A claims data approach to defining an episode of care.

OBJECTIVE: To utilize health services research techniques in developing an episode of care using an administrative data set. This method is demonstrated for an episodic clinical condition, migraine. DATA SOURCES: Medicaid administrative data set of 3,372 patients with a diagnosis of migraine (ICD-9-CM 346.0, 346.1) in the state of Pennsylvania between May 1990 and March 1992. STUDY DESIGN: The duration of a migraine episode was measured by assessing the magnitude of resource utilization and the proportion of patients with charges in the period after the index migraine as compared to the period before the index migraine. A confidence interval (CI) was developed around each measure using bootstrap techniques. DATA COLLECTION METHODS: All charge data were extracted daily for a 113-day observation period surrounding each index migraine in order to observe the duration of impact of a migraine diagnosis on resource utilization. PRINCIPAL FINDINGS: The lower limits of both the 95% and 99% CIs for the difference in charges are greater than 0 for three weeks. The lower limits of both CIs for the difference in the proportion of patients with charges are above 0 for six weeks. CONCLUSIONS: Our analysis demonstrates that a health services research framework can be used to define an episode of care for a chronic disease category such as migraine. This method can be used to evaluate episodes of care for clinical studies of limited or episodic conditions and to complement clinical expertise in developing time horizons for clinical trials.