to experience a smooth transition of care from children's hospitals, falling into a medical abyss.15

Therefore, increasingly, hospitalists are managing adults with SCD, rather than adult hematologyoncology, pain, or palliative care specialists. Adults with SCD experience negative opinions, bilateral lack of trust, and conflict in the doctorpatient relationship, frequently cited in studies of SCD adults and providers in the literature.16, 17

Evidence Base

General guidelines for SCD management have been published by the National Institutes of Health (NIH)18 and the Agency for Healthcare Policy and Research.19 But one of us (K.L.H.) found evidence lacking with regard to SCD pain management.20 Published guidelines on general pain management, such as the World Health Organization's Analgesic Ladder,21 do not address SCD. A Cochrane Review of pain management in SCD found only 9 randomized controlled trials, all with small numbers of patients, addressing acute SCD pain only.22 As well, American and British consensus SCD pain guidelines23, 24 admit, and subsequent publications emphasize,25, 26 the lack of evidence for what to do or not do for SCD pain management. At least 1 well‐done summary of the SCD evidence base intended for hospitalists has been published, but it focuses on management of issues other than pain.27

Motivations and Fears

It is not surprising then that hospitalists may bring great fear and apprehension with them into their care of SCD patients. One of us (W.R.S.), a general internist, has been called by his own and 3 other academic medical centers, 2 with active Federally‐funded SCD research programs, to address the problems of high‐utilizing adults with SCD, including counseling hospitalists frustrated with the management of pain in these patients.

Hospitalists may be motivated to provide efficient inpatient management (Table 2), and be aware of pain as the primary symptom of SCD inpatients. But they may carry knowledge gaps and biases into their relationships with SCD inpatients. They may fear opioid administration (opiophobia), loss of licensure or governmental reprisals because of high‐dose prescription of opioids, or may believe that SCD patients are more often addicted than most.17, 28 Consequently, more troublesome hospital stays may occur when patients are not rapidly and adequately titrated to appropriate analgesic doses, or when unnecessary deleterious side effects result from opioid and other analgesics. We therefore offer answers to frequently asked questions (FAQs) about pain management by hospitalists caring for adults with SCD. We address FAQs arising during the prototypical situationa patient with SCD admitted for a painful exacerbation, and little or no acute comorbidity. We refer the reader to the aforementioned articles and guidelines to address other treatment issues in adults with SCD.

FAQS

• 1

  Is there any objective way to tell when SCD patients really are in a crisis?

  Although the term crisis is used as if it were an objectively definable biological entity, no one has proposed a standard definition of a crisis based on pain intensity level, clinical features, or biomarkers. Measures of vaso‐occlusion are correlated with ischemic pain, including pain that is often called a crisis.2932 However, neither ischemic pain from SCD, nor the underlying vaso‐occlusive cascade that is associated with this pain, is a sudden, present‐or‐absent phenomenon. Instead, these are continua that can be measured using pain scales or various biomarkers.

  There is, however, correlative evidence of the intensity of SCD pain associated with various distinctive health states (admitted/not admitted, in crisis/not in crisis). The most visible measure of a crisis, health care utilization, was a strong predictor of mortality in the Cooperative Study of Sickle Cell Disease. Patients with 3 or more admissions per year had a lower 5‐year survival rate.33 In contrast, crisis in the landmark Pain in Sickle Cell Epidemiology Study (PiSCES) was self‐defined by patients.34 Despite being in pain on over half of their days on average, and despite a third of patients being in pain daily, most pain in PiSCES was not considered a crisis, and less than 5% of patients' days were spent in emergency departments (EDs) or hospitals. Ambulatory pain intensity reports were correlated with opioid use.35 A substantial minority (35%) of PiSCES patients made at least 3 ED visits per year. However, these high ED utilizers had worse laboratory values, more pain, more distress, and a lower quality of life.36

  Importantly, sometimes adults with SCD may have severe comorbidities which may not be addressed or may be mistakenly managed as an acute vaso‐occlusive episode without further investigation or timely specialist consultation. Although pain is primarily the individual's chief complaint, any potential relationship between the presence of medical comorbidities and pain should be examined when patients are admitted.

• 2

  How can one know when opioid dosages should be changed, or when SCD pain is appropriately controlled to allow discharge?

  We recommend, as a standard of care, that SCD pain assessment and pain therapy be interwoven, despite a systematic review finding no evidence that directly linked the timing, frequency, or method of pain assessment with outcomes or safety in medical inpatients, and concluding that the safety and effectiveness of patient‐controlled analgesia (PCA) in medical patients had not been adequately studied.37 Hospitalists should focus the first 24 hours of inpatient SCD pain management on cycles of recurrent pain assessment and opioid dose titration as frequently as every 1 to 2 hours, to assure safe and rapidly efficacious analgesia. Pain intensity, duration, and character should be assessed directly. Intensity is often assessed using a visual analog scale (VAS) or numeric rating scale.38 Treating physicians should themselves directly assess pain during discussions of therapy with the patient, even though some assessment usually is done in hospitals during each nursing shift. Pain and pain relief can be assessed indirectly by monitoring opioid use.

  We recommend PCA for inpatients with SCD, administered as an intermittent demand dose (patient must push a button) of opioid, with or without a background opioid constant infusion.39 We usually set the interval between doses, or lockout, to 6 to 10 minutes. Both the lockout and the sedation from delivered doses prevent patients from pushing the demand button repetitively to the point of overdose. Use of a low‐level constant infusion (basal) may sustain pain control during times when the patient is asleep, avoiding recrudescent pain and lost ground due to inadequate analgesia during rest. Alternatively, long‐acting oral opioids may be continued if already used at home, or newly introduced to provide adequate baseline pain control which is augmented by the demand dosing. Most PCA pumps monitor hourly opioid dose demand (number of pushes), as well as hourly doses delivered. Both hourly opioid dose demand and hourly dose‐per‐demand ratio are measures of PCA efficacy or futility. Pumps record this data, and can be interrogated at the patient's bedside for up to several days of prior use. Physicians should combine pump interrogation with direct pain assessment to guide demand‐dose titration. Demand doses should be increased to 1.5 to 2 times the previous demand dose after several hours of failed reduction of pain intensity and duration, and/or persistently futile dose‐per‐demand ratios.

