Patient‐centered care has been recognized by the Institute of Medicine as an essential aim of the US healthcare system.[1] A fundamental component of patient‐centered care is to engage patients and caregivers in establishing preferences, needs, values, and overall goals regarding their care.[1] Prior studies have shown that delivering high‐quality patient‐centered care is associated with improved health outcomes, and in some cases, reduced costs.[2, 3, 4, 5, 6, 7] Payors, including the Centers for Medicare and Medicaid Services under the Hospital Value‐Based Purchasing program, are increasingly tying payments to measures of patient experience.[8, 9] As more emphasis is placed on public reporting of these patient‐reported outcomes, healthcare organizations are investing in efforts to engage patients and caregivers, including efforts at establishing patients' preferences for care.[10]

In the acute care setting, a prerequisite for high‐quality patient‐centered care is identifying a patient's primary goal for recovery and then delivering care consistent with that goal.[11, 12, 13] Haberle et al. previously validated patients' most common goals for recovery in the hospital setting into 7 broad categories: (1) be cured, (2) live longer, (3) improve or maintain health, (4) be comfortable, (5) accomplish a particular life goal, (6) provide support for a family member, or (7) other.[13] When providers' understanding of these recovery goals are not concordant with the patient's stated goals, patients may receive care inconsistent with their preferences; it is not uncommon for patients to receive aggressive curative treatments (eg, cardiopulmonary resuscitation) when they have expressed otherwise.[14] On the other hand, when patient goals and priorities are clearly established, patients may have better outcomes.[15] For example, earlier conversations about patient goals and priorities in serious illness can lead to realistic expectations of treatment, enhanced goal‐concordant care, improved quality of life, higher patient satisfaction, more and earlier hospice care, fewer hospitalizations, better patient and family coping, reduced burden of decision making for families, and improved bereavement outcomes.[16, 17, 18]

Although previous studies have suggested poor patient‐physician concordance with regard to the patient's plan of care,[19, 20, 21, 22, 23, 24] there are limited data regarding providers' understanding of the patient's primary recovery goal during hospitalization. The purpose of this study was to identify the patients' Haberle goal, and then determine the degree of concordance among patients and key hospital providers regarding this goal.

METHODS

RESULTS

Of 1436 patients (882 oncology, 554 MICU) hospitalized during the study period, 341(156 oncology, 185 MICU) were admitted for 48 hours. Of 914 potentially eligible patients (617 oncology, 297 MICU), 191 (112 oncology and 79 MICU) were approached to participate based on our sampling strategy; of these, 8 (2 oncology and 6 MICU) did not have capacity (and no proxy was available) and 2 (1 oncology and 1 MICU) declined. Of the remaining 181 patients (109 oncology and 72 MICU), we obtained a completed questionnaire from all 3 interviewees on 109 (60.2% response rate).

Of the 109 study patients, 52 (47.7%) and 57 (52.3%) were admitted to the oncology and medical intensive care units, respectively (Table 1). Patients were predominantly middle aged, Caucasian, English‐speaking, and college‐educated. Healthcare proxies were frequently interviewed on behalf of patients in the MICU. Housestaff physicians were more often interviewed in the MICU, and PAs were interviewed only on oncology units. Compared to patient responders, nonresponders tended to be male and were admitted to oncology units (see Supporting Table 1 in the online version of this article).

The frequencies of selected Haberle recovery goals by participant type across all patients are listed in Table 2. Patients (or proxies) most often selected be cured (46.8%). Assigned nurses and physicians more commonly selected improve or maintain health (38.5% and 46.8%, respectively). Be comfortable was selected by nurses and physicians more frequently than by patients (16.5%, 16.5%, and 8.3%, respectively). The rate of responses for each Haberle goal was significantly different across all respondent groups (P 0.0001). The frequencies of selected Haberle goals were not significantly different between patients or proxies (P = 0.67), or for patients admitted to the MICU compared to oncology units (P = 0.64).

Inter‐rater agreement was poor to slight for the 3 participant dyads (kappa 0.09 [0.03‐0.19], 0.19 [0.08‐0.30], and 0.20 [0.08‐0.32] for patient‐physician, patient‐nurse, and nurse‐physician, respectively). The 3 participants selected the identical recovery goal in 22 (20.2%) cases, and each selected a distinct recovery goal in 32 (29.4%) cases. Pairwise concordance between nurses and physicians was 39.4%. There were no significant differences in agreement between patients admitted to the MICU compared to oncology units (P = 0.09).

DISCUSSION

We observed poor to slight concordance among patients and key hospital providers with regard to identifying the patient's primary recovery goal during acute hospitalization. The majority of patients (or proxies), chose be cured, whereas the majority of hospital providers chose improve or maintain health. Patients were twice as likely to select be cured and half as likely to choose be comfortable compared to nurses or physicians. Strikingly, the patient (or proxy), nurse, and physician identified the same recovery goal in just 20% of cases. These findings were similar for patients admitted to either the MICU or oncology units or when healthcare proxies participated on behalf of the patient (eg, when incapacitated in the MICU).

There are many reasons why hospital providers may not correctly identify the patients' primary recovery goals. First, we do not routinely ask patients to identify recovery goals upon admission in a structured and standardized manner. In fact, clinicians often do not elicit patients' needs, concerns, and expectations regarding their care in general.[25] Second, even when recovery goals are elicited at admission, they may not be communicated effectively to all members of the care team. This could be due to geographically non‐localized teams (although we did not observe a statistically significant difference between regionalized MICU and nonregionalized oncology care units), frequent provider‐to‐provider handoffs, and siloed electronic communication (eg, email, alphanumeric pages) regarding goals of care that inevitably leaves out key providers.[26] Third, healthcare proxies who are involved in decision making on the patient's behalf may not always be available to meet with the care team in person; consequently, their input may not be considered in a timely manner or reliably communicated to all members of the care team. We observed a large discrepancy in how often patients chose be cured compared to their hospital providers. This could be explained by clinicians' unwillingness to disclose bad news or divulge accurate prognostic information that causes patients to feel depressed or lose hope, particularly for those patients with the worst prognoses.[16, 27, 28] Patients may lack sophisticated knowledge of their conditions for a variety of reasons, including low health literacy, at times choosing to hope for the best even when it is not realistic. Additionally, there may be more subtle differences in what patients and hospital providers consider the primary recovery goal in context of the main reason for hospitalization and underlying medical illness. For example, a patient with metastatic lung cancer hospitalized with recurrent postobstructive pneumonia may choose be cured as his/her primary recovery goal (thinking of the pneumonia), whereas physicians may choose improve/maintain health or comfort (thinking of the cancer). We also cannot exclude the possibility that sometimes when patients state be cured and clinicians state improve health as the primary goal, that they are really saying the same thing in different ways. However, these are 2 different constructs (cure may not be possible for many patients) that may deserve an explicit discussion for patients to have realistic expectations for their health following hospitalization.

In short, our results underscore the importance of having an open and honest dialog with patients and caregivers throughout hospitalization, and the need to provide education about the potential futility of excessive care in situations where appropriate. Simply following patients' goals without discussing their feasibility and the consequences of aggressive treatments may result in unnecessary morbidity and misuse of healthcare resources. Once goals are clearly established, communicated, and refined in hospitalized patients with serious illness, there is much reason to believe that ongoing conversation will favorably impact outcomes.[29]

We found few studies that rigorously quantified the rate of concordance of hospital recovery goals among patients and key hospital providers; however, studies that measured overall plan of care agreement have demonstrated suboptimal concordance.[20, 30, 31] Shin et al. found significant underestimation of cancer patients' needs and poor concordance between patients and oncologists in assessing perceived needs of supportive care.[20] It is also notable that nurses and physicians had low levels of concordance in our study. O'Leary and colleagues found that nurses and physicians did not reliably communicate and often did not agree on the plan of care for hospitalized patients.[30] Although geographic regionalization of care teams and multidisciplinary rounds can improve the likelihood that key members of the care team are on the same page with regard to the plan of care, there is still much room for improvement.[26, 32, 33, 34] For example, although nurses and physicians in our study independently selected individual recovery goals with similar frequencies (Table 2), we observed suboptimal concordance between nurses and providers (36.8%) for specific patients, including on our regionalized care unit (MICU). This may be due to the reasons described above.

