Setting

Departmental QI leaders developed this initiative in the Internal Medicine Residency Program at the MSH in New York City, New York. This residency program trains over 140 residents annually in categorical, preliminary, and research track positions, as well as an affiliated medicine/pediatrics program. The program's residents rotate at 3 clinic sites: a tertiary care hospital, a public safety‐net hospital, and a Veterans Affairs hospital. The QI program was only implemented at the MSH. Over 90% of the program's graduates go on to complete a subspecialty fellowship.

Intervention Description

The QI program was designed around a noon‐time quality improvement conference (QIC) occurring once every 4 weeks. In the weeks prior to the session, chief residents and a hospitalist mentor selected a case related to an inpatient care issue. Potential cases were solicited, and/or offered, from a range of sources including attending physicians, nurse managers, residents, and quality officers. Only cases from the teaching services were chosen. To ensure that participants on the case were able to recall relevant details, preference was given to more recent cases. A third‐year resident on an elective or outpatient block was chosen to investigate the case. To maximize the objectivity of the investigation, every effort was made to select a resident who was not involved in the care of the patient.

The resident was instructed to complete a limited RCA (fewer fact‐finding interviews and only 1 group meeting) and was directed to online resources.11 Each resident presenter worked closely with the chief residents and hospitalist mentor to identify appropriate strategies for collecting data and interviewing involved parties. If necessary, either due to volume of work or sensitivity of the case, the chief resident or hospitalist would assist with the data gathering. The resident contacted multiple parties involved in the patient care issue including, nurses, residents, attendings, pharmacists, social workers, and, if appropriate, the patient. The resident constructed a timeline for each case, and identified specific points in the patient care experience, where errors, near misses, or misunderstandings occurred. During the QIC, these findings were presented to Internal Medicine residents, chief residents, representatives from the Chief Medical Officer's office, attending physicians overseeing the residents on inpatient rotations, and representatives from any group (social work nursing, housekeeping, pharmacy, etc) that may have impacted patient care for the particular case being investigated. On average, 50 healthcare providers attended the QIC. Lunch was provided.

After the findings were presented, a chief resident and a lead hospitalist facilitated a group discussion on the circumstances surrounding the case. Discussions were focused on identifying system‐wide failures and proposing systems‐based solutions. Great efforts were made to remind all participants to refrain from individual blame. At the end of each QIC, participants summarized and prioritized suggestions to reduce the discussed error. Interested residents were invited to form improvement committees for cases with viable solutions. Each committee attempted to implement improvements discussed during the QIC. Committees, led by a representative from the Division of Hospital Medicine, included 2 to 4 residents as well as healthcare workers from other disciplines if appropriate. For all improvement efforts, the focus was on the interventions which appeared high yield with low cost.

Intervention Evaluation

The program was exempt from Institutional Review Board review as a part of the Department of Medicine's quality improvement and assurance portfolio.

The results of the QICs were tracked. After each case, a QI team consisting of chief residents and representatives from the Division of Hospital Medicine recorded the cases presented, and interventions suggested for each case, in an online database. After implementation, the success of each intervention was recorded. To evaluate the types of interventions suggested by residents, the 3 physician‐authors, who regularly attend these conferences and have a focused career interest in QI, grouped all suggestions into 4 broad categories: Educational, Reminder Systems, Design Changes (protocol‐based), and Design Changes (Information Technology [IT]‐based). Design change interventions (IT‐based) consisted of an adjustment to electronic systems, such as displaying specific lab results on a medication ordering system. Design changes (protocol‐based) consisted of changes made to standing protocols such as nursing protocols for reporting abnormal lab values. Reminder system interventions were endeavors such as a checklist for discharge planning. Educational interventions focused on providing additional training sessions or conferences.

The 3 physician‐authors independently reviewed each suggested intervention to determine its success. They first evaluated whether the change was attempted or not. For all attempted interventions, the reviewer then assessed if there was either objective permanent system‐wide change, subjective behavior change, or no change. To meet the objective change threshold, the intervention either had to have permanently changed provider workflow or have data demonstrating behavior change or improved outcome. Interventions with anecdotal evidence that behavior was improved or modified, but lacking systematic data, were qualified as subjective behavior change. For each assessment, 2 of the 3 reviewers needed to agree for an intervention to be recorded as a success.

Resident views on the monthly conferences were solicited via an anonymous and voluntary questionnaire. A first survey was designed to assess whether residents felt that the conferences provided them with the ability to recognize and improve systems errors which compromise patient care. This survey was administered at the conclusion of the first year of the program to residents who attended the final 2 QICs. A second survey assessed whether the tone of the conferences was constructive and blame‐free. This survey was administered at the conclusion of the second year of the program to residents who attended the year's final 2 QICs.

Lessons Learned

While this QI program has had success uncovering clinical care issues, and creating a climate and process for resident participation in improvement, there has been a number of limitations and lessons learned. Most importantly, including busy residents in any process that requires regular participation and follow‐through is difficult. A number of suggested improvements which created substantial interest and early momentum were ultimately left unfinished, as residents and even faculty facilitators became overwhelmed by clinical responsibilities. In fact, the majority of suggestions have not been successfully implemented and even fewer have created lasting change. This must be carefully monitored, as experiencing multiple failures can undermine the empowerment that such QI programs are created to foster.

Regular reflection on the successful and unsuccessful projects yielded several important insights that resulted in changes over the course of the program. Suggestions were more likely to move from idea generation to execution if the QIC was attended by administrators with decision‐making authority. Several of the suggestionsimproved medication reconciliation, better transfer documentation, and improved availability of infection control productswere able to be acted upon because conference attendees were administrators with purview over these issues. Many times, these leaders were more than willing to implement helpful suggestions, but simply needed them to be brought to their attention. As a result, we have been more attentive to inviting as many stakeholders as possible to the QICs.

It was also clear that suggestions would not be realized without a physician leader and were more successful when resident interest was substantial. After each QIC, residents who had made promising suggestions were approached to continue to participate. If the residents agreed, the projects were pursued and a faculty or chief resident leader was assigned. Lastly, we have also made use of one of the department's QI data analysts to assist with project completion. This individual has been made available to provide administrative support (organizing meetings, paperwork, etc) but also to provide data for projects, should the need arise.

Another important finding is that the tone of the QI program must be constantly monitored. Despite reminding residents at each session that the exercise was for the purpose of identifying systems barriers to delivering high quality care, there were times when residents felt targeted or blamed. At one point, a number of residents voiced their concerns that the conferences had spent too much time highlighting quality failures without recognizing the many positive performances on the teaching service. As a result, subsequent conferences often began by highlighting quality improvements made. Additionally, a part of 1 session each year had been dedicated to reading letters and e‐mails sent by patients or families which highlight memorably positive performances by the residents. Finally, care was taken to make sure invited guests to the sessions were reminded of the session's blame‐free ground rules.

Care must be taken when investigating clinical cases. On several occasions, attending physicians expressed discomfort with having residents scrutinize a clinical event. Although this process was protected under the QI umbrella and faculty names were never shared at the conferences, some faculty believed that this process was the purview of departmental or hospital QI staff, not untrained residents. Given the support provided for this program by the department chair and program director, as well as the professional nature with which the residents conducted their inquiries, there was little difficulty rejecting this line of objection. This feedback did lead supervisors to be more involved with the resident presenters, coaching them regarding data gathering and interviewing. If a case appears that it will be particularly sensitive, the hospitalist mentor or chief resident will reach out to involved residents and faculty to notify them that the case will be reviewed.

A final development secured, in part, as a result of this quality program has been more protected faculty time. At the start of this program, all faculty time was donated time on top of other administrative and patient care responsibilities. After the first 18 months of the QIC program, the residency program named an assistant program director for quality. At the time of writing this manuscript, the program further invested in quality by naming both an assistant and associate program director for quality. These positions combined amount to at least 0.4 full‐time equivalents (FTE). Of that, roughly 0.1 FTE is spent working on the QICs and subsequent project implementation.

Limitations

The evaluation of the success of the interventions potentially biased our findings. The qualitative method of using multiple reviewers, all of whom were invested in the program's outcomes, to gauge the success of initiated interventions may have resulted in an overestimate of the project's effectiveness. Furthermore, the category of subjective change lacks measurable criteria, making replication of the findings difficult.

The results presented here are from a single institution, conceived of and executed by a group of dedicated faculty. Moreover, both the chair of the department and the program director were very supportive of this endeavor. Possibly, because of these aspects, the findings presented here would not be readily replicated at another institution.

The percentage of residents who completed the feedback surveys was low. This may result in an overestimate of quality, value, and tone of the conferences, as well as potentially missing an opportunity for improving the program. We will address this issue through more rigorous quantitative and qualitative feedback at the end of the third year of the program.

Setting

Departmental QI leaders developed this initiative in the Internal Medicine Residency Program at the MSH in New York City, New York. This residency program trains over 140 residents annually in categorical, preliminary, and research track positions, as well as an affiliated medicine/pediatrics program. The program's residents rotate at 3 clinic sites: a tertiary care hospital, a public safety‐net hospital, and a Veterans Affairs hospital. The QI program was only implemented at the MSH. Over 90% of the program's graduates go on to complete a subspecialty fellowship.

Intervention Description

The QI program was designed around a noon‐time quality improvement conference (QIC) occurring once every 4 weeks. In the weeks prior to the session, chief residents and a hospitalist mentor selected a case related to an inpatient care issue. Potential cases were solicited, and/or offered, from a range of sources including attending physicians, nurse managers, residents, and quality officers. Only cases from the teaching services were chosen. To ensure that participants on the case were able to recall relevant details, preference was given to more recent cases. A third‐year resident on an elective or outpatient block was chosen to investigate the case. To maximize the objectivity of the investigation, every effort was made to select a resident who was not involved in the care of the patient.

The resident was instructed to complete a limited RCA (fewer fact‐finding interviews and only 1 group meeting) and was directed to online resources.11 Each resident presenter worked closely with the chief residents and hospitalist mentor to identify appropriate strategies for collecting data and interviewing involved parties. If necessary, either due to volume of work or sensitivity of the case, the chief resident or hospitalist would assist with the data gathering. The resident contacted multiple parties involved in the patient care issue including, nurses, residents, attendings, pharmacists, social workers, and, if appropriate, the patient. The resident constructed a timeline for each case, and identified specific points in the patient care experience, where errors, near misses, or misunderstandings occurred. During the QIC, these findings were presented to Internal Medicine residents, chief residents, representatives from the Chief Medical Officer's office, attending physicians overseeing the residents on inpatient rotations, and representatives from any group (social work nursing, housekeeping, pharmacy, etc) that may have impacted patient care for the particular case being investigated. On average, 50 healthcare providers attended the QIC. Lunch was provided.

After the findings were presented, a chief resident and a lead hospitalist facilitated a group discussion on the circumstances surrounding the case. Discussions were focused on identifying system‐wide failures and proposing systems‐based solutions. Great efforts were made to remind all participants to refrain from individual blame. At the end of each QIC, participants summarized and prioritized suggestions to reduce the discussed error. Interested residents were invited to form improvement committees for cases with viable solutions. Each committee attempted to implement improvements discussed during the QIC. Committees, led by a representative from the Division of Hospital Medicine, included 2 to 4 residents as well as healthcare workers from other disciplines if appropriate. For all improvement efforts, the focus was on the interventions which appeared high yield with low cost.

