Psychological of Stay
Stephen
M. Saravay,
M.D., Simcha
Objective: particular
As general attention has
Comorbidity in the General
Maurice Pollack,
and Length Hospital
D. Steinberg, Ph.D.,
and
M.D.,
Nancy
Barbara
Alovis,
Weinschel,
Ph.D.,
B.A.
hospitals search for ways to cut costs without sacrificing been focused on factors that may prolong hospital stay.
efficiency, The results
of prior studies that have reported an association between psychological and comorbidity and longer hospital stays have been subject to different interpretations
psychiatric because
of methodological comorbidity and
ofearlier
design flaws. The current paper length ofstay that has been designed
investigations.
versity-affihiated,
Method:
voluntary,
The study
teaching
hospital.
reports to avoid
was performed During
hospital
on a study of the methodological
at a 429-bed days
psychological problems
tertiary-care,
uni-
3 to 5, patients
were
tested as available with the Mini-Mental State examination, the Zung Depression Inventory, and the SCL-90 and were rated for physical impairment with the Karnofsky Performance Status Scale. Statistical analyses were performed for correlations between length of stay and test scores, rating scales, and demographic and discharge data from the chart. Results: Of 424 patients approached, 321 (76%) agreed to participate and 278 (65.6%) completed the test battery. Depression, anxiety, and organicity, measured by psychological tests, were significantly correlated with longer hospital stay. These correlations remained significant after
the
authors
controlled
for
degree
of physical
impairment,
emergency
versus
elective
admission, and medical versus surgical service, which were themselves correlated with longer hospital stay. Discussion: This study confirms a significant correlation between psychological comorbidity and length of stay after correcting for the methodological pitfalls found in earlier studies. The clinical, research, economic, and policy implications of these findings are discussed. (Am
J
Psychiatry
1991;
148:324-329)
I
n response to the rekindled interest in factors influencing extended length of stay and its economic implications, a number of recent studies have reexamined the association between psychiatric comonbidity and longer stay first reported by Billings (1) almost SO years ago (2-1 1). While the evidence of a strong association between psychiatric comonbidity and longer hospital stay has grown, three important issues about the nature of that association have not been sufficiently addressed by this body of work. The first is whether psychiatric comorbidity precedes and, therefore, may
Presented
at a meeting
of the American
Psychosomatic
Society,
Toronto, March 26, 1988. Received Sept. 19, 1989; revision received Aug. 8, 1990; accepted Sept. 19, 1990. From the Long Island Jewish
Address
Medical
Center;
reprint
requests
ices, Long 11042.
Island
and
Jewish
St. John’s
University, Queens, N.Y. Consultation-Liaison ServCenter, New Hyde Park, NY
to Dr. Saravay, Medical
The authors thank research volunteers Dean Cestari, Lillian Hoffman, Judy Mittleman, and Jill Schreibman for their assistance. Copyright
324
© 1991
American
Psychiatric
Association.
