In our current healthcare system, pressure to provide cost- and time-efficient care is immense. Inpatient care often focuses on assessing the patient’s presenting illness or injury and treating that condition in a manner that gets the patient on their feet and out of the hospital quickly. Because depression is not an indication for hospitalization so long as active suicidality is absent, inpatient physicians may view it as a problem best managed in the outpatient setting. Yet both psychosocial and physical factors associated with depression put patients at risk for rehospitalization.1 Furthermore, hospitalization represents an unrecognized opportunity to optimize both mental and physical health outcomes.2
Indeed, poor physical and mental health often occur together. Depressed inpatients have poorer outcomes, increased length of stay, and greater vulnerability to hospital readmission.3,4 Among elderly hospitalized patients, depression is particularly common, especially in those with poor physical health, alcoholism,5 hip fracture, and stroke.6 Yet little is known about how often depression goes unrecognized, undiagnosed, and, therefore, untreated.
The US Preventive Services Task Force (USPSTF) recommends screening for depression in the general adult population, including pregnant and postpartum women, and further suggests that screening should be implemented “with adequate systems in place to ensure accurate diagnosis, effective treatment, and appropriate follow-up.”2 The USPSTF guidelines do not distinguish between inpatient and outpatient settings. However, the preponderance of evidence for screening comes from outpatient care settings, and little is known about screening among inpatient populations.7
This study had 2 objectives. First, we sought to examine the performance of depression screening tools in inpatient settings. If depression screening were to become routine in hospital settings, screening tools would need to be sensitive and specific as well as brief and suitable for self-administration by patients or for administration by nurses, resident physicians, or hospitalists. It is also important to consider administration by mental health professionals, who may be best trained to administer such tests. We, therefore, examined 3 types of studies: (1) studies that tested a self-administered screening instrument, (2) studies that tested screening by individuals without formal training, and (3) studies that compared screening tools administered by mental health professionals. Second, we sought to describe associations between depression and clinical or utilization outcomes among hospitalized patients.
We adhered to recommendations in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement,8,9 including designing the analysis before performing the review. However, we did not post a protocol in an online registry, formally assess study quality, or perform a meta-analysis.
Data Sources and Searches
We searched PsycINFO and PubMed databases for articles published between 1990 and 2016 (as of July 31, 2016). In PubMed, 2 search term strings were used to capture studies of depression screening tools in inpatient settings. The first used the advanced search option to exclude studies related to primary care settings or children and adolescents, and the second used MeSH terms to ensure that a wide variety of studies were included. Specific search terms are included in the Appendix. A similar search was conducted in the PsycINFO database and these search terms are also included in the Appendix.
Articles were eligible if they were published in English in peer-reviewed journals, included at least 20 adults hospitalized for nonpsychiatric reasons, and described the use of at least 1 measure of depression. The studies must have either tested the validity of a depression screening tool or examined the association between depression screening and clinical or utilization outcomes. Two investigators reviewed each title, abstract, and full-text article to determine eligibility, then reached a consensus on which studies to include in this review.
Two investigators reviewed each full-text article to extract information related to study design, population, and outcomes regarding screening tool analysis or clinical results. From articles that assessed the performance of depression screening tools, we extracted information related to the nature and application of the index test, the nature and application of the reference test, the prevalence of depression, and the sensitivity and specificity of the index test compared with the reference test. For articles that focused on the association between depression screening and clinical or utilization outcomes, the data on relevant clinical outcomes included symptom severity, quality of life, and daily functioning, whereas the data on utilization outcomes included length of stay, readmission, and the cost of care.
Altogether, the search identified 3226 records. After eliminating duplicates and abstracts not suitable for inclusion (Figure), 101 articles underwent full-text review and 32 were found to be eligible. Of these, 12 focused on the association between depression and clinical or utilization outcomes, while 20 assessed the performance of depression screening tools.