Intervention Evaluation

The program was exempt from Institutional Review Board review as a part of the Department of Medicine's quality improvement and assurance portfolio.

The results of the QICs were tracked. After each case, a QI team consisting of chief residents and representatives from the Division of Hospital Medicine recorded the cases presented, and interventions suggested for each case, in an online database. After implementation, the success of each intervention was recorded. To evaluate the types of interventions suggested by residents, the 3 physician‐authors, who regularly attend these conferences and have a focused career interest in QI, grouped all suggestions into 4 broad categories: Educational, Reminder Systems, Design Changes (protocol‐based), and Design Changes (Information Technology [IT]‐based). Design change interventions (IT‐based) consisted of an adjustment to electronic systems, such as displaying specific lab results on a medication ordering system. Design changes (protocol‐based) consisted of changes made to standing protocols such as nursing protocols for reporting abnormal lab values. Reminder system interventions were endeavors such as a checklist for discharge planning. Educational interventions focused on providing additional training sessions or conferences.

The 3 physician‐authors independently reviewed each suggested intervention to determine its success. They first evaluated whether the change was attempted or not. For all attempted interventions, the reviewer then assessed if there was either objective permanent system‐wide change, subjective behavior change, or no change. To meet the objective change threshold, the intervention either had to have permanently changed provider workflow or have data demonstrating behavior change or improved outcome. Interventions with anecdotal evidence that behavior was improved or modified, but lacking systematic data, were qualified as subjective behavior change. For each assessment, 2 of the 3 reviewers needed to agree for an intervention to be recorded as a success.

Resident views on the monthly conferences were solicited via an anonymous and voluntary questionnaire. A first survey was designed to assess whether residents felt that the conferences provided them with the ability to recognize and improve systems errors which compromise patient care. This survey was administered at the conclusion of the first year of the program to residents who attended the final 2 QICs. A second survey assessed whether the tone of the conferences was constructive and blame‐free. This survey was administered at the conclusion of the second year of the program to residents who attended the year's final 2 QICs.

Lessons Learned

While this QI program has had success uncovering clinical care issues, and creating a climate and process for resident participation in improvement, there has been a number of limitations and lessons learned. Most importantly, including busy residents in any process that requires regular participation and follow‐through is difficult. A number of suggested improvements which created substantial interest and early momentum were ultimately left unfinished, as residents and even faculty facilitators became overwhelmed by clinical responsibilities. In fact, the majority of suggestions have not been successfully implemented and even fewer have created lasting change. This must be carefully monitored, as experiencing multiple failures can undermine the empowerment that such QI programs are created to foster.

Regular reflection on the successful and unsuccessful projects yielded several important insights that resulted in changes over the course of the program. Suggestions were more likely to move from idea generation to execution if the QIC was attended by administrators with decision‐making authority. Several of the suggestionsimproved medication reconciliation, better transfer documentation, and improved availability of infection control productswere able to be acted upon because conference attendees were administrators with purview over these issues. Many times, these leaders were more than willing to implement helpful suggestions, but simply needed them to be brought to their attention. As a result, we have been more attentive to inviting as many stakeholders as possible to the QICs.

It was also clear that suggestions would not be realized without a physician leader and were more successful when resident interest was substantial. After each QIC, residents who had made promising suggestions were approached to continue to participate. If the residents agreed, the projects were pursued and a faculty or chief resident leader was assigned. Lastly, we have also made use of one of the department's QI data analysts to assist with project completion. This individual has been made available to provide administrative support (organizing meetings, paperwork, etc) but also to provide data for projects, should the need arise.

Another important finding is that the tone of the QI program must be constantly monitored. Despite reminding residents at each session that the exercise was for the purpose of identifying systems barriers to delivering high quality care, there were times when residents felt targeted or blamed. At one point, a number of residents voiced their concerns that the conferences had spent too much time highlighting quality failures without recognizing the many positive performances on the teaching service. As a result, subsequent conferences often began by highlighting quality improvements made. Additionally, a part of 1 session each year had been dedicated to reading letters and e‐mails sent by patients or families which highlight memorably positive performances by the residents. Finally, care was taken to make sure invited guests to the sessions were reminded of the session's blame‐free ground rules.

Care must be taken when investigating clinical cases. On several occasions, attending physicians expressed discomfort with having residents scrutinize a clinical event. Although this process was protected under the QI umbrella and faculty names were never shared at the conferences, some faculty believed that this process was the purview of departmental or hospital QI staff, not untrained residents. Given the support provided for this program by the department chair and program director, as well as the professional nature with which the residents conducted their inquiries, there was little difficulty rejecting this line of objection. This feedback did lead supervisors to be more involved with the resident presenters, coaching them regarding data gathering and interviewing. If a case appears that it will be particularly sensitive, the hospitalist mentor or chief resident will reach out to involved residents and faculty to notify them that the case will be reviewed.

A final development secured, in part, as a result of this quality program has been more protected faculty time. At the start of this program, all faculty time was donated time on top of other administrative and patient care responsibilities. After the first 18 months of the QIC program, the residency program named an assistant program director for quality. At the time of writing this manuscript, the program further invested in quality by naming both an assistant and associate program director for quality. These positions combined amount to at least 0.4 full‐time equivalents (FTE). Of that, roughly 0.1 FTE is spent working on the QICs and subsequent project implementation.

Limitations

The evaluation of the success of the interventions potentially biased our findings. The qualitative method of using multiple reviewers, all of whom were invested in the program's outcomes, to gauge the success of initiated interventions may have resulted in an overestimate of the project's effectiveness. Furthermore, the category of subjective change lacks measurable criteria, making replication of the findings difficult.

The results presented here are from a single institution, conceived of and executed by a group of dedicated faculty. Moreover, both the chair of the department and the program director were very supportive of this endeavor. Possibly, because of these aspects, the findings presented here would not be readily replicated at another institution.

The percentage of residents who completed the feedback surveys was low. This may result in an overestimate of quality, value, and tone of the conferences, as well as potentially missing an opportunity for improving the program. We will address this issue through more rigorous quantitative and qualitative feedback at the end of the third year of the program.

Setting

Departmental QI leaders developed this initiative in the Internal Medicine Residency Program at the MSH in New York City, New York. This residency program trains over 140 residents annually in categorical, preliminary, and research track positions, as well as an affiliated medicine/pediatrics program. The program's residents rotate at 3 clinic sites: a tertiary care hospital, a public safety‐net hospital, and a Veterans Affairs hospital. The QI program was only implemented at the MSH. Over 90% of the program's graduates go on to complete a subspecialty fellowship.

Intervention Description

The QI program was designed around a noon‐time quality improvement conference (QIC) occurring once every 4 weeks. In the weeks prior to the session, chief residents and a hospitalist mentor selected a case related to an inpatient care issue. Potential cases were solicited, and/or offered, from a range of sources including attending physicians, nurse managers, residents, and quality officers. Only cases from the teaching services were chosen. To ensure that participants on the case were able to recall relevant details, preference was given to more recent cases. A third‐year resident on an elective or outpatient block was chosen to investigate the case. To maximize the objectivity of the investigation, every effort was made to select a resident who was not involved in the care of the patient.

The resident was instructed to complete a limited RCA (fewer fact‐finding interviews and only 1 group meeting) and was directed to online resources.11 Each resident presenter worked closely with the chief residents and hospitalist mentor to identify appropriate strategies for collecting data and interviewing involved parties. If necessary, either due to volume of work or sensitivity of the case, the chief resident or hospitalist would assist with the data gathering. The resident contacted multiple parties involved in the patient care issue including, nurses, residents, attendings, pharmacists, social workers, and, if appropriate, the patient. The resident constructed a timeline for each case, and identified specific points in the patient care experience, where errors, near misses, or misunderstandings occurred. During the QIC, these findings were presented to Internal Medicine residents, chief residents, representatives from the Chief Medical Officer's office, attending physicians overseeing the residents on inpatient rotations, and representatives from any group (social work nursing, housekeeping, pharmacy, etc) that may have impacted patient care for the particular case being investigated. On average, 50 healthcare providers attended the QIC. Lunch was provided.

After the findings were presented, a chief resident and a lead hospitalist facilitated a group discussion on the circumstances surrounding the case. Discussions were focused on identifying system‐wide failures and proposing systems‐based solutions. Great efforts were made to remind all participants to refrain from individual blame. At the end of each QIC, participants summarized and prioritized suggestions to reduce the discussed error. Interested residents were invited to form improvement committees for cases with viable solutions. Each committee attempted to implement improvements discussed during the QIC. Committees, led by a representative from the Division of Hospital Medicine, included 2 to 4 residents as well as healthcare workers from other disciplines if appropriate. For all improvement efforts, the focus was on the interventions which appeared high yield with low cost.

Intervention Evaluation

The program was exempt from Institutional Review Board review as a part of the Department of Medicine's quality improvement and assurance portfolio.

The results of the QICs were tracked. After each case, a QI team consisting of chief residents and representatives from the Division of Hospital Medicine recorded the cases presented, and interventions suggested for each case, in an online database. After implementation, the success of each intervention was recorded. To evaluate the types of interventions suggested by residents, the 3 physician‐authors, who regularly attend these conferences and have a focused career interest in QI, grouped all suggestions into 4 broad categories: Educational, Reminder Systems, Design Changes (protocol‐based), and Design Changes (Information Technology [IT]‐based). Design change interventions (IT‐based) consisted of an adjustment to electronic systems, such as displaying specific lab results on a medication ordering system. Design changes (protocol‐based) consisted of changes made to standing protocols such as nursing protocols for reporting abnormal lab values. Reminder system interventions were endeavors such as a checklist for discharge planning. Educational interventions focused on providing additional training sessions or conferences.

The 3 physician‐authors independently reviewed each suggested intervention to determine its success. They first evaluated whether the change was attempted or not. For all attempted interventions, the reviewer then assessed if there was either objective permanent system‐wide change, subjective behavior change, or no change. To meet the objective change threshold, the intervention either had to have permanently changed provider workflow or have data demonstrating behavior change or improved outcome. Interventions with anecdotal evidence that behavior was improved or modified, but lacking systematic data, were qualified as subjective behavior change. For each assessment, 2 of the 3 reviewers needed to agree for an intervention to be recorded as a success.

Resident views on the monthly conferences were solicited via an anonymous and voluntary questionnaire. A first survey was designed to assess whether residents felt that the conferences provided them with the ability to recognize and improve systems errors which compromise patient care. This survey was administered at the conclusion of the first year of the program to residents who attended the final 2 QICs. A second survey assessed whether the tone of the conferences was constructive and blame‐free. This survey was administered at the conclusion of the second year of the program to residents who attended the year's final 2 QICs.

Lessons Learned

While this QI program has had success uncovering clinical care issues, and creating a climate and process for resident participation in improvement, there has been a number of limitations and lessons learned. Most importantly, including busy residents in any process that requires regular participation and follow‐through is difficult. A number of suggested improvements which created substantial interest and early momentum were ultimately left unfinished, as residents and even faculty facilitators became overwhelmed by clinical responsibilities. In fact, the majority of suggestions have not been successfully implemented and even fewer have created lasting change. This must be carefully monitored, as experiencing multiple failures can undermine the empowerment that such QI programs are created to foster.

