At the turn of the 20th century, most deaths in the United States occurred at home. By the 1960s, over 70% of deaths occurred in an institutional setting, reflecting an evolution of medical technology.[1, 2, 3] With the birth of the hospice movement in the 1970s, dying patients had the opportunity to have both death at home and aggressive symptom control at the end of life. Although there has been a slow decline in the proportion of deaths that occur in the hospital over the past 2 decades,[3] the overwhelming majority of persons state that they would prefer to die at home. However, recent findings suggest that most people will die in an institutional setting.[3, 4, 5, 6]

Although good data exist describing population preferences for location of death, and we know, based on death records, where deaths occur in the United States, there are few studies that examine concordance between preferred and actual site of death at the individual patient level. Furthermore, although factors have been identified that predict death at home, factors predicting concordance between preferred and actual site of death are not well described.[3, 6, 7, 8, 9, 10, 11, 12, 13]

Regardless of where death ultimately occurs, most adults will experience multiple hospitalizations within the last years of their life. Understanding the preferences and subsequent experiences of this population is of particular relevance to hospitalist physicians who are in a unique position to elicit goals from seriously ill patients and help match patient preferences with their medical care. In this observational study, we sought to determine preferences for site of death in a cohort of adult patients admitted to the hospital for medical illness, and then follow those patients to determine where death occurred for those who died. We also sought to explore factors that may predict concordance between preferred and actual site of death. We hypothesized that ethnic diversity and lower socioeconomic status would be associated with a lower likelihood of concordance between preferred and actual site of death. We also hypothesized that advanced care planning would be associated with a higher likelihood of concordance. The Colorado Multi‐Institutional Review Board approved this study.

METHODS

Participants were recruited from 3 hospitals affiliated with the University of Colorado School of Medicine Internal Medicine Residency program, including the Denver Veterans' Administration Center (DVAMC), Denver Health Medical Center (DHMC), and University of Colorado Hospital (UCH). The DVAMC is a large urban Veterans Administration hospital, serving veterans from the Denver metro area, and is a tertiary referral center for veterans in rural Colorado, Wyoming, and parts of Montana. DHMC, the safety‐net hospital for the Denver area, serves over 25% of the residents in the city and county of Denver, including such special populations as the indigent, chronically mentally ill, and persons with polysubstance dependence. UCH had 350 licensed beds at the time of our study and serves as the Rocky Mountain region's only academic tertiary, specialty care, and referral center. At the time of this study, there was limited inpatient palliative care services at the DVAMC and UH, and no palliative care services at DHMC. Participants were screened on the first day following admission to the adult general medical service. Participants were recruited on 96 postadmission days between February 2004 and June 2006. Recruitment days varied from Monday through Friday, to include admissions from the weekend and throughout the year to reduce potential bias due to seasonal trends of diseases such as influenza. Patients were excluded if they died or were discharged within the first 24 hours of admission, were pregnant, jailed, or unable to give informed consent. All other patients were approached and invited to participate in a brief survey.

After informed consent was obtained, participants completed a bedside interview that included self‐identified ethnicity and the Berkman‐Syme Social Network Index,[14] a brief questionnaire quantifying social support from spouse or domestic partner, family, friends, and other religious or secular organizations. Baseline socioeconomic measures (eg, income, employment, home ownership, car ownership) and questions related to the last days of life were also included. Participants were asked the following question, If you were very sick, with an illness that could not be cured, and in bed most of the time, where would you spend the last days of your life if you could chose?

For each participant, we performed a detailed chart review to determine demographic data, presence of advance directives, and CARING criteria (Cancer, Admissions 2, Residence in a nursing home, Intensive care unit admit with multiorgan failure, 2 Noncancer hospice Guidelines), a set of prognostic criteria identifying patients at an index hospitalization who have a high burden of illness and are at risk for death in the following year.[15] We then followed patients for 5 years. If participants died within the follow‐up period, we collected the date and location of death using medical records, death certificates, or in a few cases when official death records were unavailable, direct contact with the family. Participants were considered alive if they had a clinic visit or MD/RN phone contact within 3 months prior to the final collection point date.

RESULTS

Preferred Site of Death

When asked where they preferred to spend the last days of their life, 75% of patients (n=343) stated they would like to be at home. In the hospital was the preferred location for 10% of patients, whereas 6% stated a nursing home and 4% a hospice inpatient facility. Two percent stated they had no preference, and 3% refused to answer (Figure 1)

Figure 1

We found that in the univariable analysis the following factors were associated with preference for site of death at a significance level of P<0.25: unstable housing, hospital setting, income level, ethnicity, CARING criteria, presence of an advance directive, education level, married, primary care provider, and presence of public insurance. Results of the logit models (home vs nursing home or hospice facility, and hospital vs nursing home or hospice facility and home vs hospital) are presented in Table 2.

Table 2.Logistic Regression Modeling of Preference for Death at Home or Hospital

	Adjusted Odds Ratio (95% Confidence Interval)
	Home vs Nursing Home/Hospice Facility	Hospital vs Nursing Home/Hospice Facility	Home vs Hospital
NOTE: Abbreviations: CARING, Cancer, Admissions 2, Residence in a nursing home, Intensive care unit admit with multiorgan failure, 2 Noncancer hospice Guidelines.
Low income	2.71 (1.305.67)	3.05 (1.019.24)	0.99 (0.422.37)
Married	2.44 (1.145.21)	2.40 (0.876.62)	0.82 (0.421.57)
CARING criteria	0.58 (0.301.14)	0.44 (0.181.09)	0.89 (0.471.66)

Patients with income <$30,000/year were more likely to prefer home (or hospital) over a nursing home or hospice facility. Being married was predictive of preferring home over nursing home or hospice facility. Patients meeting 1 of the CARING criteria trended toward being less likely (P=0.11 for home and P=0.08 for hospital) to prefer home (or hospital) vs nursing home or hospice facility. However, there were no significant predictors for a preference for home or hospital when directly comparing the 2 locations, as expected from observing similar effects of variables in the other 2 logit models.

DISCUSSION

We found, similarly to previous reports in the literature, the majority of patients preferred to die at home. We did not find a significant difference in preferences or location of death by ethnicity or illness severity. Lower‐income patients and married patients were more likely to prefer to be at home over a nursing home or a hospice facility in the last days of life. We found that the minority of patients died at their stated preferred site of death, and female gender was the only predictive variable we found to distinguish those patients who died in a place concordant with their wishes compared to those who did not.

In the literature, previous studies have reported concordance rates between preferred and actual site of death that range from 30% to 90%.[12, 13, 18, 19, 20, 21, 22, 23, 24] We found a concordance rate at the lowest end of this spectrum. In trying to understand our findings and place them in context, it is helpful to examine other studies. Many of these studies focused solely on cancer patients.[13, 18, 19, 20, 21, 22, 23] Cancer follows a more predictable trajectory of decline compared to other common life‐threatening illnesses, such as cardiac disease, emphysema, or liver failure, that often involve periods of acute deteriorations and plateaus throughout illness progression. The more predictable trajectory may explain the overall higher concordance rates found in the studies involving cancer patients.

