In recent years, much attention has been paid to concerns regarding length of stay (LOS) and safety of hospital discharges.13 Yet studies conducted in a variety of populations suggest that long stays do not wholly reflect acute medical necessity, but may also be driven by nonmedical factors.46 In a study of frail elderly patients, nonmedical factors accounted for over half of patients' hospital stay days.5 Nonmedical factors may include the availability of community and outpatient resources, inadequate patient social support, disagreement with family and/or patient decision‐making, and post‐hospital placement and care needs.7

Homebound patients are at particular risk for long stays because they are typically frail, elderly, and medically complex.8 The United States homebound population numbers at least 2 million and is expected to increase to at least 3 million by the year 2020.9 This group is medically underserved, often only receiving care for medical emergencies, and represents a costly group of health care beneficiaries.10 Although homebound primary care (HBPC) programs are structured to provide coordinated medical and supportive care in the home, the clinical complexity of patients often requires hospitalization during times of acute illness.11

Navigating the discharge process for homebound patients can prove time‐consuming for inpatient physicians whose clinical obligations encompass ensuring safe care transitions between hospital and home.12 The lack of literature on the discharge needs of the homebound population provides little guidance for physicians and health systems seeking to safely transition homebound patients from the hospital to the home in a timely and efficient manner. We performed a pilot study of an urban homebound population cared for by a single academic homebound program to identify and describe nonmedical factors associated with prolonged hospitalization.

METHODS

This retrospective descriptive study included homebound patients cared for by The Mount Sinai Visiting Doctors Program (MSVD), a primary care program affiliated with The Mount Sinai Hospital. MSVD is the largest academic homebound primary care program in the United States. The structure and patient population of MSVD patients have been described previously.13 Briefly, the program employs 8 physicians, 2 nurse practitioners, and support personnel including 2 registered nurses, 4 social workers, and 4 clerical staff members to serve over 1000 homebound patients annually. To be enrolled in the program, patients must meet the Medicare definition of homebound (ie, they must be able to leave home only with great difficulty and for infrequent or short absences). Patients are referred from a variety of sources including emergency rooms, inpatient units, and local nursing and social service agencies. Physicians visit patients on average once every 2 months, but can make more frequent home visits when a clinical need arises. Thirty‐six percent of patients in the program are hospitalized at least once per year while they are under the MSVD's care.14 Patients in the MSVD are referred by their primary care physician to outpatient social work as needed for finite interventions, though not for ongoing case management; approximately one‐third of all MSVD patients have been seen by an MSVD social worker over a 1‐year period.14

All patients enrolled in MSVD discharged from The Mount Sinai Hospital in New York from January 1, 2007, to December 31, 2007, were evaluated for inclusion in this study. The MSVD clinical database was cross‐referenced with The Mount Sinai Hospital data system to maximize reliability of recorded admission and discharge dates. Discrepancies in dates were investigated and corrected by the authors. As the study focused on admissions rather than individual patients, repeat hospitalizations and factors contributing to long stays were included and considered separately. In the event of repeat hospitalizations, factors contributing to LOS were also assessed separately.

Using the University HealthSystems Consortium (UHC) Database, the selected discharges were analyzed for LOS data. The UHC Database contains data from The Mount Sinai Hospital and 106 other academic medical centers and 233 other affiliated hospitals, representing approximately 90% of the United States' nonprofit academic medical centers. UHC members submit clinical, financial, and administrative information for the purpose of facilitating comparative data analysis among institutions.15 The model assigns an expected LOS for each patient based on a 4‐step risk adjustment methodology that adjusts for variations in patient characteristics. The regression models consider a range of independent variables including patient age, sex, race, socioeconomic status, admission source, and comorbid conditions. The expected LOS value is used to produce an LOS ratio, which is the ratio of a patient's observed LOS to expected LOS. We compared LOS ratios for all patient discharges during the study period and determined the mean LOS ratio for the group. Long‐stay patients were defined using the UHC definition of an LOS ratio greater than 2 standard deviations above the mean. These patients were selected for analysis to examine factors contributing to their long stays.

The primary author conducted a chart review for the long‐stay patients. The date of medical readiness for discharge was defined as the date when no acute hospital care needs or pending procedures (eg, intravenous medication, transfusion, invasive and noninvasive testing, surgery) were documented by the attending physician, house officer, nurse practitioner, or physician assistant in the chart. Patients with a discharge immediately following determination of medical readiness were characterized as having a medical stay (eg, patient admitted for pneumonia and discharged home after 3 days once fever, leukocytosis, and symptoms improved with a prescription for remaining antibiotic course). Patients who were classified as medically ready for discharge yet remained additional days in the hospital were categorized as having a nonmedical component to their hospitalization and comprised the nonmedical stay group (eg, patient admitted for pneumonia with improved vital signs and symptoms after 3 days, but discharged home after 10 days awaiting approval of increased home health aide hours). The primary author reviewed cases (<5%) with coauthors when categorization of patient data was unclear from physician documentation in the patient's chart. Similar methodology and terminology have been used in previous studies examining the contribution of nonmedical factors to LOS.5, 7

While literature in other fields, such as social work, often characterize similar factors as social or social care factors, we use the term nonmedical to draw the distinction between factors that acutely reflect a patient's state of health and necessitate days spent in the hospital (eg, surgery, infection), in contrast to factors that are not direct contributors to the patient's current medical status (eg, post‐hospital placement). Factors contributing to nonmedical days were determined based on previous studies and categorized as follows: nursing facility bed availability, nursing facility rejection of the patient, complications with insurance coverage, lack of patient/family agreement with discharge plan, home care service delays, and other. The category other was included to identify and explore any unexpected or unique reasons for prolongation of hospitalization in this population. When multiple nonmedical factors were identified for a given hospitalization, all contributing factors were recorded.

Demographic, clinical, and discharge characteristics were extracted from the MSVD clinical database to qualitatively describe and compare the long‐stay hospitalizations to the remainder of the sample.

RESULTS

There were a total of 479 discharges of 267 unique MSVD patients from The Mount Sinai Hospital occurring from January 1, 2007, to December 31, 2007. During this 12‐month period, the average observed LOS for all admissions was 7.84 days, with a mean UHC LOS ratio of 1.23 (SD = 3.43). Seventeen admissions were identified as long‐stays, representing 3.5% of discharges. The 17 admissions represent 17 unique patients.

As shown in Table 1, the long‐stay group (n = 17) was slightly younger, more likely to be male, and had less dementia than the nonlong‐stay group (n = 462). There was a marked difference in the location of patient discharge; long‐stay patients were more than twice as likely to be discharged to a facility and less likely to be discharged home. There were no in‐hospital deaths in the long‐stay patient group during the time period studied.

Of the 17 long‐stay patients, 8 (47%) remained in the hospital past the date they were determined to be medically ready for discharge and were defined as having a nonmedical component for the extension of their hospitalizations. The number of nonmedical days ranged from 6 to 34 days (mean, 17 days). Out of 428 total long‐stay patient days, 136 were nonmedical. This represented 31.8% of all long‐stay patient days, and 53% of the nonmedical group's total hospital days. The mean LOS ratio for the nonmedical cases was 6.04 (Table 2).

Nine patients were defined as medical stay cases (ie, no nonmedical component contributing to the long hospitalization). The mean observed LOS was 19.2 days, and the mean LOS ratio for this group was 5.07 (Table 2). There were no significant differences between primary diagnosis‐related groups (DRGs) seen in the medical and nonmedical stay groups.

The most common reason for a nonmedical stay was nursing facility placement delays (Table 3), specifically related to lack of bed availability and facility rejection of the patient leading to prolonged time waiting for long‐term placement (n = 6). Other nonmedical factors contributing to LOS were lack of patient and/or family agreement with discharge plans (eg, disagreement among family members regarding caregiving responsibilities, goals of care, or patient refusal to be discharged on a particular day) (n = 4); complications with insurance coverage for facility placement or for home care (n = 3); and home care service delays, such as patient need for increased home care hours after discharge (n = 2). Of note, 5 of the 8 nonmedical stay cases had multiple factors contributing to patients' long stays. All delays were assigned to one of the a priori defined categories. There were no other or unexpected reasons identified.

Of the nonmedical cases, all but 1 patient had been seen by an MSVD social worker prior to hospital admission, though the social work referral may have been years prior to or unrelated to the current admission.

DISCUSSION

Almost half of long‐stay patients identified in this homebound population remained hospitalized in an urban academic medical center due to at least one, and often multiple, nonmedical factors. Nonmedical factors identified in this group are similar to those described in previous studies, particularly family and patient decision‐making and post‐hospital placement and care needs.5, 7 Although this pilot study was limited to a single‐site population, it is to our knowledge the first study to describe these factors in a homebound population, and may be able to guide future research and discussion on this topic.

This study used a risk‐adjusted LOS measure to determine long stay cases. Using the UHC Database allowed for a more accurate understanding of the contribution of nonmedical factors to LOS by accounting for hospitalizations that were numerically lengthy but medically appropriate for their respective DRG. The use of the LOS ratio also allows for standardized application of these data across academic health centers. In our sample, 50% of the patients classified as LOS outliers by the UHC Database (cases with LOS in the top percentile for their respective DRG) had nonmedical stays. Conventional strategies often dismiss outliers in analyses of patient LOS data. However, in doing so there is a missed opportunity to identify underlying reasons for their disproportionately long hospitalizations that may also be impacting the broader set of patients with similar nonmedical factors affecting LOS.

The 8 nonmedical stay patients spent a combined 136 days longer in the hospital than medically necessary due to a variety of nonmedical factors, and represented over half of the nonmedical stay group's total hospital days. Using a conservative estimate for cost per hospital day of $1770,16 the nonmedical days cost the hospital almost a quarter of a million dollars ($240,720). Because this figure only accounts for long‐stay patients, the actual costs attributable to nonmedical days for the homebound population in general may be higher.

The longest patient stays, whether attributable to medical or nonmedical factors, were more likely to result in discharge to a facility than the rest of the sample hospitalizations. Facility placement was the most common nonmedical factor contributing to long stays in this sample. In contrast, home carerelated factors contributed the least to nonmedical days. This finding highlights the need for hospital‐based physicians and other inpatient staff members to be aware that despite patient enrollment in an HBPC, the possibility for homebound patients to be discharged to a nursing facility remains significant. A decreasing number of skilled nursing beds across the United States may magnify this factor in long‐stay cases.17 Increased awareness of this possibility among inpatient staff can allow the team to address facility placement considerations early in the hospital stay, potentially decreasing nonmedical days.18

Seven of the 8 nonmedical stay patients had been referred to and seen by MSVD social workers before hospitalization, a high percentage relative to the general MSVD population, of which fewer than half are seen by a social worker during their enrollment in MSVD. This finding may suggest that this group of patients already exhibited difficult social circumstances before their hospital admission, yet the current referral‐based social work model at MSVD did not mitigate their high LOS. This finding further suggests that patient enrollment in an HBPC does not mitigate the risk of high LOS and prolonged nonmedical stays, and that involvement of inpatient practitioners remains a critical part of advanced discharge planning.

