Effects of diagnostic inclusion criteria on prevalence and population characteristics in database research

Studies of serious mental illnesses that use administrative databases have employed various criteria to establish diagnoses of interest. Several studies have assessed the validity of diagnostic inclusion criteria against research diagnoses. However, no studies have examined the effect of diagnostic inclusion criteria on prevalence and population characteristics across such groups.

Administrative data for 2003-2010 from the Department of Veterans Affairs were used to calculate prevalence rates and assess effects of varying the diagnostic inclusion criteria on population composition for bipolar disorder, schizophrenia, and posttraumatic stress disorder (PTSD). Specifically, for each diagnosis, mutually exclusive subpopulations were compared on the basis of the following inclusion criteria for a given diagnosis: one treatment encounter, two outpatient encounters or one inpatient encounter, and any two encounters. For bipolar disorder and schizophrenia, effects of excluding individuals who had a competing diagnosis of, respectively, schizophrenia or bipolar disorder in the prior 12 months and since 2002 were also determined.

In 2010, moving from the broadest definitions of bipolar disorder (N=120,382), schizophrenia (N=91,977), and PTSD (N=554,028) to the most restrictive definitions reduced prevalence rates by, respectively, 28.7%, 34.9%, and 25.7%, with temporal trends for 2003-2010 paralleling results in 2010. Population composition changes with changing diagnostic inclusion criteria were variable, with predominantly small odds ratios.

Population composition was relatively robust across common diagnostic inclusion criteria for each condition. Thus choice of criteria can focus on considerations of diagnostic validity and case-finding needs. Three mechanisms for the impact of diagnostic criteria on population composition in administrative data sets are discussed.