  PCA interrogation is also useful for conversion of parenteral opioids to oral opioids, as well as to guide the recommendation for discharge home. After the first 24‐48 hours of up‐titration, if opioid dose demand decreases concordantly with pain frequency and intensity, the demand dose may be safely decreased, and eventually daily PCA requirements may be summed and converted to oral medication using standard opioid dose conversion tables. At this point, physicians may use single measures or daily averages of directly assessed pain.

  Routine PCA use in SCD is backed by some evidence.40, 41 But we find it important that patients be taught and encouraged to use the demand feature of PCA. Still, for various reasons, some patients do not use PCA pumps well. Discordant or unreliable assessments (eg, high pain intensity but low‐opioid demand doses during the same interval) may result, and PCA potentially may fail as a dosing strategy. Management is more difficult for these patients. One alternative dosing strategy is prescription of scheduled doses of a short‐acting opioid, attaching to each dose the order, patient may refuse. This is different than dosing as needed, and allows counts of dose refusals over an interval, analogous to PCA pump interrogation.

• 3

  How much is too much opioid? Should one rely on side effects, or on requests for medicine, or is there a ceiling dose?

  Addictionologists, pain specialists, oncologists, those involved in hospice care, and some hematologists caring for SCD patients agree that, in general, there should be no a priori dose limitations imposed on opioid prescribing for acute pain. Instead, titration of dose of opioid to pain relief is a central principle of acute pain management. Experts also agree that particular opioids carry particular side effects which warrant dose limitation, adjustments, or avoidance of that opioid altogether. A summary of opioids commonly used in SCD, along with warnings and implied dose limitations is found in Table 3.

  For safety, it is important to assess the history of prior opioid use to recognize a patient who is not tolerant to opioids (see below, FAQ 4), to avoid mistakenly overdosing a patient using doses often required by tolerant patients. In lieu of a pre‐written, individualized opioid dosing plan in place for the patient, the patient may be the best source of information regarding preferred medication and tolerated doses.

  The reader is referred to standard texts for a description of opioids, their pharmacokinetics and pharmacodynamics, and their addictive and abuse potential. The side‐effect profile of opioids is well‐known: nausea, vomiting, and itching frequently occur; hallucinations, respiratory suppression, and myoclonus occur infrequently.42 Meperidine may more readily cause central nervous system (CNS) dysfunction, including seizures, as compared to other commonly used opioids, because of its toxic metabolite nor‐meperidine. Use of meperidine is often avoided, especially use via PCA.43 Methadone may cause dysrhythmias, specificially corrected Q‐T interval (QT_c) prolongation and torsades de pointes on an electrocardiogram, in doses above 200 mg per day.44 Some recommend baseline and yearly electrocardiogram monitoring when giving methadone chronically.

  Recognizing the potential dangers of opioids, it is also reasonable to look for opioid‐sparing analgesic strategies. Non‐opioid analgesics such as ketorolac45 and adjuvants such as ketamine46 that are opioid‐sparing should be considered whenever feasible. Complementary and alternative therapies such as transcutaneous electrical nerve stimulation (TENS)47 have less evidence of effectiveness, but have limited risks and may be of use for some individuals.

• 4

  What are the major signs of substance abuse (opioids, street drugs) in SCD patients already on opioids, and can a hospitalist judge those signs acutely and intervene appropriately?

  Reports of underprescription of opioids in SCD have cited physician fear of abuse and addiction.48 A recent informal poll of adult sickle cell providers suggests policies vary on how potential abuse is monitored in ambulatory sickle cell patients. We note that physicians, especially upon meeting a patient for the first time, may be unable to reliably judge whether that patient is abusing opioids or street drugs. Both false‐positive and false‐negative diagnoses may be made.49 Repetitive reports of lost or stolen prescriptions or pill bottles, receipt of prescriptions from multiple providers, or repeated requests for early refills increase the suspicion of misuse or abuse, but are indirect evidence. Urine and serum monitoring may be useful, but may give incorrect information if misinterpreted or not conducted frequently enough to improve sensitivity.50

  It is important to distinguish between tolerance, the decreased analgesic response over time to repeated doses of the same drug; physical dependence, the production of withdrawal upon abrupt discontinuation of an opioid agonist or administration of an antagonist; and addiction, the psychological dependence upon opioids. Tolerance may be misperceived as true addiction. Its earliest symptom is shortening of the duration of effective analgesia. In contrast, addiction may be manifested by dose escalation in the absence of an increased pain stimulus, or by use of opioids for purposes other than pain relief.51 These are not easily distinguished during a single patient encounter.

  SCD patients' requests for specific opioid medications in specific doses, should not be taken as evidence of past or current abuse, but rather evidence of a well‐informed, self‐managing patient. Adults with SCD are clearly expected to be very knowledgeable about and tolerant to opioids if they have had a life of pain as a child, and will require higher doses of opioids than other patients treated by most hospitalists. The issue of medication abuse may be best handled in the ambulatory setting. Whenever possible, hospitalists should not rely only on data from the acute care setting to manage patients. Ambulatory providers may conduct random, unannounced urine and/or serum testing, as part of an opioid prescribing agreement that is written and filed in the patient's chart. Assays for prescribed opioids (especially long‐acting agents), as well as assays for common drugs of abuse, should be conducted. Comanagement with an addictionologist, psychiatrist, or psychologist should be considered in individuals suspected of opioid abuse.

  We do not suggest routine urine drug test monitoring of all SCD patients unless routine monitoring is done as a policy for all patients on opioids. Though the prevalence of addiction may be higher in subpopulations of patients with pain,52 and though prescription of opioids, prescription drug abuse, and accidental deaths from prescribed opioids have risen exponentially in the last several years,53 in our experience and in the published literature, drug misuse/abuse among SCD patients is no worse than among patients with other illnesses.5456 However, pseudoaddiction, the appropriate seeking of needed opioids from multiple physicians because of uncontrolled pain and opioid underprescription, may well be prevalent in SCD,57 and may be mistaken for true addiction.

