The hypothesis that depressed mood may promote functional decline is grounded in theory [4, 5] and is supported by empirical evidence of a relation between symptoms of depression and poor health status [6-22]. Much of this evidence comes from studies done in outpatient and community settings [7-17]. Many of these studies, however, provide only limited evidence of a relation between symptoms of depression and poor health status outcomes for two reasons. First, many did not clarify temporal priority (that is, whether symptoms of depression precede worsening function or whether the reverse occurs) either because they were cross-sectional or because they did not adequately adjust for baseline health status [7-13]. Second, adjustment for severity of illness was often limited [7-16]. Furthermore, this evidence in outpatients may not be generalizable to those whose clinical course may be more dominated by their somatic illness, such as hospitalized older persons or patients who have acute or severe illness.

Suggestive evidence of a relation between symptoms of depression and poor outcomes in seriously ill older persons is provided by longitudinal studies of functional outcomes in patients with hip fracture [18] and studies of survival in patients with myocardial infarction [19] or nursing home residents [20]. Symptoms of depression in older patients admitted to the hospital were recently found to be associated with worse self-care function on discharge [21] and worse health status 1 month after discharge [22].

Our primary goal was to determine whether symptoms of depression on hospital admission predict worse health status outcomes in older persons hospitalized for acute medical illness. Using data from serial prospective studies, we tested five specific hypotheses [23]: 1) Symptoms of depression on admission precede and predict poor health status outcomes, 2) this association is independent of health status and severity of illness at admission, 3) the magnitude of this association is large and shows evidence of a dose response, 4) this association is consistent over time, and 5) this association is consistent even when different measures of health status outcomes are used.

The primary outcome measure was dependence in any of five basic activities of daily living (bathing, dressing, toileting, transferring, and eating). Secondary outcome measures were dependence in three or more instrumental activities of daily living (using the telephone, using transportation, shopping, preparing meals, doing housework, taking medicines, and handling finances), fair or poor self-assessed global health, and poor self-assessed global satisfaction with life.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

Study patients were assembled in serial, prospective longitudinal studies of functional change in older hospitalized patients who were admitted to the general medical service of University Hospitals of Cleveland. The inclusion and exclusion criteria for these studies have been described elsewhere [24]. The first study enrolled 206 patients who were 75 years of age or older and were admitted between March 1990 and July 1990. The second, a controlled trial of an intervention to improve functional outcomes, enrolled 651 patients who were 70 years of age or older and were admitted between November 1990 and March 1992 [24].

Of the 857 patients enrolled in the two studies, 285 were excluded from our current study because they were too ill or confused to be interviewed at the time of admission (n = 164), were admitted from nursing homes (n = 38), were not available for interview (n = 38), refused to be interviewed (n = 27), or died before being approached (n = 18). Thus, the study sample comprised 572 patients. The 285 excluded patients were more ill on admission than were the study patients, as shown by higher Acute Physiology and Chronic Health Evaluation (APACHE) II scores (15.1 compared with 12.4). The excluded patients also had higher comorbidity scores (2.5 compared with 2.2), and more excluded patients were dependent in basic activities of daily living (81% compared with 50%). Excluded patients also had worse outcomes 90 days after discharge, as shown by a higher mortality rate (33% compared with 12%); those who survived had greater dependence in basic activities of daily living (71% of excluded patients compared with 28% of study patients had one or more dependencies).

Data Collection

To determine symptoms of depression during the week preceding hospital admission, the 15-item Geriatric Depression Scale [25, 26] was used to interview patients. Interviews were conducted within 48 hours of hospital admission. The Geriatric Depression Scale is well suited for use in acutely ill older patients because of its focus on the nonsomatic symptoms of depression. We also assessed cognitive function by using the first 21 items of the Mini-Mental State examination [27], independence in basic activities of daily living [28], independence in instrumental activities of daily living [29], global health status, and satisfaction with life.

To assess the basic and instrumental activities, interviewers gave patients an example of each activity and asked patients whether they could do the activity with no help, required assistance to do the activity, or could not do the activity at all. Patients who needed assistance or could not do the activity at all were considered dependent in that activity.