Regular reflection on the successful and unsuccessful projects yielded several important insights that resulted in changes over the course of the program. Suggestions were more likely to move from idea generation to execution if the QIC was attended by administrators with decision‐making authority. Several of the suggestionsimproved medication reconciliation, better transfer documentation, and improved availability of infection control productswere able to be acted upon because conference attendees were administrators with purview over these issues. Many times, these leaders were more than willing to implement helpful suggestions, but simply needed them to be brought to their attention. As a result, we have been more attentive to inviting as many stakeholders as possible to the QICs.

It was also clear that suggestions would not be realized without a physician leader and were more successful when resident interest was substantial. After each QIC, residents who had made promising suggestions were approached to continue to participate. If the residents agreed, the projects were pursued and a faculty or chief resident leader was assigned. Lastly, we have also made use of one of the department's QI data analysts to assist with project completion. This individual has been made available to provide administrative support (organizing meetings, paperwork, etc) but also to provide data for projects, should the need arise.

Another important finding is that the tone of the QI program must be constantly monitored. Despite reminding residents at each session that the exercise was for the purpose of identifying systems barriers to delivering high quality care, there were times when residents felt targeted or blamed. At one point, a number of residents voiced their concerns that the conferences had spent too much time highlighting quality failures without recognizing the many positive performances on the teaching service. As a result, subsequent conferences often began by highlighting quality improvements made. Additionally, a part of 1 session each year had been dedicated to reading letters and e‐mails sent by patients or families which highlight memorably positive performances by the residents. Finally, care was taken to make sure invited guests to the sessions were reminded of the session's blame‐free ground rules.

Care must be taken when investigating clinical cases. On several occasions, attending physicians expressed discomfort with having residents scrutinize a clinical event. Although this process was protected under the QI umbrella and faculty names were never shared at the conferences, some faculty believed that this process was the purview of departmental or hospital QI staff, not untrained residents. Given the support provided for this program by the department chair and program director, as well as the professional nature with which the residents conducted their inquiries, there was little difficulty rejecting this line of objection. This feedback did lead supervisors to be more involved with the resident presenters, coaching them regarding data gathering and interviewing. If a case appears that it will be particularly sensitive, the hospitalist mentor or chief resident will reach out to involved residents and faculty to notify them that the case will be reviewed.

A final development secured, in part, as a result of this quality program has been more protected faculty time. At the start of this program, all faculty time was donated time on top of other administrative and patient care responsibilities. After the first 18 months of the QIC program, the residency program named an assistant program director for quality. At the time of writing this manuscript, the program further invested in quality by naming both an assistant and associate program director for quality. These positions combined amount to at least 0.4 full‐time equivalents (FTE). Of that, roughly 0.1 FTE is spent working on the QICs and subsequent project implementation.

Limitations

The evaluation of the success of the interventions potentially biased our findings. The qualitative method of using multiple reviewers, all of whom were invested in the program's outcomes, to gauge the success of initiated interventions may have resulted in an overestimate of the project's effectiveness. Furthermore, the category of subjective change lacks measurable criteria, making replication of the findings difficult.

The results presented here are from a single institution, conceived of and executed by a group of dedicated faculty. Moreover, both the chair of the department and the program director were very supportive of this endeavor. Possibly, because of these aspects, the findings presented here would not be readily replicated at another institution.

The percentage of residents who completed the feedback surveys was low. This may result in an overestimate of quality, value, and tone of the conferences, as well as potentially missing an opportunity for improving the program. We will address this issue through more rigorous quantitative and qualitative feedback at the end of the third year of the program.

Restrictive Procedures (Laparoscopic Adjustable Gastric Band and Sleeve Gastrectomy)

These procedures produce weight loss by reducing the size of the stomach or creating an obstruction in the proximal stomach, limiting the consumption of large quantities at one time. They produce early satiety, but patients may still consume a large volume of calorie‐dense liquids compromising weight loss.

Laparoscopic Adjustable Gastric Band

Laparoscopic adjustable gastric band (LAGB; Figure 1A) is the primary form of restrictive procedures with 2 Food and Drug Administration‐approved bands (Lap Band [Allergan, Inc; Irvine, CA] and REALIZE band [Ethicon Endo‐Surgery, Inc; Cincinnati, OH]). A cuff is inflated around the proximal stomach creating a gastric pouch approximately 15‐30 mL in size. A subcutaneous reservoir is attached to the cuff allowing adjustment to the degree of restriction.8 LAGB has replaced the vertical banded gastroplasty (VBG). It is less invasive, adjustable, and reversible (0.1% operative mortality rate). Weight loss is maintained with this procedure but is generally less, with a higher failure rate compared to the more common gastric bypass procedure (Table 1).3, 9 Complications may include band dysfunction (ie, slippage, erosion, infections), esophageal dilatation, balloon failure, and port malposition, with rates approaching 3%‐5% per year requiring removal or repair.10 Patients may also experience GERD symptoms, especially if the condition was present preoperatively. Progressive GERD symptoms should be investigated with an upper gastrointestinal (GI) series to ensure there is no band slippage, esophageal dilation, or dysfunction.

Figure 1

Table 1.Comparison of Outcomes for Various Procedures

	LAGB	Roux‐en‐Y Gastric Bypass	Biliopancreatic Diversion With and Without Duodenal Switch
NOTE: References: Buchwald et al.3; Kendrick and Dakin.9 Abbreviations: LAGB, laparoscopic adjustable gastric band.
Excess weight loss	48%	62%	70%
Resolution of diabetes	48%	84%	98%

Combination Procedures (Roux‐en‐Y Gastric Bypass and Biliopancreatic Diversion With and Without Duodenal Switch)

These procedures produce weight loss by decreasing caloric intake and altering digestion and absorption.

Gastrointestinal Complaints

One of the most common causes of hospital admission any time postoperatively is abdominal pain. A differential diagnosis of abdominal pain, nausea, and/or vomiting in the post‐bariatric surgery patient should include small bowel obstruction, hernias (internal or incisional), band complications, food intolerance, dietary noncompliance, ileus, mesenteric venous thrombosis, strictures (such as outlet obstruction or anastomotic stenosis), ulcers, esophagitis, cholelithiasis, dumping syndrome, and Roux stasis syndrome.20

A thorough history targeted at the relationship between symptoms and food intake, attention to the character and location of the pain, and a thorough physical exam (specifically the presence or absence of palpable tenderness, guarding, or rebound) is essential. The physical exam may be misleading in obese patients and, if radiographic studies cannot be performed secondary to patient size, surgical exploration may be needed soon after presentation. Therefore, even lacking an obvious surgical need, the bariatric surgeon should be notified of admission.

Improper food choice, and failure to slowly and adequately chew food, can result in emesis and digestive difficulty. Physical incompatibility with the small gastric pouch and gastric outlet obstructions can be caused by nondigestible foods (ie, breads, steak, raw vegetables). This highlights the importance of ordering the appropriate hospital diet.8 Specific gastric bypass hospital diets for all consistencies should reflect the mechanical limitations and carbohydrate/protein requirements of these patients.

Increased gallstone formation is observed in patients with rapid weight loss (1.5 kg/wk), especially following RYGB and less often after LAGB procedures (40% vs 20% over 3 years). Routine use of ursodiol during rapid weight loss (6 months after RYGB) reduces this complication to <5%.8

Stenosis or ulceration at the anastomotic site for RYGB can cause abdominal pain and vomiting. The incidence of stomal stenosis has been reported at 5%‐19% and typically occurs within the first 3 postoperative months.22 This problem is often amenable to endoscopic dilatation, unless a ring was used to reinforce the anastomosis. Ulceration has been reported in 1%‐16% of patients and is usually secondary to tobacco or non‐steroidal anti‐inflammatory drug (NSAID) use, H. pylori, fistula‐induced acid exposure, reaction to foreign material, or ischemia from tension and poor tissue perfusion.23, 24 Endoscopy can diagnose the presence of ulcers, with biopsies to rule out H. pylori infection. Cessation of NSAIDs and tobacco are critical. Medical management including proton pump inhibitors and/or sucralfate is sufficient for up to 95% of patients. Surgical revision is reserved for persistent ulcers associated with obstruction, pain, and/or bleeding.25

Dumping syndrome is a complex of post‐prandial symptoms occurring most commonly in the RYGB patients. As many as 44% of RYGB patients may experience this syndrome characterized by flushing, dizziness, abdominal distension, pain, nausea, vomiting, and/or diarrhea.26 Symptoms may result from the ingestion of large amounts of sugars which empty from the altered gastric pouch at an unregulated rate. This large osmotic load causes fluid shifts and surges in peptide hormone levels, resulting in symptoms which may reinforce adherence to the prescribed postoperative diet. It occurs shortly after a meal and resolves over hours. Dietary modifications, such as increased protein and fiber intake with decreased consumption of simple sugars, will ameliorate symptoms in many patients, with most seeing resolution after the first year.8, 27 Some patients experience hyperglycemia secondary to ingestion of simple carbohydrates, with hypoglycemia approximately 2 hours later (late dumping). In our experience, limiting carbohydrate intake to 30 grams at any meal usually alleviates post‐prandial hypoglycemia.

If the patient reports an absence of bowel movements and flatus, an ileus from chronic narcotic use or a mechanical small bowel obstruction secondary to internal hernias or adhesions (see Late Surgical Complications) must be investigated. Severe or prolonged pain, lasting longer than a few hours, is cause for alarm and should prompt aggressive evaluation and possibly exploratory surgery.

Excessive Weight Loss

In diagnosing postoperative excessive weight loss, it is important to understand average anticipated weight loss parameters. Compared to the values expected for RYGB, LAGB produces less weight loss and BPD with and without DS produces more (Table 227). Patients experiencing more rapid or prolonged weight loss should be investigated for bacterial overgrowth syndrome, short bowel syndrome, or other anatomic abnormalities.

Known risk factors for bacterial overgrowth, which are prominent in this population, include decreased gastric acidity and slowed intestinal transit time (ie, narcotic use). Patients may be asymptomatic or experience weight loss, abdominal bloating and/or pain, nausea, vomiting, and diarrhea. The diagnosis can be made with a hydrogen breath test or by obtaining quantitative cultures of jejunal secretions during endoscopy. Questions remain on how the normalized values of these tests are affected by the postoperative environment, and on how this syndrome may present or be treated if it affects the excluded intestine. Bacterial overgrowth may be an incidental finding and not the cause of the gastrointestinal complaints. Although data is limited, treatment typically consists of a 7‐10 day course of rifaximin 1200 mg/day (divided doses) and/or a trial of dietary modifications.20, 2831 These may include avoiding lactose and eating a high fat, low carbohydrate, low fiber diet, so nutrients are readily absorbed and not left for bacterial consumption.32

Short bowel syndrome (<100‐200 cm of intestinal tissue remaining and subsequent malabsorption) can occur after any extensive colonic resection or bypass of the intestine.33 This condition rarely results after an initial bariatric procedure; however, subsequent procedures for small bowel obstructions or intestinal ischemia may result in short bowel syndrome. Typical presentations include diarrhea, weight loss, and symptoms of vitamin and mineral deficiencies. Short bowel can also predispose patients to the development of bacterial overgrowth, further complicating weight loss. Management consists of nutritional supplementation, occasionally parenteral nutrition, and rarely reoperation to increase the length of the common channel.34 Avoidance of further bowel resection is crucial in preventing short bowel syndrome.33, 34 In the setting of carbohydrate malabsorption with concomitant bacterial overgrowth syndrome, production of d‐lactic acid causing a metabolic acidosis with encephalopathy has been reported.35

Once medical complications have been ruled out, it is prudent to evaluate for a psychological component such as anorexia nervosa. It is helpful to involve a qualified psychologist who is familiar with this population. Addictions to alcohol, gambling, and pain medications have been reported in the post‐bariatric surgery population as a substitute for food addiction.