There are several implications of these findings. As payors continue to shift payments toward value‐based metrics, largely determined by patient experience and adequate advance care planning,[9] our findings suggest that more effort should be focused on delivering care consistent with patients' primary recovery goals. As a first step, healthcare organizations can focus on efforts to systematically identify and communicate recovery goals to all members of the care team, ensuring that patients' preferences, needs, and values are captured. In addition, as innovation in patient engagement and care delivery using Web‐based and mobile technology continues to grow,[35] using these tools to capture key goals for hospitalization and recovery can play an essential role. For example, as electronic health record vendors and institutions start to implement patient portals in the acute care setting, they should consider how to configure these tools to capture key goals for hospitalization and recovery, and then communicate them to the care team; preliminary work in this area is promising.[10]

Our study has several limitations to generalizability. First, the study was conducted on 2 services (MICU and oncology) at a single institution using a sampling strategy where research assistants enrolled 2 to 3 patients per day. Although the sampling was random, the availability of patients and proxies to be interviewed may have led to selection bias. Second, the sample size was small. Third, the patients who participated were predominantly white, English‐speaking, and well educated, possibly a consequence of our sampling strategy. However, this fact makes our findings more striking; although cultural and language barriers were generally not present in our study population, large discrepancies in goal concordance still existed. Fourth, in instances when patients were unable to participate themselves, we interviewed their healthcare proxy; therefore, it is possible that the proxies' responses did not reflect those of the patient. However, we note that concordance rates did not significantly differ between the 2 services despite the fact that the proportion of proxy interviews was much higher in the MICU. Similarly, we cannot exclude the possibility that patients altered their stated goals in the presence of proxies, but patients were given the option to be interviewed alone. Patients may also have misunderstood the timing of the goals (during this hospitalization as opposed to long term), although research assistants made every effort to clarify this during the interviews. Finally, our data‐collection instrument was previously validated in hospitalized general medicine patients and not oncology or MICU patients, and it has not been used to directly ask clinicians to identify patients' recovery goals. However, there is no reason to suspect that it could not be used for this purpose in critical care as well as noncritical care settings, as the survey was developed by a multidisciplinary team that included medical professionals and was validated by clinicians who successfully identified a single, very broad goal (eg, be cured) in each case.

CONCLUSION

We report poor to slight concordance among hospitalized patients and key hospital providers with regard to the main recovery goal. Future studies should assess whether patient satisfaction and experience is adversely impacted by patient‐provider discordance regarding key recovery goals. Additionally, institutions may consider future efforts to elicit and communicate patients' primary recovery goals more effectively to all members of the care team, and address discrepancies as soon as they are discovered.

Patient‐centered care has been recognized by the Institute of Medicine as an essential aim of the US healthcare system.[1] A fundamental component of patient‐centered care is to engage patients and caregivers in establishing preferences, needs, values, and overall goals regarding their care.[1] Prior studies have shown that delivering high‐quality patient‐centered care is associated with improved health outcomes, and in some cases, reduced costs.[2, 3, 4, 5, 6, 7] Payors, including the Centers for Medicare and Medicaid Services under the Hospital Value‐Based Purchasing program, are increasingly tying payments to measures of patient experience.[8, 9] As more emphasis is placed on public reporting of these patient‐reported outcomes, healthcare organizations are investing in efforts to engage patients and caregivers, including efforts at establishing patients' preferences for care.[10]

In the acute care setting, a prerequisite for high‐quality patient‐centered care is identifying a patient's primary goal for recovery and then delivering care consistent with that goal.[11, 12, 13] Haberle et al. previously validated patients' most common goals for recovery in the hospital setting into 7 broad categories: (1) be cured, (2) live longer, (3) improve or maintain health, (4) be comfortable, (5) accomplish a particular life goal, (6) provide support for a family member, or (7) other.[13] When providers' understanding of these recovery goals are not concordant with the patient's stated goals, patients may receive care inconsistent with their preferences; it is not uncommon for patients to receive aggressive curative treatments (eg, cardiopulmonary resuscitation) when they have expressed otherwise.[14] On the other hand, when patient goals and priorities are clearly established, patients may have better outcomes.[15] For example, earlier conversations about patient goals and priorities in serious illness can lead to realistic expectations of treatment, enhanced goal‐concordant care, improved quality of life, higher patient satisfaction, more and earlier hospice care, fewer hospitalizations, better patient and family coping, reduced burden of decision making for families, and improved bereavement outcomes.[16, 17, 18]

Although previous studies have suggested poor patient‐physician concordance with regard to the patient's plan of care,[19, 20, 21, 22, 23, 24] there are limited data regarding providers' understanding of the patient's primary recovery goal during hospitalization. The purpose of this study was to identify the patients' Haberle goal, and then determine the degree of concordance among patients and key hospital providers regarding this goal.

METHODS

RESULTS

Of 1436 patients (882 oncology, 554 MICU) hospitalized during the study period, 341(156 oncology, 185 MICU) were admitted for 48 hours. Of 914 potentially eligible patients (617 oncology, 297 MICU), 191 (112 oncology and 79 MICU) were approached to participate based on our sampling strategy; of these, 8 (2 oncology and 6 MICU) did not have capacity (and no proxy was available) and 2 (1 oncology and 1 MICU) declined. Of the remaining 181 patients (109 oncology and 72 MICU), we obtained a completed questionnaire from all 3 interviewees on 109 (60.2% response rate).

Of the 109 study patients, 52 (47.7%) and 57 (52.3%) were admitted to the oncology and medical intensive care units, respectively (Table 1). Patients were predominantly middle aged, Caucasian, English‐speaking, and college‐educated. Healthcare proxies were frequently interviewed on behalf of patients in the MICU. Housestaff physicians were more often interviewed in the MICU, and PAs were interviewed only on oncology units. Compared to patient responders, nonresponders tended to be male and were admitted to oncology units (see Supporting Table 1 in the online version of this article).

The frequencies of selected Haberle recovery goals by participant type across all patients are listed in Table 2. Patients (or proxies) most often selected be cured (46.8%). Assigned nurses and physicians more commonly selected improve or maintain health (38.5% and 46.8%, respectively). Be comfortable was selected by nurses and physicians more frequently than by patients (16.5%, 16.5%, and 8.3%, respectively). The rate of responses for each Haberle goal was significantly different across all respondent groups (P 0.0001). The frequencies of selected Haberle goals were not significantly different between patients or proxies (P = 0.67), or for patients admitted to the MICU compared to oncology units (P = 0.64).

Inter‐rater agreement was poor to slight for the 3 participant dyads (kappa 0.09 [0.03‐0.19], 0.19 [0.08‐0.30], and 0.20 [0.08‐0.32] for patient‐physician, patient‐nurse, and nurse‐physician, respectively). The 3 participants selected the identical recovery goal in 22 (20.2%) cases, and each selected a distinct recovery goal in 32 (29.4%) cases. Pairwise concordance between nurses and physicians was 39.4%. There were no significant differences in agreement between patients admitted to the MICU compared to oncology units (P = 0.09).