• 5

  How can patients' readiness for discharge be assessed? What can be done for the patient who has lengthy and/or multiple hospitalizations or frequent ED visits?

  The appropriate time for discharge in most patients is when they can manage their pain at home with oral opioids or less. Often, patients do not improve even after a few days of inpatient therapy.58 A typical pain episode may last much longer than the 6‐day average US hospital length of stay for a diagnosis of sickle cell crisis among 18‐44 year olds (Table 1).59 Patients may return and be readmitted.60, 61 But in the best cases, pain resolves or reverts to a usual chronic intensity level. As described in FAQ 2, daily or more frequent pain assessment is a bedrock for making discharge decisions. Patients well‐experienced in the use of pain intensity scales can report their usual pain intensity at home, and how close they are to their baseline pain intensity. Simply asking patients, Are you ready for discharge? is appropriate and may yield a surprising positive response. In a recent inpatient trial of PCA (manuscript in preparation), adult patients were admitted with a minimum pain intensity of 45 mm on a 100 mm horizontal VAS scale after treatment in the ED, and mean pain intensity of 76 mm 10 mm. All adults in this study were discharged with pain that was clinically significantly lower. Researchers have found a VAS change of 13.5 mm to be the minimum clinically significant change62 during treatment of vaso‐occlusive crisis.63

  Unremitting pain despite appropriate titration of opioids and prolonged hospital stays suggests the need for comprehensive evaluation for medical and psychosocial comorbidities, as is done for other patients with chronic pain syndromes. If not already done, discussion with the patient's primary care provider may reveal factors impacting on persistent pain. Consultation with a hematologist, pain or palliative care specialist, or other provider familiar with SCD may prove helpful. Implementation of adjuvant therapies as discussed in FAQ 3 and adding long‐acting oral opioids to continue postdischarge may also help. Hyperalgesia, or heightened sensitivity to pain, is normal after acute tissue injury, but is now suspected in SCD as a long‐term neuropathic pain syndrome, as a consequence either of repeated painful crises or of chronic opioid therapy.2 Only some centers have specialists qualified to test for and diagnose neuropathic pain.64

  Discharge planning should include identification of a source of outpatient follow‐up. Opioids prescribed at discharge should be sufficient to last at least until the first outpatient appointment, to avoid repeated ED or hospital visits. Communication with a primary care provider at discharge can enhance successful care transition. Otherwise, for patients without established providers, social workers and others may address barriers to follow‐up that frustrate both patient and provider.

Support for Hospitalists Managing Adults With Sickle Cell Disease

Beside the general advice on pain management in SCD mentioned above or found in the bibliography of this article, at long last, a group of adult practitioners skilled in the care of SCD has formed nationally. The Sickle Cell Adult Provider Network [http://www.scapn.net] provides non‐binding advice and support to its members via an e‐mail listserve. Topics often include pain management. This advice fills a vacuum created by the lack of evidence‐based guidelines.

Ultimately, evidence and updated guidelines will be the best support for hospitalists and others managing pain in SCD. The hope is that SCD will receive the attention it deserves, so that practitioners and patients alike do not suffer continued pain from this disease or its management.

Severe, disabling pain, often requiring opioids, is the most common medical presentation for children and adults with sickle cell disease (SCD), an autosomal recessive red blood cell disorder affecting those of African, Mediterranean, and Asian descent.1, 2 A genetically controlled hemoglobin alteration impairs oxygen binding, and enables polymerization of deoxy‐hemoglobin, resulting in, classically, sickle‐shaped erythrocytes3 and a complex cascade of ischemia and vaso‐occlusion in the microcirculation.4, 5

Dramatic gains in the treatment of SCD in childhood have resulted in markedly improved survival through adulthood.68 Thus, the need for adult SCD care is relatively new and rapidly growing. In 2005, approximately 70% of the nearly 80,000 US SCD hospitalizations occurred in adults versus children (Table 1). These hospitalizations occurred in the context of a poorly coordinated American health care system,9 despite the hopes raised by the Patient‐Centered Medical Home10 and the Chronic Care Model.11

Adults with SCD are vulnerable both because they are usually members of racial and ethnic minority groups, and because they have a Food and Drug Administration (FDA)‐defined orphan disease.12 They often do not receive the only FDA‐approved medication for SCD, life‐saving hydroxyurea,13 recommended for adults with homozygous sickle cell anemia (Hb SS) and sickle‐^othalassemia (Hb S^oThal).14 Young adults often fail to experience a smooth transition of care from children's hospitals, falling into a medical abyss.15

Therefore, increasingly, hospitalists are managing adults with SCD, rather than adult hematologyoncology, pain, or palliative care specialists. Adults with SCD experience negative opinions, bilateral lack of trust, and conflict in the doctorpatient relationship, frequently cited in studies of SCD adults and providers in the literature.16, 17

Evidence Base

General guidelines for SCD management have been published by the National Institutes of Health (NIH)18 and the Agency for Healthcare Policy and Research.19 But one of us (K.L.H.) found evidence lacking with regard to SCD pain management.20 Published guidelines on general pain management, such as the World Health Organization's Analgesic Ladder,21 do not address SCD. A Cochrane Review of pain management in SCD found only 9 randomized controlled trials, all with small numbers of patients, addressing acute SCD pain only.22 As well, American and British consensus SCD pain guidelines23, 24 admit, and subsequent publications emphasize,25, 26 the lack of evidence for what to do or not do for SCD pain management. At least 1 well‐done summary of the SCD evidence base intended for hospitalists has been published, but it focuses on management of issues other than pain.27

Motivations and Fears

It is not surprising then that hospitalists may bring great fear and apprehension with them into their care of SCD patients. One of us (W.R.S.), a general internist, has been called by his own and 3 other academic medical centers, 2 with active Federally‐funded SCD research programs, to address the problems of high‐utilizing adults with SCD, including counseling hospitalists frustrated with the management of pain in these patients.

