Variations in the associations between psychiatric comorbidity and hospital mortality according to the method of identifying psychiatric diagnoses

OBJECTIVE: Little is known about associations between psychiatric comorbidity and hospital mortality for acute medical conditions. This study examined if associations varied according to the method of identifying psychiatric comorbidity and agreement between the different methods. PATIENTS/PARTICIPANTS: The sample included 31,218 consecutive admissions to 168 Veterans Affairs facilities in 2004 with a principle diagnosis of congestive heart failure (CHF) or pneumonia. Psychiatric comorbidity was identified by: (1) secondary diagnosis codes from index admission, (2) prior outpatient diagnosis codes, (3) and prior mental health clinic visits. Generalized estimating equations (GEE) adjusted in-hospital mortality for demographics, comorbidity, and severity of illness, as measured by laboratory data. MEASUREMENTS AND MAIN RESULTS: Rates of psychiatric comorbidities were 9.0% using inpatient diagnosis codes, 27.4% using outpatient diagnosis codes, and 31.0% using mental health visits for CHF and 14.5%, 33.1%, and 34.1%, respectively, for pneumonia. Agreement was highest for outpatient codes and mental health visits (kappa=0.51 for pneumonia and 0.50 for CHF). In GEE analyses, the adjusted odds of death for patients with psychiatric comorbidity were lower when such comorbidity was identified by mental health visits for both pneumonia (odds ratio [OR]=0.85; P=.009) and CHF (OR=0.70; P=.001) and by inpatient diagnosis for pneumonia (OR=0.63; P <=.001) but not for CHF (OR=0.75; P=.128). The odds of death were similar (P>.2) for psychiatric comorbidity as identified by outpatient codes for pneumonia (OR=1.04) and CHF (OR=0.93). CONCLUSIONS: The method used to identify psychiatric comorbidities in acute medical populations has a strong influence on the rates of identification and the associations between psychiatric illnesses with hospital mortality.