Neurological Complications and Vitamin Deficiencies

Neurological complications develop months to years postoperatively, secondary to vitamin, mineral, and nutrient deficiencies that result from malabsorption or inadequate intake. An inpatient provider should be aware of the potential role these conditions may play in a hospitalized patient.

Peripheral neuropathy can develop secondary to several deficiencies, including vitamin B₁₂, thiamine, vitamin E, and copper. Their sources, deficiencies, and replacement regimens are presented in Table 3.3642 Thiamine deficiency, manifesting as Wernicke's encephalopathy, is particularly important in the postoperative patient with excessive vomiting. For prevention, we recommend all patients readmitted with vomiting and dehydration receive a banana bag or rally pack (thiamine 100 mg, folic acid 1 mg, multivitamin with iron and magnesium 3 g in one liter of D5 normal saline) over 4‐8 hours. Additional deficiencies after gastric bypass include folate, selenium, zinc, vitamin B₆, and riboflavin. A multivitamin with minerals will meet the needs of most patients. Multiple fat‐soluble vitamin deficiencies can occur with small bowel bacterial overgrowth or BPD.

Anemia

Iron deficiency affects 6%‐33% of patients after 1 year.43 Iron is preferentially absorbed in the duodenum and proximal jejunum which are bypassed postoperatively. The absence of gastric acid prevents conversion of ferric (Fe²⁺) to the absorbable ferrous (Fe³⁺) iron, further decreasing absorption.44 Ferritin reflects iron stores but is also an acute phase reactant and, therefore, may mask an underlying deficiency in an acutely ill hospitalized patient. A multivitamin with iron is recommended for all patients, but additional supplementation may be required for menstruating women.43 Parenteral administration may be necessary if oral supplements are not tolerated or are inadequately absorbed.44

Fractures and Osteomalacia

Calcium and vitamin D deficiencies are a significant problem in the bariatric surgery population, with resultant osteoporosis or osteomalacia and associated fractures.38, 43 Calcium is preferentially absorbed in the duodenum and proximal jejunum. Vitamin D is absorbed in the ileum or produced in the skin in response to ultraviolet B (UVB) radiation.45 Deficiency of vitamin D exacerbates calcium malabsorption, thereby causing secondary hyperparathyroidism, increased bone turnover, and osteomalacia. Dramatic weight loss can lead to bone loss, increasing the risk for osteoporosis and fractures.38 Hypocalcemia or osteomalacia may cause generalized bone pain, muscular weakness, tetany, and chronic musculoskeletal pain.45

Fat‐soluble vitamin deficiencies are more common in those undergoing malabsorptive versus restrictive procedures and, in the case of BPD, may be related to the length of the common channel.43 It is important to ensure that calcium and vitamin D levels are sufficient prior to surgery, and prior to starting any osteoporotic treatment such as bisphosphonates.45 We recommend at least 1200 mg of calcium citrate and 1000‐2000 IU of Vitamin D daily. Up to 50,000 IU weekly or daily may be required to correct deficiency and maintain sufficiency in this population.46, 47 Vitamin D₂ (ergocalciferol) or D₃ (cholecalciferol) can be used for supplementation. Cholecalciferol is preferred if given through a feeding tube because it is less prone to clogging the tube.46 With severe malabsorption, phototherapy may be necessary, as intravenous doses are often inadequate and intramuscular preparations require special compounding.46 Calcium carbonate requires acid for proper absorption, therefore calcium citrate may be preferred due to achlorhydria from gastric exclusion.

Restrictive Procedures (Laparoscopic Adjustable Gastric Band and Sleeve Gastrectomy)

These procedures produce weight loss by reducing the size of the stomach or creating an obstruction in the proximal stomach, limiting the consumption of large quantities at one time. They produce early satiety, but patients may still consume a large volume of calorie‐dense liquids compromising weight loss.

Laparoscopic Adjustable Gastric Band

Laparoscopic adjustable gastric band (LAGB; Figure 1A) is the primary form of restrictive procedures with 2 Food and Drug Administration‐approved bands (Lap Band [Allergan, Inc; Irvine, CA] and REALIZE band [Ethicon Endo‐Surgery, Inc; Cincinnati, OH]). A cuff is inflated around the proximal stomach creating a gastric pouch approximately 15‐30 mL in size. A subcutaneous reservoir is attached to the cuff allowing adjustment to the degree of restriction.8 LAGB has replaced the vertical banded gastroplasty (VBG). It is less invasive, adjustable, and reversible (0.1% operative mortality rate). Weight loss is maintained with this procedure but is generally less, with a higher failure rate compared to the more common gastric bypass procedure (Table 1).3, 9 Complications may include band dysfunction (ie, slippage, erosion, infections), esophageal dilatation, balloon failure, and port malposition, with rates approaching 3%‐5% per year requiring removal or repair.10 Patients may also experience GERD symptoms, especially if the condition was present preoperatively. Progressive GERD symptoms should be investigated with an upper gastrointestinal (GI) series to ensure there is no band slippage, esophageal dilation, or dysfunction.

Figure 1

Table 1.Comparison of Outcomes for Various Procedures

	LAGB	Roux‐en‐Y Gastric Bypass	Biliopancreatic Diversion With and Without Duodenal Switch
NOTE: References: Buchwald et al.3; Kendrick and Dakin.9 Abbreviations: LAGB, laparoscopic adjustable gastric band.
Excess weight loss	48%	62%	70%
Resolution of diabetes	48%	84%	98%

Combination Procedures (Roux‐en‐Y Gastric Bypass and Biliopancreatic Diversion With and Without Duodenal Switch)

These procedures produce weight loss by decreasing caloric intake and altering digestion and absorption.

Gastrointestinal Complaints

One of the most common causes of hospital admission any time postoperatively is abdominal pain. A differential diagnosis of abdominal pain, nausea, and/or vomiting in the post‐bariatric surgery patient should include small bowel obstruction, hernias (internal or incisional), band complications, food intolerance, dietary noncompliance, ileus, mesenteric venous thrombosis, strictures (such as outlet obstruction or anastomotic stenosis), ulcers, esophagitis, cholelithiasis, dumping syndrome, and Roux stasis syndrome.20

A thorough history targeted at the relationship between symptoms and food intake, attention to the character and location of the pain, and a thorough physical exam (specifically the presence or absence of palpable tenderness, guarding, or rebound) is essential. The physical exam may be misleading in obese patients and, if radiographic studies cannot be performed secondary to patient size, surgical exploration may be needed soon after presentation. Therefore, even lacking an obvious surgical need, the bariatric surgeon should be notified of admission.

Improper food choice, and failure to slowly and adequately chew food, can result in emesis and digestive difficulty. Physical incompatibility with the small gastric pouch and gastric outlet obstructions can be caused by nondigestible foods (ie, breads, steak, raw vegetables). This highlights the importance of ordering the appropriate hospital diet.8 Specific gastric bypass hospital diets for all consistencies should reflect the mechanical limitations and carbohydrate/protein requirements of these patients.

Increased gallstone formation is observed in patients with rapid weight loss (1.5 kg/wk), especially following RYGB and less often after LAGB procedures (40% vs 20% over 3 years). Routine use of ursodiol during rapid weight loss (6 months after RYGB) reduces this complication to <5%.8

Stenosis or ulceration at the anastomotic site for RYGB can cause abdominal pain and vomiting. The incidence of stomal stenosis has been reported at 5%‐19% and typically occurs within the first 3 postoperative months.22 This problem is often amenable to endoscopic dilatation, unless a ring was used to reinforce the anastomosis. Ulceration has been reported in 1%‐16% of patients and is usually secondary to tobacco or non‐steroidal anti‐inflammatory drug (NSAID) use, H. pylori, fistula‐induced acid exposure, reaction to foreign material, or ischemia from tension and poor tissue perfusion.23, 24 Endoscopy can diagnose the presence of ulcers, with biopsies to rule out H. pylori infection. Cessation of NSAIDs and tobacco are critical. Medical management including proton pump inhibitors and/or sucralfate is sufficient for up to 95% of patients. Surgical revision is reserved for persistent ulcers associated with obstruction, pain, and/or bleeding.25

Dumping syndrome is a complex of post‐prandial symptoms occurring most commonly in the RYGB patients. As many as 44% of RYGB patients may experience this syndrome characterized by flushing, dizziness, abdominal distension, pain, nausea, vomiting, and/or diarrhea.26 Symptoms may result from the ingestion of large amounts of sugars which empty from the altered gastric pouch at an unregulated rate. This large osmotic load causes fluid shifts and surges in peptide hormone levels, resulting in symptoms which may reinforce adherence to the prescribed postoperative diet. It occurs shortly after a meal and resolves over hours. Dietary modifications, such as increased protein and fiber intake with decreased consumption of simple sugars, will ameliorate symptoms in many patients, with most seeing resolution after the first year.8, 27 Some patients experience hyperglycemia secondary to ingestion of simple carbohydrates, with hypoglycemia approximately 2 hours later (late dumping). In our experience, limiting carbohydrate intake to 30 grams at any meal usually alleviates post‐prandial hypoglycemia.

If the patient reports an absence of bowel movements and flatus, an ileus from chronic narcotic use or a mechanical small bowel obstruction secondary to internal hernias or adhesions (see Late Surgical Complications) must be investigated. Severe or prolonged pain, lasting longer than a few hours, is cause for alarm and should prompt aggressive evaluation and possibly exploratory surgery.

Excessive Weight Loss

In diagnosing postoperative excessive weight loss, it is important to understand average anticipated weight loss parameters. Compared to the values expected for RYGB, LAGB produces less weight loss and BPD with and without DS produces more (Table 227). Patients experiencing more rapid or prolonged weight loss should be investigated for bacterial overgrowth syndrome, short bowel syndrome, or other anatomic abnormalities.