contribute to extended hospital stays or whether it arises later after hospitalization has been prolonged by other factors. The second is whether or not length of stay and psychiatric comorbidity are directly comelated with each other or whether they covary with the degree of physical impairment. Third, all of the studies that reported a significant correlation were based on a small subgroup of the total population of patients with psychiatric morbidity who were spontaneously identifled or referred by their medical or surgical physicians. In the first of several retrospective studies, Dvoredsky and Cooley (2) reviewed 37,000 Veterans Administration hospital discharge charts and found a significant correlation between psychiatric comorbidity and length of stay. Fulop and co-workers (4) reported that the 5% of patients with comorbid psychiatric diagnoses found in the 59,000 charts they reviewed had significantly longer hospital stays. Lyons et al. (5) found significantly longer stays for patients with head and spinal cord trauma and psychiatric comorbidity in both the acute and rehabilitation phases of their hos-
Am
J
Psychiatry
148:3,
March
1991
SARAVAY,
pitalization, while Cushman (8) found longer stays for stroke patients with depression and organic brain syndrome. Because of their retrospective designs, these studies could not determine whether psychiatric comorbidity preceded and may have contributed to extended hospitalization or occurred later during the course, perhaps as a result of prolonged hospitalization. One attempt to deal with this methodological issue statistically still left the association unclear (7). Brezel et al. (9) found that burn patients with a prehospitalization history of substance abuse or mental impairment had a significantly longer hospital stay than other patients. However, they did not control for age on flame burns, both of which were greaten in the study group. Four prospective studies addressed some of the methodological limitations described previously. Levenson et al. (3) tested 80 consecutively admitted patients with a psychological battery but failed to find significant correlations between psychological test scores and subsequent length of hospital stay. Rogers et al. (10) failed to find a significantly longer stay for patients diagnosed with delirium on the fourth day after orthopedic surgery. In another prospective study, Morris and Goldberg (1 1) failed to demonstrate a melationship between psychological or psychiatric comorbidity and length of stay in gastroenterology patients with mixed diagnoses or a single diagnosis; this finding may be due to small sample size. However, Thomas et al. (6) did find that a diagnosis of delirium was associated with a longer stay. Unfortunately, none of the cited studies with positive findings controlled for severity of physical impairment. Severity of physical illness and impairment have been found to strongly correlate with both longer stay (12) and greater psychiatric morbidity (13). In the study by Thomas et al. (6), the group of patients with delirium had a higher mortality rate than the control group. Since delirium is often seen as a harbinger of terminal states (14, 15), which may be associated with prolonged hospital stays, the longer stay may have been due to the greaten physical morbidity of these patients. A similar issue can be raised regarding the retrospective study by Fulop et al. (4). In that study, it also is not clean whether severity of physical illness might have contributed to the longer stay, since patients with psychiatric comorbidity had a greater numben of medical diagnoses, surgical consultations, and surgical procedures and a higher mortality rate. All of the cited studies with positive findings relied on psychiatric diagnoses recorded by surgical on other nonpsychiatnic physicians in the discharge summary or on spontaneous referral for psychiatric consultation. Psychiatric disorders are known to be undendiagnosed and patients with these disorders underreferned (1, 6, 16, 17) in the general hospital. One study demonstrated that only 10% of the population with psychiatric comorbidity were referred (16). Even when patients with psychiatric disorders are diagnosed and referred, the psychiatric diagnoses are frequently not
Am
J
Psychiatry
I 48:3,
March
1991
STEINBERG,
WEINSCHEL,
ET AL.
reported in the discharge summaries (18). Studies using discharge psychiatric diagnoses or spontaneous mefemral to define psychiatric comorbidity report on mesuits taken from a small subsample of patients who may be more likely to provoke referral and may not be representative of the population with psychiatric comorbidity. The current paper reports on a study of psychological comorbidity and length of stay designed to avoid the methodological problems of earlier investigations. This prospective study used psychological tests administered to patients during days 3 to S and controlled for the severity of physical impairment and other potentially confounding variables.
METHOD The study was carried out from June 1, 1985, through June 30, 1986, at Long Island Jewish Hospital, a 429-bed-tertiary cane, university-affiliated, voluntany, teaching hospital. Patients from the four medical floors, four surgical floors, and one gynecology floor were tested as available on weekdays during the third to fifth hospital days. In order to address the potential bias of nonresponders, length of stay and other demographic data were also recorded for those patients who refused to participate or who agreed to participate but were excluded for the reasons described later. The test battery included a modified version of the SCL-90 (questions 13, 47, 70, and 80 were deleted because they were not appropriate for hospital inpatients-e.g., they dealt with phobic concerns on buses and subways in the prior week), the Zung Depression Inventory, and the Mini-Mental State examination. At the time of testing, each patient’s nurse mated the degree of physical impairment according to the Kannofsky Performance Status Scale (19). After each patient’s discharge we gathered data from the chant on social and environmental factors that might affect length of stay, such as preadmission employment status, preadmission and postdischarge living arrangements, availability of postdischarge care, availability of significant others on discharge, and limitation of activity as prescribed by the physician. We also recorded type of admission, final diagnoses, and other demographic data.