DISCUSSION

We observed poor to slight concordance among patients and key hospital providers with regard to identifying the patient's primary recovery goal during acute hospitalization. The majority of patients (or proxies), chose be cured, whereas the majority of hospital providers chose improve or maintain health. Patients were twice as likely to select be cured and half as likely to choose be comfortable compared to nurses or physicians. Strikingly, the patient (or proxy), nurse, and physician identified the same recovery goal in just 20% of cases. These findings were similar for patients admitted to either the MICU or oncology units or when healthcare proxies participated on behalf of the patient (eg, when incapacitated in the MICU).

There are many reasons why hospital providers may not correctly identify the patients' primary recovery goals. First, we do not routinely ask patients to identify recovery goals upon admission in a structured and standardized manner. In fact, clinicians often do not elicit patients' needs, concerns, and expectations regarding their care in general.[25] Second, even when recovery goals are elicited at admission, they may not be communicated effectively to all members of the care team. This could be due to geographically non‐localized teams (although we did not observe a statistically significant difference between regionalized MICU and nonregionalized oncology care units), frequent provider‐to‐provider handoffs, and siloed electronic communication (eg, email, alphanumeric pages) regarding goals of care that inevitably leaves out key providers.[26] Third, healthcare proxies who are involved in decision making on the patient's behalf may not always be available to meet with the care team in person; consequently, their input may not be considered in a timely manner or reliably communicated to all members of the care team. We observed a large discrepancy in how often patients chose be cured compared to their hospital providers. This could be explained by clinicians' unwillingness to disclose bad news or divulge accurate prognostic information that causes patients to feel depressed or lose hope, particularly for those patients with the worst prognoses.[16, 27, 28] Patients may lack sophisticated knowledge of their conditions for a variety of reasons, including low health literacy, at times choosing to hope for the best even when it is not realistic. Additionally, there may be more subtle differences in what patients and hospital providers consider the primary recovery goal in context of the main reason for hospitalization and underlying medical illness. For example, a patient with metastatic lung cancer hospitalized with recurrent postobstructive pneumonia may choose be cured as his/her primary recovery goal (thinking of the pneumonia), whereas physicians may choose improve/maintain health or comfort (thinking of the cancer). We also cannot exclude the possibility that sometimes when patients state be cured and clinicians state improve health as the primary goal, that they are really saying the same thing in different ways. However, these are 2 different constructs (cure may not be possible for many patients) that may deserve an explicit discussion for patients to have realistic expectations for their health following hospitalization.

In short, our results underscore the importance of having an open and honest dialog with patients and caregivers throughout hospitalization, and the need to provide education about the potential futility of excessive care in situations where appropriate. Simply following patients' goals without discussing their feasibility and the consequences of aggressive treatments may result in unnecessary morbidity and misuse of healthcare resources. Once goals are clearly established, communicated, and refined in hospitalized patients with serious illness, there is much reason to believe that ongoing conversation will favorably impact outcomes.[29]

We found few studies that rigorously quantified the rate of concordance of hospital recovery goals among patients and key hospital providers; however, studies that measured overall plan of care agreement have demonstrated suboptimal concordance.[20, 30, 31] Shin et al. found significant underestimation of cancer patients' needs and poor concordance between patients and oncologists in assessing perceived needs of supportive care.[20] It is also notable that nurses and physicians had low levels of concordance in our study. O'Leary and colleagues found that nurses and physicians did not reliably communicate and often did not agree on the plan of care for hospitalized patients.[30] Although geographic regionalization of care teams and multidisciplinary rounds can improve the likelihood that key members of the care team are on the same page with regard to the plan of care, there is still much room for improvement.[26, 32, 33, 34] For example, although nurses and physicians in our study independently selected individual recovery goals with similar frequencies (Table 2), we observed suboptimal concordance between nurses and providers (36.8%) for specific patients, including on our regionalized care unit (MICU). This may be due to the reasons described above.

There are several implications of these findings. As payors continue to shift payments toward value‐based metrics, largely determined by patient experience and adequate advance care planning,[9] our findings suggest that more effort should be focused on delivering care consistent with patients' primary recovery goals. As a first step, healthcare organizations can focus on efforts to systematically identify and communicate recovery goals to all members of the care team, ensuring that patients' preferences, needs, and values are captured. In addition, as innovation in patient engagement and care delivery using Web‐based and mobile technology continues to grow,[35] using these tools to capture key goals for hospitalization and recovery can play an essential role. For example, as electronic health record vendors and institutions start to implement patient portals in the acute care setting, they should consider how to configure these tools to capture key goals for hospitalization and recovery, and then communicate them to the care team; preliminary work in this area is promising.[10]

Our study has several limitations to generalizability. First, the study was conducted on 2 services (MICU and oncology) at a single institution using a sampling strategy where research assistants enrolled 2 to 3 patients per day. Although the sampling was random, the availability of patients and proxies to be interviewed may have led to selection bias. Second, the sample size was small. Third, the patients who participated were predominantly white, English‐speaking, and well educated, possibly a consequence of our sampling strategy. However, this fact makes our findings more striking; although cultural and language barriers were generally not present in our study population, large discrepancies in goal concordance still existed. Fourth, in instances when patients were unable to participate themselves, we interviewed their healthcare proxy; therefore, it is possible that the proxies' responses did not reflect those of the patient. However, we note that concordance rates did not significantly differ between the 2 services despite the fact that the proportion of proxy interviews was much higher in the MICU. Similarly, we cannot exclude the possibility that patients altered their stated goals in the presence of proxies, but patients were given the option to be interviewed alone. Patients may also have misunderstood the timing of the goals (during this hospitalization as opposed to long term), although research assistants made every effort to clarify this during the interviews. Finally, our data‐collection instrument was previously validated in hospitalized general medicine patients and not oncology or MICU patients, and it has not been used to directly ask clinicians to identify patients' recovery goals. However, there is no reason to suspect that it could not be used for this purpose in critical care as well as noncritical care settings, as the survey was developed by a multidisciplinary team that included medical professionals and was validated by clinicians who successfully identified a single, very broad goal (eg, be cured) in each case.

CONCLUSION

We report poor to slight concordance among hospitalized patients and key hospital providers with regard to the main recovery goal. Future studies should assess whether patient satisfaction and experience is adversely impacted by patient‐provider discordance regarding key recovery goals. Additionally, institutions may consider future efforts to elicit and communicate patients' primary recovery goals more effectively to all members of the care team, and address discrepancies as soon as they are discovered.

Patient‐centered care has been recognized by the Institute of Medicine as an essential aim of the US healthcare system.[1] A fundamental component of patient‐centered care is to engage patients and caregivers in establishing preferences, needs, values, and overall goals regarding their care.[1] Prior studies have shown that delivering high‐quality patient‐centered care is associated with improved health outcomes, and in some cases, reduced costs.[2, 3, 4, 5, 6, 7] Payors, including the Centers for Medicare and Medicaid Services under the Hospital Value‐Based Purchasing program, are increasingly tying payments to measures of patient experience.[8, 9] As more emphasis is placed on public reporting of these patient‐reported outcomes, healthcare organizations are investing in efforts to engage patients and caregivers, including efforts at establishing patients' preferences for care.[10]

In the acute care setting, a prerequisite for high‐quality patient‐centered care is identifying a patient's primary goal for recovery and then delivering care consistent with that goal.[11, 12, 13] Haberle et al. previously validated patients' most common goals for recovery in the hospital setting into 7 broad categories: (1) be cured, (2) live longer, (3) improve or maintain health, (4) be comfortable, (5) accomplish a particular life goal, (6) provide support for a family member, or (7) other.[13] When providers' understanding of these recovery goals are not concordant with the patient's stated goals, patients may receive care inconsistent with their preferences; it is not uncommon for patients to receive aggressive curative treatments (eg, cardiopulmonary resuscitation) when they have expressed otherwise.[14] On the other hand, when patient goals and priorities are clearly established, patients may have better outcomes.[15] For example, earlier conversations about patient goals and priorities in serious illness can lead to realistic expectations of treatment, enhanced goal‐concordant care, improved quality of life, higher patient satisfaction, more and earlier hospice care, fewer hospitalizations, better patient and family coping, reduced burden of decision making for families, and improved bereavement outcomes.[16, 17, 18]

Although previous studies have suggested poor patient‐physician concordance with regard to the patient's plan of care,[19, 20, 21, 22, 23, 24] there are limited data regarding providers' understanding of the patient's primary recovery goal during hospitalization. The purpose of this study was to identify the patients' Haberle goal, and then determine the degree of concordance among patients and key hospital providers regarding this goal.