This pilot study found that 32% of all long‐stay hospital days were due to nonmedical factors, suggesting that these factors play a greater role in the homebound population than for general medical patients. A recent study at an academic medical center examined 3574 patient‐days on a general medicine service, and noted that 11% of all days were felt to be medically unnecessary by the treating hospitalists.19 Hospitalists are well situated to participate in and lead improvement efforts given their expertise in managing complex dispositions and advancing collaborative strategies for care of patients with high overall acuity.20 These efforts will be needed to target those patients at highest risk for prolonged LOS with the greatest social care needs. Because this study did not pilot strategies to reduce LOS, we cannot offer evidence‐based suggestions for an enhanced multidisciplinary approach or other avenues for improvement. However, we believe that the study findings provide the basis for future research to test strategies to reduce excess LOS by focusing on nonmedical factors and a multidisciplinary approach. This will become especially relevant as health care systems bear increasing financial responsibility for inefficient and/or unnecessary hospitalizations and readmissions.

The involvement of social work before hospitalization for most of the homebound population with prolonged hospitalization suggests a need for greater team‐based efforts across venues. Though hospital interdisciplinary rounds aim to increase collaboration and reduce LOS, costs, and readmissions, these rounds do not typically include outpatient care providers.21 Improved communication and collaboration between social work with both inpatient and outpatient care teams to address nonmedical issues contributing to long stays are likely to improve care and transitions, though rigorous studies examining specific communication models across venues are lacking. This study found that delay in nursing facility placement was the most common reason for prolonged hospitalization for long‐stay cases. This finding emphasizes the need for communication between inpatient and outpatient staff to convey prior conversations or preparations for placement, identify patients who need post‐discharge facility placement early in hospitalization, and prompt timely discussions with patients and families.

The finding that prolonged hospitalization for the homebound population was due to nonmedical factors for almost one‐half of patients with long hospital stays has important implications for policymakers and other key stakeholders. For example, accountable care organizations are being developed to align members of the health care sector to provide higher quality care in a more efficient manner. These study data suggest that this alignment should include hospitals, nursing homes, and home health care agencies to ensure that discharge delays are minimized and unnecessary societal costs are avoided. Future research will need to confirm and build upon these findings of nonmedical reasons for excessive LOS to further inform the process of implementation of health care reform measures. Recent plans to cut Medicaid funding to nursing homes may further limit bed availability, increasing the risk of prolonged LOS and related costs to the health care system. This potential concern highlights the importance of care coordination and communication between inpatient and outpatient care providers to proactively address nursing home placement needs before hospitalization occurs, and/or to identify alternative safe discharge plans if a previously homebound patient is hospitalized.

There are several limitations to this descriptive study. Admissions included in this analysis were only captured for those admitted to The Mount Sinai Hospital. While MSVD providers report that more than 90% of hospitalizations for MSVD patients occur at The Mount Sinai Hospital, patients may also be admitted to one of many New York City metropolitan area hospitals closer to the patient's residence. It is possible that additional factors contributing to high LOS might be revealed if these admissions were included in the analysis. The urban homebound population served by MSVD may have more access to supplementary home care services (e.g. home attendants, meal services) than populations in more rural and less service‐intensive areas. Thus, it may be difficult to generalize these findings to programs serving less urban constituencies or with more restrictive policies regarding home care services. Additionally, as New York registers one of the highest nursing facility occupancy rates (in 2008, 92.2% versus the national average of 82.9%), patients in other markets may face a shorter wait time for a bed, decreasing the number of nonmedical days attributable to nursing home bed supply.17 The small total number of long‐stay patients also prevented statistical analysis comparing those patients with the rest of the sample. This pilot study may inform the design of future studies that may be able to include multiple HBPC programs or study homebound patients over a longer period to increase sample size.

Identifying the significant contribution of nonmedical days to patient stay is an important initial step to avoiding costly and medically unnecessary days for the patient and the hospital. As has been demonstrated in other interdisciplinary efforts, increased collaboration among physicians, social workers, discharge planners, and other disciplines may help address current gaps in patient care with regard to LOS.20, 21 Future studies should determine which homebound patients are at highest risk for prolonged hospitalization due to nonmedical factors to help design focused strategies and interventions for this vulnerable population.

In recent years, much attention has been paid to concerns regarding length of stay (LOS) and safety of hospital discharges.13 Yet studies conducted in a variety of populations suggest that long stays do not wholly reflect acute medical necessity, but may also be driven by nonmedical factors.46 In a study of frail elderly patients, nonmedical factors accounted for over half of patients' hospital stay days.5 Nonmedical factors may include the availability of community and outpatient resources, inadequate patient social support, disagreement with family and/or patient decision‐making, and post‐hospital placement and care needs.7

Homebound patients are at particular risk for long stays because they are typically frail, elderly, and medically complex.8 The United States homebound population numbers at least 2 million and is expected to increase to at least 3 million by the year 2020.9 This group is medically underserved, often only receiving care for medical emergencies, and represents a costly group of health care beneficiaries.10 Although homebound primary care (HBPC) programs are structured to provide coordinated medical and supportive care in the home, the clinical complexity of patients often requires hospitalization during times of acute illness.11

Navigating the discharge process for homebound patients can prove time‐consuming for inpatient physicians whose clinical obligations encompass ensuring safe care transitions between hospital and home.12 The lack of literature on the discharge needs of the homebound population provides little guidance for physicians and health systems seeking to safely transition homebound patients from the hospital to the home in a timely and efficient manner. We performed a pilot study of an urban homebound population cared for by a single academic homebound program to identify and describe nonmedical factors associated with prolonged hospitalization.

METHODS

This retrospective descriptive study included homebound patients cared for by The Mount Sinai Visiting Doctors Program (MSVD), a primary care program affiliated with The Mount Sinai Hospital. MSVD is the largest academic homebound primary care program in the United States. The structure and patient population of MSVD patients have been described previously.13 Briefly, the program employs 8 physicians, 2 nurse practitioners, and support personnel including 2 registered nurses, 4 social workers, and 4 clerical staff members to serve over 1000 homebound patients annually. To be enrolled in the program, patients must meet the Medicare definition of homebound (ie, they must be able to leave home only with great difficulty and for infrequent or short absences). Patients are referred from a variety of sources including emergency rooms, inpatient units, and local nursing and social service agencies. Physicians visit patients on average once every 2 months, but can make more frequent home visits when a clinical need arises. Thirty‐six percent of patients in the program are hospitalized at least once per year while they are under the MSVD's care.14 Patients in the MSVD are referred by their primary care physician to outpatient social work as needed for finite interventions, though not for ongoing case management; approximately one‐third of all MSVD patients have been seen by an MSVD social worker over a 1‐year period.14

All patients enrolled in MSVD discharged from The Mount Sinai Hospital in New York from January 1, 2007, to December 31, 2007, were evaluated for inclusion in this study. The MSVD clinical database was cross‐referenced with The Mount Sinai Hospital data system to maximize reliability of recorded admission and discharge dates. Discrepancies in dates were investigated and corrected by the authors. As the study focused on admissions rather than individual patients, repeat hospitalizations and factors contributing to long stays were included and considered separately. In the event of repeat hospitalizations, factors contributing to LOS were also assessed separately.

Using the University HealthSystems Consortium (UHC) Database, the selected discharges were analyzed for LOS data. The UHC Database contains data from The Mount Sinai Hospital and 106 other academic medical centers and 233 other affiliated hospitals, representing approximately 90% of the United States' nonprofit academic medical centers. UHC members submit clinical, financial, and administrative information for the purpose of facilitating comparative data analysis among institutions.15 The model assigns an expected LOS for each patient based on a 4‐step risk adjustment methodology that adjusts for variations in patient characteristics. The regression models consider a range of independent variables including patient age, sex, race, socioeconomic status, admission source, and comorbid conditions. The expected LOS value is used to produce an LOS ratio, which is the ratio of a patient's observed LOS to expected LOS. We compared LOS ratios for all patient discharges during the study period and determined the mean LOS ratio for the group. Long‐stay patients were defined using the UHC definition of an LOS ratio greater than 2 standard deviations above the mean. These patients were selected for analysis to examine factors contributing to their long stays.

The primary author conducted a chart review for the long‐stay patients. The date of medical readiness for discharge was defined as the date when no acute hospital care needs or pending procedures (eg, intravenous medication, transfusion, invasive and noninvasive testing, surgery) were documented by the attending physician, house officer, nurse practitioner, or physician assistant in the chart. Patients with a discharge immediately following determination of medical readiness were characterized as having a medical stay (eg, patient admitted for pneumonia and discharged home after 3 days once fever, leukocytosis, and symptoms improved with a prescription for remaining antibiotic course). Patients who were classified as medically ready for discharge yet remained additional days in the hospital were categorized as having a nonmedical component to their hospitalization and comprised the nonmedical stay group (eg, patient admitted for pneumonia with improved vital signs and symptoms after 3 days, but discharged home after 10 days awaiting approval of increased home health aide hours). The primary author reviewed cases (<5%) with coauthors when categorization of patient data was unclear from physician documentation in the patient's chart. Similar methodology and terminology have been used in previous studies examining the contribution of nonmedical factors to LOS.5, 7

While literature in other fields, such as social work, often characterize similar factors as social or social care factors, we use the term nonmedical to draw the distinction between factors that acutely reflect a patient's state of health and necessitate days spent in the hospital (eg, surgery, infection), in contrast to factors that are not direct contributors to the patient's current medical status (eg, post‐hospital placement). Factors contributing to nonmedical days were determined based on previous studies and categorized as follows: nursing facility bed availability, nursing facility rejection of the patient, complications with insurance coverage, lack of patient/family agreement with discharge plan, home care service delays, and other. The category other was included to identify and explore any unexpected or unique reasons for prolongation of hospitalization in this population. When multiple nonmedical factors were identified for a given hospitalization, all contributing factors were recorded.

Demographic, clinical, and discharge characteristics were extracted from the MSVD clinical database to qualitatively describe and compare the long‐stay hospitalizations to the remainder of the sample.