Hospitalists may be motivated to provide efficient inpatient management (Table 2), and be aware of pain as the primary symptom of SCD inpatients. But they may carry knowledge gaps and biases into their relationships with SCD inpatients. They may fear opioid administration (opiophobia), loss of licensure or governmental reprisals because of high‐dose prescription of opioids, or may believe that SCD patients are more often addicted than most.17, 28 Consequently, more troublesome hospital stays may occur when patients are not rapidly and adequately titrated to appropriate analgesic doses, or when unnecessary deleterious side effects result from opioid and other analgesics. We therefore offer answers to frequently asked questions (FAQs) about pain management by hospitalists caring for adults with SCD. We address FAQs arising during the prototypical situationa patient with SCD admitted for a painful exacerbation, and little or no acute comorbidity. We refer the reader to the aforementioned articles and guidelines to address other treatment issues in adults with SCD.

FAQS

• 1

  Is there any objective way to tell when SCD patients really are in a crisis?

  Although the term crisis is used as if it were an objectively definable biological entity, no one has proposed a standard definition of a crisis based on pain intensity level, clinical features, or biomarkers. Measures of vaso‐occlusion are correlated with ischemic pain, including pain that is often called a crisis.2932 However, neither ischemic pain from SCD, nor the underlying vaso‐occlusive cascade that is associated with this pain, is a sudden, present‐or‐absent phenomenon. Instead, these are continua that can be measured using pain scales or various biomarkers.

  There is, however, correlative evidence of the intensity of SCD pain associated with various distinctive health states (admitted/not admitted, in crisis/not in crisis). The most visible measure of a crisis, health care utilization, was a strong predictor of mortality in the Cooperative Study of Sickle Cell Disease. Patients with 3 or more admissions per year had a lower 5‐year survival rate.33 In contrast, crisis in the landmark Pain in Sickle Cell Epidemiology Study (PiSCES) was self‐defined by patients.34 Despite being in pain on over half of their days on average, and despite a third of patients being in pain daily, most pain in PiSCES was not considered a crisis, and less than 5% of patients' days were spent in emergency departments (EDs) or hospitals. Ambulatory pain intensity reports were correlated with opioid use.35 A substantial minority (35%) of PiSCES patients made at least 3 ED visits per year. However, these high ED utilizers had worse laboratory values, more pain, more distress, and a lower quality of life.36

  Importantly, sometimes adults with SCD may have severe comorbidities which may not be addressed or may be mistakenly managed as an acute vaso‐occlusive episode without further investigation or timely specialist consultation. Although pain is primarily the individual's chief complaint, any potential relationship between the presence of medical comorbidities and pain should be examined when patients are admitted.

• 2

  How can one know when opioid dosages should be changed, or when SCD pain is appropriately controlled to allow discharge?

  We recommend, as a standard of care, that SCD pain assessment and pain therapy be interwoven, despite a systematic review finding no evidence that directly linked the timing, frequency, or method of pain assessment with outcomes or safety in medical inpatients, and concluding that the safety and effectiveness of patient‐controlled analgesia (PCA) in medical patients had not been adequately studied.37 Hospitalists should focus the first 24 hours of inpatient SCD pain management on cycles of recurrent pain assessment and opioid dose titration as frequently as every 1 to 2 hours, to assure safe and rapidly efficacious analgesia. Pain intensity, duration, and character should be assessed directly. Intensity is often assessed using a visual analog scale (VAS) or numeric rating scale.38 Treating physicians should themselves directly assess pain during discussions of therapy with the patient, even though some assessment usually is done in hospitals during each nursing shift. Pain and pain relief can be assessed indirectly by monitoring opioid use.

  We recommend PCA for inpatients with SCD, administered as an intermittent demand dose (patient must push a button) of opioid, with or without a background opioid constant infusion.39 We usually set the interval between doses, or lockout, to 6 to 10 minutes. Both the lockout and the sedation from delivered doses prevent patients from pushing the demand button repetitively to the point of overdose. Use of a low‐level constant infusion (basal) may sustain pain control during times when the patient is asleep, avoiding recrudescent pain and lost ground due to inadequate analgesia during rest. Alternatively, long‐acting oral opioids may be continued if already used at home, or newly introduced to provide adequate baseline pain control which is augmented by the demand dosing. Most PCA pumps monitor hourly opioid dose demand (number of pushes), as well as hourly doses delivered. Both hourly opioid dose demand and hourly dose‐per‐demand ratio are measures of PCA efficacy or futility. Pumps record this data, and can be interrogated at the patient's bedside for up to several days of prior use. Physicians should combine pump interrogation with direct pain assessment to guide demand‐dose titration. Demand doses should be increased to 1.5 to 2 times the previous demand dose after several hours of failed reduction of pain intensity and duration, and/or persistently futile dose‐per‐demand ratios.

  PCA interrogation is also useful for conversion of parenteral opioids to oral opioids, as well as to guide the recommendation for discharge home. After the first 24‐48 hours of up‐titration, if opioid dose demand decreases concordantly with pain frequency and intensity, the demand dose may be safely decreased, and eventually daily PCA requirements may be summed and converted to oral medication using standard opioid dose conversion tables. At this point, physicians may use single measures or daily averages of directly assessed pain.

  Routine PCA use in SCD is backed by some evidence.40, 41 But we find it important that patients be taught and encouraged to use the demand feature of PCA. Still, for various reasons, some patients do not use PCA pumps well. Discordant or unreliable assessments (eg, high pain intensity but low‐opioid demand doses during the same interval) may result, and PCA potentially may fail as a dosing strategy. Management is more difficult for these patients. One alternative dosing strategy is prescription of scheduled doses of a short‐acting opioid, attaching to each dose the order, patient may refuse. This is different than dosing as needed, and allows counts of dose refusals over an interval, analogous to PCA pump interrogation.

• 3

  How much is too much opioid? Should one rely on side effects, or on requests for medicine, or is there a ceiling dose?

  Addictionologists, pain specialists, oncologists, those involved in hospice care, and some hematologists caring for SCD patients agree that, in general, there should be no a priori dose limitations imposed on opioid prescribing for acute pain. Instead, titration of dose of opioid to pain relief is a central principle of acute pain management. Experts also agree that particular opioids carry particular side effects which warrant dose limitation, adjustments, or avoidance of that opioid altogether. A summary of opioids commonly used in SCD, along with warnings and implied dose limitations is found in Table 3.