The majority of studies in the literature examining concordance between preferred and actual site of death recruited the study sample from home health or home palliative care programs that were providing support and care to participants.[10, 12, 13, 18, 22, 25, 26, 27] The high concordance rates reported may be the result of the patients in the sample receiving services at home aimed at eliciting preferences and providing support at home. Our observational study is unique in that we elicited patient preferences from a diverse group of hospitalized adults. Patients had a broad range of medical illness and were at various stages in their disease trajectory. This allowed our findings to be more generalizable, a major strength of our study.

The only variable associated with concordance that we identified to predict concordance between preferred and actual site of death was female gender. Women have been shown to be more active in medical decision making, which may explain our findings.[28] Female gender and illness severity (as measured by the CARING criteria) were found to be associated with concordance when the preference is for death at home. For persons with more advanced medical illness, they may have had more opportunity to consider their preferences and talk about these preferences. It is even possible that our interview prompted some participants to have discussions with their families or providers.

Variables with high face validity, such as high social support, higher education, and completing an advance directive, did not demonstrate any effect on the outcome of concordance. Other studies have shown that low functional status, Caucasian ethnicity, home care, higher education, and social support have been associated with a greater likelihood for a home death.[3, 6, 9] However, although studies specifically examining concordance between preferred and actual site of death have looked at predictors for home death, we were unable to find predictors for concordance across all preferences in the literature. We can conclude from our findings that the factors that influence concordance of preferences for site of death are extremely complex and difficult to capture and measure. This is extremely unsatisfying in the face of the low concordance rate of 30% we identified.

Latino ethnicity showed a trend toward having a negative association with concordance between preferred and actual site of death. This trend persisted whether it was concordance overall or for concordance with those who preferred a death at home. In the literature, Latinos have been found to be less likely to complete advance directives, use hospice services at the end of life, and are more likely to experience a hospital death.[29, 30, 31, 32, 33] As care at the end of life continues to improve, careful attention should be paid to ensure that these kinds of gaps do not widen any further.

We interviewed patients at an index hospitalization. Patients had an acute medical illness or an exacerbation of a chronic medical illness and required at least 24 hours of hospitalization to be eligible for inclusion. Our bedside interview made use of an opportune time to question patients, a time when it may have been easier for patients to visualize severe illness at the end of life, rather than asking this question during a time of wellness. Although participants overwhelmingly stated they preferred to be at home, for those who died, decisions were made in their care that did not allow for this preference.

Our follow‐up after the initial bedside interview only included death records of where and when participants died. We do not have the details and narrative of the conversations that may have taken place that led to the care decisions that determined participants' actual place of death. We do not know if preferences were elicited or discussed, and care decisions then negotiated, to best meet the goals and preferences expressed at that time. We also do not know if the conversations did not occur and the default of medical intervention and cure‐focused care dictated how participants spent the last days of their life. There is evidence that when conversations about goals and preferences do occur, concordance between preferences and care received are high.[12, 21]

We were unable to identify any predictors beyond gender in this cohort of adults hospitalized with a broad spectrum of severe medical illness to predict concordance with stated preferences and actual site of death. We can conclude then, based on our findings and supported by the literature, that the default trends toward institutional end‐of‐life experiences. To shift to a more patient‐centered approach, away from the default, healthcare providers need to embrace a palliative approach and incorporate preferences and goals into the discussions about next steps of care to facilitate the peaceful death that the majority of patients imagine for themselves. Hospitalist physicians have a unique opportunity at an index hospitalization to start the conversation about preferences for care including where patients would want to spend the last days of their life.

Our study does have some limitations. We elicited preferences at a single point in time, at an index hospitalization. It is possible that participants' preferences changed over the course of their illness. However, in Agar et al.'s study of longitudinal patient preferences for site of death and place of care, most preferences remained stable over time.[18] We also did not have data that included palliative care involvement, homecare or hospice utilization, or cause of death. All of these variables may be important predictors of concordance. Issues of symptom management and lack of caregiver may also dictate place of death, even when goals and care are aligned. We do not have data to address these components of end‐of‐life decision making.

CONCLUSION

Patients continue to express a preference for death at home. However, the majority of patients experienced an institutional death. Furthermore, few participants achieved concordance with where they preferred to die and where they actually died. Female gender was the sole factor associated with concordance between preferred and actual site of death. Incorporating a palliative approach that elicits goals and helps match goals to care, may offer the best opportunity to help people die where they chose.

Disclosures: This research was supported by the Brookdale National Fellowship Award and the NIA/Beeson grant 5K23AG028957. All authors have seen and agree with the contents of the article. This submission was not under review by any other publication. The authors have no financial interest or other potential conflicts of interest.

At the turn of the 20th century, most deaths in the United States occurred at home. By the 1960s, over 70% of deaths occurred in an institutional setting, reflecting an evolution of medical technology.[1, 2, 3] With the birth of the hospice movement in the 1970s, dying patients had the opportunity to have both death at home and aggressive symptom control at the end of life. Although there has been a slow decline in the proportion of deaths that occur in the hospital over the past 2 decades,[3] the overwhelming majority of persons state that they would prefer to die at home. However, recent findings suggest that most people will die in an institutional setting.[3, 4, 5, 6]

Although good data exist describing population preferences for location of death, and we know, based on death records, where deaths occur in the United States, there are few studies that examine concordance between preferred and actual site of death at the individual patient level. Furthermore, although factors have been identified that predict death at home, factors predicting concordance between preferred and actual site of death are not well described.[3, 6, 7, 8, 9, 10, 11, 12, 13]

Regardless of where death ultimately occurs, most adults will experience multiple hospitalizations within the last years of their life. Understanding the preferences and subsequent experiences of this population is of particular relevance to hospitalist physicians who are in a unique position to elicit goals from seriously ill patients and help match patient preferences with their medical care. In this observational study, we sought to determine preferences for site of death in a cohort of adult patients admitted to the hospital for medical illness, and then follow those patients to determine where death occurred for those who died. We also sought to explore factors that may predict concordance between preferred and actual site of death. We hypothesized that ethnic diversity and lower socioeconomic status would be associated with a lower likelihood of concordance between preferred and actual site of death. We also hypothesized that advanced care planning would be associated with a higher likelihood of concordance. The Colorado Multi‐Institutional Review Board approved this study.

METHODS

Participants were recruited from 3 hospitals affiliated with the University of Colorado School of Medicine Internal Medicine Residency program, including the Denver Veterans' Administration Center (DVAMC), Denver Health Medical Center (DHMC), and University of Colorado Hospital (UCH). The DVAMC is a large urban Veterans Administration hospital, serving veterans from the Denver metro area, and is a tertiary referral center for veterans in rural Colorado, Wyoming, and parts of Montana. DHMC, the safety‐net hospital for the Denver area, serves over 25% of the residents in the city and county of Denver, including such special populations as the indigent, chronically mentally ill, and persons with polysubstance dependence. UCH had 350 licensed beds at the time of our study and serves as the Rocky Mountain region's only academic tertiary, specialty care, and referral center. At the time of this study, there was limited inpatient palliative care services at the DVAMC and UH, and no palliative care services at DHMC. Participants were screened on the first day following admission to the adult general medical service. Participants were recruited on 96 postadmission days between February 2004 and June 2006. Recruitment days varied from Monday through Friday, to include admissions from the weekend and throughout the year to reduce potential bias due to seasonal trends of diseases such as influenza. Patients were excluded if they died or were discharged within the first 24 hours of admission, were pregnant, jailed, or unable to give informed consent. All other patients were approached and invited to participate in a brief survey.