Known risk factors for bacterial overgrowth, which are prominent in this population, include decreased gastric acidity and slowed intestinal transit time (ie, narcotic use). Patients may be asymptomatic or experience weight loss, abdominal bloating and/or pain, nausea, vomiting, and diarrhea. The diagnosis can be made with a hydrogen breath test or by obtaining quantitative cultures of jejunal secretions during endoscopy. Questions remain on how the normalized values of these tests are affected by the postoperative environment, and on how this syndrome may present or be treated if it affects the excluded intestine. Bacterial overgrowth may be an incidental finding and not the cause of the gastrointestinal complaints. Although data is limited, treatment typically consists of a 7‐10 day course of rifaximin 1200 mg/day (divided doses) and/or a trial of dietary modifications.20, 2831 These may include avoiding lactose and eating a high fat, low carbohydrate, low fiber diet, so nutrients are readily absorbed and not left for bacterial consumption.32

Short bowel syndrome (<100‐200 cm of intestinal tissue remaining and subsequent malabsorption) can occur after any extensive colonic resection or bypass of the intestine.33 This condition rarely results after an initial bariatric procedure; however, subsequent procedures for small bowel obstructions or intestinal ischemia may result in short bowel syndrome. Typical presentations include diarrhea, weight loss, and symptoms of vitamin and mineral deficiencies. Short bowel can also predispose patients to the development of bacterial overgrowth, further complicating weight loss. Management consists of nutritional supplementation, occasionally parenteral nutrition, and rarely reoperation to increase the length of the common channel.34 Avoidance of further bowel resection is crucial in preventing short bowel syndrome.33, 34 In the setting of carbohydrate malabsorption with concomitant bacterial overgrowth syndrome, production of d‐lactic acid causing a metabolic acidosis with encephalopathy has been reported.35

Once medical complications have been ruled out, it is prudent to evaluate for a psychological component such as anorexia nervosa. It is helpful to involve a qualified psychologist who is familiar with this population. Addictions to alcohol, gambling, and pain medications have been reported in the post‐bariatric surgery population as a substitute for food addiction.

Neurological Complications and Vitamin Deficiencies

Neurological complications develop months to years postoperatively, secondary to vitamin, mineral, and nutrient deficiencies that result from malabsorption or inadequate intake. An inpatient provider should be aware of the potential role these conditions may play in a hospitalized patient.

Peripheral neuropathy can develop secondary to several deficiencies, including vitamin B₁₂, thiamine, vitamin E, and copper. Their sources, deficiencies, and replacement regimens are presented in Table 3.3642 Thiamine deficiency, manifesting as Wernicke's encephalopathy, is particularly important in the postoperative patient with excessive vomiting. For prevention, we recommend all patients readmitted with vomiting and dehydration receive a banana bag or rally pack (thiamine 100 mg, folic acid 1 mg, multivitamin with iron and magnesium 3 g in one liter of D5 normal saline) over 4‐8 hours. Additional deficiencies after gastric bypass include folate, selenium, zinc, vitamin B₆, and riboflavin. A multivitamin with minerals will meet the needs of most patients. Multiple fat‐soluble vitamin deficiencies can occur with small bowel bacterial overgrowth or BPD.

Anemia

Iron deficiency affects 6%‐33% of patients after 1 year.43 Iron is preferentially absorbed in the duodenum and proximal jejunum which are bypassed postoperatively. The absence of gastric acid prevents conversion of ferric (Fe²⁺) to the absorbable ferrous (Fe³⁺) iron, further decreasing absorption.44 Ferritin reflects iron stores but is also an acute phase reactant and, therefore, may mask an underlying deficiency in an acutely ill hospitalized patient. A multivitamin with iron is recommended for all patients, but additional supplementation may be required for menstruating women.43 Parenteral administration may be necessary if oral supplements are not tolerated or are inadequately absorbed.44

Fractures and Osteomalacia

Calcium and vitamin D deficiencies are a significant problem in the bariatric surgery population, with resultant osteoporosis or osteomalacia and associated fractures.38, 43 Calcium is preferentially absorbed in the duodenum and proximal jejunum. Vitamin D is absorbed in the ileum or produced in the skin in response to ultraviolet B (UVB) radiation.45 Deficiency of vitamin D exacerbates calcium malabsorption, thereby causing secondary hyperparathyroidism, increased bone turnover, and osteomalacia. Dramatic weight loss can lead to bone loss, increasing the risk for osteoporosis and fractures.38 Hypocalcemia or osteomalacia may cause generalized bone pain, muscular weakness, tetany, and chronic musculoskeletal pain.45

Fat‐soluble vitamin deficiencies are more common in those undergoing malabsorptive versus restrictive procedures and, in the case of BPD, may be related to the length of the common channel.43 It is important to ensure that calcium and vitamin D levels are sufficient prior to surgery, and prior to starting any osteoporotic treatment such as bisphosphonates.45 We recommend at least 1200 mg of calcium citrate and 1000‐2000 IU of Vitamin D daily. Up to 50,000 IU weekly or daily may be required to correct deficiency and maintain sufficiency in this population.46, 47 Vitamin D₂ (ergocalciferol) or D₃ (cholecalciferol) can be used for supplementation. Cholecalciferol is preferred if given through a feeding tube because it is less prone to clogging the tube.46 With severe malabsorption, phototherapy may be necessary, as intravenous doses are often inadequate and intramuscular preparations require special compounding.46 Calcium carbonate requires acid for proper absorption, therefore calcium citrate may be preferred due to achlorhydria from gastric exclusion.

Restrictive Procedures (Laparoscopic Adjustable Gastric Band and Sleeve Gastrectomy)

These procedures produce weight loss by reducing the size of the stomach or creating an obstruction in the proximal stomach, limiting the consumption of large quantities at one time. They produce early satiety, but patients may still consume a large volume of calorie‐dense liquids compromising weight loss.

Laparoscopic Adjustable Gastric Band

Laparoscopic adjustable gastric band (LAGB; Figure 1A) is the primary form of restrictive procedures with 2 Food and Drug Administration‐approved bands (Lap Band [Allergan, Inc; Irvine, CA] and REALIZE band [Ethicon Endo‐Surgery, Inc; Cincinnati, OH]). A cuff is inflated around the proximal stomach creating a gastric pouch approximately 15‐30 mL in size. A subcutaneous reservoir is attached to the cuff allowing adjustment to the degree of restriction.8 LAGB has replaced the vertical banded gastroplasty (VBG). It is less invasive, adjustable, and reversible (0.1% operative mortality rate). Weight loss is maintained with this procedure but is generally less, with a higher failure rate compared to the more common gastric bypass procedure (Table 1).3, 9 Complications may include band dysfunction (ie, slippage, erosion, infections), esophageal dilatation, balloon failure, and port malposition, with rates approaching 3%‐5% per year requiring removal or repair.10 Patients may also experience GERD symptoms, especially if the condition was present preoperatively. Progressive GERD symptoms should be investigated with an upper gastrointestinal (GI) series to ensure there is no band slippage, esophageal dilation, or dysfunction.

Figure 1

Table 1.Comparison of Outcomes for Various Procedures

	LAGB	Roux‐en‐Y Gastric Bypass	Biliopancreatic Diversion With and Without Duodenal Switch
NOTE: References: Buchwald et al.3; Kendrick and Dakin.9 Abbreviations: LAGB, laparoscopic adjustable gastric band.
Excess weight loss	48%	62%	70%
Resolution of diabetes	48%	84%	98%

Combination Procedures (Roux‐en‐Y Gastric Bypass and Biliopancreatic Diversion With and Without Duodenal Switch)

These procedures produce weight loss by decreasing caloric intake and altering digestion and absorption.

Gastrointestinal Complaints

One of the most common causes of hospital admission any time postoperatively is abdominal pain. A differential diagnosis of abdominal pain, nausea, and/or vomiting in the post‐bariatric surgery patient should include small bowel obstruction, hernias (internal or incisional), band complications, food intolerance, dietary noncompliance, ileus, mesenteric venous thrombosis, strictures (such as outlet obstruction or anastomotic stenosis), ulcers, esophagitis, cholelithiasis, dumping syndrome, and Roux stasis syndrome.20

A thorough history targeted at the relationship between symptoms and food intake, attention to the character and location of the pain, and a thorough physical exam (specifically the presence or absence of palpable tenderness, guarding, or rebound) is essential. The physical exam may be misleading in obese patients and, if radiographic studies cannot be performed secondary to patient size, surgical exploration may be needed soon after presentation. Therefore, even lacking an obvious surgical need, the bariatric surgeon should be notified of admission.

Improper food choice, and failure to slowly and adequately chew food, can result in emesis and digestive difficulty. Physical incompatibility with the small gastric pouch and gastric outlet obstructions can be caused by nondigestible foods (ie, breads, steak, raw vegetables). This highlights the importance of ordering the appropriate hospital diet.8 Specific gastric bypass hospital diets for all consistencies should reflect the mechanical limitations and carbohydrate/protein requirements of these patients.

Increased gallstone formation is observed in patients with rapid weight loss (1.5 kg/wk), especially following RYGB and less often after LAGB procedures (40% vs 20% over 3 years). Routine use of ursodiol during rapid weight loss (6 months after RYGB) reduces this complication to <5%.8

Stenosis or ulceration at the anastomotic site for RYGB can cause abdominal pain and vomiting. The incidence of stomal stenosis has been reported at 5%‐19% and typically occurs within the first 3 postoperative months.22 This problem is often amenable to endoscopic dilatation, unless a ring was used to reinforce the anastomosis. Ulceration has been reported in 1%‐16% of patients and is usually secondary to tobacco or non‐steroidal anti‐inflammatory drug (NSAID) use, H. pylori, fistula‐induced acid exposure, reaction to foreign material, or ischemia from tension and poor tissue perfusion.23, 24 Endoscopy can diagnose the presence of ulcers, with biopsies to rule out H. pylori infection. Cessation of NSAIDs and tobacco are critical. Medical management including proton pump inhibitors and/or sucralfate is sufficient for up to 95% of patients. Surgical revision is reserved for persistent ulcers associated with obstruction, pain, and/or bleeding.25

Dumping syndrome is a complex of post‐prandial symptoms occurring most commonly in the RYGB patients. As many as 44% of RYGB patients may experience this syndrome characterized by flushing, dizziness, abdominal distension, pain, nausea, vomiting, and/or diarrhea.26 Symptoms may result from the ingestion of large amounts of sugars which empty from the altered gastric pouch at an unregulated rate. This large osmotic load causes fluid shifts and surges in peptide hormone levels, resulting in symptoms which may reinforce adherence to the prescribed postoperative diet. It occurs shortly after a meal and resolves over hours. Dietary modifications, such as increased protein and fiber intake with decreased consumption of simple sugars, will ameliorate symptoms in many patients, with most seeing resolution after the first year.8, 27 Some patients experience hyperglycemia secondary to ingestion of simple carbohydrates, with hypoglycemia approximately 2 hours later (late dumping). In our experience, limiting carbohydrate intake to 30 grams at any meal usually alleviates post‐prandial hypoglycemia.

If the patient reports an absence of bowel movements and flatus, an ileus from chronic narcotic use or a mechanical small bowel obstruction secondary to internal hernias or adhesions (see Late Surgical Complications) must be investigated. Severe or prolonged pain, lasting longer than a few hours, is cause for alarm and should prompt aggressive evaluation and possibly exploratory surgery.

Excessive Weight Loss

In diagnosing postoperative excessive weight loss, it is important to understand average anticipated weight loss parameters. Compared to the values expected for RYGB, LAGB produces less weight loss and BPD with and without DS produces more (Table 227). Patients experiencing more rapid or prolonged weight loss should be investigated for bacterial overgrowth syndrome, short bowel syndrome, or other anatomic abnormalities.

