Session number: 1148
Abstract title: Prevalence of Depression After Stroke: What You Find Depends on Where You Look
Author(s):
LS Williams - Roudebush VAMC HSR&D
S Ghose - Substance Abuse Mental Health Services Administration
RW Swindle - Outcomes Research, USMD, Eli Lilly, Inc.
Department of Psychology, Indiana University, Bloomington
Objectives: Depression occurs in 20-40% of stroke survivors and has been linked to worse functional recovery and increased mortality. Administrative databases may be useful to examine the relationships between depression and other discrete outcome events but are likely to underestimate the prevalence of depression. We sought to examine various VA data sources to determine which sources maximized case ascertainment of post-stroke depression (PSD).
Methods: We examined three databases: 1) a national VA database containing inpatient records and diagnoses of patients hospitalized with ischemic stroke between 1990-1997 (N = 56,093), 2) a local VA VISTA database with inpatient data, outpatient diagnoses, and pharmacy data of patients hospitalized with ischemic stroke (N = 525), and 3) a prospective database of veterans with stroke who received systematic depression screening (N = 139). In the national database, ischemic stroke was defined by primary position discharge diagnosis ICD-9 code 434 or 436 and depression was defined using standard discharge ICD-9 codes in any position. In the local VISTA database, depression was identified by examining outpatient diagnosis lists and pharmacy data. Patients on therapeutic doses of antidepressant medications with no other documented reason for antidepressants were counted as depressed. Veterans in the prospective database were considered depressed if their CES-D score was > 16.
Results: Only 2,405 (4.3%) of the national sample had an ICD-9 diagnosis of depression after stroke. PSD prevalence was similar in the local database when only inpatient ICD-9 diagnoses were examined (21 cases, 4%). Examination of local electronic VISTA data added 25 additional cases. 23 of these were identified from outpatient clinic diagnoses and only 2 solely from pharmacy data. PSD prevalence based on prospective screening was highest at 63 cases (45% of the prospective sample). Even if a CES-D cutpoint of 20 was used, 41 (30%) of patients screened positive for depression.
Conclusions: PSD is common in veterans surviving ischemic stroke. Using inpatient ICD-9 codes for case ascertainment significantly underestimates the prevalence of PSD. Use of outpatient diagnoses in VISTA maximizes electronic case ascertainment, but still misses many veterans with PSD.
Impact statement: Post-stroke depression affects 20-40% of veterans after ischemic stroke. Researchers wishing to identify cases of post-stroke depression to study disease risk factors or patient outcomes should use outpatient DHCP diagnoses to maximize case ascertainment. Even with this strategy, however, as many as half of the cases may be missed.