After informed consent was obtained, participants completed a bedside interview that included self‐identified ethnicity and the Berkman‐Syme Social Network Index,[14] a brief questionnaire quantifying social support from spouse or domestic partner, family, friends, and other religious or secular organizations. Baseline socioeconomic measures (eg, income, employment, home ownership, car ownership) and questions related to the last days of life were also included. Participants were asked the following question, If you were very sick, with an illness that could not be cured, and in bed most of the time, where would you spend the last days of your life if you could chose?

For each participant, we performed a detailed chart review to determine demographic data, presence of advance directives, and CARING criteria (Cancer, Admissions 2, Residence in a nursing home, Intensive care unit admit with multiorgan failure, 2 Noncancer hospice Guidelines), a set of prognostic criteria identifying patients at an index hospitalization who have a high burden of illness and are at risk for death in the following year.[15] We then followed patients for 5 years. If participants died within the follow‐up period, we collected the date and location of death using medical records, death certificates, or in a few cases when official death records were unavailable, direct contact with the family. Participants were considered alive if they had a clinic visit or MD/RN phone contact within 3 months prior to the final collection point date.

RESULTS

Preferred Site of Death

When asked where they preferred to spend the last days of their life, 75% of patients (n=343) stated they would like to be at home. In the hospital was the preferred location for 10% of patients, whereas 6% stated a nursing home and 4% a hospice inpatient facility. Two percent stated they had no preference, and 3% refused to answer (Figure 1)

Figure 1

We found that in the univariable analysis the following factors were associated with preference for site of death at a significance level of P<0.25: unstable housing, hospital setting, income level, ethnicity, CARING criteria, presence of an advance directive, education level, married, primary care provider, and presence of public insurance. Results of the logit models (home vs nursing home or hospice facility, and hospital vs nursing home or hospice facility and home vs hospital) are presented in Table 2.

Table 2.Logistic Regression Modeling of Preference for Death at Home or Hospital

	Adjusted Odds Ratio (95% Confidence Interval)
	Home vs Nursing Home/Hospice Facility	Hospital vs Nursing Home/Hospice Facility	Home vs Hospital
NOTE: Abbreviations: CARING, Cancer, Admissions 2, Residence in a nursing home, Intensive care unit admit with multiorgan failure, 2 Noncancer hospice Guidelines.
Low income	2.71 (1.305.67)	3.05 (1.019.24)	0.99 (0.422.37)
Married	2.44 (1.145.21)	2.40 (0.876.62)	0.82 (0.421.57)
CARING criteria	0.58 (0.301.14)	0.44 (0.181.09)	0.89 (0.471.66)

Patients with income <$30,000/year were more likely to prefer home (or hospital) over a nursing home or hospice facility. Being married was predictive of preferring home over nursing home or hospice facility. Patients meeting 1 of the CARING criteria trended toward being less likely (P=0.11 for home and P=0.08 for hospital) to prefer home (or hospital) vs nursing home or hospice facility. However, there were no significant predictors for a preference for home or hospital when directly comparing the 2 locations, as expected from observing similar effects of variables in the other 2 logit models.

DISCUSSION

We found, similarly to previous reports in the literature, the majority of patients preferred to die at home. We did not find a significant difference in preferences or location of death by ethnicity or illness severity. Lower‐income patients and married patients were more likely to prefer to be at home over a nursing home or a hospice facility in the last days of life. We found that the minority of patients died at their stated preferred site of death, and female gender was the only predictive variable we found to distinguish those patients who died in a place concordant with their wishes compared to those who did not.

In the literature, previous studies have reported concordance rates between preferred and actual site of death that range from 30% to 90%.[12, 13, 18, 19, 20, 21, 22, 23, 24] We found a concordance rate at the lowest end of this spectrum. In trying to understand our findings and place them in context, it is helpful to examine other studies. Many of these studies focused solely on cancer patients.[13, 18, 19, 20, 21, 22, 23] Cancer follows a more predictable trajectory of decline compared to other common life‐threatening illnesses, such as cardiac disease, emphysema, or liver failure, that often involve periods of acute deteriorations and plateaus throughout illness progression. The more predictable trajectory may explain the overall higher concordance rates found in the studies involving cancer patients.

The majority of studies in the literature examining concordance between preferred and actual site of death recruited the study sample from home health or home palliative care programs that were providing support and care to participants.[10, 12, 13, 18, 22, 25, 26, 27] The high concordance rates reported may be the result of the patients in the sample receiving services at home aimed at eliciting preferences and providing support at home. Our observational study is unique in that we elicited patient preferences from a diverse group of hospitalized adults. Patients had a broad range of medical illness and were at various stages in their disease trajectory. This allowed our findings to be more generalizable, a major strength of our study.

The only variable associated with concordance that we identified to predict concordance between preferred and actual site of death was female gender. Women have been shown to be more active in medical decision making, which may explain our findings.[28] Female gender and illness severity (as measured by the CARING criteria) were found to be associated with concordance when the preference is for death at home. For persons with more advanced medical illness, they may have had more opportunity to consider their preferences and talk about these preferences. It is even possible that our interview prompted some participants to have discussions with their families or providers.

Variables with high face validity, such as high social support, higher education, and completing an advance directive, did not demonstrate any effect on the outcome of concordance. Other studies have shown that low functional status, Caucasian ethnicity, home care, higher education, and social support have been associated with a greater likelihood for a home death.[3, 6, 9] However, although studies specifically examining concordance between preferred and actual site of death have looked at predictors for home death, we were unable to find predictors for concordance across all preferences in the literature. We can conclude from our findings that the factors that influence concordance of preferences for site of death are extremely complex and difficult to capture and measure. This is extremely unsatisfying in the face of the low concordance rate of 30% we identified.

Latino ethnicity showed a trend toward having a negative association with concordance between preferred and actual site of death. This trend persisted whether it was concordance overall or for concordance with those who preferred a death at home. In the literature, Latinos have been found to be less likely to complete advance directives, use hospice services at the end of life, and are more likely to experience a hospital death.[29, 30, 31, 32, 33] As care at the end of life continues to improve, careful attention should be paid to ensure that these kinds of gaps do not widen any further.

We interviewed patients at an index hospitalization. Patients had an acute medical illness or an exacerbation of a chronic medical illness and required at least 24 hours of hospitalization to be eligible for inclusion. Our bedside interview made use of an opportune time to question patients, a time when it may have been easier for patients to visualize severe illness at the end of life, rather than asking this question during a time of wellness. Although participants overwhelmingly stated they preferred to be at home, for those who died, decisions were made in their care that did not allow for this preference.

Our follow‐up after the initial bedside interview only included death records of where and when participants died. We do not have the details and narrative of the conversations that may have taken place that led to the care decisions that determined participants' actual place of death. We do not know if preferences were elicited or discussed, and care decisions then negotiated, to best meet the goals and preferences expressed at that time. We also do not know if the conversations did not occur and the default of medical intervention and cure‐focused care dictated how participants spent the last days of their life. There is evidence that when conversations about goals and preferences do occur, concordance between preferences and care received are high.[12, 21]

We were unable to identify any predictors beyond gender in this cohort of adults hospitalized with a broad spectrum of severe medical illness to predict concordance with stated preferences and actual site of death. We can conclude then, based on our findings and supported by the literature, that the default trends toward institutional end‐of‐life experiences. To shift to a more patient‐centered approach, away from the default, healthcare providers need to embrace a palliative approach and incorporate preferences and goals into the discussions about next steps of care to facilitate the peaceful death that the majority of patients imagine for themselves. Hospitalist physicians have a unique opportunity at an index hospitalization to start the conversation about preferences for care including where patients would want to spend the last days of their life.