RESULTS

There were a total of 479 discharges of 267 unique MSVD patients from The Mount Sinai Hospital occurring from January 1, 2007, to December 31, 2007. During this 12‐month period, the average observed LOS for all admissions was 7.84 days, with a mean UHC LOS ratio of 1.23 (SD = 3.43). Seventeen admissions were identified as long‐stays, representing 3.5% of discharges. The 17 admissions represent 17 unique patients.

As shown in Table 1, the long‐stay group (n = 17) was slightly younger, more likely to be male, and had less dementia than the nonlong‐stay group (n = 462). There was a marked difference in the location of patient discharge; long‐stay patients were more than twice as likely to be discharged to a facility and less likely to be discharged home. There were no in‐hospital deaths in the long‐stay patient group during the time period studied.

Of the 17 long‐stay patients, 8 (47%) remained in the hospital past the date they were determined to be medically ready for discharge and were defined as having a nonmedical component for the extension of their hospitalizations. The number of nonmedical days ranged from 6 to 34 days (mean, 17 days). Out of 428 total long‐stay patient days, 136 were nonmedical. This represented 31.8% of all long‐stay patient days, and 53% of the nonmedical group's total hospital days. The mean LOS ratio for the nonmedical cases was 6.04 (Table 2).

Nine patients were defined as medical stay cases (ie, no nonmedical component contributing to the long hospitalization). The mean observed LOS was 19.2 days, and the mean LOS ratio for this group was 5.07 (Table 2). There were no significant differences between primary diagnosis‐related groups (DRGs) seen in the medical and nonmedical stay groups.

The most common reason for a nonmedical stay was nursing facility placement delays (Table 3), specifically related to lack of bed availability and facility rejection of the patient leading to prolonged time waiting for long‐term placement (n = 6). Other nonmedical factors contributing to LOS were lack of patient and/or family agreement with discharge plans (eg, disagreement among family members regarding caregiving responsibilities, goals of care, or patient refusal to be discharged on a particular day) (n = 4); complications with insurance coverage for facility placement or for home care (n = 3); and home care service delays, such as patient need for increased home care hours after discharge (n = 2). Of note, 5 of the 8 nonmedical stay cases had multiple factors contributing to patients' long stays. All delays were assigned to one of the a priori defined categories. There were no other or unexpected reasons identified.

Of the nonmedical cases, all but 1 patient had been seen by an MSVD social worker prior to hospital admission, though the social work referral may have been years prior to or unrelated to the current admission.

DISCUSSION

Almost half of long‐stay patients identified in this homebound population remained hospitalized in an urban academic medical center due to at least one, and often multiple, nonmedical factors. Nonmedical factors identified in this group are similar to those described in previous studies, particularly family and patient decision‐making and post‐hospital placement and care needs.5, 7 Although this pilot study was limited to a single‐site population, it is to our knowledge the first study to describe these factors in a homebound population, and may be able to guide future research and discussion on this topic.

This study used a risk‐adjusted LOS measure to determine long stay cases. Using the UHC Database allowed for a more accurate understanding of the contribution of nonmedical factors to LOS by accounting for hospitalizations that were numerically lengthy but medically appropriate for their respective DRG. The use of the LOS ratio also allows for standardized application of these data across academic health centers. In our sample, 50% of the patients classified as LOS outliers by the UHC Database (cases with LOS in the top percentile for their respective DRG) had nonmedical stays. Conventional strategies often dismiss outliers in analyses of patient LOS data. However, in doing so there is a missed opportunity to identify underlying reasons for their disproportionately long hospitalizations that may also be impacting the broader set of patients with similar nonmedical factors affecting LOS.

The 8 nonmedical stay patients spent a combined 136 days longer in the hospital than medically necessary due to a variety of nonmedical factors, and represented over half of the nonmedical stay group's total hospital days. Using a conservative estimate for cost per hospital day of $1770,16 the nonmedical days cost the hospital almost a quarter of a million dollars ($240,720). Because this figure only accounts for long‐stay patients, the actual costs attributable to nonmedical days for the homebound population in general may be higher.

The longest patient stays, whether attributable to medical or nonmedical factors, were more likely to result in discharge to a facility than the rest of the sample hospitalizations. Facility placement was the most common nonmedical factor contributing to long stays in this sample. In contrast, home carerelated factors contributed the least to nonmedical days. This finding highlights the need for hospital‐based physicians and other inpatient staff members to be aware that despite patient enrollment in an HBPC, the possibility for homebound patients to be discharged to a nursing facility remains significant. A decreasing number of skilled nursing beds across the United States may magnify this factor in long‐stay cases.17 Increased awareness of this possibility among inpatient staff can allow the team to address facility placement considerations early in the hospital stay, potentially decreasing nonmedical days.18

Seven of the 8 nonmedical stay patients had been referred to and seen by MSVD social workers before hospitalization, a high percentage relative to the general MSVD population, of which fewer than half are seen by a social worker during their enrollment in MSVD. This finding may suggest that this group of patients already exhibited difficult social circumstances before their hospital admission, yet the current referral‐based social work model at MSVD did not mitigate their high LOS. This finding further suggests that patient enrollment in an HBPC does not mitigate the risk of high LOS and prolonged nonmedical stays, and that involvement of inpatient practitioners remains a critical part of advanced discharge planning.

This pilot study found that 32% of all long‐stay hospital days were due to nonmedical factors, suggesting that these factors play a greater role in the homebound population than for general medical patients. A recent study at an academic medical center examined 3574 patient‐days on a general medicine service, and noted that 11% of all days were felt to be medically unnecessary by the treating hospitalists.19 Hospitalists are well situated to participate in and lead improvement efforts given their expertise in managing complex dispositions and advancing collaborative strategies for care of patients with high overall acuity.20 These efforts will be needed to target those patients at highest risk for prolonged LOS with the greatest social care needs. Because this study did not pilot strategies to reduce LOS, we cannot offer evidence‐based suggestions for an enhanced multidisciplinary approach or other avenues for improvement. However, we believe that the study findings provide the basis for future research to test strategies to reduce excess LOS by focusing on nonmedical factors and a multidisciplinary approach. This will become especially relevant as health care systems bear increasing financial responsibility for inefficient and/or unnecessary hospitalizations and readmissions.

The involvement of social work before hospitalization for most of the homebound population with prolonged hospitalization suggests a need for greater team‐based efforts across venues. Though hospital interdisciplinary rounds aim to increase collaboration and reduce LOS, costs, and readmissions, these rounds do not typically include outpatient care providers.21 Improved communication and collaboration between social work with both inpatient and outpatient care teams to address nonmedical issues contributing to long stays are likely to improve care and transitions, though rigorous studies examining specific communication models across venues are lacking. This study found that delay in nursing facility placement was the most common reason for prolonged hospitalization for long‐stay cases. This finding emphasizes the need for communication between inpatient and outpatient staff to convey prior conversations or preparations for placement, identify patients who need post‐discharge facility placement early in hospitalization, and prompt timely discussions with patients and families.

The finding that prolonged hospitalization for the homebound population was due to nonmedical factors for almost one‐half of patients with long hospital stays has important implications for policymakers and other key stakeholders. For example, accountable care organizations are being developed to align members of the health care sector to provide higher quality care in a more efficient manner. These study data suggest that this alignment should include hospitals, nursing homes, and home health care agencies to ensure that discharge delays are minimized and unnecessary societal costs are avoided. Future research will need to confirm and build upon these findings of nonmedical reasons for excessive LOS to further inform the process of implementation of health care reform measures. Recent plans to cut Medicaid funding to nursing homes may further limit bed availability, increasing the risk of prolonged LOS and related costs to the health care system. This potential concern highlights the importance of care coordination and communication between inpatient and outpatient care providers to proactively address nursing home placement needs before hospitalization occurs, and/or to identify alternative safe discharge plans if a previously homebound patient is hospitalized.

There are several limitations to this descriptive study. Admissions included in this analysis were only captured for those admitted to The Mount Sinai Hospital. While MSVD providers report that more than 90% of hospitalizations for MSVD patients occur at The Mount Sinai Hospital, patients may also be admitted to one of many New York City metropolitan area hospitals closer to the patient's residence. It is possible that additional factors contributing to high LOS might be revealed if these admissions were included in the analysis. The urban homebound population served by MSVD may have more access to supplementary home care services (e.g. home attendants, meal services) than populations in more rural and less service‐intensive areas. Thus, it may be difficult to generalize these findings to programs serving less urban constituencies or with more restrictive policies regarding home care services. Additionally, as New York registers one of the highest nursing facility occupancy rates (in 2008, 92.2% versus the national average of 82.9%), patients in other markets may face a shorter wait time for a bed, decreasing the number of nonmedical days attributable to nursing home bed supply.17 The small total number of long‐stay patients also prevented statistical analysis comparing those patients with the rest of the sample. This pilot study may inform the design of future studies that may be able to include multiple HBPC programs or study homebound patients over a longer period to increase sample size.

Identifying the significant contribution of nonmedical days to patient stay is an important initial step to avoiding costly and medically unnecessary days for the patient and the hospital. As has been demonstrated in other interdisciplinary efforts, increased collaboration among physicians, social workers, discharge planners, and other disciplines may help address current gaps in patient care with regard to LOS.20, 21 Future studies should determine which homebound patients are at highest risk for prolonged hospitalization due to nonmedical factors to help design focused strategies and interventions for this vulnerable population.

In recent years, much attention has been paid to concerns regarding length of stay (LOS) and safety of hospital discharges.13 Yet studies conducted in a variety of populations suggest that long stays do not wholly reflect acute medical necessity, but may also be driven by nonmedical factors.46 In a study of frail elderly patients, nonmedical factors accounted for over half of patients' hospital stay days.5 Nonmedical factors may include the availability of community and outpatient resources, inadequate patient social support, disagreement with family and/or patient decision‐making, and post‐hospital placement and care needs.7

Homebound patients are at particular risk for long stays because they are typically frail, elderly, and medically complex.8 The United States homebound population numbers at least 2 million and is expected to increase to at least 3 million by the year 2020.9 This group is medically underserved, often only receiving care for medical emergencies, and represents a costly group of health care beneficiaries.10 Although homebound primary care (HBPC) programs are structured to provide coordinated medical and supportive care in the home, the clinical complexity of patients often requires hospitalization during times of acute illness.11

Navigating the discharge process for homebound patients can prove time‐consuming for inpatient physicians whose clinical obligations encompass ensuring safe care transitions between hospital and home.12 The lack of literature on the discharge needs of the homebound population provides little guidance for physicians and health systems seeking to safely transition homebound patients from the hospital to the home in a timely and efficient manner. We performed a pilot study of an urban homebound population cared for by a single academic homebound program to identify and describe nonmedical factors associated with prolonged hospitalization.