Known risk factors for bacterial overgrowth, which are prominent in this population, include decreased gastric acidity and slowed intestinal transit time (ie, narcotic use). Patients may be asymptomatic or experience weight loss, abdominal bloating and/or pain, nausea, vomiting, and diarrhea. The diagnosis can be made with a hydrogen breath test or by obtaining quantitative cultures of jejunal secretions during endoscopy. Questions remain on how the normalized values of these tests are affected by the postoperative environment, and on how this syndrome may present or be treated if it affects the excluded intestine. Bacterial overgrowth may be an incidental finding and not the cause of the gastrointestinal complaints. Although data is limited, treatment typically consists of a 7‐10 day course of rifaximin 1200 mg/day (divided doses) and/or a trial of dietary modifications.20, 2831 These may include avoiding lactose and eating a high fat, low carbohydrate, low fiber diet, so nutrients are readily absorbed and not left for bacterial consumption.32

Short bowel syndrome (<100‐200 cm of intestinal tissue remaining and subsequent malabsorption) can occur after any extensive colonic resection or bypass of the intestine.33 This condition rarely results after an initial bariatric procedure; however, subsequent procedures for small bowel obstructions or intestinal ischemia may result in short bowel syndrome. Typical presentations include diarrhea, weight loss, and symptoms of vitamin and mineral deficiencies. Short bowel can also predispose patients to the development of bacterial overgrowth, further complicating weight loss. Management consists of nutritional supplementation, occasionally parenteral nutrition, and rarely reoperation to increase the length of the common channel.34 Avoidance of further bowel resection is crucial in preventing short bowel syndrome.33, 34 In the setting of carbohydrate malabsorption with concomitant bacterial overgrowth syndrome, production of d‐lactic acid causing a metabolic acidosis with encephalopathy has been reported.35

Once medical complications have been ruled out, it is prudent to evaluate for a psychological component such as anorexia nervosa. It is helpful to involve a qualified psychologist who is familiar with this population. Addictions to alcohol, gambling, and pain medications have been reported in the post‐bariatric surgery population as a substitute for food addiction.

Neurological Complications and Vitamin Deficiencies

Neurological complications develop months to years postoperatively, secondary to vitamin, mineral, and nutrient deficiencies that result from malabsorption or inadequate intake. An inpatient provider should be aware of the potential role these conditions may play in a hospitalized patient.

Peripheral neuropathy can develop secondary to several deficiencies, including vitamin B₁₂, thiamine, vitamin E, and copper. Their sources, deficiencies, and replacement regimens are presented in Table 3.3642 Thiamine deficiency, manifesting as Wernicke's encephalopathy, is particularly important in the postoperative patient with excessive vomiting. For prevention, we recommend all patients readmitted with vomiting and dehydration receive a banana bag or rally pack (thiamine 100 mg, folic acid 1 mg, multivitamin with iron and magnesium 3 g in one liter of D5 normal saline) over 4‐8 hours. Additional deficiencies after gastric bypass include folate, selenium, zinc, vitamin B₆, and riboflavin. A multivitamin with minerals will meet the needs of most patients. Multiple fat‐soluble vitamin deficiencies can occur with small bowel bacterial overgrowth or BPD.

Anemia

Iron deficiency affects 6%‐33% of patients after 1 year.43 Iron is preferentially absorbed in the duodenum and proximal jejunum which are bypassed postoperatively. The absence of gastric acid prevents conversion of ferric (Fe²⁺) to the absorbable ferrous (Fe³⁺) iron, further decreasing absorption.44 Ferritin reflects iron stores but is also an acute phase reactant and, therefore, may mask an underlying deficiency in an acutely ill hospitalized patient. A multivitamin with iron is recommended for all patients, but additional supplementation may be required for menstruating women.43 Parenteral administration may be necessary if oral supplements are not tolerated or are inadequately absorbed.44

Fractures and Osteomalacia

Calcium and vitamin D deficiencies are a significant problem in the bariatric surgery population, with resultant osteoporosis or osteomalacia and associated fractures.38, 43 Calcium is preferentially absorbed in the duodenum and proximal jejunum. Vitamin D is absorbed in the ileum or produced in the skin in response to ultraviolet B (UVB) radiation.45 Deficiency of vitamin D exacerbates calcium malabsorption, thereby causing secondary hyperparathyroidism, increased bone turnover, and osteomalacia. Dramatic weight loss can lead to bone loss, increasing the risk for osteoporosis and fractures.38 Hypocalcemia or osteomalacia may cause generalized bone pain, muscular weakness, tetany, and chronic musculoskeletal pain.45

Fat‐soluble vitamin deficiencies are more common in those undergoing malabsorptive versus restrictive procedures and, in the case of BPD, may be related to the length of the common channel.43 It is important to ensure that calcium and vitamin D levels are sufficient prior to surgery, and prior to starting any osteoporotic treatment such as bisphosphonates.45 We recommend at least 1200 mg of calcium citrate and 1000‐2000 IU of Vitamin D daily. Up to 50,000 IU weekly or daily may be required to correct deficiency and maintain sufficiency in this population.46, 47 Vitamin D₂ (ergocalciferol) or D₃ (cholecalciferol) can be used for supplementation. Cholecalciferol is preferred if given through a feeding tube because it is less prone to clogging the tube.46 With severe malabsorption, phototherapy may be necessary, as intravenous doses are often inadequate and intramuscular preparations require special compounding.46 Calcium carbonate requires acid for proper absorption, therefore calcium citrate may be preferred due to achlorhydria from gastric exclusion.

Inclusion and Exclusion Criteria

We performed a retrospective cohort study of patients at the Veterans Affairs (VA) Greater Los Angeles Healthcare System during 2 time periods (May 1, 2005‐April 30, 2006 and Jan 1, 2007‐Dec 31, 2007) when hospital guidelines recommended different vancomycin dosing regimens based on indication. During the first time period, the recommended target trough level was 10 mg/L, regardless of indication. In May 2006, target troughs were changed according to the following institutional guidelines: 8‐12 mg/L for cellulitis, urinary tract infection (UTI), and uncomplicated bacteremia; 10‐15 mg/L for endocarditis, osteomyelitis, and visceral abscesses; and 15‐20 mg/L for bacterial meningitis and pneumonia. The vancomycin manufacturers (American Pharmaceutical Partners (Schaumburg, IL) and Baxter (Deerfield, IL)) were the same during both time periods. Patient data was collected from the VA Computerized Patient Records System (CPRS) by 2 trained reviewers (K.K.P. and T.P.). All inpatients who received 5 days of intravenous vancomycin therapy during these time periods were identified via electronic pharmacy records. We then excluded all patients with serum creatinine >2.0 mg/L prior to starting vancomycin, no serum creatinine collected before or during receipt of vancomycin, no trough levels drawn while on vancomycin (or for patients experiencing nephrotoxicity, no trough levels drawn prior to nephrotoxicity onset), nephrotoxicity occurring before day 5 of vancomycin therapy, and receipt of concomitant amphotericin B.

Data Collection and Study Definitions

In patients who received multiple courses of vancomycin during the specified time period, only the first course starting on, or after, May 1, 2005 and lasting 5 days was analyzed. Data collected for each patient included age, sex, race, and comorbidities (diabetes mellitus, liver dysfunction, and active malignancy). Diabetes mellitus was defined as 2 fasting blood glucose levels >125, or receipt of insulin or other hypoglycemic medications during vancomycin treatment. Patients were considered to have liver disease if they had a prior diagnosis of cirrhosis, hepatic encephalopathy, or hepatic insufficiency, or if 2 of the following criteria were met: total bilirubin >2 mg/L, aspartate aminotransferase (AST) or alanine aminotransferase (ALT) >2 the upper limit of normal, or serum albumin <3 g/dL. Receipt of 1 dose of potentially nephrotoxic agents, including aminoglycosides, intravenous furosemide, intravenous trimethoprim‐sulfamethoxazole, intravenous contrast dye, potentially nephrotoxic chemotherapy, and vasopressors, were recorded beginning 72 hours prior to vancomycin therapy until onset of nephrotoxicity, or, if nephrotoxicity did not occur, the final vancomycin dose. Angiotensin‐converting enzyme inhibitors (ACE‐I) and non‐steroidal anti‐inflammatory drugs (NSAIDs) or aspirin were considered potentially nephrotoxic if they were newly started within 72 hours of vancomycin.

For each patient, the serum creatinine was recorded upon admission, within 24 hours of starting vancomycin, during vancomycin treatment, and at 24 hours and 72 hours following the final vancomycin dose. Serum creatinine was typically measured daily. Per institutional protocol, vancomycin trough levels were drawn 30‐60 minutes prior to the fourth dose, and again in 5‐7 days or with any large change in renal function. Extrapolated troughs were calculated by a pharmacist if levels were drawn outside of the 60‐minute time period. The highest trough and duration of therapy was documented for each patient. The mean trough was equal to the arithmetic mean of all troughs obtained during vancomycin administration until 72 hours following the final dose.

Outcome Analysis

The primary end point was the development of nephrotoxicity, which was defined as an increase in serum creatinine by either 0.5 mg/dL or 50% for at least 2 consecutive days after receipt of vancomycin, up to 72 hours after the final dose, compared to the last creatinine measured prior to vancomycin initiation. Patients who had a documented isolated increase in serum creatinine that resolved upon recheck within 24 hours were not classified as experiencing nephrotoxicity. In patients who developed nephrotoxicity, mean troughs, maximum troughs, duration of vancomycin treatment, and receipt of concomitant nephrotoxins were ascertained using data collected only before nephrotoxicity onset. Bivariate and multivariate models were subsequently constructed in order to determine risk factors for nephrotoxicity, using either mean or maximum trough achieved prior to nephrotoxicity for each patient.

Statistical Methods

Comparisons between the 2005‐2006 and 2007 groups were made using Student t test for continuous variables, Wilcoxon rank‐sum test for ordinal variables, and Fisher's exact test for nominal variables. Association of clinical variables with nephrotoxicity was assessed using bivariate logistic regression with subsequent multivariable logistic regression. We initially decided to use maximum vancomycin trough 15 mg/L as the vancomycin exposure variable of interest to include in multivariable models, as we felt that (1) trough 15 mg/L is clinically relevant given current guidelines that recommend aiming for trough 15 mg/L for treatment of most invasive staphylococcal disease,31 and (2) prior studies identified a single trough 15 mg/L as a possible risk factor for nephrotoxicity.9, 27, 29, 31 However, we also generated other multivariable models that included either maximum vancomycin trough 20 mg/L, mean vancomycin trough 15 mg/L, or mean vancomycin trough 20 mg/L, and models in which maximum and mean vancomycin troughs were treated as continuous variables. All variables were initially included in multivariable models; nonsignificant variables were removed from the models in a backwards stepwise fashion until likelihood ratio testing determined that removal of any variable was associated with likelihood ratio test P value <0.20 in comparing the full to reduced model. All calculated P values are two‐sided. All calculations were performed with STATA, version 10 (StataCorp, College Station, TX). This study was approved via expedited review by the Institutional Review Board of the VA Greater Los Angeles Healthcare System.