RESULTS Of 424 patients approached, 321 (76%) agreed to participate in the study. Thirty-four (10.6%) of the patients who agreed to participate had to be excluded before testing-21 because they were too physically ill, two because they were too emotionally ill (one was confused and one schizophrenic), eight because of language problems, and three for miscellaneous reasons. An additional nine (2.8%) were excluded after they failed to complete the test battery-three because they
325
COMORBIDITY
AND
LENGTH
OF
HOSPITAL
STAY
were too physically ill, one because of emotional upset, and five for miscellaneous reasons. A total of 278 patients, 65.6% of those initially approached, completed the test battery and made up the group of subjects studied. Subjects ranged in age from 18 to 83 years (mean± SD=48.1±16.2). The mean length of stay was 8.7± 6.7 days (nange=3-61). Fifty-four percent (N=1SO) were women; 89% (N=247) were white; 42% (N= 117) were Catholic, 35% (N97) were Jewish, and 16% (N=44) were Protestant; 67% (N= 186) were married; and 9% (N=25) were unemployed or disabled, 28% (N=78) were retired or homemakers, and 53% (N= 147) were employed. Ninety-six percent (N=267) lived in their own home or apartment, and only 11% (N=31) lived alone. The mean Zung depression score of the study group was 45.8 ± 10.4; 9% (N=25) had moderate to severe depression (scores of 60 and above), while 34% (N= 95) had minimal to severe depression (scones of SO and above). The mean Mini-Mental State score was 27.3± 2.3. We used a cutoff score of 23 on below (previously shown to have an 87% sensitivity and 82% specificity [20]) and found that 8% (N=22) of the patients had organic impairment according to the Mini-Mental State examination. The mean Kamnofsky scone was 72.9± 15.7 (out of a possible 100); this score indicates that the patient can care for himself or herself but is unable to carry on normal activity or do active work. Since SCL-90 scones were not compatible with norms because of the deletion of four items inappropriate for hospitalized patients, only correlations will be meported for these data. The group of patients who mefused to participate on who agreed to participate but had to be excluded for the reasons described earlier were significantly olden than the study group (61.2± 17.5 versus 48.1±16.2 years; t=7.62, dfr379, p< 0.001) and stayed significantly longer (13.8±14.5 yensus 8.7±6.7 days; t2.52, df=378, p
Stephen
M. Saravay,
M.D., Simcha
Objective: particular
As general attention has
Comorbidity in the General
Maurice Pollack,
and Length Hospital
D. Steinberg, Ph.D.,
and
M.D.,
Nancy
Barbara
Alovis,
Weinschel,
Ph.D.,
B.A.
hospitals search for ways to cut costs without sacrificing been focused on factors that may prolong hospital stay.
efficiency, The results
of prior studies that have reported an association between psychological and comorbidity and longer hospital stays have been subject to different interpretations
psychiatric because
of methodological comorbidity and
ofearlier
design flaws. The current paper length ofstay that has been designed
investigations.
versity-affihiated,
Method:
voluntary,
The study
teaching
hospital.
reports to avoid
was performed During
hospital
on a study of the methodological
at a 429-bed days
psychological problems
tertiary-care,
uni-
3 to 5, patients
were
tested as available with the Mini-Mental State examination, the Zung Depression Inventory, and the SCL-90 and were rated for physical impairment with the Karnofsky Performance Status Scale. Statistical analyses were performed for correlations between length of stay and test scores, rating scales, and demographic and discharge data from the chart. Results: Of 424 patients approached, 321 (76%) agreed to participate and 278 (65.6%) completed the test battery. Depression, anxiety, and organicity, measured by psychological tests, were significantly correlated with longer hospital stay. These correlations remained significant after
the
authors
controlled
for
degree
of physical
impairment,
emergency
versus
elective
admission, and medical versus surgical service, which were themselves correlated with longer hospital stay. Discussion: This study confirms a significant correlation between psychological comorbidity and length of stay after correcting for the methodological pitfalls found in earlier studies. The clinical, research, economic, and policy implications of these findings are discussed. (Am
J
Psychiatry
1991;
148:324-329)
I
n response to the rekindled interest in factors influencing extended length of stay and its economic implications, a number of recent studies have reexamined the association between psychiatric comonbidity and longer stay first reported by Billings (1) almost SO years ago (2-1 1). While the evidence of a strong association between psychiatric comonbidity and longer hospital stay has grown, three important issues about the nature of that association have not been sufficiently addressed by this body of work. The first is whether psychiatric comorbidity precedes and, therefore, may
Presented
at a meeting
of the American
Psychosomatic
Society,
Toronto, March 26, 1988. Received Sept. 19, 1989; revision received Aug. 8, 1990; accepted Sept. 19, 1990. From the Long Island Jewish
Address
Medical
Center;
reprint
requests
ices, Long 11042.