Our study does have some limitations. We elicited preferences at a single point in time, at an index hospitalization. It is possible that participants' preferences changed over the course of their illness. However, in Agar et al.'s study of longitudinal patient preferences for site of death and place of care, most preferences remained stable over time.[18] We also did not have data that included palliative care involvement, homecare or hospice utilization, or cause of death. All of these variables may be important predictors of concordance. Issues of symptom management and lack of caregiver may also dictate place of death, even when goals and care are aligned. We do not have data to address these components of end‐of‐life decision making.

CONCLUSION

Patients continue to express a preference for death at home. However, the majority of patients experienced an institutional death. Furthermore, few participants achieved concordance with where they preferred to die and where they actually died. Female gender was the sole factor associated with concordance between preferred and actual site of death. Incorporating a palliative approach that elicits goals and helps match goals to care, may offer the best opportunity to help people die where they chose.

Disclosures: This research was supported by the Brookdale National Fellowship Award and the NIA/Beeson grant 5K23AG028957. All authors have seen and agree with the contents of the article. This submission was not under review by any other publication. The authors have no financial interest or other potential conflicts of interest.

At the turn of the 20th century, most deaths in the United States occurred at home. By the 1960s, over 70% of deaths occurred in an institutional setting, reflecting an evolution of medical technology.[1, 2, 3] With the birth of the hospice movement in the 1970s, dying patients had the opportunity to have both death at home and aggressive symptom control at the end of life. Although there has been a slow decline in the proportion of deaths that occur in the hospital over the past 2 decades,[3] the overwhelming majority of persons state that they would prefer to die at home. However, recent findings suggest that most people will die in an institutional setting.[3, 4, 5, 6]

Although good data exist describing population preferences for location of death, and we know, based on death records, where deaths occur in the United States, there are few studies that examine concordance between preferred and actual site of death at the individual patient level. Furthermore, although factors have been identified that predict death at home, factors predicting concordance between preferred and actual site of death are not well described.[3, 6, 7, 8, 9, 10, 11, 12, 13]

Regardless of where death ultimately occurs, most adults will experience multiple hospitalizations within the last years of their life. Understanding the preferences and subsequent experiences of this population is of particular relevance to hospitalist physicians who are in a unique position to elicit goals from seriously ill patients and help match patient preferences with their medical care. In this observational study, we sought to determine preferences for site of death in a cohort of adult patients admitted to the hospital for medical illness, and then follow those patients to determine where death occurred for those who died. We also sought to explore factors that may predict concordance between preferred and actual site of death. We hypothesized that ethnic diversity and lower socioeconomic status would be associated with a lower likelihood of concordance between preferred and actual site of death. We also hypothesized that advanced care planning would be associated with a higher likelihood of concordance. The Colorado Multi‐Institutional Review Board approved this study.

METHODS

Participants were recruited from 3 hospitals affiliated with the University of Colorado School of Medicine Internal Medicine Residency program, including the Denver Veterans' Administration Center (DVAMC), Denver Health Medical Center (DHMC), and University of Colorado Hospital (UCH). The DVAMC is a large urban Veterans Administration hospital, serving veterans from the Denver metro area, and is a tertiary referral center for veterans in rural Colorado, Wyoming, and parts of Montana. DHMC, the safety‐net hospital for the Denver area, serves over 25% of the residents in the city and county of Denver, including such special populations as the indigent, chronically mentally ill, and persons with polysubstance dependence. UCH had 350 licensed beds at the time of our study and serves as the Rocky Mountain region's only academic tertiary, specialty care, and referral center. At the time of this study, there was limited inpatient palliative care services at the DVAMC and UH, and no palliative care services at DHMC. Participants were screened on the first day following admission to the adult general medical service. Participants were recruited on 96 postadmission days between February 2004 and June 2006. Recruitment days varied from Monday through Friday, to include admissions from the weekend and throughout the year to reduce potential bias due to seasonal trends of diseases such as influenza. Patients were excluded if they died or were discharged within the first 24 hours of admission, were pregnant, jailed, or unable to give informed consent. All other patients were approached and invited to participate in a brief survey.

After informed consent was obtained, participants completed a bedside interview that included self‐identified ethnicity and the Berkman‐Syme Social Network Index,[14] a brief questionnaire quantifying social support from spouse or domestic partner, family, friends, and other religious or secular organizations. Baseline socioeconomic measures (eg, income, employment, home ownership, car ownership) and questions related to the last days of life were also included. Participants were asked the following question, If you were very sick, with an illness that could not be cured, and in bed most of the time, where would you spend the last days of your life if you could chose?

For each participant, we performed a detailed chart review to determine demographic data, presence of advance directives, and CARING criteria (Cancer, Admissions 2, Residence in a nursing home, Intensive care unit admit with multiorgan failure, 2 Noncancer hospice Guidelines), a set of prognostic criteria identifying patients at an index hospitalization who have a high burden of illness and are at risk for death in the following year.[15] We then followed patients for 5 years. If participants died within the follow‐up period, we collected the date and location of death using medical records, death certificates, or in a few cases when official death records were unavailable, direct contact with the family. Participants were considered alive if they had a clinic visit or MD/RN phone contact within 3 months prior to the final collection point date.

RESULTS

Preferred Site of Death

When asked where they preferred to spend the last days of their life, 75% of patients (n=343) stated they would like to be at home. In the hospital was the preferred location for 10% of patients, whereas 6% stated a nursing home and 4% a hospice inpatient facility. Two percent stated they had no preference, and 3% refused to answer (Figure 1)

Figure 1

We found that in the univariable analysis the following factors were associated with preference for site of death at a significance level of P<0.25: unstable housing, hospital setting, income level, ethnicity, CARING criteria, presence of an advance directive, education level, married, primary care provider, and presence of public insurance. Results of the logit models (home vs nursing home or hospice facility, and hospital vs nursing home or hospice facility and home vs hospital) are presented in Table 2.

Table 2.Logistic Regression Modeling of Preference for Death at Home or Hospital

	Adjusted Odds Ratio (95% Confidence Interval)
	Home vs Nursing Home/Hospice Facility	Hospital vs Nursing Home/Hospice Facility	Home vs Hospital
NOTE: Abbreviations: CARING, Cancer, Admissions 2, Residence in a nursing home, Intensive care unit admit with multiorgan failure, 2 Noncancer hospice Guidelines.
Low income	2.71 (1.305.67)	3.05 (1.019.24)	0.99 (0.422.37)
Married	2.44 (1.145.21)	2.40 (0.876.62)	0.82 (0.421.57)
CARING criteria	0.58 (0.301.14)	0.44 (0.181.09)	0.89 (0.471.66)

Patients with income <$30,000/year were more likely to prefer home (or hospital) over a nursing home or hospice facility. Being married was predictive of preferring home over nursing home or hospice facility. Patients meeting 1 of the CARING criteria trended toward being less likely (P=0.11 for home and P=0.08 for hospital) to prefer home (or hospital) vs nursing home or hospice facility. However, there were no significant predictors for a preference for home or hospital when directly comparing the 2 locations, as expected from observing similar effects of variables in the other 2 logit models.