Comparison of 2005‐2006 Versus 2007 Cohorts

Of the 705 patients who were identified by pharmacy records to have received intravenous vancomycin, 348 patients remained after exclusion criteria were applied; the vast majority of patients were excluded because they received <5 days of vancomycin therapy. Of the 348 patients included in the study, 201 received vancomycin in 2005‐2006, and 147 received vancomycin in 2007 (Table 1). Mean vancomycin trough was significantly higher in 2007 than 2005‐2006 (average mean trough 13.2 mg/L 4.3 vs 9.7 mg/L 3.6; P < 0.0001), although median (8 vs 9 days) and mean (11.2 vs 12.2 days) duration of therapy was 1 day shorter in 2007 versus 2005‐2006. Age, sex, race, comorbidities, and indication for vancomycin use were similar between the 2 groups. The receipt of concomitant nephrotoxins was largely similar between the 2 time periods, with the primary exception being that a higher proportion of patients received intravenous contrast dye in 2007 (19%) than in 2005‐2006 (8.0%) (P = 0.003), and a lower proportion of patients received amikacin in 2007 (7.5%) than in 2005‐2006 (15%) (P = 0.043), though overall receipt of aminoglycosides was similar. Overall, nephrotoxicity was noted in 31 patients (8.9%), with similar incidence in 2005‐2006 (8.0%) and 2007 (10.2%) (P = 0.57). The median time to onset of nephrotoxicity was 7 days, with a median peak serum creatinine of 1.8 mg/dL.

Determination of Clinical Factors for Nephrotoxicity

Results of bivariate and multivariate analysis of clinical factors potentially associated with nephrotoxicity are displayed in Table 2. Among the 31 patients experiencing nephrotoxicity, the mean maximum vancomycin trough prior to nephrotoxicity onset was 14.9 mg/L, compared to 13.3 mg/L among those not experiencing nephrotoxicity (OR 1.03 for each 1 mg/L increment in mean trough, 95% confidence interval [CI] 0.98‐1.09; P = 0.21). While there was a trend toward patients with nephrotoxicity having a maximum trough 15 mg/L, it was not significant in either bivariate (OR 2.18, 95% CI 0.85‐5.63; P = 0.11) or multivariate (OR 2.05, 95% CI 0.91‐4.61; P = 0.082) analysis. The duration of vancomycin therapy was also not significantly associated with nephrotoxicity, both when evaluated as a continuous variable and when prolonged courses (14 days) were compared to short courses (between 5 and 14 days) of therapy. Other multivariable models were constructed that included maximum trough 20 mg/L, mean trough 15 mg/L, mean trough 20 mg/L, and maximum and mean trough as continuous variables; in all of these models, the vancomycin exposure variable of interest was not significant enough to remain in the final model after backwards elimination. The only factor significantly associated with nephrotoxicity in either bivariate or multivariate analysis was receipt of intravenous contrast dye (OR 3.64, 95% CI 1.52‐8.68; P = 0.004 in multivariate analysis).

Reversibility of Nephrotoxicity

Of the 31 patients with nephrotoxicity, 20 (64.5%) patients still met criteria for nephrotoxicity at the time of vancomycin discontinuation. Nephrotoxicity subsequently resolved in 10 of the 16 patients that were still nephrotoxic at the time of vancomycin discontinuation (4 patients did not have follow‐up creatinine checked within 72 hours of vancomycin discontinuation). Thus, overall reversibility of nephrotoxicity either prior to, or within, 72 hours of vancomycin discontinuation was 77.8% (21/27 patients). Of the 6 patients who remained persistently nephrotoxic at 72 hours, all had received concomitant nephrotoxins prior to the onset of nephrotoxicity, as compared to 15/21 (71.4%) patients whose nephrotoxicity resolved (P = 0.28 by Fisher's exact test). Only 1 persistently nephrotoxic patient required dialysis: a critically ill patient with multiorgan failure for whom care was withdrawn within 4 days of vancomycin discontinuation.

Inclusion and Exclusion Criteria

We performed a retrospective cohort study of patients at the Veterans Affairs (VA) Greater Los Angeles Healthcare System during 2 time periods (May 1, 2005‐April 30, 2006 and Jan 1, 2007‐Dec 31, 2007) when hospital guidelines recommended different vancomycin dosing regimens based on indication. During the first time period, the recommended target trough level was 10 mg/L, regardless of indication. In May 2006, target troughs were changed according to the following institutional guidelines: 8‐12 mg/L for cellulitis, urinary tract infection (UTI), and uncomplicated bacteremia; 10‐15 mg/L for endocarditis, osteomyelitis, and visceral abscesses; and 15‐20 mg/L for bacterial meningitis and pneumonia. The vancomycin manufacturers (American Pharmaceutical Partners (Schaumburg, IL) and Baxter (Deerfield, IL)) were the same during both time periods. Patient data was collected from the VA Computerized Patient Records System (CPRS) by 2 trained reviewers (K.K.P. and T.P.). All inpatients who received 5 days of intravenous vancomycin therapy during these time periods were identified via electronic pharmacy records. We then excluded all patients with serum creatinine >2.0 mg/L prior to starting vancomycin, no serum creatinine collected before or during receipt of vancomycin, no trough levels drawn while on vancomycin (or for patients experiencing nephrotoxicity, no trough levels drawn prior to nephrotoxicity onset), nephrotoxicity occurring before day 5 of vancomycin therapy, and receipt of concomitant amphotericin B.

Data Collection and Study Definitions

In patients who received multiple courses of vancomycin during the specified time period, only the first course starting on, or after, May 1, 2005 and lasting 5 days was analyzed. Data collected for each patient included age, sex, race, and comorbidities (diabetes mellitus, liver dysfunction, and active malignancy). Diabetes mellitus was defined as 2 fasting blood glucose levels >125, or receipt of insulin or other hypoglycemic medications during vancomycin treatment. Patients were considered to have liver disease if they had a prior diagnosis of cirrhosis, hepatic encephalopathy, or hepatic insufficiency, or if 2 of the following criteria were met: total bilirubin >2 mg/L, aspartate aminotransferase (AST) or alanine aminotransferase (ALT) >2 the upper limit of normal, or serum albumin <3 g/dL. Receipt of 1 dose of potentially nephrotoxic agents, including aminoglycosides, intravenous furosemide, intravenous trimethoprim‐sulfamethoxazole, intravenous contrast dye, potentially nephrotoxic chemotherapy, and vasopressors, were recorded beginning 72 hours prior to vancomycin therapy until onset of nephrotoxicity, or, if nephrotoxicity did not occur, the final vancomycin dose. Angiotensin‐converting enzyme inhibitors (ACE‐I) and non‐steroidal anti‐inflammatory drugs (NSAIDs) or aspirin were considered potentially nephrotoxic if they were newly started within 72 hours of vancomycin.

For each patient, the serum creatinine was recorded upon admission, within 24 hours of starting vancomycin, during vancomycin treatment, and at 24 hours and 72 hours following the final vancomycin dose. Serum creatinine was typically measured daily. Per institutional protocol, vancomycin trough levels were drawn 30‐60 minutes prior to the fourth dose, and again in 5‐7 days or with any large change in renal function. Extrapolated troughs were calculated by a pharmacist if levels were drawn outside of the 60‐minute time period. The highest trough and duration of therapy was documented for each patient. The mean trough was equal to the arithmetic mean of all troughs obtained during vancomycin administration until 72 hours following the final dose.

Outcome Analysis

The primary end point was the development of nephrotoxicity, which was defined as an increase in serum creatinine by either 0.5 mg/dL or 50% for at least 2 consecutive days after receipt of vancomycin, up to 72 hours after the final dose, compared to the last creatinine measured prior to vancomycin initiation. Patients who had a documented isolated increase in serum creatinine that resolved upon recheck within 24 hours were not classified as experiencing nephrotoxicity. In patients who developed nephrotoxicity, mean troughs, maximum troughs, duration of vancomycin treatment, and receipt of concomitant nephrotoxins were ascertained using data collected only before nephrotoxicity onset. Bivariate and multivariate models were subsequently constructed in order to determine risk factors for nephrotoxicity, using either mean or maximum trough achieved prior to nephrotoxicity for each patient.

Statistical Methods

Comparisons between the 2005‐2006 and 2007 groups were made using Student t test for continuous variables, Wilcoxon rank‐sum test for ordinal variables, and Fisher's exact test for nominal variables. Association of clinical variables with nephrotoxicity was assessed using bivariate logistic regression with subsequent multivariable logistic regression. We initially decided to use maximum vancomycin trough 15 mg/L as the vancomycin exposure variable of interest to include in multivariable models, as we felt that (1) trough 15 mg/L is clinically relevant given current guidelines that recommend aiming for trough 15 mg/L for treatment of most invasive staphylococcal disease,31 and (2) prior studies identified a single trough 15 mg/L as a possible risk factor for nephrotoxicity.9, 27, 29, 31 However, we also generated other multivariable models that included either maximum vancomycin trough 20 mg/L, mean vancomycin trough 15 mg/L, or mean vancomycin trough 20 mg/L, and models in which maximum and mean vancomycin troughs were treated as continuous variables. All variables were initially included in multivariable models; nonsignificant variables were removed from the models in a backwards stepwise fashion until likelihood ratio testing determined that removal of any variable was associated with likelihood ratio test P value <0.20 in comparing the full to reduced model. All calculated P values are two‐sided. All calculations were performed with STATA, version 10 (StataCorp, College Station, TX). This study was approved via expedited review by the Institutional Review Board of the VA Greater Los Angeles Healthcare System.

Comparison of 2005‐2006 Versus 2007 Cohorts

Of the 705 patients who were identified by pharmacy records to have received intravenous vancomycin, 348 patients remained after exclusion criteria were applied; the vast majority of patients were excluded because they received <5 days of vancomycin therapy. Of the 348 patients included in the study, 201 received vancomycin in 2005‐2006, and 147 received vancomycin in 2007 (Table 1). Mean vancomycin trough was significantly higher in 2007 than 2005‐2006 (average mean trough 13.2 mg/L 4.3 vs 9.7 mg/L 3.6; P < 0.0001), although median (8 vs 9 days) and mean (11.2 vs 12.2 days) duration of therapy was 1 day shorter in 2007 versus 2005‐2006. Age, sex, race, comorbidities, and indication for vancomycin use were similar between the 2 groups. The receipt of concomitant nephrotoxins was largely similar between the 2 time periods, with the primary exception being that a higher proportion of patients received intravenous contrast dye in 2007 (19%) than in 2005‐2006 (8.0%) (P = 0.003), and a lower proportion of patients received amikacin in 2007 (7.5%) than in 2005‐2006 (15%) (P = 0.043), though overall receipt of aminoglycosides was similar. Overall, nephrotoxicity was noted in 31 patients (8.9%), with similar incidence in 2005‐2006 (8.0%) and 2007 (10.2%) (P = 0.57). The median time to onset of nephrotoxicity was 7 days, with a median peak serum creatinine of 1.8 mg/dL.