For the assessment of global health status, patients were asked, "How do you rate your health today?" Response categories were "excellent," "good," "fair," or "poor." The question used to assess satisfaction with life was, "How satisfied are you with your life as a whole today?" Response categories ranged from "very satisfied" to "very dissatisfied." At discharge and 30 and 90 days after discharge, interviews were conducted to reassess dependence in basic and instrumental activities of daily living, global health status, and satisfaction with life. For dependence in basic and instrumental activities of daily living, data are based on patients' responses to the interview questions unless we could not interview the patient at a particular time point. If this was the case, we obtained the data from a surrogate (a nurse or family member). For dependence in basic activities of daily living, we obtained data from surrogates for 0% of patients at admission, 9% of patients at discharge, 16% of patients 30 days after discharge, and 17% of patients 90 days after discharge. These rates were similar for dependence in instrumental activities of daily living. For global health status and satisfaction with life, we used only data provided by the patients.

Data gathered from medical records included reasons for admission, components of the APACHE II score [30], and components of the weighted comorbidity index of Charlson [31].

Analytical Strategy

Several variables were categorized according to their distributions or clinically sensible thresholds. Patients were grouped into tertiles of similar size on the basis of the number of symptoms of depression. The primary outcome measure, basic activity of daily living function, was classified as independent (patient could perform all five basic activities of daily living without assistance) or dependent (patient needed assistance with 1 basic activity of daily living). Instrumental activity of daily living function was dichotomized as dependent in three or more activities compared with dependent in two or less. Global health status was dichotomized as excellent or good compared with fair or poor. Satisfaction with life was classified as satisfied or very satisfied compared with equally satisfied and dissatisfied, dissatisfied, or very dissatisfied.

We did the analysis in two stages. First, we determined the cross-sectional relation, on admission, of symptoms of depression to dependence in basic activities of daily living and to each of the three secondary health status measures. Bivariate associations were tested using chi-square tests; modification for trend was done when appropriate. To determine the independent contribution of symptoms of depression to each measure on admission, logistic regression analyses adjusted for severity of illness (APACHE II score), comorbidity (Charlson score), mental status score, and demographic characteristics (age, race, sex, living arrangement, and marital status).

Second, we determined whether symptoms of depression on admission predicted dependence in basic activities of daily living and worse secondary outcomes at discharge and 30 and 90 days after discharge. These analyses included only patients who survived for 90 days after discharge and for whom data on the outcome variable of interest were collected at each time point. Complete data on basic activities of daily living were available for 467 of the 503 (93%) surviving patients, and data on health status were available for 347 of the 503 patients (69%). Because data on instrumental activities of daily living and satisfaction with life were collected at all time points in the second study only, the potential analytical sample for these outcomes was limited to the 359 patients in the second study who survived for 3 months. Complete data on instrumental activities of daily living were available for 336 of the 359 patients (94%); complete data on satisfaction with life were available for 223 of the 359 patients (62%).

To decrease confounding by health status at admission, the bivariate analysis for each outcome measure was stratified by the admission health status measure of interest (for example, the bivariate analysis assessing the influence of symptoms of depression on dependence in basic activities of daily living was stratified by whether the patient was dependent in basic activities of daily living on admission). The multivariable models for each health status outcome adjusted for APACHE II, Charlson, and mental status scores at admission; demographic characteristics; and health status at admission. Thus, all of the analyses of outcomes at discharge and thereafter assess whether symptoms of depression predict worse future health status after adjustment for the possibility that patients with more symptoms of depression have worse health status on admission.

All analyses were done using SAS software, version 6.08 (SAS Institute, Cary, North Carolina) [32].

Additional Analyses

To test the robustness of our analyses, we did three additional analyses. First, for logistic regression models of function, we constructed ordinal logistic regression models in which the outcome variable was the number of independent basic activities of daily living or independent instrumental activities of daily living and the number of symptoms of depression was a predictor variable. Second, we identified four items on the Geriatric Depression Scale that were likely to be attributable to acute physical illness (not full of energy, preferred to stay at home rather than doing new things, dropped many activities and interests, and had more problems with memory than most persons [Table 1]). After these four symptoms were removed, patients were reclassified into tertiles according to their responses to the remaining 11 items. All analyses were repeated on the basis of the reclassification. Third, analyses of dependence in basic and instrumental activities of daily living were repeated by using only patients who themselves reported functional outcomes at all time points. Results of these additional analyses were similar to the results of the primary analyses and are not reported separately.

Results

Top Methods Results Discussion Author & Article Info References

The mean (±SD) age of the 572 patients was 79.9 ± 6.4 years (range, 70 to 101 years). Sixty-eight percent of patients were women, 61% were white, and 39% were black. Forty-three percent of patients lived alone, 34% percent lived with their spouse, and 23% lived with another adult. The main reasons for hospital admission were cardiovascular disorders (22% of patients), pulmonary disorders (20% of patients), gastrointestinal disorders (21% of patients), neurologic disorders (11% of patients), infectious disorders (9% of patients), metabolic disorders (8% of patients), and other reasons (9% of patients). On admission, 50% of patients were dependent in at least one of the five basic activities of daily living and 67% were dependent in three or more instrumental activities of daily living. The mean Charlson comorbidity index score at admission was 2.2 ± 2.2, and the mean 21-item Mini-Mental State examination score at admission was 17.1 ± 3.8. The mean length of stay was 7.4 ± 6.2 days.