DISCUSSION

We found, similarly to previous reports in the literature, the majority of patients preferred to die at home. We did not find a significant difference in preferences or location of death by ethnicity or illness severity. Lower‐income patients and married patients were more likely to prefer to be at home over a nursing home or a hospice facility in the last days of life. We found that the minority of patients died at their stated preferred site of death, and female gender was the only predictive variable we found to distinguish those patients who died in a place concordant with their wishes compared to those who did not.

In the literature, previous studies have reported concordance rates between preferred and actual site of death that range from 30% to 90%.[12, 13, 18, 19, 20, 21, 22, 23, 24] We found a concordance rate at the lowest end of this spectrum. In trying to understand our findings and place them in context, it is helpful to examine other studies. Many of these studies focused solely on cancer patients.[13, 18, 19, 20, 21, 22, 23] Cancer follows a more predictable trajectory of decline compared to other common life‐threatening illnesses, such as cardiac disease, emphysema, or liver failure, that often involve periods of acute deteriorations and plateaus throughout illness progression. The more predictable trajectory may explain the overall higher concordance rates found in the studies involving cancer patients.

The majority of studies in the literature examining concordance between preferred and actual site of death recruited the study sample from home health or home palliative care programs that were providing support and care to participants.[10, 12, 13, 18, 22, 25, 26, 27] The high concordance rates reported may be the result of the patients in the sample receiving services at home aimed at eliciting preferences and providing support at home. Our observational study is unique in that we elicited patient preferences from a diverse group of hospitalized adults. Patients had a broad range of medical illness and were at various stages in their disease trajectory. This allowed our findings to be more generalizable, a major strength of our study.

The only variable associated with concordance that we identified to predict concordance between preferred and actual site of death was female gender. Women have been shown to be more active in medical decision making, which may explain our findings.[28] Female gender and illness severity (as measured by the CARING criteria) were found to be associated with concordance when the preference is for death at home. For persons with more advanced medical illness, they may have had more opportunity to consider their preferences and talk about these preferences. It is even possible that our interview prompted some participants to have discussions with their families or providers.

Variables with high face validity, such as high social support, higher education, and completing an advance directive, did not demonstrate any effect on the outcome of concordance. Other studies have shown that low functional status, Caucasian ethnicity, home care, higher education, and social support have been associated with a greater likelihood for a home death.[3, 6, 9] However, although studies specifically examining concordance between preferred and actual site of death have looked at predictors for home death, we were unable to find predictors for concordance across all preferences in the literature. We can conclude from our findings that the factors that influence concordance of preferences for site of death are extremely complex and difficult to capture and measure. This is extremely unsatisfying in the face of the low concordance rate of 30% we identified.

Latino ethnicity showed a trend toward having a negative association with concordance between preferred and actual site of death. This trend persisted whether it was concordance overall or for concordance with those who preferred a death at home. In the literature, Latinos have been found to be less likely to complete advance directives, use hospice services at the end of life, and are more likely to experience a hospital death.[29, 30, 31, 32, 33] As care at the end of life continues to improve, careful attention should be paid to ensure that these kinds of gaps do not widen any further.

We interviewed patients at an index hospitalization. Patients had an acute medical illness or an exacerbation of a chronic medical illness and required at least 24 hours of hospitalization to be eligible for inclusion. Our bedside interview made use of an opportune time to question patients, a time when it may have been easier for patients to visualize severe illness at the end of life, rather than asking this question during a time of wellness. Although participants overwhelmingly stated they preferred to be at home, for those who died, decisions were made in their care that did not allow for this preference.

Our follow‐up after the initial bedside interview only included death records of where and when participants died. We do not have the details and narrative of the conversations that may have taken place that led to the care decisions that determined participants' actual place of death. We do not know if preferences were elicited or discussed, and care decisions then negotiated, to best meet the goals and preferences expressed at that time. We also do not know if the conversations did not occur and the default of medical intervention and cure‐focused care dictated how participants spent the last days of their life. There is evidence that when conversations about goals and preferences do occur, concordance between preferences and care received are high.[12, 21]

We were unable to identify any predictors beyond gender in this cohort of adults hospitalized with a broad spectrum of severe medical illness to predict concordance with stated preferences and actual site of death. We can conclude then, based on our findings and supported by the literature, that the default trends toward institutional end‐of‐life experiences. To shift to a more patient‐centered approach, away from the default, healthcare providers need to embrace a palliative approach and incorporate preferences and goals into the discussions about next steps of care to facilitate the peaceful death that the majority of patients imagine for themselves. Hospitalist physicians have a unique opportunity at an index hospitalization to start the conversation about preferences for care including where patients would want to spend the last days of their life.

Our study does have some limitations. We elicited preferences at a single point in time, at an index hospitalization. It is possible that participants' preferences changed over the course of their illness. However, in Agar et al.'s study of longitudinal patient preferences for site of death and place of care, most preferences remained stable over time.[18] We also did not have data that included palliative care involvement, homecare or hospice utilization, or cause of death. All of these variables may be important predictors of concordance. Issues of symptom management and lack of caregiver may also dictate place of death, even when goals and care are aligned. We do not have data to address these components of end‐of‐life decision making.

CONCLUSION

Patients continue to express a preference for death at home. However, the majority of patients experienced an institutional death. Furthermore, few participants achieved concordance with where they preferred to die and where they actually died. Female gender was the sole factor associated with concordance between preferred and actual site of death. Incorporating a palliative approach that elicits goals and helps match goals to care, may offer the best opportunity to help people die where they chose.

Disclosures: This research was supported by the Brookdale National Fellowship Award and the NIA/Beeson grant 5K23AG028957. All authors have seen and agree with the contents of the article. This submission was not under review by any other publication. The authors have no financial interest or other potential conflicts of interest.

Timely interventions are essential in the management of complex medical conditions such as new‐onset sepsis in order to prevent rapid progression to severe sepsis and septic shock.[1, 2, 3, 4, 5] Similarly, rapid identification and appropriate treatment of other medical and surgical conditions have been associated with improved outcomes.[6, 7, 8] We previously developed a real‐time, computerized prediction tool (PT) using recursive partitioning regression tree analysis for the identification of impending sepsis for use on general hospital wards.[9] We also showed that implementation of a real‐time computerized sepsis alert on hospital wards based on the PT resulted in increased use of early interventions, including antibiotic escalation, intravenous fluids, oxygen therapy, and diagnostics in patients identified as at risk.[10]

The first goal of this study was to develop an updated PT for use on hospital wards that could be used to predict subsequent global clinical deterioration and the need for a higher level of care. The second goal was to determine whether simply providing a real‐time alert to nursing staff based on the updated PT resulted in any demonstrable changes in patient outcomes.

METHODS

RESULTS

DISCUSSION

We have demonstrated that a relatively simple hospital‐specific method for generating a PT derived from routine laboratory and hemodynamic values is capable of predicting clinical deterioration and the need for ICU transfer, as well as hospital mortality, in non‐ICU patients admitted to general hospital wards. We also found that the PT identified a sicker patient population as manifest by longer hospital LOS. The methods used in generating this real‐time PT are relatively simple and easily executed with the use of an electronic medical record (EMR) system. However, our data also showed that simply providing an alert to nursing units based on the PT did not result in any demonstrable improvement in patient outcomes. Moreover, our PT and intervention in their current form have substantial limitations, including low sensitivity and positive predictive value, high possibility of alert fatigue, and no clear clinical impact. These limitations suggest that this approach has limited applicability in its current form.