Determination of Clinical Factors for Nephrotoxicity

Results of bivariate and multivariate analysis of clinical factors potentially associated with nephrotoxicity are displayed in Table 2. Among the 31 patients experiencing nephrotoxicity, the mean maximum vancomycin trough prior to nephrotoxicity onset was 14.9 mg/L, compared to 13.3 mg/L among those not experiencing nephrotoxicity (OR 1.03 for each 1 mg/L increment in mean trough, 95% confidence interval [CI] 0.98‐1.09; P = 0.21). While there was a trend toward patients with nephrotoxicity having a maximum trough 15 mg/L, it was not significant in either bivariate (OR 2.18, 95% CI 0.85‐5.63; P = 0.11) or multivariate (OR 2.05, 95% CI 0.91‐4.61; P = 0.082) analysis. The duration of vancomycin therapy was also not significantly associated with nephrotoxicity, both when evaluated as a continuous variable and when prolonged courses (14 days) were compared to short courses (between 5 and 14 days) of therapy. Other multivariable models were constructed that included maximum trough 20 mg/L, mean trough 15 mg/L, mean trough 20 mg/L, and maximum and mean trough as continuous variables; in all of these models, the vancomycin exposure variable of interest was not significant enough to remain in the final model after backwards elimination. The only factor significantly associated with nephrotoxicity in either bivariate or multivariate analysis was receipt of intravenous contrast dye (OR 3.64, 95% CI 1.52‐8.68; P = 0.004 in multivariate analysis).

Reversibility of Nephrotoxicity

Of the 31 patients with nephrotoxicity, 20 (64.5%) patients still met criteria for nephrotoxicity at the time of vancomycin discontinuation. Nephrotoxicity subsequently resolved in 10 of the 16 patients that were still nephrotoxic at the time of vancomycin discontinuation (4 patients did not have follow‐up creatinine checked within 72 hours of vancomycin discontinuation). Thus, overall reversibility of nephrotoxicity either prior to, or within, 72 hours of vancomycin discontinuation was 77.8% (21/27 patients). Of the 6 patients who remained persistently nephrotoxic at 72 hours, all had received concomitant nephrotoxins prior to the onset of nephrotoxicity, as compared to 15/21 (71.4%) patients whose nephrotoxicity resolved (P = 0.28 by Fisher's exact test). Only 1 persistently nephrotoxic patient required dialysis: a critically ill patient with multiorgan failure for whom care was withdrawn within 4 days of vancomycin discontinuation.

Inclusion and Exclusion Criteria

We performed a retrospective cohort study of patients at the Veterans Affairs (VA) Greater Los Angeles Healthcare System during 2 time periods (May 1, 2005‐April 30, 2006 and Jan 1, 2007‐Dec 31, 2007) when hospital guidelines recommended different vancomycin dosing regimens based on indication. During the first time period, the recommended target trough level was 10 mg/L, regardless of indication. In May 2006, target troughs were changed according to the following institutional guidelines: 8‐12 mg/L for cellulitis, urinary tract infection (UTI), and uncomplicated bacteremia; 10‐15 mg/L for endocarditis, osteomyelitis, and visceral abscesses; and 15‐20 mg/L for bacterial meningitis and pneumonia. The vancomycin manufacturers (American Pharmaceutical Partners (Schaumburg, IL) and Baxter (Deerfield, IL)) were the same during both time periods. Patient data was collected from the VA Computerized Patient Records System (CPRS) by 2 trained reviewers (K.K.P. and T.P.). All inpatients who received 5 days of intravenous vancomycin therapy during these time periods were identified via electronic pharmacy records. We then excluded all patients with serum creatinine >2.0 mg/L prior to starting vancomycin, no serum creatinine collected before or during receipt of vancomycin, no trough levels drawn while on vancomycin (or for patients experiencing nephrotoxicity, no trough levels drawn prior to nephrotoxicity onset), nephrotoxicity occurring before day 5 of vancomycin therapy, and receipt of concomitant amphotericin B.

Data Collection and Study Definitions

In patients who received multiple courses of vancomycin during the specified time period, only the first course starting on, or after, May 1, 2005 and lasting 5 days was analyzed. Data collected for each patient included age, sex, race, and comorbidities (diabetes mellitus, liver dysfunction, and active malignancy). Diabetes mellitus was defined as 2 fasting blood glucose levels >125, or receipt of insulin or other hypoglycemic medications during vancomycin treatment. Patients were considered to have liver disease if they had a prior diagnosis of cirrhosis, hepatic encephalopathy, or hepatic insufficiency, or if 2 of the following criteria were met: total bilirubin >2 mg/L, aspartate aminotransferase (AST) or alanine aminotransferase (ALT) >2 the upper limit of normal, or serum albumin <3 g/dL. Receipt of 1 dose of potentially nephrotoxic agents, including aminoglycosides, intravenous furosemide, intravenous trimethoprim‐sulfamethoxazole, intravenous contrast dye, potentially nephrotoxic chemotherapy, and vasopressors, were recorded beginning 72 hours prior to vancomycin therapy until onset of nephrotoxicity, or, if nephrotoxicity did not occur, the final vancomycin dose. Angiotensin‐converting enzyme inhibitors (ACE‐I) and non‐steroidal anti‐inflammatory drugs (NSAIDs) or aspirin were considered potentially nephrotoxic if they were newly started within 72 hours of vancomycin.

For each patient, the serum creatinine was recorded upon admission, within 24 hours of starting vancomycin, during vancomycin treatment, and at 24 hours and 72 hours following the final vancomycin dose. Serum creatinine was typically measured daily. Per institutional protocol, vancomycin trough levels were drawn 30‐60 minutes prior to the fourth dose, and again in 5‐7 days or with any large change in renal function. Extrapolated troughs were calculated by a pharmacist if levels were drawn outside of the 60‐minute time period. The highest trough and duration of therapy was documented for each patient. The mean trough was equal to the arithmetic mean of all troughs obtained during vancomycin administration until 72 hours following the final dose.

Outcome Analysis

The primary end point was the development of nephrotoxicity, which was defined as an increase in serum creatinine by either 0.5 mg/dL or 50% for at least 2 consecutive days after receipt of vancomycin, up to 72 hours after the final dose, compared to the last creatinine measured prior to vancomycin initiation. Patients who had a documented isolated increase in serum creatinine that resolved upon recheck within 24 hours were not classified as experiencing nephrotoxicity. In patients who developed nephrotoxicity, mean troughs, maximum troughs, duration of vancomycin treatment, and receipt of concomitant nephrotoxins were ascertained using data collected only before nephrotoxicity onset. Bivariate and multivariate models were subsequently constructed in order to determine risk factors for nephrotoxicity, using either mean or maximum trough achieved prior to nephrotoxicity for each patient.

Statistical Methods

Comparisons between the 2005‐2006 and 2007 groups were made using Student t test for continuous variables, Wilcoxon rank‐sum test for ordinal variables, and Fisher's exact test for nominal variables. Association of clinical variables with nephrotoxicity was assessed using bivariate logistic regression with subsequent multivariable logistic regression. We initially decided to use maximum vancomycin trough 15 mg/L as the vancomycin exposure variable of interest to include in multivariable models, as we felt that (1) trough 15 mg/L is clinically relevant given current guidelines that recommend aiming for trough 15 mg/L for treatment of most invasive staphylococcal disease,31 and (2) prior studies identified a single trough 15 mg/L as a possible risk factor for nephrotoxicity.9, 27, 29, 31 However, we also generated other multivariable models that included either maximum vancomycin trough 20 mg/L, mean vancomycin trough 15 mg/L, or mean vancomycin trough 20 mg/L, and models in which maximum and mean vancomycin troughs were treated as continuous variables. All variables were initially included in multivariable models; nonsignificant variables were removed from the models in a backwards stepwise fashion until likelihood ratio testing determined that removal of any variable was associated with likelihood ratio test P value <0.20 in comparing the full to reduced model. All calculated P values are two‐sided. All calculations were performed with STATA, version 10 (StataCorp, College Station, TX). This study was approved via expedited review by the Institutional Review Board of the VA Greater Los Angeles Healthcare System.

Comparison of 2005‐2006 Versus 2007 Cohorts

Of the 705 patients who were identified by pharmacy records to have received intravenous vancomycin, 348 patients remained after exclusion criteria were applied; the vast majority of patients were excluded because they received <5 days of vancomycin therapy. Of the 348 patients included in the study, 201 received vancomycin in 2005‐2006, and 147 received vancomycin in 2007 (Table 1). Mean vancomycin trough was significantly higher in 2007 than 2005‐2006 (average mean trough 13.2 mg/L 4.3 vs 9.7 mg/L 3.6; P < 0.0001), although median (8 vs 9 days) and mean (11.2 vs 12.2 days) duration of therapy was 1 day shorter in 2007 versus 2005‐2006. Age, sex, race, comorbidities, and indication for vancomycin use were similar between the 2 groups. The receipt of concomitant nephrotoxins was largely similar between the 2 time periods, with the primary exception being that a higher proportion of patients received intravenous contrast dye in 2007 (19%) than in 2005‐2006 (8.0%) (P = 0.003), and a lower proportion of patients received amikacin in 2007 (7.5%) than in 2005‐2006 (15%) (P = 0.043), though overall receipt of aminoglycosides was similar. Overall, nephrotoxicity was noted in 31 patients (8.9%), with similar incidence in 2005‐2006 (8.0%) and 2007 (10.2%) (P = 0.57). The median time to onset of nephrotoxicity was 7 days, with a median peak serum creatinine of 1.8 mg/dL.

Determination of Clinical Factors for Nephrotoxicity

Results of bivariate and multivariate analysis of clinical factors potentially associated with nephrotoxicity are displayed in Table 2. Among the 31 patients experiencing nephrotoxicity, the mean maximum vancomycin trough prior to nephrotoxicity onset was 14.9 mg/L, compared to 13.3 mg/L among those not experiencing nephrotoxicity (OR 1.03 for each 1 mg/L increment in mean trough, 95% confidence interval [CI] 0.98‐1.09; P = 0.21). While there was a trend toward patients with nephrotoxicity having a maximum trough 15 mg/L, it was not significant in either bivariate (OR 2.18, 95% CI 0.85‐5.63; P = 0.11) or multivariate (OR 2.05, 95% CI 0.91‐4.61; P = 0.082) analysis. The duration of vancomycin therapy was also not significantly associated with nephrotoxicity, both when evaluated as a continuous variable and when prolonged courses (14 days) were compared to short courses (between 5 and 14 days) of therapy. Other multivariable models were constructed that included maximum trough 20 mg/L, mean trough 15 mg/L, mean trough 20 mg/L, and maximum and mean trough as continuous variables; in all of these models, the vancomycin exposure variable of interest was not significant enough to remain in the final model after backwards elimination. The only factor significantly associated with nephrotoxicity in either bivariate or multivariate analysis was receipt of intravenous contrast dye (OR 3.64, 95% CI 1.52‐8.68; P = 0.004 in multivariate analysis).

Reversibility of Nephrotoxicity

Of the 31 patients with nephrotoxicity, 20 (64.5%) patients still met criteria for nephrotoxicity at the time of vancomycin discontinuation. Nephrotoxicity subsequently resolved in 10 of the 16 patients that were still nephrotoxic at the time of vancomycin discontinuation (4 patients did not have follow‐up creatinine checked within 72 hours of vancomycin discontinuation). Thus, overall reversibility of nephrotoxicity either prior to, or within, 72 hours of vancomycin discontinuation was 77.8% (21/27 patients). Of the 6 patients who remained persistently nephrotoxic at 72 hours, all had received concomitant nephrotoxins prior to the onset of nephrotoxicity, as compared to 15/21 (71.4%) patients whose nephrotoxicity resolved (P = 0.28 by Fisher's exact test). Only 1 persistently nephrotoxic patient required dialysis: a critically ill patient with multiorgan failure for whom care was withdrawn within 4 days of vancomycin discontinuation.