Each symptom of depression was reported by at least 15% of patients, and eight of the symptoms were reported by more than 25% of patients (Table 1). Figure 1 shows the frequencies of the total number of symptoms of depression, with patients divided into tertiles.

View larger version (21K): [in this window] [in a new window]   Figure 1. Frequency distribution of the total number of symptoms of depression on hospital admission. Patients were divided into tertiles on the basis of the frequency distribution.

Relation between Number of Symptoms of Depression and Health Status at Admission

On admission, patients with more symptoms of depression had worse health status, as indicated by greater dependence in basic activities of daily living and each of the complementary health status measures (Table 2). Patients with more symptoms of depression also had higher mean APACHE II scores (11.9 in patients with zero to two symptoms, 12.7 in patients with three to five symptoms, and 12.7 in patients with six or more symptoms; P = 0.07) and higher mean comorbidity scores (1.8, 2.2, and 2.6 in the three groups, respectively; P = 0.004). The number of symptoms of depression was not associated (P > 0.1) with age, sex, race, marital status, living arrangement, reason for admission, or mental status on admission. In multivariable models that controlled for all of these potentially confounding factors, more symptoms of depression continued to be independently associated with dependence in basic and instrumental activities of daily living, less satisfaction with life, and worse global health status (Table 2).

Relation between Symptoms of Depression on Admission and Health Status Outcomes at Discharge and 30 and 90 Days after Discharge

Figure 2 stratifies patients according to whether they were dependent in one or more basic activities of daily living on admission. Patients with more symptoms of depression on admission were more likely to be dependent in at least one basic activity of daily living at discharge, 30 days after discharge, and 90 days after discharge. The stratified analysis shows that the worse outcomes in patients with more symptoms of depression are explained by 1) a greater likelihood that patients who were not dependent in any basic activity of daily living on admission would decline and become dependent in function at discharge and 30 and 90 days after discharge (Figure 2, top) and 2) a greater likelihood that patients who were dependent in function on admission would not improve and would remain dependent in function at discharge and thereafter (Figure 2, bottom). For example, among patients who were not dependent in any basic activity of daily living on admission (Figure 2, top), 32% of patients with six or more symptoms became dependent in at least one basic activity of daily living on discharge. In contrast, only 16% of patients with zero to two symptoms became dependent. Among patients who were dependent in at least one basic activity of daily living on admission (Figure 2, bottom), 71% of those with six or more symptoms remained dependent at discharge (that is, 29% improved and became independent) whereas 35% of those with zero to two symptoms remained dependent (that is, 65% improved and became independent). These differences in dependence in basic activities of daily living were smaller 30 days after discharge and continued to narrow through day 90. At each of the three time points, however, older patients with more symptoms of depression were significantly more likely to be dependent in at least one basic activity of daily living.

View larger version (27K): [in this window] [in a new window]   Figure 2. Frequency of dependence in one or more basic activities of daily living at discharge and 30 and 90 days after discharge, according to the number of symptoms of depression on admission. To reduce confounding between dependence in basic activities of daily living on admission and subsequent dependence, patients were stratified by whether they were dependent in one or more basic activities of daily living on admission. Of the 503 patients who survived for 90 days, the 36 patients for whom data on dependence in basic activities of daily living were not available for all three time points were excluded from this analysis. Top. Frequency of dependence in one or more basic activities of daily living in 248 patients who were not dependent in any activity on admission. Bottom. Frequency of dependence in one or more basic activities of daily living in 219 patients who were dependent in at least one basic activity of daily living on admission. At each of the three time points, the association between symptoms of depression on admission and dependence in basic activities of daily living was significant in stratified analyses that controlled for dependence in basic activity of daily living on admission (P < 0.01).

Patients with more symptoms of depression were also more likely to remain or become dependent in three or more instrumental activities of daily living, to remain or become dissatisfied with life, or to remain in or develop fair or poor global health (data on instrumental activities of daily living are shown in Figure 3; other data are not shown but are available from the authors on request).