Unplanned ICU transfers occurring as early as within 8 hours of hospitalization are relatively common and associated with increased mortality.[13] Bapoje et al evaluated a total of 152 patients over 1 year who had unplanned ICU transfers.[14] The most common reason was worsening of the problem for which the patient was admitted (48%). Other investigators have also attempted to identify predictors for clinical deterioration resulting in unplanned ICU transfer that could be employed in a PT or early warning system (EWS). Keller et al evaluated 50 consecutive general medical patients with unplanned ICU transfers between 2003 and 2004.[15] Using a case‐control methodology, these investigators found shock index values>0.85 to be the best predictor for subsequent unplanned ICU transfer (P<0.02; odds ratio: 3.0).

Organizations such as the Institute for Healthcare Improvement have called for the development and implementation of EWSs in order to direct the activities of RRTs and improve outcomes.[16] Escobar et al carried out a retrospective case‐control study using as the unit of analysis 12‐hour patient shifts on hospital wards.[17] Using logistic regression and split validation, they developed a PT for ICU transfer from clinical variables available in their EMR. The EMR derived PT had a C‐statistic of 0.845 in the derivation dataset and 0.775 in the validation dataset, concluding that EMR‐based detection of impending deterioration outside the ICU is feasible in integrated healthcare delivery systems.

We found that simply providing an alert to nursing units did not result in any demonstrable improvements in the outcomes of high‐risk patients identified by our PT. This may have been due to simply relying on the alerted nursing staff to make phone calls to physicians and not linking a specific and effective patient‐directed intervention to the PT. Other investigators have similarly observed that the use of an EWS or PT may not result in outcome improvements.[18] Gao et al performed an analysis of 31 studies describing hospital track and trigger EWSs.[19] They found little evidence of reliability, validity, and utility of these systems. Peebles et al showed that even when high‐risk non‐ICU patients are identified, delays in providing appropriate therapies occur, which may explain the lack of efficacy of EWSs and RRTs.[20] These observations suggest that there is currently a paucity of validated interventions available to improve outcome in deteriorating patients, despite our ability to identify patients who are at risk for such deterioration.

As a result of mandates from quality‐improvement organizations, most US hospitals currently employ RRTs for emergent mobilization of resources when a clinically deteriorating patient is identified on a hospital ward.[21] However, as noted above, there is limited evidence to suggest that RRTs contribute to improved patient outcomes.[22, 23, 24, 25, 26, 27] The potential importance of this is reflected in a recent report suggesting that 2900 US hospitals now have rapid‐response systems in place without clear demonstration of their overall efficacy.[28] Linking rapid‐response interventions with a validated real‐time alert may represent a way of improving the effectiveness of such interventions.[29, 30, 31, 32, 33, 34] Our data showed that hospital LOS was statistically longer among alerted patients compared with nonalerted patients. This supports the conclusion that the alerts helped identify a sicker group of patients, but the nursing alerts did not appear to change outcomes. This finding also seems to refute the hypothesis that simply linking an intervention to a PT will improve outcomes, albeit the intervention we employed may not have been robust enough to influence patient outcomes.

The development of accurate real‐time EWSs holds the potential to identify patients at risk for clinical deterioration at an earlier point in time when rescue interventions can be implemented in a potentially more effective manner in both adults and children.[35] Unfortunately, the ideal intervention to be applied in this situation is unknown. Our experience suggests that successful interventions will require a more integrated approach than simply providing an alert with general management principles. As a result of our experience, we are undertaking a randomized clinical trial in 2013 to determine whether linking a patient‐specific intervention to a PT will result in improved outcomes. The intervention we will be testing is to have the RRT immediately notified about alerted patients so as to formally evaluate them and to determine the need for therapeutic interventions, and to administer such interventions as needed and/or transfer the alerted patients to a higher level of care as deemed necessary. Additionally, we are updating our PT with more temporal data to determine if this will improve its accuracy. One of these updates will include linking the PT to wirelessly obtained continuous oximetry and heart‐rate data, using minimally intrusive sensors, to establish a 2‐tiered EWS.[11]

Our study has several important limitations. First, the PT was developed using local data, and thus the results may not be applicable to other settings. However, our model shares many of the characteristics identified in other clinical‐deterioration PTs.[15, 17] Second, the positive prediction value of 15.2% for ICU transfer may not be clinically useful due to the large number of false‐positive results. Moreover, the large number of false positives could result in alert fatigue, causing alerts to be ignored. Third, although the charge nurses were supposed to call the responsible physicians for the alerted patients, we did not determine whether all these calls occurred or whether they resulted in any meaningful changes in monitoring or patient treatment. This is important because lack of an effective intervention or treatment would make the intervention group much more like our control group. Future studies are needed to assess the impact of an integrated intervention (eg, notification of experienced RRT members with adequate resource access) to determine if patient outcomes can be impacted by the use of an EWS. Finally, we did not compare the performance of our PT to other models such as the modified early warning score (MEWS).

An additional limitation to consider is that our PT offered no new information to the nurse manager, or the PT did not change the opinions of the charge nurses. This is supported by a recent study of 63 serious adverse outcomes in a Belgian teaching hospital where death was the final outcome.[36] Survey results revealed that nurses were often unaware that their patients were deteriorating before the crisis. Nurses also reported threshold levels for concern for abnormal vital signs that suggested they would call for assistance relatively late in clinical crises. The limited ability of nursing staff to identify deteriorating patients is also supported by a recent simulation study demonstrating that nurses did identify that patients were deteriorating, but as each patient deteriorated staff performance declined, with a reduction in all observational records and actions.[37]

In summary, we have demonstrated that a relatively simple hospital‐specific PT could accurately identify patients on general medicine wards who subsequently developed clinical deterioration and the need for ICU transfer, as well as hospital mortality. However, no improvements in patient outcomes were found from reporting this information to nursing wards on a real‐time basis. The low positive predictive value of the alerts, local development of the EWS, and absence of improved outcomes substantially limits the broader application of this system in its current form. Continued efforts are needed to identify and implement systems that will not only accurately identify high‐risk patients on general wards but also intervene to improve their outcomes.

Timely interventions are essential in the management of complex medical conditions such as new‐onset sepsis in order to prevent rapid progression to severe sepsis and septic shock.[1, 2, 3, 4, 5] Similarly, rapid identification and appropriate treatment of other medical and surgical conditions have been associated with improved outcomes.[6, 7, 8] We previously developed a real‐time, computerized prediction tool (PT) using recursive partitioning regression tree analysis for the identification of impending sepsis for use on general hospital wards.[9] We also showed that implementation of a real‐time computerized sepsis alert on hospital wards based on the PT resulted in increased use of early interventions, including antibiotic escalation, intravenous fluids, oxygen therapy, and diagnostics in patients identified as at risk.[10]

The first goal of this study was to develop an updated PT for use on hospital wards that could be used to predict subsequent global clinical deterioration and the need for a higher level of care. The second goal was to determine whether simply providing a real‐time alert to nursing staff based on the updated PT resulted in any demonstrable changes in patient outcomes.