METHODS

This retrospective descriptive study included homebound patients cared for by The Mount Sinai Visiting Doctors Program (MSVD), a primary care program affiliated with The Mount Sinai Hospital. MSVD is the largest academic homebound primary care program in the United States. The structure and patient population of MSVD patients have been described previously.13 Briefly, the program employs 8 physicians, 2 nurse practitioners, and support personnel including 2 registered nurses, 4 social workers, and 4 clerical staff members to serve over 1000 homebound patients annually. To be enrolled in the program, patients must meet the Medicare definition of homebound (ie, they must be able to leave home only with great difficulty and for infrequent or short absences). Patients are referred from a variety of sources including emergency rooms, inpatient units, and local nursing and social service agencies. Physicians visit patients on average once every 2 months, but can make more frequent home visits when a clinical need arises. Thirty‐six percent of patients in the program are hospitalized at least once per year while they are under the MSVD's care.14 Patients in the MSVD are referred by their primary care physician to outpatient social work as needed for finite interventions, though not for ongoing case management; approximately one‐third of all MSVD patients have been seen by an MSVD social worker over a 1‐year period.14

All patients enrolled in MSVD discharged from The Mount Sinai Hospital in New York from January 1, 2007, to December 31, 2007, were evaluated for inclusion in this study. The MSVD clinical database was cross‐referenced with The Mount Sinai Hospital data system to maximize reliability of recorded admission and discharge dates. Discrepancies in dates were investigated and corrected by the authors. As the study focused on admissions rather than individual patients, repeat hospitalizations and factors contributing to long stays were included and considered separately. In the event of repeat hospitalizations, factors contributing to LOS were also assessed separately.

Using the University HealthSystems Consortium (UHC) Database, the selected discharges were analyzed for LOS data. The UHC Database contains data from The Mount Sinai Hospital and 106 other academic medical centers and 233 other affiliated hospitals, representing approximately 90% of the United States' nonprofit academic medical centers. UHC members submit clinical, financial, and administrative information for the purpose of facilitating comparative data analysis among institutions.15 The model assigns an expected LOS for each patient based on a 4‐step risk adjustment methodology that adjusts for variations in patient characteristics. The regression models consider a range of independent variables including patient age, sex, race, socioeconomic status, admission source, and comorbid conditions. The expected LOS value is used to produce an LOS ratio, which is the ratio of a patient's observed LOS to expected LOS. We compared LOS ratios for all patient discharges during the study period and determined the mean LOS ratio for the group. Long‐stay patients were defined using the UHC definition of an LOS ratio greater than 2 standard deviations above the mean. These patients were selected for analysis to examine factors contributing to their long stays.

The primary author conducted a chart review for the long‐stay patients. The date of medical readiness for discharge was defined as the date when no acute hospital care needs or pending procedures (eg, intravenous medication, transfusion, invasive and noninvasive testing, surgery) were documented by the attending physician, house officer, nurse practitioner, or physician assistant in the chart. Patients with a discharge immediately following determination of medical readiness were characterized as having a medical stay (eg, patient admitted for pneumonia and discharged home after 3 days once fever, leukocytosis, and symptoms improved with a prescription for remaining antibiotic course). Patients who were classified as medically ready for discharge yet remained additional days in the hospital were categorized as having a nonmedical component to their hospitalization and comprised the nonmedical stay group (eg, patient admitted for pneumonia with improved vital signs and symptoms after 3 days, but discharged home after 10 days awaiting approval of increased home health aide hours). The primary author reviewed cases (<5%) with coauthors when categorization of patient data was unclear from physician documentation in the patient's chart. Similar methodology and terminology have been used in previous studies examining the contribution of nonmedical factors to LOS.5, 7

While literature in other fields, such as social work, often characterize similar factors as social or social care factors, we use the term nonmedical to draw the distinction between factors that acutely reflect a patient's state of health and necessitate days spent in the hospital (eg, surgery, infection), in contrast to factors that are not direct contributors to the patient's current medical status (eg, post‐hospital placement). Factors contributing to nonmedical days were determined based on previous studies and categorized as follows: nursing facility bed availability, nursing facility rejection of the patient, complications with insurance coverage, lack of patient/family agreement with discharge plan, home care service delays, and other. The category other was included to identify and explore any unexpected or unique reasons for prolongation of hospitalization in this population. When multiple nonmedical factors were identified for a given hospitalization, all contributing factors were recorded.

Demographic, clinical, and discharge characteristics were extracted from the MSVD clinical database to qualitatively describe and compare the long‐stay hospitalizations to the remainder of the sample.

RESULTS

There were a total of 479 discharges of 267 unique MSVD patients from The Mount Sinai Hospital occurring from January 1, 2007, to December 31, 2007. During this 12‐month period, the average observed LOS for all admissions was 7.84 days, with a mean UHC LOS ratio of 1.23 (SD = 3.43). Seventeen admissions were identified as long‐stays, representing 3.5% of discharges. The 17 admissions represent 17 unique patients.

As shown in Table 1, the long‐stay group (n = 17) was slightly younger, more likely to be male, and had less dementia than the nonlong‐stay group (n = 462). There was a marked difference in the location of patient discharge; long‐stay patients were more than twice as likely to be discharged to a facility and less likely to be discharged home. There were no in‐hospital deaths in the long‐stay patient group during the time period studied.

Of the 17 long‐stay patients, 8 (47%) remained in the hospital past the date they were determined to be medically ready for discharge and were defined as having a nonmedical component for the extension of their hospitalizations. The number of nonmedical days ranged from 6 to 34 days (mean, 17 days). Out of 428 total long‐stay patient days, 136 were nonmedical. This represented 31.8% of all long‐stay patient days, and 53% of the nonmedical group's total hospital days. The mean LOS ratio for the nonmedical cases was 6.04 (Table 2).

Nine patients were defined as medical stay cases (ie, no nonmedical component contributing to the long hospitalization). The mean observed LOS was 19.2 days, and the mean LOS ratio for this group was 5.07 (Table 2). There were no significant differences between primary diagnosis‐related groups (DRGs) seen in the medical and nonmedical stay groups.

The most common reason for a nonmedical stay was nursing facility placement delays (Table 3), specifically related to lack of bed availability and facility rejection of the patient leading to prolonged time waiting for long‐term placement (n = 6). Other nonmedical factors contributing to LOS were lack of patient and/or family agreement with discharge plans (eg, disagreement among family members regarding caregiving responsibilities, goals of care, or patient refusal to be discharged on a particular day) (n = 4); complications with insurance coverage for facility placement or for home care (n = 3); and home care service delays, such as patient need for increased home care hours after discharge (n = 2). Of note, 5 of the 8 nonmedical stay cases had multiple factors contributing to patients' long stays. All delays were assigned to one of the a priori defined categories. There were no other or unexpected reasons identified.

Of the nonmedical cases, all but 1 patient had been seen by an MSVD social worker prior to hospital admission, though the social work referral may have been years prior to or unrelated to the current admission.

DISCUSSION

Almost half of long‐stay patients identified in this homebound population remained hospitalized in an urban academic medical center due to at least one, and often multiple, nonmedical factors. Nonmedical factors identified in this group are similar to those described in previous studies, particularly family and patient decision‐making and post‐hospital placement and care needs.5, 7 Although this pilot study was limited to a single‐site population, it is to our knowledge the first study to describe these factors in a homebound population, and may be able to guide future research and discussion on this topic.

This study used a risk‐adjusted LOS measure to determine long stay cases. Using the UHC Database allowed for a more accurate understanding of the contribution of nonmedical factors to LOS by accounting for hospitalizations that were numerically lengthy but medically appropriate for their respective DRG. The use of the LOS ratio also allows for standardized application of these data across academic health centers. In our sample, 50% of the patients classified as LOS outliers by the UHC Database (cases with LOS in the top percentile for their respective DRG) had nonmedical stays. Conventional strategies often dismiss outliers in analyses of patient LOS data. However, in doing so there is a missed opportunity to identify underlying reasons for their disproportionately long hospitalizations that may also be impacting the broader set of patients with similar nonmedical factors affecting LOS.

The 8 nonmedical stay patients spent a combined 136 days longer in the hospital than medically necessary due to a variety of nonmedical factors, and represented over half of the nonmedical stay group's total hospital days. Using a conservative estimate for cost per hospital day of $1770,16 the nonmedical days cost the hospital almost a quarter of a million dollars ($240,720). Because this figure only accounts for long‐stay patients, the actual costs attributable to nonmedical days for the homebound population in general may be higher.

The longest patient stays, whether attributable to medical or nonmedical factors, were more likely to result in discharge to a facility than the rest of the sample hospitalizations. Facility placement was the most common nonmedical factor contributing to long stays in this sample. In contrast, home carerelated factors contributed the least to nonmedical days. This finding highlights the need for hospital‐based physicians and other inpatient staff members to be aware that despite patient enrollment in an HBPC, the possibility for homebound patients to be discharged to a nursing facility remains significant. A decreasing number of skilled nursing beds across the United States may magnify this factor in long‐stay cases.17 Increased awareness of this possibility among inpatient staff can allow the team to address facility placement considerations early in the hospital stay, potentially decreasing nonmedical days.18

Seven of the 8 nonmedical stay patients had been referred to and seen by MSVD social workers before hospitalization, a high percentage relative to the general MSVD population, of which fewer than half are seen by a social worker during their enrollment in MSVD. This finding may suggest that this group of patients already exhibited difficult social circumstances before their hospital admission, yet the current referral‐based social work model at MSVD did not mitigate their high LOS. This finding further suggests that patient enrollment in an HBPC does not mitigate the risk of high LOS and prolonged nonmedical stays, and that involvement of inpatient practitioners remains a critical part of advanced discharge planning.

This pilot study found that 32% of all long‐stay hospital days were due to nonmedical factors, suggesting that these factors play a greater role in the homebound population than for general medical patients. A recent study at an academic medical center examined 3574 patient‐days on a general medicine service, and noted that 11% of all days were felt to be medically unnecessary by the treating hospitalists.19 Hospitalists are well situated to participate in and lead improvement efforts given their expertise in managing complex dispositions and advancing collaborative strategies for care of patients with high overall acuity.20 These efforts will be needed to target those patients at highest risk for prolonged LOS with the greatest social care needs. Because this study did not pilot strategies to reduce LOS, we cannot offer evidence‐based suggestions for an enhanced multidisciplinary approach or other avenues for improvement. However, we believe that the study findings provide the basis for future research to test strategies to reduce excess LOS by focusing on nonmedical factors and a multidisciplinary approach. This will become especially relevant as health care systems bear increasing financial responsibility for inefficient and/or unnecessary hospitalizations and readmissions.