View larger version (32K): [in this window] [in a new window]   Figure 3. Frequency of dependence in instrumental activities of daily living at discharge and 30 and 90 days after discharge, according to the number of symptoms of depression on admission. To reduce confounding between dependence in instrumental activities of daily living on admission and subsequent dependence, patients were stratified by whether they were dependent in three or more activities on admission. Of the 359 eligible patients who survived for 90 days, the 23 patients for whom data on dependence in instrumental activities of daily living were not available for all three time points were excluded from this analysis. The eligible sample size is smaller for this analysis than for the analysis of basic activities of daily living because data on the former at admission were collected only during the latter part of the study period. Top. Frequency of dependence in three or more instrumental activities of daily living in 116 patients who were dependent in fewer than three activities on admission. Bottom. Frequency of dependence in three or more instrumental activities of daily living in 220 patients who were dependent in at least three activities on admission. At each of the three time points, the association between symptoms of depression on admission and dependence in instrumental activities of daily living was significant in stratified analyses that controlled for dependence on admission (P < 0.01).

In multivariable logistic regression models controlling for potentially confounding factors, such as severity of acute illness, chronic comorbid conditions, and admission health status (Table 3), patients with six or more symptoms of depression on admission were consistently more likely to be dependent in basic and instrumental activities of daily living, have less satisfaction with life, and have worse global health at discharge and 30 and 90 days after discharge. In these models, patients with three to five symptoms of depression were also consistently more likely to have worse health status at discharge and thereafter than were patients with zero to two symptoms. In general, however, these differences were not statistically significant.

Ninety days after discharge, mortality rates in patients with zero to two, three to five, and six or more symptoms of depression were 10%, 12%, and 14%, respectively (P > 0.2 [odds ratio comparing patients who had six or more symptoms with those who had zero to two symptoms was 1.36 {CI, 0.73 to 2.55}]). Although these differences are not statistically significant, our study did not have sufficient power to detect clinically important differences in mortality.

Depression scores at discharge were obtained for 349 patients. Although mean depression scores decreased slightly between admission and discharge (from 4.5 to 3.7 symptoms), admission and discharge depression scores were strongly correlated (r = 0.64). Of patients in the upper tertile of symptoms of depression on admission, 69% remained in the upper tertile on discharge and 85% had scores greater than the median.

Discussion

Top Methods Results Discussion Author & Article Info References

We emphasize two findings from this prospective longitudinal study of older patients hospitalized because of acute medical illness. First, symptoms of depression on admission were associated with worse health status on admission in four complementary domains, including dependence in basic activities of daily living. In multivariable analyses, these associations were independent of several potentially confounding factors, including the physiologic severity of illness and comorbidity.

Second, after controlling for relevant measures of health status on admission, as well as the physiologic severity of illness and comorbidity, we found that symptoms of depression on admission were associated with worse health status at discharge and thereafter. Thus, the poor functional outcomes seen in patients with more symptoms of depression are probably not attributable to more severe somatic illness or to worse health status on admission.

Symptoms of depression identified a particularly vulnerable group of hospitalized older patients. These patients began their hospitalization with worse health status and were at higher risk for further deterioration in health status. These results are consistent with the general hypothesis that symptoms of depression on hospital admission are associated with poor subsequent health status outcomes and that depressed patients are more likely to not recover health status during hospitalization. This hypothesis is supported by the temporal priority of the relation (symptoms of depression precede the poor health status outcomes), the magnitude of the associations, the persistence of the association after controlling for health status and severity of illness at admission, and the consistency of the associations in multivariable models for different measures of health status at several time points. Evidence also suggested a dose-response relation. Although differences in rates of adverse health status outcomes between patients with zero to two symptoms of depression and patients with three to five symptoms generally did not reach conventional definitions of statistical significance, a strong and consistent trend toward worse outcomes was seen in patients with three to five symptoms.

Our results support and extend other research findings that suggest a strong relation between symptoms of depression and poor health status [7-22]. Much of this work was done in community-based cohorts and did not control for baseline health status [7-13]. Some prospective studies were limited to older patients who had excellent baseline health [17]. Other research suggests that symptoms of depression are predictors of adverse outcomes in such groups as institutionalized older persons and patients hospitalized because of myocardial infarction or hip fracture [18-20]. We are aware of only one other investigation, in a smaller cohort, that showed a prospective relation between symptoms of depression and health status outcomes in general medical patients [22]. Thus, our study extends previous findings on the relation between depression and health outcomes to a broad group of hospitalized older medical patients, a group at high risk for functional deterioration. Our results suggest that symptoms of depression are important prognostic markers and, as such, deserve the attention of clinicians even in the presence of competing acute illness and substantial medical comorbidity.