Depression Screening Tools
Table 1 describes the index and reference instruments as well as methods of administration, the prevalence of depression, and the sensitivity and specificity of the index instruments relative to the reference instruments. Across the 20 studies, the prevalence of depression ranged from 15% to 60%, with a median of 34%.10–29 This finding may reflect different methods of screening or variation among diverse hospitalized populations. Many of the studies excluded patients with cognitive impairment or communication barriers.
The included studies tested a wide range of unique instruments, and compared them with diverse reference standards. Five studies examined instruments that were self-administered by patients10–14; 9 studies assessed instruments administered by nurses, physicians, or research staff members without formal psychiatric training15–23; and 6 studies evaluated instruments administered by mental health professionals.24–29 Four studies compared different instruments that were administered in the same manner (eg, both self-administered by patients).12–14,22 In the remaining studies, both instruments and methods of administration differed between the index and reference conditions.
Eight studies tested brief instruments with 5 or fewer items, most of which exhibited good sensitivity (range 38%–91%) and specificity (range 68%–86%) relative to longer instruments.12,14–19,22 In 2 of these studies, instruments were self-administered. In 1 case, a single self-administered item from the STOP-D instrument (“Over the past 2 weeks, how much have you been bothered by feeling sad, down, or uninterested in life?”) performed nearly as well as the 14-item Hospital Anxiety and Depression Scale.14 In the other 6 studies testing brief instruments, the instruments were administered by individuals without formal training.15–19,22 In 1 such study, geriatricians asking 2 questions about depressed mood and anhedonia performed well compared with a formal psychiatric interview.17
Four studies tested variations of the Geriatric Depression Scale (GDS).12,18,21,23 In 3 of these studies, abbreviated versions of the GDS exhibited relatively high sensitivity and specificity.12,18,21 However, a study comparing the 15-item GDS (GDS-15) with the GDS-4 found that GDS-15 correctly classified 10% more patients with suspected depression.12 Two studies examined variations of the Patient Health Questionnaire (PHQ). One study found that both the PHQ-2 and PHQ-9 obtained by staff nurses performed well relative to a comprehensive assessment by a trained advanced practice nurse.13,19
When reported, positive predictive value, negative predictive value, and area under the receiver-operator curve were generally high.
Depression and Clinical or Utilization Outcomes
Of the 12 studies that reported either clinical or utilization outcomes for depression screening in an inpatient setting,4,30–40 3 measured rates of rehospitalization.4,31,39 The other 9 studies tested for associations between symptoms of depression and either health or treatment outcomes. Table 2 provides a more detailed description of the study designs and results.
Other studies found that depression was associated with reduced functional abilities such as mobility and self-care,30,32–34 and increased hospital readmission31 as well as physical and mental health deficits.37 Interestingly, although 1 study did not find that depression and hospital readmission were closely linked (frequency at 19%), it found that comorbid illness and previous hospitalizations predicted readmission.4
We also evaluated the associations between depression diagnosed in the inpatient studies and 2 types of outcomes. The first type includes clinical outcomes including symptom severity, quality of life, and daily functioning. Most studies we identified assessed clinical outcomes, and all detected an association between depression and worse clinical outcomes. The second type includes healthcare utilization, which can be measured with the patients’ length of hospital stay, readmission and cost of care. In 1 such study, Mitchell aet al.31 reported a 54% increase in readmission within 30 days of discharge among patients who screened positive for depression.31 Additionally, Cully et al.30 found that depression may impinge on the recovery process of acute rehabilitation patients.