Island
and
Jewish
St. John’s
University, Queens, N.Y. Consultation-Liaison ServCenter, New Hyde Park, NY
to Dr. Saravay, Medical
The authors thank research volunteers Dean Cestari, Lillian Hoffman, Judy Mittleman, and Jill Schreibman for their assistance. Copyright
324
© 1991
American
Psychiatric
Association.
contribute to extended hospital stays or whether it arises later after hospitalization has been prolonged by other factors. The second is whether or not length of stay and psychiatric comorbidity are directly comelated with each other or whether they covary with the degree of physical impairment. Third, all of the studies that reported a significant correlation were based on a small subgroup of the total population of patients with psychiatric morbidity who were spontaneously identifled or referred by their medical or surgical physicians. In the first of several retrospective studies, Dvoredsky and Cooley (2) reviewed 37,000 Veterans Administration hospital discharge charts and found a significant correlation between psychiatric comorbidity and length of stay. Fulop and co-workers (4) reported that the 5% of patients with comorbid psychiatric diagnoses found in the 59,000 charts they reviewed had significantly longer hospital stays. Lyons et al. (5) found significantly longer stays for patients with head and spinal cord trauma and psychiatric comorbidity in both the acute and rehabilitation phases of their hos-
Am
J
Psychiatry
148:3,
March
1991
SARAVAY,
pitalization, while Cushman (8) found longer stays for stroke patients with depression and organic brain syndrome. Because of their retrospective designs, these studies could not determine whether psychiatric comorbidity preceded and may have contributed to extended hospitalization or occurred later during the course, perhaps as a result of prolonged hospitalization. One attempt to deal with this methodological issue statistically still left the association unclear (7). Brezel et al. (9) found that burn patients with a prehospitalization history of substance abuse or mental impairment had a significantly longer hospital stay than other patients. However, they did not control for age on flame burns, both of which were greaten in the study group. Four prospective studies addressed some of the methodological limitations described previously. Levenson et al. (3) tested 80 consecutively admitted patients with a psychological battery but failed to find significant correlations between psychological test scores and subsequent length of hospital stay. Rogers et al. (10) failed to find a significantly longer stay for patients diagnosed with delirium on the fourth day after orthopedic surgery. In another prospective study, Morris and Goldberg (1 1) failed to demonstrate a melationship between psychological or psychiatric comorbidity and length of stay in gastroenterology patients with mixed diagnoses or a single diagnosis; this finding may be due to small sample size. However, Thomas et al. (6) did find that a diagnosis of delirium was associated with a longer stay. Unfortunately, none of the cited studies with positive findings controlled for severity of physical impairment. Severity of physical illness and impairment have been found to strongly correlate with both longer stay (12) and greater psychiatric morbidity (13). In the study by Thomas et al. (6), the group of patients with delirium had a higher mortality rate than the control group. Since delirium is often seen as a harbinger of terminal states (14, 15), which may be associated with prolonged hospital stays, the longer stay may have been due to the greaten physical morbidity of these patients. A similar issue can be raised regarding the retrospective study by Fulop et al. (4). In that study, it also is not clean whether severity of physical illness might have contributed to the longer stay, since patients with psychiatric comorbidity had a greater numben of medical diagnoses, surgical consultations, and surgical procedures and a higher mortality rate. All of the cited studies with positive findings relied on psychiatric diagnoses recorded by surgical on other nonpsychiatnic physicians in the discharge summary or on spontaneous referral for psychiatric consultation. Psychiatric disorders are known to be undendiagnosed and patients with these disorders underreferned (1, 6, 16, 17) in the general hospital. One study demonstrated that only 10% of the population with psychiatric comorbidity were referred (16). Even when patients with psychiatric disorders are diagnosed and referred, the psychiatric diagnoses are frequently not
Am
J
Psychiatry
I 48:3,
March
1991
STEINBERG,
WEINSCHEL,
ET AL.