Strengths of our study include a diverse sample of hospitalized medical patients, the use of validated methods to control for severity of illness and comorbidity, a longitudinal design, and adjustment for health status at admission. The latter two methods provide more compelling evidence of an association between symptoms of depression and poor health status outcomes than do cross-sectional studies or longitudinal studies that do not control for baseline health status.

Limitations in our methods should also be recognized. First, measures of health status were based on patient reports; it is possible that patients with symptoms of depression systematically report their function to be worse than it actually is. Nonetheless, such a reporting bias, if stable over time, could not explain the longitudinal relation observed between symptoms of depression on admission and subsequent health status. Because the postadmission analyses adjusted for the self-report of health status at admission, reporting bias would have had to increase markedly between hospital admission and discharge to explain our results. Second, the relation between symptoms of depression and health status may have been confounded by unmeasured factors, such as social support and nutritional status, both of which are associated with depression and poor outcomes [33, 34].

Third, some of the most ill hospitalized patients were not included in our study because we could not directly measure symptoms of depression in patients whose cognitive or medical status precluded interview. It is not clear whether our results can be generalized beyond community-dwelling patients who could be interviewed. Fourth, in some cases, we obtained functional outcome data from surrogates. This may have biased our results because surrogates sometimes report levels of function that are different from those the patients would report [35]. However, because our results were similar after we excluded surrogate-provided data, this bias is probably minimal. Fifth, we divided scores for symptoms of depression into tertiles and dichotomized most of our outcome measures. The association between symptoms of depression and function persisted when we analyzed each of these variables as continuous variables; this suggests that our general conclusions are not affected by our categorizations. However, the sizes of the effects probably differ with different categorizations.

Sixth, we began collecting data in 1990. Changes in inpatient care over the ensuing years may have affected some of our conclusions. Finally, we have no data on symptoms of depression 30 and 90 days after discharge. We therefore do not know whether the narrowing of differences in health status outcomes between patients with more symptoms of depression and patients with fewer symptoms was accompanied by a narrowing of differences in types or numbers of symptoms of depression.

Our results show that a simple, easily administered depression symptom inventory identifies older patients at risk for poor health status outcomes. However, because we did not conduct structured psychiatric interviews with patients, we do not know how many patients had major depression or another depressive disorder, as defined by standard diagnostic criteria [36]. Thus, we do not know how much of the poor health status associated with symptoms of depression is attributable to major depression and how much is attributable to subthreshold depression. It is possible that many of the patients with poor outcomes would not have been recognized as having a depressive illness by standard diagnostic criteria, despite the presence of clinically significant symptoms of depression. This issue is important because evidence from outpatient populations suggests that major depression accounts for less than half of the community-wide morbidity caused by symptoms of depression [10]. Further, it has been suggested that use of the standard criteria for major depression may lead to the underdiagnosis of clinically important depressive disorders in older persons [37, 38]. Thus, it is likely that symptoms of depression, even in the absence of major depression, predict and possibly promote worse outcomes in hospitalized older patients.

The relation between symptoms of depression and declining health status is probably complex and reciprocal [4]. Although our results show a strong association between symptoms of depression and subsequent declines in health status, they do not explain the mechanism of such a relation. Depression may contribute to poor physical health through biological mechanisms, perhaps mediated by a link between mood and immunologic function [4, 39]. Psychological illness may also reduce higher-order capacities, including motivation, energy, self-efficacy, concentration, regulation of affect, and interpersonal skills, thereby reducing a person's ability to adapt to physical impairments or physical symptoms [9, 17, 40].

Our results show that screening hospitalized older patients for symptoms of depression can identify a particularly vulnerable group of patients at risk for undesirable health status outcomes. Physicians should consider routinely screening older hospitalized patients for symptoms of depression as a method of identifying these vulnerable patients. It is unknown, however, whether identification and treatment of these symptoms in hospitalized older patients will prevent functional decline and restore health status. Because symptoms of depression, loss of functional independence, and declining health status are common among acutely ill hospitalized older patients [1-341, 42], determining whether treatment of depression will improve health status outcomes in these patients is an important priority.

From Case Western Reserve University, University Hospitals of Cleveland, and the Cleveland Veterans Affairs Medical Center, Cleveland, Ohio.

Dr. Palmer: Cleveland Clinic Foundation, 9500 Euclid Avenue, Desk A72, Cleveland, OH 44195.

Ms. Kresevic: University Hospitals of Cleveland, 11100 Euclid Avenue, Cleveland, OH 44106.

Dr. Landefeld: Cleveland Veterans Affairs Medical Center, 10701 East Boulevard, Cleveland, OH 44106.