METHODS

RESULTS

Of 1436 patients (882 oncology, 554 MICU) hospitalized during the study period, 341(156 oncology, 185 MICU) were admitted for 48 hours. Of 914 potentially eligible patients (617 oncology, 297 MICU), 191 (112 oncology and 79 MICU) were approached to participate based on our sampling strategy; of these, 8 (2 oncology and 6 MICU) did not have capacity (and no proxy was available) and 2 (1 oncology and 1 MICU) declined. Of the remaining 181 patients (109 oncology and 72 MICU), we obtained a completed questionnaire from all 3 interviewees on 109 (60.2% response rate).

Of the 109 study patients, 52 (47.7%) and 57 (52.3%) were admitted to the oncology and medical intensive care units, respectively (Table 1). Patients were predominantly middle aged, Caucasian, English‐speaking, and college‐educated. Healthcare proxies were frequently interviewed on behalf of patients in the MICU. Housestaff physicians were more often interviewed in the MICU, and PAs were interviewed only on oncology units. Compared to patient responders, nonresponders tended to be male and were admitted to oncology units (see Supporting Table 1 in the online version of this article).

The frequencies of selected Haberle recovery goals by participant type across all patients are listed in Table 2. Patients (or proxies) most often selected be cured (46.8%). Assigned nurses and physicians more commonly selected improve or maintain health (38.5% and 46.8%, respectively). Be comfortable was selected by nurses and physicians more frequently than by patients (16.5%, 16.5%, and 8.3%, respectively). The rate of responses for each Haberle goal was significantly different across all respondent groups (P 0.0001). The frequencies of selected Haberle goals were not significantly different between patients or proxies (P = 0.67), or for patients admitted to the MICU compared to oncology units (P = 0.64).

Inter‐rater agreement was poor to slight for the 3 participant dyads (kappa 0.09 [0.03‐0.19], 0.19 [0.08‐0.30], and 0.20 [0.08‐0.32] for patient‐physician, patient‐nurse, and nurse‐physician, respectively). The 3 participants selected the identical recovery goal in 22 (20.2%) cases, and each selected a distinct recovery goal in 32 (29.4%) cases. Pairwise concordance between nurses and physicians was 39.4%. There were no significant differences in agreement between patients admitted to the MICU compared to oncology units (P = 0.09).

DISCUSSION

We observed poor to slight concordance among patients and key hospital providers with regard to identifying the patient's primary recovery goal during acute hospitalization. The majority of patients (or proxies), chose be cured, whereas the majority of hospital providers chose improve or maintain health. Patients were twice as likely to select be cured and half as likely to choose be comfortable compared to nurses or physicians. Strikingly, the patient (or proxy), nurse, and physician identified the same recovery goal in just 20% of cases. These findings were similar for patients admitted to either the MICU or oncology units or when healthcare proxies participated on behalf of the patient (eg, when incapacitated in the MICU).

There are many reasons why hospital providers may not correctly identify the patients' primary recovery goals. First, we do not routinely ask patients to identify recovery goals upon admission in a structured and standardized manner. In fact, clinicians often do not elicit patients' needs, concerns, and expectations regarding their care in general.[25] Second, even when recovery goals are elicited at admission, they may not be communicated effectively to all members of the care team. This could be due to geographically non‐localized teams (although we did not observe a statistically significant difference between regionalized MICU and nonregionalized oncology care units), frequent provider‐to‐provider handoffs, and siloed electronic communication (eg, email, alphanumeric pages) regarding goals of care that inevitably leaves out key providers.[26] Third, healthcare proxies who are involved in decision making on the patient's behalf may not always be available to meet with the care team in person; consequently, their input may not be considered in a timely manner or reliably communicated to all members of the care team. We observed a large discrepancy in how often patients chose be cured compared to their hospital providers. This could be explained by clinicians' unwillingness to disclose bad news or divulge accurate prognostic information that causes patients to feel depressed or lose hope, particularly for those patients with the worst prognoses.[16, 27, 28] Patients may lack sophisticated knowledge of their conditions for a variety of reasons, including low health literacy, at times choosing to hope for the best even when it is not realistic. Additionally, there may be more subtle differences in what patients and hospital providers consider the primary recovery goal in context of the main reason for hospitalization and underlying medical illness. For example, a patient with metastatic lung cancer hospitalized with recurrent postobstructive pneumonia may choose be cured as his/her primary recovery goal (thinking of the pneumonia), whereas physicians may choose improve/maintain health or comfort (thinking of the cancer). We also cannot exclude the possibility that sometimes when patients state be cured and clinicians state improve health as the primary goal, that they are really saying the same thing in different ways. However, these are 2 different constructs (cure may not be possible for many patients) that may deserve an explicit discussion for patients to have realistic expectations for their health following hospitalization.

In short, our results underscore the importance of having an open and honest dialog with patients and caregivers throughout hospitalization, and the need to provide education about the potential futility of excessive care in situations where appropriate. Simply following patients' goals without discussing their feasibility and the consequences of aggressive treatments may result in unnecessary morbidity and misuse of healthcare resources. Once goals are clearly established, communicated, and refined in hospitalized patients with serious illness, there is much reason to believe that ongoing conversation will favorably impact outcomes.[29]

We found few studies that rigorously quantified the rate of concordance of hospital recovery goals among patients and key hospital providers; however, studies that measured overall plan of care agreement have demonstrated suboptimal concordance.[20, 30, 31] Shin et al. found significant underestimation of cancer patients' needs and poor concordance between patients and oncologists in assessing perceived needs of supportive care.[20] It is also notable that nurses and physicians had low levels of concordance in our study. O'Leary and colleagues found that nurses and physicians did not reliably communicate and often did not agree on the plan of care for hospitalized patients.[30] Although geographic regionalization of care teams and multidisciplinary rounds can improve the likelihood that key members of the care team are on the same page with regard to the plan of care, there is still much room for improvement.[26, 32, 33, 34] For example, although nurses and physicians in our study independently selected individual recovery goals with similar frequencies (Table 2), we observed suboptimal concordance between nurses and providers (36.8%) for specific patients, including on our regionalized care unit (MICU). This may be due to the reasons described above.