The involvement of social work before hospitalization for most of the homebound population with prolonged hospitalization suggests a need for greater team‐based efforts across venues. Though hospital interdisciplinary rounds aim to increase collaboration and reduce LOS, costs, and readmissions, these rounds do not typically include outpatient care providers.21 Improved communication and collaboration between social work with both inpatient and outpatient care teams to address nonmedical issues contributing to long stays are likely to improve care and transitions, though rigorous studies examining specific communication models across venues are lacking. This study found that delay in nursing facility placement was the most common reason for prolonged hospitalization for long‐stay cases. This finding emphasizes the need for communication between inpatient and outpatient staff to convey prior conversations or preparations for placement, identify patients who need post‐discharge facility placement early in hospitalization, and prompt timely discussions with patients and families.

The finding that prolonged hospitalization for the homebound population was due to nonmedical factors for almost one‐half of patients with long hospital stays has important implications for policymakers and other key stakeholders. For example, accountable care organizations are being developed to align members of the health care sector to provide higher quality care in a more efficient manner. These study data suggest that this alignment should include hospitals, nursing homes, and home health care agencies to ensure that discharge delays are minimized and unnecessary societal costs are avoided. Future research will need to confirm and build upon these findings of nonmedical reasons for excessive LOS to further inform the process of implementation of health care reform measures. Recent plans to cut Medicaid funding to nursing homes may further limit bed availability, increasing the risk of prolonged LOS and related costs to the health care system. This potential concern highlights the importance of care coordination and communication between inpatient and outpatient care providers to proactively address nursing home placement needs before hospitalization occurs, and/or to identify alternative safe discharge plans if a previously homebound patient is hospitalized.

There are several limitations to this descriptive study. Admissions included in this analysis were only captured for those admitted to The Mount Sinai Hospital. While MSVD providers report that more than 90% of hospitalizations for MSVD patients occur at The Mount Sinai Hospital, patients may also be admitted to one of many New York City metropolitan area hospitals closer to the patient's residence. It is possible that additional factors contributing to high LOS might be revealed if these admissions were included in the analysis. The urban homebound population served by MSVD may have more access to supplementary home care services (e.g. home attendants, meal services) than populations in more rural and less service‐intensive areas. Thus, it may be difficult to generalize these findings to programs serving less urban constituencies or with more restrictive policies regarding home care services. Additionally, as New York registers one of the highest nursing facility occupancy rates (in 2008, 92.2% versus the national average of 82.9%), patients in other markets may face a shorter wait time for a bed, decreasing the number of nonmedical days attributable to nursing home bed supply.17 The small total number of long‐stay patients also prevented statistical analysis comparing those patients with the rest of the sample. This pilot study may inform the design of future studies that may be able to include multiple HBPC programs or study homebound patients over a longer period to increase sample size.

Identifying the significant contribution of nonmedical days to patient stay is an important initial step to avoiding costly and medically unnecessary days for the patient and the hospital. As has been demonstrated in other interdisciplinary efforts, increased collaboration among physicians, social workers, discharge planners, and other disciplines may help address current gaps in patient care with regard to LOS.20, 21 Future studies should determine which homebound patients are at highest risk for prolonged hospitalization due to nonmedical factors to help design focused strategies and interventions for this vulnerable population.

Physician assistants (PA) have rapidly become an integral component in the United States health care delivery system, including in the field of Hospital Medicine, the fastest growing medical field in the United States.1, 2 Since its induction in 1997, hospitalist providers in North America have increased by 30‐fold.3 Correlating with this, the number of PAs practicing in the field of hospital medicine has also increased greatly in recent years. According to the American Academy of Physician Assistants (AAPA) census reports, Hospital Medicine first appeared as one of the specialty choices in the 2006 census (response rate, 33% of all individuals eligible to practice as PAs) when it was selected as the primary specialty by 239 PAs (1.1% of respondents). In the 2008 report (response rate, 35%), the number grew to 421 (1.7%) PAs.2

PA training programs emphasize primary care and offer limited exposure to inpatient medicine. After PA students complete their first 12 months of training in didactic coursework that teach the basic sciences, they typically spend the next year on clinical rotations, largely rooted in outpatient care.2, 4 Upon graduation, PAs do not have to pursue postgraduate training before beginning to practice in their preferred specialty areas. Thus, a majority of PAs going into specialty areas are trained on the job. This is not an exception in the field of hospital medicine.

In recent years, despite an increase in the number of PAs in Hospital Medicine, some medical centers have chosen to phase out the use of midlevel hospitalist providers (including PAs) with the purposeful decision to not hire new midlevel providers.5 The rationale for this strategy is that there is thought to be a steep learning curve that requires much time to overcome before these providers feel comfortable across the breadth of clinical cases. Before they become experienced and confident in caring for a highly complex heterogeneous patient population, they cannot operate autonomously and are not a cost‐effective alternative to physicians. The complexities associated with practicing in this field were clarified in 2006 when the Society of Hospital Medicine identified 51 core competencies in hospital medicine.3, 6 Some hospitalist programs are willing to provide their PAs with on‐the‐job training, but many programs do not have the educational expertise or the resources to make this happen. Structured and focused postgraduate training in hospital medicine seems like a reasonable solution to prepare newly graduating PAs that are interested in pursuing hospitalist careers, but such opportunities are very limited.7

To date, there is no available information about the learning needs of PAs working in hospital medicine settings. We hypothesized that understanding the learning needs of PA hospitalists would inform the development of more effective and efficient training programs. We studied PAs with experience working in hospital medicine to (1) identify self‐perceived gaps in their skills and knowledge upon starting their hospitalist careers and (2) understand their views about optimal training for careers in hospital medicine.

METHODS

RESULTS

Sixty‐nine PAs responded (response rate, 67%). Table 1 provides demographic characteristics of the respondents. The majority of respondents were 2635 years old and had worked as hospitalists for a mean of 4.3 years.

DISCUSSION

Our survey addresses a wide range of topics related to PA hospitalists' learning needs including their experience with the Core Competencies in Hospital Medicine and their views on the benefits of PA training following graduation. Although self‐efficacy is not assessed, our study revealed that PAs who are choosing hospitalist careers have limited prior clinical experience treating many medical conditions that are managed in inpatient settings, such as sepsis syndrome. This inexperience with commonly seen clinical conditions, such as sepsis, wherein following guidelines can both reduce costs and improve outcomes, is problematic. More experience and training with such conditions would almost certainly reduce variability, improve skills, and augment confidence. These observed variations in experience in caring for conditions that often prompt admission to the hospital among PAs starting their hospitalist careers emphasizes the need to be learner‐centered when training PAs, so as to provide tailored guidance and oversight.

Only a few other empiric research articles have focused on PA hospitalists. One article described a postgraduate training program for PAs in hospital medicine that was launched in 2008. The curriculum was developed based on the Core Competencies in Hospital Medicine, and the authors explained that after 12 months of training, their first graduate functioned at the level of a PA with 4 years of experience under her belt.7 Several articles describe experiences using midlevel providers (including PAs) in general surgery, primary care medicine, cardiology, emergency medicine, critical care, pediatrics, and hospital medicine settings.5, 820 Many of these articles reported favorable results showing that using midlevel providers was either superior or just as effective in terms of cost and quality measures to physician‐only models. Many of these papers alluded to the ways in which PAs have enabled graduate medical education training programs to comply with residents' duty‐hour restrictions. A recent analysis that compared outcomes related to inpatient care provided by a hospitalist‐PA model versus a traditional resident‐based model revealed a slightly longer length of stay on the PA team but similar charges, readmission rates, and mortality.19 Yet another paper revealed that patients admitted to a residents' service, compared with the nonteaching hospitalist service that uses PAs and nurse practitioners, were different, having higher comorbidity burdens and higher acuity diagnoses.20 The authors suggested that this variance might be explained by the difference in their training, abilities, and goals of the groups. There was no research article that sought to capture the perspectives of practicing hospitalist PAs.

Our study revealed that although half of respondents became hospitalists immediately after graduating from PA school, a majority agreed that additional clinical training in inpatient settings would have been welcomed and helpful. This study's results reveal that although there is a fair amount of perceived interest in postgraduate training programs in hospital medicine, there are very few training opportunities for PAs in hospital medicine.7, 21 The American Academy of Physician Assistants, the Society of Hospital Medicine, and the American Academy of Nurse Practitioners cosponsor Adult Hospital Medicine Boot Camp for PAs and nurse practitioners annually to facilitate knowledge acquisition, but this course is truly an orientation rather than a comprehensive training program.22 Our findings suggest that more rigorous and thorough training in hospital medicine would be valued and appreciated by PA hospitalists.

Several limitations of this study should be considered. First, our survey respondents may not represent the entire spectrum of practicing PA hospitalists. However, the demographic data of 421 PAs who indicated their specialty as hospital medicine in the 2008 National Physician Assistants Census Report were not dissimilar from our informants; 65% were women, and their mean number of years in hospital medicine was 3.9 years.2 Second, our study sample was small. It was difficult to identify a national sample of hospitalist PAs, and we had to resort to a creative use of social media to find a national sample. Third, the study relied exclusively on self‐report, and since we asked about their perceived learning needs when they started working as hospitalists, recall bias cannot be excluded. However, the questions addressing attitudes and beliefs can only be ascertained from the informants themselves. That said, the input from hospitalist physicians about training needs for the PAs who they are supervising would have strengthened the reliability of the data, but this was not possible given the sampling strategy that we elected to use. Finally, our survey instrument was developed based on the Core Competencies in Hospital Medicine, which is a blueprint to develop standardized curricula for teaching hospital medicine in medical school, postgraduate training programs (ie, residency, fellowship), and continuing medical education programs. It is not clear whether the same competencies should be expected of PA hospitalists who may have different job descriptions from physician hospitalists.

In conclusion, we present the first national data on self‐perceived learning needs of PAs working in hospital medicine settings. This study collates the perceptions of PAs working in hospital medicine and highlights the fact that training in PA school does not adequately prepare them to care for hospitalized patients. Hospitalist groups may use this study's findings to coach and instruct newly hired or inexperienced hospitalist PAs, particularly until postgraduate training opportunities become more prevalent. PA schools may consider the results of this study for modifying their curricula in hopes of emphasizing the clinical content that may be most relevant for a proportion of their graduates.