  For safety, it is important to assess the history of prior opioid use to recognize a patient who is not tolerant to opioids (see below, FAQ 4), to avoid mistakenly overdosing a patient using doses often required by tolerant patients. In lieu of a pre‐written, individualized opioid dosing plan in place for the patient, the patient may be the best source of information regarding preferred medication and tolerated doses.

  The reader is referred to standard texts for a description of opioids, their pharmacokinetics and pharmacodynamics, and their addictive and abuse potential. The side‐effect profile of opioids is well‐known: nausea, vomiting, and itching frequently occur; hallucinations, respiratory suppression, and myoclonus occur infrequently.42 Meperidine may more readily cause central nervous system (CNS) dysfunction, including seizures, as compared to other commonly used opioids, because of its toxic metabolite nor‐meperidine. Use of meperidine is often avoided, especially use via PCA.43 Methadone may cause dysrhythmias, specificially corrected Q‐T interval (QT_c) prolongation and torsades de pointes on an electrocardiogram, in doses above 200 mg per day.44 Some recommend baseline and yearly electrocardiogram monitoring when giving methadone chronically.

  Recognizing the potential dangers of opioids, it is also reasonable to look for opioid‐sparing analgesic strategies. Non‐opioid analgesics such as ketorolac45 and adjuvants such as ketamine46 that are opioid‐sparing should be considered whenever feasible. Complementary and alternative therapies such as transcutaneous electrical nerve stimulation (TENS)47 have less evidence of effectiveness, but have limited risks and may be of use for some individuals.

• 4

  What are the major signs of substance abuse (opioids, street drugs) in SCD patients already on opioids, and can a hospitalist judge those signs acutely and intervene appropriately?

  Reports of underprescription of opioids in SCD have cited physician fear of abuse and addiction.48 A recent informal poll of adult sickle cell providers suggests policies vary on how potential abuse is monitored in ambulatory sickle cell patients. We note that physicians, especially upon meeting a patient for the first time, may be unable to reliably judge whether that patient is abusing opioids or street drugs. Both false‐positive and false‐negative diagnoses may be made.49 Repetitive reports of lost or stolen prescriptions or pill bottles, receipt of prescriptions from multiple providers, or repeated requests for early refills increase the suspicion of misuse or abuse, but are indirect evidence. Urine and serum monitoring may be useful, but may give incorrect information if misinterpreted or not conducted frequently enough to improve sensitivity.50

  It is important to distinguish between tolerance, the decreased analgesic response over time to repeated doses of the same drug; physical dependence, the production of withdrawal upon abrupt discontinuation of an opioid agonist or administration of an antagonist; and addiction, the psychological dependence upon opioids. Tolerance may be misperceived as true addiction. Its earliest symptom is shortening of the duration of effective analgesia. In contrast, addiction may be manifested by dose escalation in the absence of an increased pain stimulus, or by use of opioids for purposes other than pain relief.51 These are not easily distinguished during a single patient encounter.

  SCD patients' requests for specific opioid medications in specific doses, should not be taken as evidence of past or current abuse, but rather evidence of a well‐informed, self‐managing patient. Adults with SCD are clearly expected to be very knowledgeable about and tolerant to opioids if they have had a life of pain as a child, and will require higher doses of opioids than other patients treated by most hospitalists. The issue of medication abuse may be best handled in the ambulatory setting. Whenever possible, hospitalists should not rely only on data from the acute care setting to manage patients. Ambulatory providers may conduct random, unannounced urine and/or serum testing, as part of an opioid prescribing agreement that is written and filed in the patient's chart. Assays for prescribed opioids (especially long‐acting agents), as well as assays for common drugs of abuse, should be conducted. Comanagement with an addictionologist, psychiatrist, or psychologist should be considered in individuals suspected of opioid abuse.

  We do not suggest routine urine drug test monitoring of all SCD patients unless routine monitoring is done as a policy for all patients on opioids. Though the prevalence of addiction may be higher in subpopulations of patients with pain,52 and though prescription of opioids, prescription drug abuse, and accidental deaths from prescribed opioids have risen exponentially in the last several years,53 in our experience and in the published literature, drug misuse/abuse among SCD patients is no worse than among patients with other illnesses.5456 However, pseudoaddiction, the appropriate seeking of needed opioids from multiple physicians because of uncontrolled pain and opioid underprescription, may well be prevalent in SCD,57 and may be mistaken for true addiction.

• 5

  How can patients' readiness for discharge be assessed? What can be done for the patient who has lengthy and/or multiple hospitalizations or frequent ED visits?

  The appropriate time for discharge in most patients is when they can manage their pain at home with oral opioids or less. Often, patients do not improve even after a few days of inpatient therapy.58 A typical pain episode may last much longer than the 6‐day average US hospital length of stay for a diagnosis of sickle cell crisis among 18‐44 year olds (Table 1).59 Patients may return and be readmitted.60, 61 But in the best cases, pain resolves or reverts to a usual chronic intensity level. As described in FAQ 2, daily or more frequent pain assessment is a bedrock for making discharge decisions. Patients well‐experienced in the use of pain intensity scales can report their usual pain intensity at home, and how close they are to their baseline pain intensity. Simply asking patients, Are you ready for discharge? is appropriate and may yield a surprising positive response. In a recent inpatient trial of PCA (manuscript in preparation), adult patients were admitted with a minimum pain intensity of 45 mm on a 100 mm horizontal VAS scale after treatment in the ED, and mean pain intensity of 76 mm 10 mm. All adults in this study were discharged with pain that was clinically significantly lower. Researchers have found a VAS change of 13.5 mm to be the minimum clinically significant change62 during treatment of vaso‐occlusive crisis.63

