Abstract :
PURPOSE: Together, the sophisticated computerized information systems and defined, stable populations within Health Maintenance Organizations (HMOs) can reduce imprecision associated with modeling disease incidence and prevalence. Such data are useful for resource allocation, intervention planning, epidemiological study and surveillance. Identifying all cases of illness within such a defined population is vital to these aims, yet detailed diagnostic information is lacking in many pre-paid plans. In this presentation, diagnostic and prescription data are compared for their ability to assess the treatment prevalence for 28 chronic medical conditions.
METHODS: Group Health Cooperative is a mixed model HMO in Western WA state. Automated records for calendar year 1999 were examined for 182,174 commercially insured patients age 20–64. Prescription drug data and International Classification of Disease—9th Clinical Modification codes (ICD-9CM) were used to identify 28 chronic medical conditions among patients. Chi square tests were performed to compare results produced by the two methods.
RESULTS: The diagnostic method worked best for conditions with no clear pharmacologic treatment. However, the pharmacy-based method better identified patients with chronic illnesses initially diagnosed prior to 1999, and with disorders for which standard diagnostic codes are inconsistently utilized, such as mental health conditions.
CONCLUSIONS: While both diagnostic and pharmacy-based methods perform well in identifying a limited set of conditions, each performs poorly under certain circumstances. A hybrid method utilizing both diagnostic and pharmacy data may be more appropriate for an HMO setting. Additional research is needed to develop and refine such a method.