There are several implications of these findings. As payors continue to shift payments toward value‐based metrics, largely determined by patient experience and adequate advance care planning,[9] our findings suggest that more effort should be focused on delivering care consistent with patients' primary recovery goals. As a first step, healthcare organizations can focus on efforts to systematically identify and communicate recovery goals to all members of the care team, ensuring that patients' preferences, needs, and values are captured. In addition, as innovation in patient engagement and care delivery using Web‐based and mobile technology continues to grow,[35] using these tools to capture key goals for hospitalization and recovery can play an essential role. For example, as electronic health record vendors and institutions start to implement patient portals in the acute care setting, they should consider how to configure these tools to capture key goals for hospitalization and recovery, and then communicate them to the care team; preliminary work in this area is promising.[10]

Our study has several limitations to generalizability. First, the study was conducted on 2 services (MICU and oncology) at a single institution using a sampling strategy where research assistants enrolled 2 to 3 patients per day. Although the sampling was random, the availability of patients and proxies to be interviewed may have led to selection bias. Second, the sample size was small. Third, the patients who participated were predominantly white, English‐speaking, and well educated, possibly a consequence of our sampling strategy. However, this fact makes our findings more striking; although cultural and language barriers were generally not present in our study population, large discrepancies in goal concordance still existed. Fourth, in instances when patients were unable to participate themselves, we interviewed their healthcare proxy; therefore, it is possible that the proxies' responses did not reflect those of the patient. However, we note that concordance rates did not significantly differ between the 2 services despite the fact that the proportion of proxy interviews was much higher in the MICU. Similarly, we cannot exclude the possibility that patients altered their stated goals in the presence of proxies, but patients were given the option to be interviewed alone. Patients may also have misunderstood the timing of the goals (during this hospitalization as opposed to long term), although research assistants made every effort to clarify this during the interviews. Finally, our data‐collection instrument was previously validated in hospitalized general medicine patients and not oncology or MICU patients, and it has not been used to directly ask clinicians to identify patients' recovery goals. However, there is no reason to suspect that it could not be used for this purpose in critical care as well as noncritical care settings, as the survey was developed by a multidisciplinary team that included medical professionals and was validated by clinicians who successfully identified a single, very broad goal (eg, be cured) in each case.

CONCLUSION

We report poor to slight concordance among hospitalized patients and key hospital providers with regard to the main recovery goal. Future studies should assess whether patient satisfaction and experience is adversely impacted by patient‐provider discordance regarding key recovery goals. Additionally, institutions may consider future efforts to elicit and communicate patients' primary recovery goals more effectively to all members of the care team, and address discrepancies as soon as they are discovered.

In recent years, US hospitals have focused on decreasing readmission rates, incented by reimbursement penalties to hospitals having excessive readmissions.[1] Gaps in the quality of care provided during transitions likely contribute to preventable readmissions.[2] One compelling quality assessment in this setting is measuring patients' discharge preparedness, using key dimensions such as understanding their instructions for medication use and follow‐up. Patient‐reported preparedness for discharge may also be useful to identify risk of readmission.

Several patient‐reported measures of preparedness for discharge exist, and herein we describe 2 measures of interest. First, the Brief‐PREPARED (B‐PREPARED) measure was derived from the longer PREPARED instrument (Prescriptions, Ready to re‐enter community, Education, Placement, Assurance of safety, Realistic expectations, Empowerment, Directed to appropriate services), which reflects the patient's perceived needs at discharge. In previous research, the B‐PREPARED measure predicted emergency department (ED) visits for patients who had been recently hospitalized and had a high risk for readmission.[3] Second, the Care Transitions Measure‐3 (CTM‐3) was developed by Coleman et al. as a patient‐reported measure to discriminate between patients who were more likely to have an ED visit or readmission from those who did not. CTM‐3 has also been used to evaluate hospitals' level of care coordination and for public reporting purposes.[4, 5, 6] It has been endorsed by the National Quality Forum and incorporated into the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey provided to samples of recently hospitalized US patients.[7] However, recent evidence from an inpatient cohort of cardiovascular patients suggests the CTM‐3 overinflates care transition scores compared to the longer 15‐item CTM. In that cohort, the CTM‐3 could not differentiate between patients who did or did not have repeat ED visits or readmission.[8] Thus far, the B‐PREPARED and CTM‐3 measures have not been compared to one another directly.

In addition to the development of patient‐reported measures, hospitals increasingly employ administrative algorithms to predict likelihood of readmission.[9] A commonly used measure is the LACE index (Length of stay, Acuity, Comorbidity, and Emergency department use).[10] The LACE index predicted readmission and death within 30 days of discharge in a large cohort in Canada. In 2 retrospective studies of recently hospitalized patients in the United States, the LACE index's ability to discriminate between patients readmitted or not ranged from slightly better than chance to moderate (C statistic 0.56‐0.77).[11, 12]

It is unknown whether adding patient‐reported preparedness measures to commonly used readmission prediction scores increases the ability to predict readmission risk. We sought to determine whether the B‐PREPARED and CTM‐3 measures were predictive of readmission or death, as compared to the LACE index, in a large cohort of cardiovascular patients. In addition, we sought to determine the additional predictive and discriminative ability gained from administering the B‐PREPARED and CTM‐3 measures, while adjusting for the LACE index and other clinical factors. We hypothesized that: (1) higher preparedness scores on both measures would predict lower risk of readmission or death in a cohort of patients hospitalized with cardiac diagnoses; and (2) because it provides more specific and actionable information, the B‐PREPARED would discriminate readmission more accurately than CTM‐3, after controlling for clinical factors.

METHODS

RESULTS

From the cohort of 1239 patients (Figure 1), 64%, 28%, and 7% of patients were hospitalized with ACS, ADHF, or both, respectively (Table 1). Nearly 45% of patients were female, 83% were white, and the median age was 61 years (IQR 5269). The median length of stay was 3 days (IQR 25). The median preparedness scores were high for both B‐PREPARED (21, IQR 1822) and CTM‐3 (77.8, IQR 66.7100). A total of 211 (17%) and 380 (31%) were readmitted or died within 30 and 90 days, respectively. The completion rate for the postdischarge phone calls was 88%.

Figure 1

B‐PREPARED and CTM‐3 were moderately correlated with one another (Spearman's = 0.40, P 0.001). In bivariate analyses (Table 1), the association between B‐PREPARED and readmission or death was significant at 90 days (P = 0.030) but not 30 days. The CTM‐3 showed no significant association with readmission or death at either time point. The LACE score was significantly associated with rates of readmission at 30 and 90 days (P 0.001).

DISCUSSION/CONCLUSION

In this cohort of patients hospitalized with cardiovascular disease, we compared 2 patient‐reported measures of preparedness for discharge, their association with time to death or readmission at 30 and 90 days, and their ability to discriminate patients who were or were not readmitted or died. Higher preparedness as measured by higher B‐PREPARED scores was associated with lower risk of readmission or death at 30 and 90 days after discharge in unadjusted models, and at 90 days in adjusted models. CTM‐3 was not associated with the outcome in any analyses. Lastly, the individual preparedness measures were not as strongly associated with readmission or death compared to the LACE readmission index alone.

How do our findings relate to the measurement of care transition quality? We consider 2 scenarios. First, if hospitals utilize the LACE index to predict readmission, then neither self‐reported measure of preparedness adds meaningfully to its predictive ability. However, hospital management may still find the B‐PREPARED and CTM‐3 useful as a means to direct care transition quality‐improvement efforts. These measures can instruct hospitals as to what areas their patients express the greatest difficulty or lack of preparedness and closely attend to patient needs with appropriate resources. Furthermore, the patient's perception of being prepared for discharge may be different than their actual preparedness. Their perceived preparedness may be affected by cognitive impairment, dissatisfaction with medical care, depression, lower health‐related quality of life, and lower educational attainment as demonstrated by Lau et al.[16] If a patient's perception of preparedness were low, it would behoove the clinician to investigate these other issues and address those that are mutable. Additionally, perceived preparedness may not correlate with the patient's understanding of their medical conditions, so it is imperative that clinicians provide prospective guidance about their probable postdischarge trajectory. If hospitals are not utilizing the LACE index, then perhaps using the B‐PREPARED, but not the CTM‐3, may be beneficial for predicting readmission.