METHODS

RESULTS

DISCUSSION

We have demonstrated that a relatively simple hospital‐specific method for generating a PT derived from routine laboratory and hemodynamic values is capable of predicting clinical deterioration and the need for ICU transfer, as well as hospital mortality, in non‐ICU patients admitted to general hospital wards. We also found that the PT identified a sicker patient population as manifest by longer hospital LOS. The methods used in generating this real‐time PT are relatively simple and easily executed with the use of an electronic medical record (EMR) system. However, our data also showed that simply providing an alert to nursing units based on the PT did not result in any demonstrable improvement in patient outcomes. Moreover, our PT and intervention in their current form have substantial limitations, including low sensitivity and positive predictive value, high possibility of alert fatigue, and no clear clinical impact. These limitations suggest that this approach has limited applicability in its current form.

Unplanned ICU transfers occurring as early as within 8 hours of hospitalization are relatively common and associated with increased mortality.[13] Bapoje et al evaluated a total of 152 patients over 1 year who had unplanned ICU transfers.[14] The most common reason was worsening of the problem for which the patient was admitted (48%). Other investigators have also attempted to identify predictors for clinical deterioration resulting in unplanned ICU transfer that could be employed in a PT or early warning system (EWS). Keller et al evaluated 50 consecutive general medical patients with unplanned ICU transfers between 2003 and 2004.[15] Using a case‐control methodology, these investigators found shock index values>0.85 to be the best predictor for subsequent unplanned ICU transfer (P<0.02; odds ratio: 3.0).

Organizations such as the Institute for Healthcare Improvement have called for the development and implementation of EWSs in order to direct the activities of RRTs and improve outcomes.[16] Escobar et al carried out a retrospective case‐control study using as the unit of analysis 12‐hour patient shifts on hospital wards.[17] Using logistic regression and split validation, they developed a PT for ICU transfer from clinical variables available in their EMR. The EMR derived PT had a C‐statistic of 0.845 in the derivation dataset and 0.775 in the validation dataset, concluding that EMR‐based detection of impending deterioration outside the ICU is feasible in integrated healthcare delivery systems.

We found that simply providing an alert to nursing units did not result in any demonstrable improvements in the outcomes of high‐risk patients identified by our PT. This may have been due to simply relying on the alerted nursing staff to make phone calls to physicians and not linking a specific and effective patient‐directed intervention to the PT. Other investigators have similarly observed that the use of an EWS or PT may not result in outcome improvements.[18] Gao et al performed an analysis of 31 studies describing hospital track and trigger EWSs.[19] They found little evidence of reliability, validity, and utility of these systems. Peebles et al showed that even when high‐risk non‐ICU patients are identified, delays in providing appropriate therapies occur, which may explain the lack of efficacy of EWSs and RRTs.[20] These observations suggest that there is currently a paucity of validated interventions available to improve outcome in deteriorating patients, despite our ability to identify patients who are at risk for such deterioration.

As a result of mandates from quality‐improvement organizations, most US hospitals currently employ RRTs for emergent mobilization of resources when a clinically deteriorating patient is identified on a hospital ward.[21] However, as noted above, there is limited evidence to suggest that RRTs contribute to improved patient outcomes.[22, 23, 24, 25, 26, 27] The potential importance of this is reflected in a recent report suggesting that 2900 US hospitals now have rapid‐response systems in place without clear demonstration of their overall efficacy.[28] Linking rapid‐response interventions with a validated real‐time alert may represent a way of improving the effectiveness of such interventions.[29, 30, 31, 32, 33, 34] Our data showed that hospital LOS was statistically longer among alerted patients compared with nonalerted patients. This supports the conclusion that the alerts helped identify a sicker group of patients, but the nursing alerts did not appear to change outcomes. This finding also seems to refute the hypothesis that simply linking an intervention to a PT will improve outcomes, albeit the intervention we employed may not have been robust enough to influence patient outcomes.

The development of accurate real‐time EWSs holds the potential to identify patients at risk for clinical deterioration at an earlier point in time when rescue interventions can be implemented in a potentially more effective manner in both adults and children.[35] Unfortunately, the ideal intervention to be applied in this situation is unknown. Our experience suggests that successful interventions will require a more integrated approach than simply providing an alert with general management principles. As a result of our experience, we are undertaking a randomized clinical trial in 2013 to determine whether linking a patient‐specific intervention to a PT will result in improved outcomes. The intervention we will be testing is to have the RRT immediately notified about alerted patients so as to formally evaluate them and to determine the need for therapeutic interventions, and to administer such interventions as needed and/or transfer the alerted patients to a higher level of care as deemed necessary. Additionally, we are updating our PT with more temporal data to determine if this will improve its accuracy. One of these updates will include linking the PT to wirelessly obtained continuous oximetry and heart‐rate data, using minimally intrusive sensors, to establish a 2‐tiered EWS.[11]

Our study has several important limitations. First, the PT was developed using local data, and thus the results may not be applicable to other settings. However, our model shares many of the characteristics identified in other clinical‐deterioration PTs.[15, 17] Second, the positive prediction value of 15.2% for ICU transfer may not be clinically useful due to the large number of false‐positive results. Moreover, the large number of false positives could result in alert fatigue, causing alerts to be ignored. Third, although the charge nurses were supposed to call the responsible physicians for the alerted patients, we did not determine whether all these calls occurred or whether they resulted in any meaningful changes in monitoring or patient treatment. This is important because lack of an effective intervention or treatment would make the intervention group much more like our control group. Future studies are needed to assess the impact of an integrated intervention (eg, notification of experienced RRT members with adequate resource access) to determine if patient outcomes can be impacted by the use of an EWS. Finally, we did not compare the performance of our PT to other models such as the modified early warning score (MEWS).

An additional limitation to consider is that our PT offered no new information to the nurse manager, or the PT did not change the opinions of the charge nurses. This is supported by a recent study of 63 serious adverse outcomes in a Belgian teaching hospital where death was the final outcome.[36] Survey results revealed that nurses were often unaware that their patients were deteriorating before the crisis. Nurses also reported threshold levels for concern for abnormal vital signs that suggested they would call for assistance relatively late in clinical crises. The limited ability of nursing staff to identify deteriorating patients is also supported by a recent simulation study demonstrating that nurses did identify that patients were deteriorating, but as each patient deteriorated staff performance declined, with a reduction in all observational records and actions.[37]

In summary, we have demonstrated that a relatively simple hospital‐specific PT could accurately identify patients on general medicine wards who subsequently developed clinical deterioration and the need for ICU transfer, as well as hospital mortality. However, no improvements in patient outcomes were found from reporting this information to nursing wards on a real‐time basis. The low positive predictive value of the alerts, local development of the EWS, and absence of improved outcomes substantially limits the broader application of this system in its current form. Continued efforts are needed to identify and implement systems that will not only accurately identify high‐risk patients on general wards but also intervene to improve their outcomes.

Timely interventions are essential in the management of complex medical conditions such as new‐onset sepsis in order to prevent rapid progression to severe sepsis and septic shock.[1, 2, 3, 4, 5] Similarly, rapid identification and appropriate treatment of other medical and surgical conditions have been associated with improved outcomes.[6, 7, 8] We previously developed a real‐time, computerized prediction tool (PT) using recursive partitioning regression tree analysis for the identification of impending sepsis for use on general hospital wards.[9] We also showed that implementation of a real‐time computerized sepsis alert on hospital wards based on the PT resulted in increased use of early interventions, including antibiotic escalation, intravenous fluids, oxygen therapy, and diagnostics in patients identified as at risk.[10]

The first goal of this study was to develop an updated PT for use on hospital wards that could be used to predict subsequent global clinical deterioration and the need for a higher level of care. The second goal was to determine whether simply providing a real‐time alert to nursing staff based on the updated PT resulted in any demonstrable changes in patient outcomes.