Physician assistants (PA) have rapidly become an integral component in the United States health care delivery system, including in the field of Hospital Medicine, the fastest growing medical field in the United States.1, 2 Since its induction in 1997, hospitalist providers in North America have increased by 30‐fold.3 Correlating with this, the number of PAs practicing in the field of hospital medicine has also increased greatly in recent years. According to the American Academy of Physician Assistants (AAPA) census reports, Hospital Medicine first appeared as one of the specialty choices in the 2006 census (response rate, 33% of all individuals eligible to practice as PAs) when it was selected as the primary specialty by 239 PAs (1.1% of respondents). In the 2008 report (response rate, 35%), the number grew to 421 (1.7%) PAs.2

PA training programs emphasize primary care and offer limited exposure to inpatient medicine. After PA students complete their first 12 months of training in didactic coursework that teach the basic sciences, they typically spend the next year on clinical rotations, largely rooted in outpatient care.2, 4 Upon graduation, PAs do not have to pursue postgraduate training before beginning to practice in their preferred specialty areas. Thus, a majority of PAs going into specialty areas are trained on the job. This is not an exception in the field of hospital medicine.

In recent years, despite an increase in the number of PAs in Hospital Medicine, some medical centers have chosen to phase out the use of midlevel hospitalist providers (including PAs) with the purposeful decision to not hire new midlevel providers.5 The rationale for this strategy is that there is thought to be a steep learning curve that requires much time to overcome before these providers feel comfortable across the breadth of clinical cases. Before they become experienced and confident in caring for a highly complex heterogeneous patient population, they cannot operate autonomously and are not a cost‐effective alternative to physicians. The complexities associated with practicing in this field were clarified in 2006 when the Society of Hospital Medicine identified 51 core competencies in hospital medicine.3, 6 Some hospitalist programs are willing to provide their PAs with on‐the‐job training, but many programs do not have the educational expertise or the resources to make this happen. Structured and focused postgraduate training in hospital medicine seems like a reasonable solution to prepare newly graduating PAs that are interested in pursuing hospitalist careers, but such opportunities are very limited.7

To date, there is no available information about the learning needs of PAs working in hospital medicine settings. We hypothesized that understanding the learning needs of PA hospitalists would inform the development of more effective and efficient training programs. We studied PAs with experience working in hospital medicine to (1) identify self‐perceived gaps in their skills and knowledge upon starting their hospitalist careers and (2) understand their views about optimal training for careers in hospital medicine.

METHODS

RESULTS

Sixty‐nine PAs responded (response rate, 67%). Table 1 provides demographic characteristics of the respondents. The majority of respondents were 2635 years old and had worked as hospitalists for a mean of 4.3 years.

DISCUSSION

Our survey addresses a wide range of topics related to PA hospitalists' learning needs including their experience with the Core Competencies in Hospital Medicine and their views on the benefits of PA training following graduation. Although self‐efficacy is not assessed, our study revealed that PAs who are choosing hospitalist careers have limited prior clinical experience treating many medical conditions that are managed in inpatient settings, such as sepsis syndrome. This inexperience with commonly seen clinical conditions, such as sepsis, wherein following guidelines can both reduce costs and improve outcomes, is problematic. More experience and training with such conditions would almost certainly reduce variability, improve skills, and augment confidence. These observed variations in experience in caring for conditions that often prompt admission to the hospital among PAs starting their hospitalist careers emphasizes the need to be learner‐centered when training PAs, so as to provide tailored guidance and oversight.

Only a few other empiric research articles have focused on PA hospitalists. One article described a postgraduate training program for PAs in hospital medicine that was launched in 2008. The curriculum was developed based on the Core Competencies in Hospital Medicine, and the authors explained that after 12 months of training, their first graduate functioned at the level of a PA with 4 years of experience under her belt.7 Several articles describe experiences using midlevel providers (including PAs) in general surgery, primary care medicine, cardiology, emergency medicine, critical care, pediatrics, and hospital medicine settings.5, 820 Many of these articles reported favorable results showing that using midlevel providers was either superior or just as effective in terms of cost and quality measures to physician‐only models. Many of these papers alluded to the ways in which PAs have enabled graduate medical education training programs to comply with residents' duty‐hour restrictions. A recent analysis that compared outcomes related to inpatient care provided by a hospitalist‐PA model versus a traditional resident‐based model revealed a slightly longer length of stay on the PA team but similar charges, readmission rates, and mortality.19 Yet another paper revealed that patients admitted to a residents' service, compared with the nonteaching hospitalist service that uses PAs and nurse practitioners, were different, having higher comorbidity burdens and higher acuity diagnoses.20 The authors suggested that this variance might be explained by the difference in their training, abilities, and goals of the groups. There was no research article that sought to capture the perspectives of practicing hospitalist PAs.

Our study revealed that although half of respondents became hospitalists immediately after graduating from PA school, a majority agreed that additional clinical training in inpatient settings would have been welcomed and helpful. This study's results reveal that although there is a fair amount of perceived interest in postgraduate training programs in hospital medicine, there are very few training opportunities for PAs in hospital medicine.7, 21 The American Academy of Physician Assistants, the Society of Hospital Medicine, and the American Academy of Nurse Practitioners cosponsor Adult Hospital Medicine Boot Camp for PAs and nurse practitioners annually to facilitate knowledge acquisition, but this course is truly an orientation rather than a comprehensive training program.22 Our findings suggest that more rigorous and thorough training in hospital medicine would be valued and appreciated by PA hospitalists.

Several limitations of this study should be considered. First, our survey respondents may not represent the entire spectrum of practicing PA hospitalists. However, the demographic data of 421 PAs who indicated their specialty as hospital medicine in the 2008 National Physician Assistants Census Report were not dissimilar from our informants; 65% were women, and their mean number of years in hospital medicine was 3.9 years.2 Second, our study sample was small. It was difficult to identify a national sample of hospitalist PAs, and we had to resort to a creative use of social media to find a national sample. Third, the study relied exclusively on self‐report, and since we asked about their perceived learning needs when they started working as hospitalists, recall bias cannot be excluded. However, the questions addressing attitudes and beliefs can only be ascertained from the informants themselves. That said, the input from hospitalist physicians about training needs for the PAs who they are supervising would have strengthened the reliability of the data, but this was not possible given the sampling strategy that we elected to use. Finally, our survey instrument was developed based on the Core Competencies in Hospital Medicine, which is a blueprint to develop standardized curricula for teaching hospital medicine in medical school, postgraduate training programs (ie, residency, fellowship), and continuing medical education programs. It is not clear whether the same competencies should be expected of PA hospitalists who may have different job descriptions from physician hospitalists.

In conclusion, we present the first national data on self‐perceived learning needs of PAs working in hospital medicine settings. This study collates the perceptions of PAs working in hospital medicine and highlights the fact that training in PA school does not adequately prepare them to care for hospitalized patients. Hospitalist groups may use this study's findings to coach and instruct newly hired or inexperienced hospitalist PAs, particularly until postgraduate training opportunities become more prevalent. PA schools may consider the results of this study for modifying their curricula in hopes of emphasizing the clinical content that may be most relevant for a proportion of their graduates.

Physician assistants (PA) have rapidly become an integral component in the United States health care delivery system, including in the field of Hospital Medicine, the fastest growing medical field in the United States.1, 2 Since its induction in 1997, hospitalist providers in North America have increased by 30‐fold.3 Correlating with this, the number of PAs practicing in the field of hospital medicine has also increased greatly in recent years. According to the American Academy of Physician Assistants (AAPA) census reports, Hospital Medicine first appeared as one of the specialty choices in the 2006 census (response rate, 33% of all individuals eligible to practice as PAs) when it was selected as the primary specialty by 239 PAs (1.1% of respondents). In the 2008 report (response rate, 35%), the number grew to 421 (1.7%) PAs.2

PA training programs emphasize primary care and offer limited exposure to inpatient medicine. After PA students complete their first 12 months of training in didactic coursework that teach the basic sciences, they typically spend the next year on clinical rotations, largely rooted in outpatient care.2, 4 Upon graduation, PAs do not have to pursue postgraduate training before beginning to practice in their preferred specialty areas. Thus, a majority of PAs going into specialty areas are trained on the job. This is not an exception in the field of hospital medicine.

In recent years, despite an increase in the number of PAs in Hospital Medicine, some medical centers have chosen to phase out the use of midlevel hospitalist providers (including PAs) with the purposeful decision to not hire new midlevel providers.5 The rationale for this strategy is that there is thought to be a steep learning curve that requires much time to overcome before these providers feel comfortable across the breadth of clinical cases. Before they become experienced and confident in caring for a highly complex heterogeneous patient population, they cannot operate autonomously and are not a cost‐effective alternative to physicians. The complexities associated with practicing in this field were clarified in 2006 when the Society of Hospital Medicine identified 51 core competencies in hospital medicine.3, 6 Some hospitalist programs are willing to provide their PAs with on‐the‐job training, but many programs do not have the educational expertise or the resources to make this happen. Structured and focused postgraduate training in hospital medicine seems like a reasonable solution to prepare newly graduating PAs that are interested in pursuing hospitalist careers, but such opportunities are very limited.7

To date, there is no available information about the learning needs of PAs working in hospital medicine settings. We hypothesized that understanding the learning needs of PA hospitalists would inform the development of more effective and efficient training programs. We studied PAs with experience working in hospital medicine to (1) identify self‐perceived gaps in their skills and knowledge upon starting their hospitalist careers and (2) understand their views about optimal training for careers in hospital medicine.

METHODS

RESULTS

Sixty‐nine PAs responded (response rate, 67%). Table 1 provides demographic characteristics of the respondents. The majority of respondents were 2635 years old and had worked as hospitalists for a mean of 4.3 years.

DISCUSSION

Our survey addresses a wide range of topics related to PA hospitalists' learning needs including their experience with the Core Competencies in Hospital Medicine and their views on the benefits of PA training following graduation. Although self‐efficacy is not assessed, our study revealed that PAs who are choosing hospitalist careers have limited prior clinical experience treating many medical conditions that are managed in inpatient settings, such as sepsis syndrome. This inexperience with commonly seen clinical conditions, such as sepsis, wherein following guidelines can both reduce costs and improve outcomes, is problematic. More experience and training with such conditions would almost certainly reduce variability, improve skills, and augment confidence. These observed variations in experience in caring for conditions that often prompt admission to the hospital among PAs starting their hospitalist careers emphasizes the need to be learner‐centered when training PAs, so as to provide tailored guidance and oversight.