  Unremitting pain despite appropriate titration of opioids and prolonged hospital stays suggests the need for comprehensive evaluation for medical and psychosocial comorbidities, as is done for other patients with chronic pain syndromes. If not already done, discussion with the patient's primary care provider may reveal factors impacting on persistent pain. Consultation with a hematologist, pain or palliative care specialist, or other provider familiar with SCD may prove helpful. Implementation of adjuvant therapies as discussed in FAQ 3 and adding long‐acting oral opioids to continue postdischarge may also help. Hyperalgesia, or heightened sensitivity to pain, is normal after acute tissue injury, but is now suspected in SCD as a long‐term neuropathic pain syndrome, as a consequence either of repeated painful crises or of chronic opioid therapy.2 Only some centers have specialists qualified to test for and diagnose neuropathic pain.64

  Discharge planning should include identification of a source of outpatient follow‐up. Opioids prescribed at discharge should be sufficient to last at least until the first outpatient appointment, to avoid repeated ED or hospital visits. Communication with a primary care provider at discharge can enhance successful care transition. Otherwise, for patients without established providers, social workers and others may address barriers to follow‐up that frustrate both patient and provider.

Support for Hospitalists Managing Adults With Sickle Cell Disease

Beside the general advice on pain management in SCD mentioned above or found in the bibliography of this article, at long last, a group of adult practitioners skilled in the care of SCD has formed nationally. The Sickle Cell Adult Provider Network [http://www.scapn.net] provides non‐binding advice and support to its members via an e‐mail listserve. Topics often include pain management. This advice fills a vacuum created by the lack of evidence‐based guidelines.

Ultimately, evidence and updated guidelines will be the best support for hospitalists and others managing pain in SCD. The hope is that SCD will receive the attention it deserves, so that practitioners and patients alike do not suffer continued pain from this disease or its management.

Severe, disabling pain, often requiring opioids, is the most common medical presentation for children and adults with sickle cell disease (SCD), an autosomal recessive red blood cell disorder affecting those of African, Mediterranean, and Asian descent.1, 2 A genetically controlled hemoglobin alteration impairs oxygen binding, and enables polymerization of deoxy‐hemoglobin, resulting in, classically, sickle‐shaped erythrocytes3 and a complex cascade of ischemia and vaso‐occlusion in the microcirculation.4, 5

Dramatic gains in the treatment of SCD in childhood have resulted in markedly improved survival through adulthood.68 Thus, the need for adult SCD care is relatively new and rapidly growing. In 2005, approximately 70% of the nearly 80,000 US SCD hospitalizations occurred in adults versus children (Table 1). These hospitalizations occurred in the context of a poorly coordinated American health care system,9 despite the hopes raised by the Patient‐Centered Medical Home10 and the Chronic Care Model.11

Adults with SCD are vulnerable both because they are usually members of racial and ethnic minority groups, and because they have a Food and Drug Administration (FDA)‐defined orphan disease.12 They often do not receive the only FDA‐approved medication for SCD, life‐saving hydroxyurea,13 recommended for adults with homozygous sickle cell anemia (Hb SS) and sickle‐^othalassemia (Hb S^oThal).14 Young adults often fail to experience a smooth transition of care from children's hospitals, falling into a medical abyss.15

Therefore, increasingly, hospitalists are managing adults with SCD, rather than adult hematologyoncology, pain, or palliative care specialists. Adults with SCD experience negative opinions, bilateral lack of trust, and conflict in the doctorpatient relationship, frequently cited in studies of SCD adults and providers in the literature.16, 17

Evidence Base

General guidelines for SCD management have been published by the National Institutes of Health (NIH)18 and the Agency for Healthcare Policy and Research.19 But one of us (K.L.H.) found evidence lacking with regard to SCD pain management.20 Published guidelines on general pain management, such as the World Health Organization's Analgesic Ladder,21 do not address SCD. A Cochrane Review of pain management in SCD found only 9 randomized controlled trials, all with small numbers of patients, addressing acute SCD pain only.22 As well, American and British consensus SCD pain guidelines23, 24 admit, and subsequent publications emphasize,25, 26 the lack of evidence for what to do or not do for SCD pain management. At least 1 well‐done summary of the SCD evidence base intended for hospitalists has been published, but it focuses on management of issues other than pain.27

Motivations and Fears

It is not surprising then that hospitalists may bring great fear and apprehension with them into their care of SCD patients. One of us (W.R.S.), a general internist, has been called by his own and 3 other academic medical centers, 2 with active Federally‐funded SCD research programs, to address the problems of high‐utilizing adults with SCD, including counseling hospitalists frustrated with the management of pain in these patients.

Hospitalists may be motivated to provide efficient inpatient management (Table 2), and be aware of pain as the primary symptom of SCD inpatients. But they may carry knowledge gaps and biases into their relationships with SCD inpatients. They may fear opioid administration (opiophobia), loss of licensure or governmental reprisals because of high‐dose prescription of opioids, or may believe that SCD patients are more often addicted than most.17, 28 Consequently, more troublesome hospital stays may occur when patients are not rapidly and adequately titrated to appropriate analgesic doses, or when unnecessary deleterious side effects result from opioid and other analgesics. We therefore offer answers to frequently asked questions (FAQs) about pain management by hospitalists caring for adults with SCD. We address FAQs arising during the prototypical situationa patient with SCD admitted for a painful exacerbation, and little or no acute comorbidity. We refer the reader to the aforementioned articles and guidelines to address other treatment issues in adults with SCD.

FAQS

• 1

  Is there any objective way to tell when SCD patients really are in a crisis?

  Although the term crisis is used as if it were an objectively definable biological entity, no one has proposed a standard definition of a crisis based on pain intensity level, clinical features, or biomarkers. Measures of vaso‐occlusion are correlated with ischemic pain, including pain that is often called a crisis.2932 However, neither ischemic pain from SCD, nor the underlying vaso‐occlusive cascade that is associated with this pain, is a sudden, present‐or‐absent phenomenon. Instead, these are continua that can be measured using pain scales or various biomarkers.