METHODS

RESULTS

DISCUSSION

We have demonstrated that a relatively simple hospital‐specific method for generating a PT derived from routine laboratory and hemodynamic values is capable of predicting clinical deterioration and the need for ICU transfer, as well as hospital mortality, in non‐ICU patients admitted to general hospital wards. We also found that the PT identified a sicker patient population as manifest by longer hospital LOS. The methods used in generating this real‐time PT are relatively simple and easily executed with the use of an electronic medical record (EMR) system. However, our data also showed that simply providing an alert to nursing units based on the PT did not result in any demonstrable improvement in patient outcomes. Moreover, our PT and intervention in their current form have substantial limitations, including low sensitivity and positive predictive value, high possibility of alert fatigue, and no clear clinical impact. These limitations suggest that this approach has limited applicability in its current form.

Unplanned ICU transfers occurring as early as within 8 hours of hospitalization are relatively common and associated with increased mortality.[13] Bapoje et al evaluated a total of 152 patients over 1 year who had unplanned ICU transfers.[14] The most common reason was worsening of the problem for which the patient was admitted (48%). Other investigators have also attempted to identify predictors for clinical deterioration resulting in unplanned ICU transfer that could be employed in a PT or early warning system (EWS). Keller et al evaluated 50 consecutive general medical patients with unplanned ICU transfers between 2003 and 2004.[15] Using a case‐control methodology, these investigators found shock index values>0.85 to be the best predictor for subsequent unplanned ICU transfer (P<0.02; odds ratio: 3.0).

Organizations such as the Institute for Healthcare Improvement have called for the development and implementation of EWSs in order to direct the activities of RRTs and improve outcomes.[16] Escobar et al carried out a retrospective case‐control study using as the unit of analysis 12‐hour patient shifts on hospital wards.[17] Using logistic regression and split validation, they developed a PT for ICU transfer from clinical variables available in their EMR. The EMR derived PT had a C‐statistic of 0.845 in the derivation dataset and 0.775 in the validation dataset, concluding that EMR‐based detection of impending deterioration outside the ICU is feasible in integrated healthcare delivery systems.

We found that simply providing an alert to nursing units did not result in any demonstrable improvements in the outcomes of high‐risk patients identified by our PT. This may have been due to simply relying on the alerted nursing staff to make phone calls to physicians and not linking a specific and effective patient‐directed intervention to the PT. Other investigators have similarly observed that the use of an EWS or PT may not result in outcome improvements.[18] Gao et al performed an analysis of 31 studies describing hospital track and trigger EWSs.[19] They found little evidence of reliability, validity, and utility of these systems. Peebles et al showed that even when high‐risk non‐ICU patients are identified, delays in providing appropriate therapies occur, which may explain the lack of efficacy of EWSs and RRTs.[20] These observations suggest that there is currently a paucity of validated interventions available to improve outcome in deteriorating patients, despite our ability to identify patients who are at risk for such deterioration.

As a result of mandates from quality‐improvement organizations, most US hospitals currently employ RRTs for emergent mobilization of resources when a clinically deteriorating patient is identified on a hospital ward.[21] However, as noted above, there is limited evidence to suggest that RRTs contribute to improved patient outcomes.[22, 23, 24, 25, 26, 27] The potential importance of this is reflected in a recent report suggesting that 2900 US hospitals now have rapid‐response systems in place without clear demonstration of their overall efficacy.[28] Linking rapid‐response interventions with a validated real‐time alert may represent a way of improving the effectiveness of such interventions.[29, 30, 31, 32, 33, 34] Our data showed that hospital LOS was statistically longer among alerted patients compared with nonalerted patients. This supports the conclusion that the alerts helped identify a sicker group of patients, but the nursing alerts did not appear to change outcomes. This finding also seems to refute the hypothesis that simply linking an intervention to a PT will improve outcomes, albeit the intervention we employed may not have been robust enough to influence patient outcomes.

The development of accurate real‐time EWSs holds the potential to identify patients at risk for clinical deterioration at an earlier point in time when rescue interventions can be implemented in a potentially more effective manner in both adults and children.[35] Unfortunately, the ideal intervention to be applied in this situation is unknown. Our experience suggests that successful interventions will require a more integrated approach than simply providing an alert with general management principles. As a result of our experience, we are undertaking a randomized clinical trial in 2013 to determine whether linking a patient‐specific intervention to a PT will result in improved outcomes. The intervention we will be testing is to have the RRT immediately notified about alerted patients so as to formally evaluate them and to determine the need for therapeutic interventions, and to administer such interventions as needed and/or transfer the alerted patients to a higher level of care as deemed necessary. Additionally, we are updating our PT with more temporal data to determine if this will improve its accuracy. One of these updates will include linking the PT to wirelessly obtained continuous oximetry and heart‐rate data, using minimally intrusive sensors, to establish a 2‐tiered EWS.[11]

Our study has several important limitations. First, the PT was developed using local data, and thus the results may not be applicable to other settings. However, our model shares many of the characteristics identified in other clinical‐deterioration PTs.[15, 17] Second, the positive prediction value of 15.2% for ICU transfer may not be clinically useful due to the large number of false‐positive results. Moreover, the large number of false positives could result in alert fatigue, causing alerts to be ignored. Third, although the charge nurses were supposed to call the responsible physicians for the alerted patients, we did not determine whether all these calls occurred or whether they resulted in any meaningful changes in monitoring or patient treatment. This is important because lack of an effective intervention or treatment would make the intervention group much more like our control group. Future studies are needed to assess the impact of an integrated intervention (eg, notification of experienced RRT members with adequate resource access) to determine if patient outcomes can be impacted by the use of an EWS. Finally, we did not compare the performance of our PT to other models such as the modified early warning score (MEWS).

An additional limitation to consider is that our PT offered no new information to the nurse manager, or the PT did not change the opinions of the charge nurses. This is supported by a recent study of 63 serious adverse outcomes in a Belgian teaching hospital where death was the final outcome.[36] Survey results revealed that nurses were often unaware that their patients were deteriorating before the crisis. Nurses also reported threshold levels for concern for abnormal vital signs that suggested they would call for assistance relatively late in clinical crises. The limited ability of nursing staff to identify deteriorating patients is also supported by a recent simulation study demonstrating that nurses did identify that patients were deteriorating, but as each patient deteriorated staff performance declined, with a reduction in all observational records and actions.[37]

In summary, we have demonstrated that a relatively simple hospital‐specific PT could accurately identify patients on general medicine wards who subsequently developed clinical deterioration and the need for ICU transfer, as well as hospital mortality. However, no improvements in patient outcomes were found from reporting this information to nursing wards on a real‐time basis. The low positive predictive value of the alerts, local development of the EWS, and absence of improved outcomes substantially limits the broader application of this system in its current form. Continued efforts are needed to identify and implement systems that will not only accurately identify high‐risk patients on general wards but also intervene to improve their outcomes.