How do our results fit with evidence from prior studies, and what do they mean in the context of care transitions quality? First, in the psychometric evaluation of the B‐PREPARED measure in a cohort of recently hospitalized patients, the mean score was 17.3, lower than the median of 21 in our cohort.[3] Numerous studies have utilized the CTM‐3 and the longer‐version CTM‐15. Though we cannot make a direct comparison, the median in our cohort (77.8) was on par with the means from other studies, which ranged from 63 to 82.[5, 17, 18, 19] Several studies also note ceiling effects with clusters of scores at the upper end of the scale, as did we. We conjecture that our cohort's preparedness scores may be higher because our institution has made concerted efforts to improve the discharge education for cardiovascular patients.

In a comparable patient population, the TRACE‐CORE (Transitions, Risks, and Actions in Coronary Events Center for Outcomes Research and Education) study is a cohort of more than 2200 patients with ACS who were administered the CTM‐15 within 1 month of discharge.[8] In that study, the median CTM‐15 score was 66.6, which is lower than our cohort. With regard to the predictive ability of the CTM‐3, they note that CTM‐3 scores did not differentiate between patients who were or were not readmitted or had emergency department visits. Our results support their concern that the CTM‐15 and by extension the CTM‐3, though adopted widely as part of HCAHPS, may not have sufficient ability to discriminate differences in patient outcomes or the quality of care transitions.

More recently, patient‐reported preparedness for discharge was assessed in a prospective cohort in Canada.[16] Lau et al. administered a single‐item measure of readiness at the time of discharge to general medicine patients, and found that lower readiness scores were also not associated with readmission or death at 30 days, when adjusted for the LACE index as we did.

We must acknowledge the limitations of our findings. First, our sample of recently discharged patients with cardiovascular disease is different than the community‐dwelling, underserved Americans hospitalized in the prior year, which served as the sample for reducing the CTM‐15 to 3 items.[5] This fact may explain why we did not find the CTM‐3 to be associated with readmission in our sample. Second, our analyses did not include extensive adjustment for patient‐related factors. Rather, our intention was to see how well the preparedness measures performed independently and compare their abilities to predict readmission, which is particularly relevant for clinicians who may not have all possible covariates in predicting readmission. Finally, because we limited the analyses to the patients who completed the B‐PREPARED and CTM‐3 measures (88% completion rate), we may not have data for: (1) very ill patients, who had a higher risk of readmission and least prepared, and were not able to answer the postdischarge phone call; and (2) very functional patients, who had a lower risk of readmission and were too busy to answer the postdischarge phone call. This may have limited the extremes in the spectrum of our sample.

Importantly, our study has several strengths. We report on the largest sample to date with results of both B‐PREPARED and CTM‐3. Moreover, we examined how these measures compared to a widely used readmission prediction tool, the LACE index. We had very high postdischarge phone call completion rates in the week following discharge. Furthermore, we had thorough assessment of readmission data through patient report, electronic medical record documentation, and collection of outside medical records.

Further research is needed to elucidate: (1) the ideal administration time of the patient‐reported measures of preparedness (before or after discharge), and (2) the challenges to the implementation of measures in healthcare systems. Remaining research questions center on the tradeoffs and barriers to implementing a longer measure like the 11‐item B‐PREPARED compared to a shorter measure like the CTM‐3. We do not know whether longer measures preclude their use by busy clinicians, though it provides more specific information about what patients feel they need at hospital discharge. Additionally, studies need to demonstrate the mutability of preparedness and the response of measures to interventions designed to improve the hospital discharge process.

In our sample of recently hospitalized cardiovascular patients, there was a statistically significant association between patient‐reported preparedness for discharged, as measured by B‐PREPARED, and readmissions/death at 30 and 90 days, but the magnitude of the association was very small. Furthermore, another patient‐reported preparedness measure, CTM‐3, was not associated with readmissions or death at either 30 or 90 days. Lastly, neither measure discriminated well between patients who were readmitted or not, and neither measure added meaningfully to the LACE index in terms of predicting 30‐ or 90‐day readmissions.

In recent years, US hospitals have focused on decreasing readmission rates, incented by reimbursement penalties to hospitals having excessive readmissions.[1] Gaps in the quality of care provided during transitions likely contribute to preventable readmissions.[2] One compelling quality assessment in this setting is measuring patients' discharge preparedness, using key dimensions such as understanding their instructions for medication use and follow‐up. Patient‐reported preparedness for discharge may also be useful to identify risk of readmission.

Several patient‐reported measures of preparedness for discharge exist, and herein we describe 2 measures of interest. First, the Brief‐PREPARED (B‐PREPARED) measure was derived from the longer PREPARED instrument (Prescriptions, Ready to re‐enter community, Education, Placement, Assurance of safety, Realistic expectations, Empowerment, Directed to appropriate services), which reflects the patient's perceived needs at discharge. In previous research, the B‐PREPARED measure predicted emergency department (ED) visits for patients who had been recently hospitalized and had a high risk for readmission.[3] Second, the Care Transitions Measure‐3 (CTM‐3) was developed by Coleman et al. as a patient‐reported measure to discriminate between patients who were more likely to have an ED visit or readmission from those who did not. CTM‐3 has also been used to evaluate hospitals' level of care coordination and for public reporting purposes.[4, 5, 6] It has been endorsed by the National Quality Forum and incorporated into the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey provided to samples of recently hospitalized US patients.[7] However, recent evidence from an inpatient cohort of cardiovascular patients suggests the CTM‐3 overinflates care transition scores compared to the longer 15‐item CTM. In that cohort, the CTM‐3 could not differentiate between patients who did or did not have repeat ED visits or readmission.[8] Thus far, the B‐PREPARED and CTM‐3 measures have not been compared to one another directly.

In addition to the development of patient‐reported measures, hospitals increasingly employ administrative algorithms to predict likelihood of readmission.[9] A commonly used measure is the LACE index (Length of stay, Acuity, Comorbidity, and Emergency department use).[10] The LACE index predicted readmission and death within 30 days of discharge in a large cohort in Canada. In 2 retrospective studies of recently hospitalized patients in the United States, the LACE index's ability to discriminate between patients readmitted or not ranged from slightly better than chance to moderate (C statistic 0.56‐0.77).[11, 12]

It is unknown whether adding patient‐reported preparedness measures to commonly used readmission prediction scores increases the ability to predict readmission risk. We sought to determine whether the B‐PREPARED and CTM‐3 measures were predictive of readmission or death, as compared to the LACE index, in a large cohort of cardiovascular patients. In addition, we sought to determine the additional predictive and discriminative ability gained from administering the B‐PREPARED and CTM‐3 measures, while adjusting for the LACE index and other clinical factors. We hypothesized that: (1) higher preparedness scores on both measures would predict lower risk of readmission or death in a cohort of patients hospitalized with cardiac diagnoses; and (2) because it provides more specific and actionable information, the B‐PREPARED would discriminate readmission more accurately than CTM‐3, after controlling for clinical factors.

METHODS

RESULTS

From the cohort of 1239 patients (Figure 1), 64%, 28%, and 7% of patients were hospitalized with ACS, ADHF, or both, respectively (Table 1). Nearly 45% of patients were female, 83% were white, and the median age was 61 years (IQR 5269). The median length of stay was 3 days (IQR 25). The median preparedness scores were high for both B‐PREPARED (21, IQR 1822) and CTM‐3 (77.8, IQR 66.7100). A total of 211 (17%) and 380 (31%) were readmitted or died within 30 and 90 days, respectively. The completion rate for the postdischarge phone calls was 88%.

Figure 1

B‐PREPARED and CTM‐3 were moderately correlated with one another (Spearman's = 0.40, P 0.001). In bivariate analyses (Table 1), the association between B‐PREPARED and readmission or death was significant at 90 days (P = 0.030) but not 30 days. The CTM‐3 showed no significant association with readmission or death at either time point. The LACE score was significantly associated with rates of readmission at 30 and 90 days (P 0.001).