  There is, however, correlative evidence of the intensity of SCD pain associated with various distinctive health states (admitted/not admitted, in crisis/not in crisis). The most visible measure of a crisis, health care utilization, was a strong predictor of mortality in the Cooperative Study of Sickle Cell Disease. Patients with 3 or more admissions per year had a lower 5‐year survival rate.33 In contrast, crisis in the landmark Pain in Sickle Cell Epidemiology Study (PiSCES) was self‐defined by patients.34 Despite being in pain on over half of their days on average, and despite a third of patients being in pain daily, most pain in PiSCES was not considered a crisis, and less than 5% of patients' days were spent in emergency departments (EDs) or hospitals. Ambulatory pain intensity reports were correlated with opioid use.35 A substantial minority (35%) of PiSCES patients made at least 3 ED visits per year. However, these high ED utilizers had worse laboratory values, more pain, more distress, and a lower quality of life.36

  Importantly, sometimes adults with SCD may have severe comorbidities which may not be addressed or may be mistakenly managed as an acute vaso‐occlusive episode without further investigation or timely specialist consultation. Although pain is primarily the individual's chief complaint, any potential relationship between the presence of medical comorbidities and pain should be examined when patients are admitted.

• 2

  How can one know when opioid dosages should be changed, or when SCD pain is appropriately controlled to allow discharge?

  We recommend, as a standard of care, that SCD pain assessment and pain therapy be interwoven, despite a systematic review finding no evidence that directly linked the timing, frequency, or method of pain assessment with outcomes or safety in medical inpatients, and concluding that the safety and effectiveness of patient‐controlled analgesia (PCA) in medical patients had not been adequately studied.37 Hospitalists should focus the first 24 hours of inpatient SCD pain management on cycles of recurrent pain assessment and opioid dose titration as frequently as every 1 to 2 hours, to assure safe and rapidly efficacious analgesia. Pain intensity, duration, and character should be assessed directly. Intensity is often assessed using a visual analog scale (VAS) or numeric rating scale.38 Treating physicians should themselves directly assess pain during discussions of therapy with the patient, even though some assessment usually is done in hospitals during each nursing shift. Pain and pain relief can be assessed indirectly by monitoring opioid use.

  We recommend PCA for inpatients with SCD, administered as an intermittent demand dose (patient must push a button) of opioid, with or without a background opioid constant infusion.39 We usually set the interval between doses, or lockout, to 6 to 10 minutes. Both the lockout and the sedation from delivered doses prevent patients from pushing the demand button repetitively to the point of overdose. Use of a low‐level constant infusion (basal) may sustain pain control during times when the patient is asleep, avoiding recrudescent pain and lost ground due to inadequate analgesia during rest. Alternatively, long‐acting oral opioids may be continued if already used at home, or newly introduced to provide adequate baseline pain control which is augmented by the demand dosing. Most PCA pumps monitor hourly opioid dose demand (number of pushes), as well as hourly doses delivered. Both hourly opioid dose demand and hourly dose‐per‐demand ratio are measures of PCA efficacy or futility. Pumps record this data, and can be interrogated at the patient's bedside for up to several days of prior use. Physicians should combine pump interrogation with direct pain assessment to guide demand‐dose titration. Demand doses should be increased to 1.5 to 2 times the previous demand dose after several hours of failed reduction of pain intensity and duration, and/or persistently futile dose‐per‐demand ratios.

  PCA interrogation is also useful for conversion of parenteral opioids to oral opioids, as well as to guide the recommendation for discharge home. After the first 24‐48 hours of up‐titration, if opioid dose demand decreases concordantly with pain frequency and intensity, the demand dose may be safely decreased, and eventually daily PCA requirements may be summed and converted to oral medication using standard opioid dose conversion tables. At this point, physicians may use single measures or daily averages of directly assessed pain.

  Routine PCA use in SCD is backed by some evidence.40, 41 But we find it important that patients be taught and encouraged to use the demand feature of PCA. Still, for various reasons, some patients do not use PCA pumps well. Discordant or unreliable assessments (eg, high pain intensity but low‐opioid demand doses during the same interval) may result, and PCA potentially may fail as a dosing strategy. Management is more difficult for these patients. One alternative dosing strategy is prescription of scheduled doses of a short‐acting opioid, attaching to each dose the order, patient may refuse. This is different than dosing as needed, and allows counts of dose refusals over an interval, analogous to PCA pump interrogation.

• 3

  How much is too much opioid? Should one rely on side effects, or on requests for medicine, or is there a ceiling dose?

  Addictionologists, pain specialists, oncologists, those involved in hospice care, and some hematologists caring for SCD patients agree that, in general, there should be no a priori dose limitations imposed on opioid prescribing for acute pain. Instead, titration of dose of opioid to pain relief is a central principle of acute pain management. Experts also agree that particular opioids carry particular side effects which warrant dose limitation, adjustments, or avoidance of that opioid altogether. A summary of opioids commonly used in SCD, along with warnings and implied dose limitations is found in Table 3.

  For safety, it is important to assess the history of prior opioid use to recognize a patient who is not tolerant to opioids (see below, FAQ 4), to avoid mistakenly overdosing a patient using doses often required by tolerant patients. In lieu of a pre‐written, individualized opioid dosing plan in place for the patient, the patient may be the best source of information regarding preferred medication and tolerated doses.

  The reader is referred to standard texts for a description of opioids, their pharmacokinetics and pharmacodynamics, and their addictive and abuse potential. The side‐effect profile of opioids is well‐known: nausea, vomiting, and itching frequently occur; hallucinations, respiratory suppression, and myoclonus occur infrequently.42 Meperidine may more readily cause central nervous system (CNS) dysfunction, including seizures, as compared to other commonly used opioids, because of its toxic metabolite nor‐meperidine. Use of meperidine is often avoided, especially use via PCA.43 Methadone may cause dysrhythmias, specificially corrected Q‐T interval (QT_c) prolongation and torsades de pointes on an electrocardiogram, in doses above 200 mg per day.44 Some recommend baseline and yearly electrocardiogram monitoring when giving methadone chronically.

  Recognizing the potential dangers of opioids, it is also reasonable to look for opioid‐sparing analgesic strategies. Non‐opioid analgesics such as ketorolac45 and adjuvants such as ketamine46 that are opioid‐sparing should be considered whenever feasible. Complementary and alternative therapies such as transcutaneous electrical nerve stimulation (TENS)47 have less evidence of effectiveness, but have limited risks and may be of use for some individuals.