reported in the discharge summaries (18). Studies using discharge psychiatric diagnoses or spontaneous mefemral to define psychiatric comorbidity report on mesuits taken from a small subsample of patients who may be more likely to provoke referral and may not be representative of the population with psychiatric comorbidity. The current paper reports on a study of psychological comorbidity and length of stay designed to avoid the methodological problems of earlier investigations. This prospective study used psychological tests administered to patients during days 3 to S and controlled for the severity of physical impairment and other potentially confounding variables.
METHOD The study was carried out from June 1, 1985, through June 30, 1986, at Long Island Jewish Hospital, a 429-bed-tertiary cane, university-affiliated, voluntany, teaching hospital. Patients from the four medical floors, four surgical floors, and one gynecology floor were tested as available on weekdays during the third to fifth hospital days. In order to address the potential bias of nonresponders, length of stay and other demographic data were also recorded for those patients who refused to participate or who agreed to participate but were excluded for the reasons described later. The test battery included a modified version of the SCL-90 (questions 13, 47, 70, and 80 were deleted because they were not appropriate for hospital inpatients-e.g., they dealt with phobic concerns on buses and subways in the prior week), the Zung Depression Inventory, and the Mini-Mental State examination. At the time of testing, each patient’s nurse mated the degree of physical impairment according to the Kannofsky Performance Status Scale (19). After each patient’s discharge we gathered data from the chant on social and environmental factors that might affect length of stay, such as preadmission employment status, preadmission and postdischarge living arrangements, availability of postdischarge care, availability of significant others on discharge, and limitation of activity as prescribed by the physician. We also recorded type of admission, final diagnoses, and other demographic data.
RESULTS Of 424 patients approached, 321 (76%) agreed to participate in the study. Thirty-four (10.6%) of the patients who agreed to participate had to be excluded before testing-21 because they were too physically ill, two because they were too emotionally ill (one was confused and one schizophrenic), eight because of language problems, and three for miscellaneous reasons. An additional nine (2.8%) were excluded after they failed to complete the test battery-three because they
325
COMORBIDITY
AND
LENGTH
OF
HOSPITAL
STAY
were too physically ill, one because of emotional upset, and five for miscellaneous reasons. A total of 278 patients, 65.6% of those initially approached, completed the test battery and made up the group of subjects studied. Subjects ranged in age from 18 to 83 years (mean± SD=48.1±16.2). The mean length of stay was 8.7± 6.7 days (nange=3-61). Fifty-four percent (N=1SO) were women; 89% (N=247) were white; 42% (N= 117) were Catholic, 35% (N97) were Jewish, and 16% (N=44) were Protestant; 67% (N= 186) were married; and 9% (N=25) were unemployed or disabled, 28% (N=78) were retired or homemakers, and 53% (N= 147) were employed. Ninety-six percent (N=267) lived in their own home or apartment, and only 11% (N=31) lived alone. The mean Zung depression score of the study group was 45.8 ± 10.4; 9% (N=25) had moderate to severe depression (scores of 60 and above), while 34% (N= 95) had minimal to severe depression (scones of SO and above). The mean Mini-Mental State score was 27.3± 2.3. We used a cutoff score of 23 on below (previously shown to have an 87% sensitivity and 82% specificity [20]) and found that 8% (N=22) of the patients had organic impairment according to the Mini-Mental State examination. The mean Kamnofsky scone was 72.9± 15.7 (out of a possible 100); this score indicates that the patient can care for himself or herself but is unable to carry on normal activity or do active work. Since SCL-90 scones were not compatible with norms because of the deletion of four items inappropriate for hospitalized patients, only correlations will be meported for these data. The group of patients who mefused to participate on who agreed to participate but had to be excluded for the reasons described earlier were significantly olden than the study group (61.2± 17.5 versus 48.1±16.2 years; t=7.62, dfr379, p< 0.001) and stayed significantly longer (13.8±14.5 yensus 8.7±6.7 days; t2.52, df=378, p