The purpose of this study was to describe the feasibility and performance of depression screening tools in inpatient medical settings, as well as associations between depression diagnosed in the inpatient setting and clinical and utilization outcomes. The median rate at which depression was detected among inpatients was 33%, ranging from 5% to 60%. Studies from several individual hospitals indicated that depression can be associated with higher healthcare utilization, including return to the hospital after discharge, as well as worse clinical outcomes. To detect undiagnosed depression among inpatients, screening appears feasible. Depression screening instruments generally exhibited good sensitivity and specificity relative to comprehensive clinical evaluations by mental health professionals. Furthermore, several self-administered and brief instruments had good performance. Prior authors have reported that screening for depression among inpatients may not be particularly burdensome to patients or staff members.41
The studies we reviewed used diverse screening instruments. Further research is needed to determine which tools are preferable in which patient populations, and to confirm that brief instruments are adequate for screening. The GDS is widely used, and many patients hospitalized in the United States fall into the geriatric group. The PHQ has been validated for self-administration and is widely used among outpatients42; it may be more suitable for younger populations. We found that several abbreviated versions of these and other screening instruments have exhibited good sensitivity and specificity among inpatients. However, many of the studies excluded patients with cognitive impairment or communication barriers. For individuals with auditory impairment, the Brief Assessment Schedule Depression Cards (BASDEC) might be an option. Used in 2 studies, the BASDEC involves showing patients a deck of 19 easy-to-read cards. The time required to administer the BASDEC is modest.15,23 Sets of smiley face diagrams might also be suitable for some patients with communication barriers or cognitive impairment. An ineligible study among stroke survivors found that selecting a sad face had a sensitivity of 76% and specificity of 77% relative to a formal diagnostic evaluation for depression.43
In considering the instruments that may be most suitable for inpatients, the role of somatic symptoms is also important because these can overlap between depression and the medical conditions that lead to hospitalization.44–46 Prior investigators found, for example, that 47% of Beck Depression Inventory (BDI) scores were attributable to somatic symptoms among patients hospitalized after myocardial infarction, whereas 37% of BDI scores were attributable to somatic symptoms among depressed outpatients.47 Future research is needed to determine the significance of somatic symptoms among inpatients, including whether they should be considered during screening, add prognostic value, or warrant specific treatment. In addition, although positive and negative predictive values were generally high among the screening instruments we evaluated, confirming the diagnosis of depression with a thorough clinical assessment is likely to be necessary.44,45
Despite the high prevalence of depression, associations with suboptimal outcomes, and the good performance of screening tools to date, screening for depression in the inpatient setting has received little attention. Prior authors have questioned whether hospital-based screening is an efficient and effective way to detect depression, and have raised valid concerns regarding false-positive diagnoses and unnecessary treatment, as well as a lack of randomized controlled trials.7,48,49 Whereas some studies suggest that depression is associated with greater healthcare utilization,3,4 little information exists regarding whether screening during hospitalization and treating previously undiagnosed depression improves clinical outcomes or reduces healthcare utilization.
Several important questions remain. What is the pathophysiology of depressed mood during hospitalization? How often does depressed mood during hospitalization reflect longstanding undiagnosed depression, longstanding undertreated depression, an acute stress disorder, or a normal if unpleasant short-term reaction to the stress of acute illnesses? Do the manifestations and effects of depressed mood differ among these situations? What is the prognosis of depressed mood occurring during hospitalization, and how many patients continue to have depression after recovery from acute illness; what factors affect prognosis? In a small sample of hospitalized patients, nearly 50% of those who had been depressed at intake remained depressed 1 month after discharge.50 Given that most antidepressant medications have to be taken for several weeks before effects can be detected, what, if any, approach to treatment should be taken? More research is needed on the effectiveness and cost-effectiveness of diagnosing and treating depression in the inpatient setting.
This work has several limitations. We found relatively few studies meeting eligibility criteria, particularly studies assessing clinical and utilization outcomes among depressed inpatients. Among the screening tools that were studied in the hospital setting, the highly diverse instruments and modes of administration precluded a quantitative synthesis such as meta-analysis. Prior meta-analyses on specific screening tools have focused on outpatient populations.51–53 Furthermore, we did not evaluate study quality or risk of bias.
In conclusion, screening for depression in the inpatient setting via patient self-assessment or assessment by hospital staff appears feasible. Several brief screening tools are available that have good sensitivity and specificity relative to diagnoses made by mental health professionals. Limited evidence suggests that screening tools for depression may be ready to integrate into inpatient care.41 Yet, although depression appears to be common and associated with worse clinical outcomes and higher healthcare utilization, more research is needed on the benefits, risks, and potential costs of adding depression screening in the inpatient healthcare setting.
The authors report no conflicts of interest.