• 4

  What are the major signs of substance abuse (opioids, street drugs) in SCD patients already on opioids, and can a hospitalist judge those signs acutely and intervene appropriately?

  Reports of underprescription of opioids in SCD have cited physician fear of abuse and addiction.48 A recent informal poll of adult sickle cell providers suggests policies vary on how potential abuse is monitored in ambulatory sickle cell patients. We note that physicians, especially upon meeting a patient for the first time, may be unable to reliably judge whether that patient is abusing opioids or street drugs. Both false‐positive and false‐negative diagnoses may be made.49 Repetitive reports of lost or stolen prescriptions or pill bottles, receipt of prescriptions from multiple providers, or repeated requests for early refills increase the suspicion of misuse or abuse, but are indirect evidence. Urine and serum monitoring may be useful, but may give incorrect information if misinterpreted or not conducted frequently enough to improve sensitivity.50

  It is important to distinguish between tolerance, the decreased analgesic response over time to repeated doses of the same drug; physical dependence, the production of withdrawal upon abrupt discontinuation of an opioid agonist or administration of an antagonist; and addiction, the psychological dependence upon opioids. Tolerance may be misperceived as true addiction. Its earliest symptom is shortening of the duration of effective analgesia. In contrast, addiction may be manifested by dose escalation in the absence of an increased pain stimulus, or by use of opioids for purposes other than pain relief.51 These are not easily distinguished during a single patient encounter.

  SCD patients' requests for specific opioid medications in specific doses, should not be taken as evidence of past or current abuse, but rather evidence of a well‐informed, self‐managing patient. Adults with SCD are clearly expected to be very knowledgeable about and tolerant to opioids if they have had a life of pain as a child, and will require higher doses of opioids than other patients treated by most hospitalists. The issue of medication abuse may be best handled in the ambulatory setting. Whenever possible, hospitalists should not rely only on data from the acute care setting to manage patients. Ambulatory providers may conduct random, unannounced urine and/or serum testing, as part of an opioid prescribing agreement that is written and filed in the patient's chart. Assays for prescribed opioids (especially long‐acting agents), as well as assays for common drugs of abuse, should be conducted. Comanagement with an addictionologist, psychiatrist, or psychologist should be considered in individuals suspected of opioid abuse.

  We do not suggest routine urine drug test monitoring of all SCD patients unless routine monitoring is done as a policy for all patients on opioids. Though the prevalence of addiction may be higher in subpopulations of patients with pain,52 and though prescription of opioids, prescription drug abuse, and accidental deaths from prescribed opioids have risen exponentially in the last several years,53 in our experience and in the published literature, drug misuse/abuse among SCD patients is no worse than among patients with other illnesses.5456 However, pseudoaddiction, the appropriate seeking of needed opioids from multiple physicians because of uncontrolled pain and opioid underprescription, may well be prevalent in SCD,57 and may be mistaken for true addiction.

• 5

  How can patients' readiness for discharge be assessed? What can be done for the patient who has lengthy and/or multiple hospitalizations or frequent ED visits?

  The appropriate time for discharge in most patients is when they can manage their pain at home with oral opioids or less. Often, patients do not improve even after a few days of inpatient therapy.58 A typical pain episode may last much longer than the 6‐day average US hospital length of stay for a diagnosis of sickle cell crisis among 18‐44 year olds (Table 1).59 Patients may return and be readmitted.60, 61 But in the best cases, pain resolves or reverts to a usual chronic intensity level. As described in FAQ 2, daily or more frequent pain assessment is a bedrock for making discharge decisions. Patients well‐experienced in the use of pain intensity scales can report their usual pain intensity at home, and how close they are to their baseline pain intensity. Simply asking patients, Are you ready for discharge? is appropriate and may yield a surprising positive response. In a recent inpatient trial of PCA (manuscript in preparation), adult patients were admitted with a minimum pain intensity of 45 mm on a 100 mm horizontal VAS scale after treatment in the ED, and mean pain intensity of 76 mm 10 mm. All adults in this study were discharged with pain that was clinically significantly lower. Researchers have found a VAS change of 13.5 mm to be the minimum clinically significant change62 during treatment of vaso‐occlusive crisis.63

  Unremitting pain despite appropriate titration of opioids and prolonged hospital stays suggests the need for comprehensive evaluation for medical and psychosocial comorbidities, as is done for other patients with chronic pain syndromes. If not already done, discussion with the patient's primary care provider may reveal factors impacting on persistent pain. Consultation with a hematologist, pain or palliative care specialist, or other provider familiar with SCD may prove helpful. Implementation of adjuvant therapies as discussed in FAQ 3 and adding long‐acting oral opioids to continue postdischarge may also help. Hyperalgesia, or heightened sensitivity to pain, is normal after acute tissue injury, but is now suspected in SCD as a long‐term neuropathic pain syndrome, as a consequence either of repeated painful crises or of chronic opioid therapy.2 Only some centers have specialists qualified to test for and diagnose neuropathic pain.64

  Discharge planning should include identification of a source of outpatient follow‐up. Opioids prescribed at discharge should be sufficient to last at least until the first outpatient appointment, to avoid repeated ED or hospital visits. Communication with a primary care provider at discharge can enhance successful care transition. Otherwise, for patients without established providers, social workers and others may address barriers to follow‐up that frustrate both patient and provider.

Support for Hospitalists Managing Adults With Sickle Cell Disease

Beside the general advice on pain management in SCD mentioned above or found in the bibliography of this article, at long last, a group of adult practitioners skilled in the care of SCD has formed nationally. The Sickle Cell Adult Provider Network [http://www.scapn.net] provides non‐binding advice and support to its members via an e‐mail listserve. Topics often include pain management. This advice fills a vacuum created by the lack of evidence‐based guidelines.

Ultimately, evidence and updated guidelines will be the best support for hospitalists and others managing pain in SCD. The hope is that SCD will receive the attention it deserves, so that practitioners and patients alike do not suffer continued pain from this disease or its management.