DISCUSSION/CONCLUSION

In this cohort of patients hospitalized with cardiovascular disease, we compared 2 patient‐reported measures of preparedness for discharge, their association with time to death or readmission at 30 and 90 days, and their ability to discriminate patients who were or were not readmitted or died. Higher preparedness as measured by higher B‐PREPARED scores was associated with lower risk of readmission or death at 30 and 90 days after discharge in unadjusted models, and at 90 days in adjusted models. CTM‐3 was not associated with the outcome in any analyses. Lastly, the individual preparedness measures were not as strongly associated with readmission or death compared to the LACE readmission index alone.

How do our findings relate to the measurement of care transition quality? We consider 2 scenarios. First, if hospitals utilize the LACE index to predict readmission, then neither self‐reported measure of preparedness adds meaningfully to its predictive ability. However, hospital management may still find the B‐PREPARED and CTM‐3 useful as a means to direct care transition quality‐improvement efforts. These measures can instruct hospitals as to what areas their patients express the greatest difficulty or lack of preparedness and closely attend to patient needs with appropriate resources. Furthermore, the patient's perception of being prepared for discharge may be different than their actual preparedness. Their perceived preparedness may be affected by cognitive impairment, dissatisfaction with medical care, depression, lower health‐related quality of life, and lower educational attainment as demonstrated by Lau et al.[16] If a patient's perception of preparedness were low, it would behoove the clinician to investigate these other issues and address those that are mutable. Additionally, perceived preparedness may not correlate with the patient's understanding of their medical conditions, so it is imperative that clinicians provide prospective guidance about their probable postdischarge trajectory. If hospitals are not utilizing the LACE index, then perhaps using the B‐PREPARED, but not the CTM‐3, may be beneficial for predicting readmission.

How do our results fit with evidence from prior studies, and what do they mean in the context of care transitions quality? First, in the psychometric evaluation of the B‐PREPARED measure in a cohort of recently hospitalized patients, the mean score was 17.3, lower than the median of 21 in our cohort.[3] Numerous studies have utilized the CTM‐3 and the longer‐version CTM‐15. Though we cannot make a direct comparison, the median in our cohort (77.8) was on par with the means from other studies, which ranged from 63 to 82.[5, 17, 18, 19] Several studies also note ceiling effects with clusters of scores at the upper end of the scale, as did we. We conjecture that our cohort's preparedness scores may be higher because our institution has made concerted efforts to improve the discharge education for cardiovascular patients.

In a comparable patient population, the TRACE‐CORE (Transitions, Risks, and Actions in Coronary Events Center for Outcomes Research and Education) study is a cohort of more than 2200 patients with ACS who were administered the CTM‐15 within 1 month of discharge.[8] In that study, the median CTM‐15 score was 66.6, which is lower than our cohort. With regard to the predictive ability of the CTM‐3, they note that CTM‐3 scores did not differentiate between patients who were or were not readmitted or had emergency department visits. Our results support their concern that the CTM‐15 and by extension the CTM‐3, though adopted widely as part of HCAHPS, may not have sufficient ability to discriminate differences in patient outcomes or the quality of care transitions.

More recently, patient‐reported preparedness for discharge was assessed in a prospective cohort in Canada.[16] Lau et al. administered a single‐item measure of readiness at the time of discharge to general medicine patients, and found that lower readiness scores were also not associated with readmission or death at 30 days, when adjusted for the LACE index as we did.

We must acknowledge the limitations of our findings. First, our sample of recently discharged patients with cardiovascular disease is different than the community‐dwelling, underserved Americans hospitalized in the prior year, which served as the sample for reducing the CTM‐15 to 3 items.[5] This fact may explain why we did not find the CTM‐3 to be associated with readmission in our sample. Second, our analyses did not include extensive adjustment for patient‐related factors. Rather, our intention was to see how well the preparedness measures performed independently and compare their abilities to predict readmission, which is particularly relevant for clinicians who may not have all possible covariates in predicting readmission. Finally, because we limited the analyses to the patients who completed the B‐PREPARED and CTM‐3 measures (88% completion rate), we may not have data for: (1) very ill patients, who had a higher risk of readmission and least prepared, and were not able to answer the postdischarge phone call; and (2) very functional patients, who had a lower risk of readmission and were too busy to answer the postdischarge phone call. This may have limited the extremes in the spectrum of our sample.

Importantly, our study has several strengths. We report on the largest sample to date with results of both B‐PREPARED and CTM‐3. Moreover, we examined how these measures compared to a widely used readmission prediction tool, the LACE index. We had very high postdischarge phone call completion rates in the week following discharge. Furthermore, we had thorough assessment of readmission data through patient report, electronic medical record documentation, and collection of outside medical records.

Further research is needed to elucidate: (1) the ideal administration time of the patient‐reported measures of preparedness (before or after discharge), and (2) the challenges to the implementation of measures in healthcare systems. Remaining research questions center on the tradeoffs and barriers to implementing a longer measure like the 11‐item B‐PREPARED compared to a shorter measure like the CTM‐3. We do not know whether longer measures preclude their use by busy clinicians, though it provides more specific information about what patients feel they need at hospital discharge. Additionally, studies need to demonstrate the mutability of preparedness and the response of measures to interventions designed to improve the hospital discharge process.

In our sample of recently hospitalized cardiovascular patients, there was a statistically significant association between patient‐reported preparedness for discharged, as measured by B‐PREPARED, and readmissions/death at 30 and 90 days, but the magnitude of the association was very small. Furthermore, another patient‐reported preparedness measure, CTM‐3, was not associated with readmissions or death at either 30 or 90 days. Lastly, neither measure discriminated well between patients who were readmitted or not, and neither measure added meaningfully to the LACE index in terms of predicting 30‐ or 90‐day readmissions.

In recent years, US hospitals have focused on decreasing readmission rates, incented by reimbursement penalties to hospitals having excessive readmissions.[1] Gaps in the quality of care provided during transitions likely contribute to preventable readmissions.[2] One compelling quality assessment in this setting is measuring patients' discharge preparedness, using key dimensions such as understanding their instructions for medication use and follow‐up. Patient‐reported preparedness for discharge may also be useful to identify risk of readmission.

Several patient‐reported measures of preparedness for discharge exist, and herein we describe 2 measures of interest. First, the Brief‐PREPARED (B‐PREPARED) measure was derived from the longer PREPARED instrument (Prescriptions, Ready to re‐enter community, Education, Placement, Assurance of safety, Realistic expectations, Empowerment, Directed to appropriate services), which reflects the patient's perceived needs at discharge. In previous research, the B‐PREPARED measure predicted emergency department (ED) visits for patients who had been recently hospitalized and had a high risk for readmission.[3] Second, the Care Transitions Measure‐3 (CTM‐3) was developed by Coleman et al. as a patient‐reported measure to discriminate between patients who were more likely to have an ED visit or readmission from those who did not. CTM‐3 has also been used to evaluate hospitals' level of care coordination and for public reporting purposes.[4, 5, 6] It has been endorsed by the National Quality Forum and incorporated into the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey provided to samples of recently hospitalized US patients.[7] However, recent evidence from an inpatient cohort of cardiovascular patients suggests the CTM‐3 overinflates care transition scores compared to the longer 15‐item CTM. In that cohort, the CTM‐3 could not differentiate between patients who did or did not have repeat ED visits or readmission.[8] Thus far, the B‐PREPARED and CTM‐3 measures have not been compared to one another directly.