Only a few other empiric research articles have focused on PA hospitalists. One article described a postgraduate training program for PAs in hospital medicine that was launched in 2008. The curriculum was developed based on the Core Competencies in Hospital Medicine, and the authors explained that after 12 months of training, their first graduate functioned at the level of a PA with 4 years of experience under her belt.7 Several articles describe experiences using midlevel providers (including PAs) in general surgery, primary care medicine, cardiology, emergency medicine, critical care, pediatrics, and hospital medicine settings.5, 820 Many of these articles reported favorable results showing that using midlevel providers was either superior or just as effective in terms of cost and quality measures to physician‐only models. Many of these papers alluded to the ways in which PAs have enabled graduate medical education training programs to comply with residents' duty‐hour restrictions. A recent analysis that compared outcomes related to inpatient care provided by a hospitalist‐PA model versus a traditional resident‐based model revealed a slightly longer length of stay on the PA team but similar charges, readmission rates, and mortality.19 Yet another paper revealed that patients admitted to a residents' service, compared with the nonteaching hospitalist service that uses PAs and nurse practitioners, were different, having higher comorbidity burdens and higher acuity diagnoses.20 The authors suggested that this variance might be explained by the difference in their training, abilities, and goals of the groups. There was no research article that sought to capture the perspectives of practicing hospitalist PAs.

Our study revealed that although half of respondents became hospitalists immediately after graduating from PA school, a majority agreed that additional clinical training in inpatient settings would have been welcomed and helpful. This study's results reveal that although there is a fair amount of perceived interest in postgraduate training programs in hospital medicine, there are very few training opportunities for PAs in hospital medicine.7, 21 The American Academy of Physician Assistants, the Society of Hospital Medicine, and the American Academy of Nurse Practitioners cosponsor Adult Hospital Medicine Boot Camp for PAs and nurse practitioners annually to facilitate knowledge acquisition, but this course is truly an orientation rather than a comprehensive training program.22 Our findings suggest that more rigorous and thorough training in hospital medicine would be valued and appreciated by PA hospitalists.

Several limitations of this study should be considered. First, our survey respondents may not represent the entire spectrum of practicing PA hospitalists. However, the demographic data of 421 PAs who indicated their specialty as hospital medicine in the 2008 National Physician Assistants Census Report were not dissimilar from our informants; 65% were women, and their mean number of years in hospital medicine was 3.9 years.2 Second, our study sample was small. It was difficult to identify a national sample of hospitalist PAs, and we had to resort to a creative use of social media to find a national sample. Third, the study relied exclusively on self‐report, and since we asked about their perceived learning needs when they started working as hospitalists, recall bias cannot be excluded. However, the questions addressing attitudes and beliefs can only be ascertained from the informants themselves. That said, the input from hospitalist physicians about training needs for the PAs who they are supervising would have strengthened the reliability of the data, but this was not possible given the sampling strategy that we elected to use. Finally, our survey instrument was developed based on the Core Competencies in Hospital Medicine, which is a blueprint to develop standardized curricula for teaching hospital medicine in medical school, postgraduate training programs (ie, residency, fellowship), and continuing medical education programs. It is not clear whether the same competencies should be expected of PA hospitalists who may have different job descriptions from physician hospitalists.

In conclusion, we present the first national data on self‐perceived learning needs of PAs working in hospital medicine settings. This study collates the perceptions of PAs working in hospital medicine and highlights the fact that training in PA school does not adequately prepare them to care for hospitalized patients. Hospitalist groups may use this study's findings to coach and instruct newly hired or inexperienced hospitalist PAs, particularly until postgraduate training opportunities become more prevalent. PA schools may consider the results of this study for modifying their curricula in hopes of emphasizing the clinical content that may be most relevant for a proportion of their graduates.

Vital signs constitute a fundamental component of the physical examination and serve key diagnostic and monitoring purposes. The brain is as vital to life as the cardiovascular, respiratory, and immune/thermoregulatory systems, yet currently no vital sign exists that would allow rapid, reliable, and easily reproducible assessment of cognition.1 As a result, acute mental status changes frequently go undetected and untreated.24 Delirium is defined as an acute change in attention with fluctuations in cognition, thought, and/or consciousness throughout the course of the day.5 Because delirium in older patients is common and is associated with increased morbidity, mortality, functional decline, and costs,69 development and validation of a rapid, objective screening assessment could be used by nursing staff to identify patients at high risk for delirium.

Current recommendations for inpatient delirium monitoring usually involve daily cognitive screening with a standardized screening instrument.6 Because this process is often time‐consuming (8‐12 minutes), most patients do not undergo routine screening. To facilitate clinical implementation, we focused on developing a brief (<30‐second) inpatient screening measure of a feature of mental status that could be administered serially. The purpose of this study was to (1) develop a brief screening tool for a core feature of mental status and (2) validate this screening tool for delirium in an older inpatient population.

METHODS

Consensus Panel

In June 2009, the Veterans Administration sponsored an interdisciplinary conference that solicited input on identifying the most targetable components of delirium and discussing potential clinical instruments. Following this, a consensus panel of 8 representatives from medicine, geriatrics, nursing, psychiatry, and psychology used a modified Delphi method to target characteristic features of delirium and identify instruments that could best capture mental status change. While inattention was agreed upon as the core cognitive feature of delirium, the group came to consensus that capturing the acute onset and fluctuating nature of delirium was better suited as a vital sign. To meet these criteria, the group modified the Richmond Agitation Sedation Scale (RASS).10

The RASS is an observational instrument that has been validated in the intesive care unit setting for objectively determining level of sedation. It has been shown to be highly reliable and associated with delirium.11 The RASS is a quick, objective scale of consciousness with a scoring system that captures both hyperactive and hypoactive levels of consciousness. A disadvantage of using the RASS includes its limited attention assessment. The Consensus panel modified the RASS to improve its assessment of attention, using a brief open‐ended question that was asked before scoring (Figure 1).

Figure 1

RESULTS

Characteristics of the study population are presented in Table 1. Because this was a VA population, the vast majority (94%) of participants were men, with a mean age of 81 years (range, 66‐96 years), and 89% were white. This population had a high Charlson Comorbidity Index (mean SD, 4.0 2.4), which was reflected in functional assessments, with 37% reporting difficulty with activities of daily living and 58% reporting difficulty with instrumental activities of daily living. Despite the age and comorbidity, delirium prevalence was 11% (n = 10) and incidence was 14% (n = 13). Interrater reliability of the mRASS yielded 98% agreement with a weighted kappa of 0.48 (P < 0.001).

When the mRASS was analyzed as a single‐day independent assessment, any abnormal score (ie, a score 0) had a sensitivity of 64% and a specificity of 93% for delirium relative to the expert evaluation (Table 2). With an abnormal mRASS as 2 or 2, the sensitivity fell to 34%, while the specificity increased to 99.6%.

When the mRASS was used longitudinally to detect change in delirium during the hospital stay among all participants, it had a sensitivity of 74% and specificity of 92% for any change. Increasing the stringency of the criteria by looking at a change of 2 mRASS points decreased the sensitivity (22%) and increased the specificity (100%).

To capture the potential of the mRASS administered on a longitudinal basis as a diagnostic aid, the prevalent cases of delirium were excluded. In this analysis, any change in the mRASS had a sensitivity of 85% and a specificity of 92% for incident delirium. With more stringent criteria of a change of 2 points, the sensitivity was 23% and the specificity was 100%.

DISCUSSION

In this study, we developed a modified RASS (mRASS) for serial mental status assessment. Whereas a single measurement of the mRASS had modest sensitivity and good specificity for delirium, longitudinal measurement increased the sensitivity with no loss in specificity. Importantly, the <30 seconds required for the mRASS could be incorporated into daily workflow and provides an objective measure of consciousness. As such, we believe the mRASS can potentially serve as a longitudinal measure of consciousnessmuch like a vital sign for mental status.

Altered consciousness is a clinical and diagnostic feature of delirium,5, 13 and fluctuation in mental status is a diagnostic feature of delirium. As such, a screening instrument able to quantify the level of consciousness longitudinally and allow comparison to prior and subsequent determinations has face validity as a delirium screening instrument.

The mRASS has other features that make it appropriate for serial measurement in a manner similar to a vital sign. First, it objectively described consciousness on a scale, which is an improvement relative to many of the subjective descriptions clinically used. Consistent with other studies of the RASS,10, 11 the mRASS has good interrater reliability, allowing a common language to be used to describe level of consciousness across health care settings that can become the basis for a systematic and standardized monitor of cognitive change, improving continuity of care and communication between providers. It can be further used to objectively establish a patient's baseline and monitor change longitudinally.

The current study is limited by the lack of diversity and small size of the study population, which limits external validity (generalizability). Additional studies evaluating the utility of the mRASS by a variety of health care team members in a larger, more ethnically/racially diverse and heterogeneous population should be completed before we can determine if it can perform as a mental status vital sign, and if it is associated with better patient outcomes. Additionally, this study selected patients who were physically and cognitively capable of enrolling and excluded patients with severe cognitive and sensory impairment who were unable to provide consent to participate. Thus, some of the sickest, frailest, and most cognitively impaired patients were excluded. Unfortunately, this study therefore excluded a population significantly more vulnerable to the development of delirium.

Because a change in mental status (such as delirium) is common, morbid, and costly, a brief tool that can reliably and effectively assess mental status is needed. The mRASS used in this study provided an objective measurement of consciousness, a key component of mental status, and was demonstrated to reliably screen for presence or absence of delirium when administered longitudinally. Further study in diverse populations with administration by a variety of health care team members is needed to determine whether the mRASS can accurately serve as a mental status vital sign. If adopted widely, the mRASS could be used alongside the traditional vital signs to establish patient baselines, monitor change, improve provider communication, and potentially improve patient outcomes.

Vital signs constitute a fundamental component of the physical examination and serve key diagnostic and monitoring purposes. The brain is as vital to life as the cardiovascular, respiratory, and immune/thermoregulatory systems, yet currently no vital sign exists that would allow rapid, reliable, and easily reproducible assessment of cognition.1 As a result, acute mental status changes frequently go undetected and untreated.24 Delirium is defined as an acute change in attention with fluctuations in cognition, thought, and/or consciousness throughout the course of the day.5 Because delirium in older patients is common and is associated with increased morbidity, mortality, functional decline, and costs,69 development and validation of a rapid, objective screening assessment could be used by nursing staff to identify patients at high risk for delirium.

Current recommendations for inpatient delirium monitoring usually involve daily cognitive screening with a standardized screening instrument.6 Because this process is often time‐consuming (8‐12 minutes), most patients do not undergo routine screening. To facilitate clinical implementation, we focused on developing a brief (<30‐second) inpatient screening measure of a feature of mental status that could be administered serially. The purpose of this study was to (1) develop a brief screening tool for a core feature of mental status and (2) validate this screening tool for delirium in an older inpatient population.