In addition to the development of patient‐reported measures, hospitals increasingly employ administrative algorithms to predict likelihood of readmission.[9] A commonly used measure is the LACE index (Length of stay, Acuity, Comorbidity, and Emergency department use).[10] The LACE index predicted readmission and death within 30 days of discharge in a large cohort in Canada. In 2 retrospective studies of recently hospitalized patients in the United States, the LACE index's ability to discriminate between patients readmitted or not ranged from slightly better than chance to moderate (C statistic 0.56‐0.77).[11, 12]

It is unknown whether adding patient‐reported preparedness measures to commonly used readmission prediction scores increases the ability to predict readmission risk. We sought to determine whether the B‐PREPARED and CTM‐3 measures were predictive of readmission or death, as compared to the LACE index, in a large cohort of cardiovascular patients. In addition, we sought to determine the additional predictive and discriminative ability gained from administering the B‐PREPARED and CTM‐3 measures, while adjusting for the LACE index and other clinical factors. We hypothesized that: (1) higher preparedness scores on both measures would predict lower risk of readmission or death in a cohort of patients hospitalized with cardiac diagnoses; and (2) because it provides more specific and actionable information, the B‐PREPARED would discriminate readmission more accurately than CTM‐3, after controlling for clinical factors.

METHODS

RESULTS

From the cohort of 1239 patients (Figure 1), 64%, 28%, and 7% of patients were hospitalized with ACS, ADHF, or both, respectively (Table 1). Nearly 45% of patients were female, 83% were white, and the median age was 61 years (IQR 5269). The median length of stay was 3 days (IQR 25). The median preparedness scores were high for both B‐PREPARED (21, IQR 1822) and CTM‐3 (77.8, IQR 66.7100). A total of 211 (17%) and 380 (31%) were readmitted or died within 30 and 90 days, respectively. The completion rate for the postdischarge phone calls was 88%.

Figure 1

B‐PREPARED and CTM‐3 were moderately correlated with one another (Spearman's = 0.40, P 0.001). In bivariate analyses (Table 1), the association between B‐PREPARED and readmission or death was significant at 90 days (P = 0.030) but not 30 days. The CTM‐3 showed no significant association with readmission or death at either time point. The LACE score was significantly associated with rates of readmission at 30 and 90 days (P 0.001).

DISCUSSION/CONCLUSION

In this cohort of patients hospitalized with cardiovascular disease, we compared 2 patient‐reported measures of preparedness for discharge, their association with time to death or readmission at 30 and 90 days, and their ability to discriminate patients who were or were not readmitted or died. Higher preparedness as measured by higher B‐PREPARED scores was associated with lower risk of readmission or death at 30 and 90 days after discharge in unadjusted models, and at 90 days in adjusted models. CTM‐3 was not associated with the outcome in any analyses. Lastly, the individual preparedness measures were not as strongly associated with readmission or death compared to the LACE readmission index alone.

How do our findings relate to the measurement of care transition quality? We consider 2 scenarios. First, if hospitals utilize the LACE index to predict readmission, then neither self‐reported measure of preparedness adds meaningfully to its predictive ability. However, hospital management may still find the B‐PREPARED and CTM‐3 useful as a means to direct care transition quality‐improvement efforts. These measures can instruct hospitals as to what areas their patients express the greatest difficulty or lack of preparedness and closely attend to patient needs with appropriate resources. Furthermore, the patient's perception of being prepared for discharge may be different than their actual preparedness. Their perceived preparedness may be affected by cognitive impairment, dissatisfaction with medical care, depression, lower health‐related quality of life, and lower educational attainment as demonstrated by Lau et al.[16] If a patient's perception of preparedness were low, it would behoove the clinician to investigate these other issues and address those that are mutable. Additionally, perceived preparedness may not correlate with the patient's understanding of their medical conditions, so it is imperative that clinicians provide prospective guidance about their probable postdischarge trajectory. If hospitals are not utilizing the LACE index, then perhaps using the B‐PREPARED, but not the CTM‐3, may be beneficial for predicting readmission.

How do our results fit with evidence from prior studies, and what do they mean in the context of care transitions quality? First, in the psychometric evaluation of the B‐PREPARED measure in a cohort of recently hospitalized patients, the mean score was 17.3, lower than the median of 21 in our cohort.[3] Numerous studies have utilized the CTM‐3 and the longer‐version CTM‐15. Though we cannot make a direct comparison, the median in our cohort (77.8) was on par with the means from other studies, which ranged from 63 to 82.[5, 17, 18, 19] Several studies also note ceiling effects with clusters of scores at the upper end of the scale, as did we. We conjecture that our cohort's preparedness scores may be higher because our institution has made concerted efforts to improve the discharge education for cardiovascular patients.

In a comparable patient population, the TRACE‐CORE (Transitions, Risks, and Actions in Coronary Events Center for Outcomes Research and Education) study is a cohort of more than 2200 patients with ACS who were administered the CTM‐15 within 1 month of discharge.[8] In that study, the median CTM‐15 score was 66.6, which is lower than our cohort. With regard to the predictive ability of the CTM‐3, they note that CTM‐3 scores did not differentiate between patients who were or were not readmitted or had emergency department visits. Our results support their concern that the CTM‐15 and by extension the CTM‐3, though adopted widely as part of HCAHPS, may not have sufficient ability to discriminate differences in patient outcomes or the quality of care transitions.

More recently, patient‐reported preparedness for discharge was assessed in a prospective cohort in Canada.[16] Lau et al. administered a single‐item measure of readiness at the time of discharge to general medicine patients, and found that lower readiness scores were also not associated with readmission or death at 30 days, when adjusted for the LACE index as we did.

We must acknowledge the limitations of our findings. First, our sample of recently discharged patients with cardiovascular disease is different than the community‐dwelling, underserved Americans hospitalized in the prior year, which served as the sample for reducing the CTM‐15 to 3 items.[5] This fact may explain why we did not find the CTM‐3 to be associated with readmission in our sample. Second, our analyses did not include extensive adjustment for patient‐related factors. Rather, our intention was to see how well the preparedness measures performed independently and compare their abilities to predict readmission, which is particularly relevant for clinicians who may not have all possible covariates in predicting readmission. Finally, because we limited the analyses to the patients who completed the B‐PREPARED and CTM‐3 measures (88% completion rate), we may not have data for: (1) very ill patients, who had a higher risk of readmission and least prepared, and were not able to answer the postdischarge phone call; and (2) very functional patients, who had a lower risk of readmission and were too busy to answer the postdischarge phone call. This may have limited the extremes in the spectrum of our sample.

Importantly, our study has several strengths. We report on the largest sample to date with results of both B‐PREPARED and CTM‐3. Moreover, we examined how these measures compared to a widely used readmission prediction tool, the LACE index. We had very high postdischarge phone call completion rates in the week following discharge. Furthermore, we had thorough assessment of readmission data through patient report, electronic medical record documentation, and collection of outside medical records.

Further research is needed to elucidate: (1) the ideal administration time of the patient‐reported measures of preparedness (before or after discharge), and (2) the challenges to the implementation of measures in healthcare systems. Remaining research questions center on the tradeoffs and barriers to implementing a longer measure like the 11‐item B‐PREPARED compared to a shorter measure like the CTM‐3. We do not know whether longer measures preclude their use by busy clinicians, though it provides more specific information about what patients feel they need at hospital discharge. Additionally, studies need to demonstrate the mutability of preparedness and the response of measures to interventions designed to improve the hospital discharge process.

In our sample of recently hospitalized cardiovascular patients, there was a statistically significant association between patient‐reported preparedness for discharged, as measured by B‐PREPARED, and readmissions/death at 30 and 90 days, but the magnitude of the association was very small. Furthermore, another patient‐reported preparedness measure, CTM‐3, was not associated with readmissions or death at either 30 or 90 days. Lastly, neither measure discriminated well between patients who were readmitted or not, and neither measure added meaningfully to the LACE index in terms of predicting 30‐ or 90‐day readmissions.