METHODS

Consensus Panel

In June 2009, the Veterans Administration sponsored an interdisciplinary conference that solicited input on identifying the most targetable components of delirium and discussing potential clinical instruments. Following this, a consensus panel of 8 representatives from medicine, geriatrics, nursing, psychiatry, and psychology used a modified Delphi method to target characteristic features of delirium and identify instruments that could best capture mental status change. While inattention was agreed upon as the core cognitive feature of delirium, the group came to consensus that capturing the acute onset and fluctuating nature of delirium was better suited as a vital sign. To meet these criteria, the group modified the Richmond Agitation Sedation Scale (RASS).10

The RASS is an observational instrument that has been validated in the intesive care unit setting for objectively determining level of sedation. It has been shown to be highly reliable and associated with delirium.11 The RASS is a quick, objective scale of consciousness with a scoring system that captures both hyperactive and hypoactive levels of consciousness. A disadvantage of using the RASS includes its limited attention assessment. The Consensus panel modified the RASS to improve its assessment of attention, using a brief open‐ended question that was asked before scoring (Figure 1).

Figure 1

RESULTS

Characteristics of the study population are presented in Table 1. Because this was a VA population, the vast majority (94%) of participants were men, with a mean age of 81 years (range, 66‐96 years), and 89% were white. This population had a high Charlson Comorbidity Index (mean SD, 4.0 2.4), which was reflected in functional assessments, with 37% reporting difficulty with activities of daily living and 58% reporting difficulty with instrumental activities of daily living. Despite the age and comorbidity, delirium prevalence was 11% (n = 10) and incidence was 14% (n = 13). Interrater reliability of the mRASS yielded 98% agreement with a weighted kappa of 0.48 (P < 0.001).

When the mRASS was analyzed as a single‐day independent assessment, any abnormal score (ie, a score 0) had a sensitivity of 64% and a specificity of 93% for delirium relative to the expert evaluation (Table 2). With an abnormal mRASS as 2 or 2, the sensitivity fell to 34%, while the specificity increased to 99.6%.

When the mRASS was used longitudinally to detect change in delirium during the hospital stay among all participants, it had a sensitivity of 74% and specificity of 92% for any change. Increasing the stringency of the criteria by looking at a change of 2 mRASS points decreased the sensitivity (22%) and increased the specificity (100%).

To capture the potential of the mRASS administered on a longitudinal basis as a diagnostic aid, the prevalent cases of delirium were excluded. In this analysis, any change in the mRASS had a sensitivity of 85% and a specificity of 92% for incident delirium. With more stringent criteria of a change of 2 points, the sensitivity was 23% and the specificity was 100%.

DISCUSSION

In this study, we developed a modified RASS (mRASS) for serial mental status assessment. Whereas a single measurement of the mRASS had modest sensitivity and good specificity for delirium, longitudinal measurement increased the sensitivity with no loss in specificity. Importantly, the <30 seconds required for the mRASS could be incorporated into daily workflow and provides an objective measure of consciousness. As such, we believe the mRASS can potentially serve as a longitudinal measure of consciousnessmuch like a vital sign for mental status.

Altered consciousness is a clinical and diagnostic feature of delirium,5, 13 and fluctuation in mental status is a diagnostic feature of delirium. As such, a screening instrument able to quantify the level of consciousness longitudinally and allow comparison to prior and subsequent determinations has face validity as a delirium screening instrument.

The mRASS has other features that make it appropriate for serial measurement in a manner similar to a vital sign. First, it objectively described consciousness on a scale, which is an improvement relative to many of the subjective descriptions clinically used. Consistent with other studies of the RASS,10, 11 the mRASS has good interrater reliability, allowing a common language to be used to describe level of consciousness across health care settings that can become the basis for a systematic and standardized monitor of cognitive change, improving continuity of care and communication between providers. It can be further used to objectively establish a patient's baseline and monitor change longitudinally.

The current study is limited by the lack of diversity and small size of the study population, which limits external validity (generalizability). Additional studies evaluating the utility of the mRASS by a variety of health care team members in a larger, more ethnically/racially diverse and heterogeneous population should be completed before we can determine if it can perform as a mental status vital sign, and if it is associated with better patient outcomes. Additionally, this study selected patients who were physically and cognitively capable of enrolling and excluded patients with severe cognitive and sensory impairment who were unable to provide consent to participate. Thus, some of the sickest, frailest, and most cognitively impaired patients were excluded. Unfortunately, this study therefore excluded a population significantly more vulnerable to the development of delirium.

Because a change in mental status (such as delirium) is common, morbid, and costly, a brief tool that can reliably and effectively assess mental status is needed. The mRASS used in this study provided an objective measurement of consciousness, a key component of mental status, and was demonstrated to reliably screen for presence or absence of delirium when administered longitudinally. Further study in diverse populations with administration by a variety of health care team members is needed to determine whether the mRASS can accurately serve as a mental status vital sign. If adopted widely, the mRASS could be used alongside the traditional vital signs to establish patient baselines, monitor change, improve provider communication, and potentially improve patient outcomes.

Vital signs constitute a fundamental component of the physical examination and serve key diagnostic and monitoring purposes. The brain is as vital to life as the cardiovascular, respiratory, and immune/thermoregulatory systems, yet currently no vital sign exists that would allow rapid, reliable, and easily reproducible assessment of cognition.1 As a result, acute mental status changes frequently go undetected and untreated.24 Delirium is defined as an acute change in attention with fluctuations in cognition, thought, and/or consciousness throughout the course of the day.5 Because delirium in older patients is common and is associated with increased morbidity, mortality, functional decline, and costs,69 development and validation of a rapid, objective screening assessment could be used by nursing staff to identify patients at high risk for delirium.

Current recommendations for inpatient delirium monitoring usually involve daily cognitive screening with a standardized screening instrument.6 Because this process is often time‐consuming (8‐12 minutes), most patients do not undergo routine screening. To facilitate clinical implementation, we focused on developing a brief (<30‐second) inpatient screening measure of a feature of mental status that could be administered serially. The purpose of this study was to (1) develop a brief screening tool for a core feature of mental status and (2) validate this screening tool for delirium in an older inpatient population.

METHODS

Consensus Panel

In June 2009, the Veterans Administration sponsored an interdisciplinary conference that solicited input on identifying the most targetable components of delirium and discussing potential clinical instruments. Following this, a consensus panel of 8 representatives from medicine, geriatrics, nursing, psychiatry, and psychology used a modified Delphi method to target characteristic features of delirium and identify instruments that could best capture mental status change. While inattention was agreed upon as the core cognitive feature of delirium, the group came to consensus that capturing the acute onset and fluctuating nature of delirium was better suited as a vital sign. To meet these criteria, the group modified the Richmond Agitation Sedation Scale (RASS).10

The RASS is an observational instrument that has been validated in the intesive care unit setting for objectively determining level of sedation. It has been shown to be highly reliable and associated with delirium.11 The RASS is a quick, objective scale of consciousness with a scoring system that captures both hyperactive and hypoactive levels of consciousness. A disadvantage of using the RASS includes its limited attention assessment. The Consensus panel modified the RASS to improve its assessment of attention, using a brief open‐ended question that was asked before scoring (Figure 1).

Figure 1

RESULTS

Characteristics of the study population are presented in Table 1. Because this was a VA population, the vast majority (94%) of participants were men, with a mean age of 81 years (range, 66‐96 years), and 89% were white. This population had a high Charlson Comorbidity Index (mean SD, 4.0 2.4), which was reflected in functional assessments, with 37% reporting difficulty with activities of daily living and 58% reporting difficulty with instrumental activities of daily living. Despite the age and comorbidity, delirium prevalence was 11% (n = 10) and incidence was 14% (n = 13). Interrater reliability of the mRASS yielded 98% agreement with a weighted kappa of 0.48 (P < 0.001).

When the mRASS was analyzed as a single‐day independent assessment, any abnormal score (ie, a score 0) had a sensitivity of 64% and a specificity of 93% for delirium relative to the expert evaluation (Table 2). With an abnormal mRASS as 2 or 2, the sensitivity fell to 34%, while the specificity increased to 99.6%.

When the mRASS was used longitudinally to detect change in delirium during the hospital stay among all participants, it had a sensitivity of 74% and specificity of 92% for any change. Increasing the stringency of the criteria by looking at a change of 2 mRASS points decreased the sensitivity (22%) and increased the specificity (100%).

To capture the potential of the mRASS administered on a longitudinal basis as a diagnostic aid, the prevalent cases of delirium were excluded. In this analysis, any change in the mRASS had a sensitivity of 85% and a specificity of 92% for incident delirium. With more stringent criteria of a change of 2 points, the sensitivity was 23% and the specificity was 100%.

DISCUSSION

In this study, we developed a modified RASS (mRASS) for serial mental status assessment. Whereas a single measurement of the mRASS had modest sensitivity and good specificity for delirium, longitudinal measurement increased the sensitivity with no loss in specificity. Importantly, the <30 seconds required for the mRASS could be incorporated into daily workflow and provides an objective measure of consciousness. As such, we believe the mRASS can potentially serve as a longitudinal measure of consciousnessmuch like a vital sign for mental status.

Altered consciousness is a clinical and diagnostic feature of delirium,5, 13 and fluctuation in mental status is a diagnostic feature of delirium. As such, a screening instrument able to quantify the level of consciousness longitudinally and allow comparison to prior and subsequent determinations has face validity as a delirium screening instrument.

The mRASS has other features that make it appropriate for serial measurement in a manner similar to a vital sign. First, it objectively described consciousness on a scale, which is an improvement relative to many of the subjective descriptions clinically used. Consistent with other studies of the RASS,10, 11 the mRASS has good interrater reliability, allowing a common language to be used to describe level of consciousness across health care settings that can become the basis for a systematic and standardized monitor of cognitive change, improving continuity of care and communication between providers. It can be further used to objectively establish a patient's baseline and monitor change longitudinally.

The current study is limited by the lack of diversity and small size of the study population, which limits external validity (generalizability). Additional studies evaluating the utility of the mRASS by a variety of health care team members in a larger, more ethnically/racially diverse and heterogeneous population should be completed before we can determine if it can perform as a mental status vital sign, and if it is associated with better patient outcomes. Additionally, this study selected patients who were physically and cognitively capable of enrolling and excluded patients with severe cognitive and sensory impairment who were unable to provide consent to participate. Thus, some of the sickest, frailest, and most cognitively impaired patients were excluded. Unfortunately, this study therefore excluded a population significantly more vulnerable to the development of delirium.

Because a change in mental status (such as delirium) is common, morbid, and costly, a brief tool that can reliably and effectively assess mental status is needed. The mRASS used in this study provided an objective measurement of consciousness, a key component of mental status, and was demonstrated to reliably screen for presence or absence of delirium when administered longitudinally. Further study in diverse populations with administration by a variety of health care team members is needed to determine whether the mRASS can accurately serve as a mental status vital sign. If adopted widely, the mRASS could be used alongside the traditional vital signs to establish patient baselines, monitor change, improve provider communication, and potentially improve patient outcomes.