Misclassification of incident conditions using claims data: impact of varying the period used to exclude pre-existing disease

BACKGROUND: Estimating the incidence of medical conditions using claims data often requires constructing a prevalence period that predates an event of interest, for instance the diagnosis of cancer, to exclude those with pre-existing conditions from the incidence risk set. Those conditions missed during the prevalence period may be misclassified as incident conditions (false positives) after the event of interest. Using Medicare claims, we examined the impact of selecting shorter versus longer prevalence periods on the incidence and misclassification of 12 relatively common conditions in older persons. METHODS: The source of data for this study was the National Cancer Institute’s Surveillance, Epidemiology, and End Results cancer registry linked to Medicare claims. Two cohorts of women were included: 33,731 diagnosed with breast cancer between 2000 and 2002, who had ≥ 36 months of Medicare eligibility prior to cancer, the event of interest; and 101,649 without cancer meeting the same Medicare eligibility criterion. Cancer patients were followed from 36 months before cancer diagnosis (prevalence period) up to 3 months after diagnosis (incidence period). Non-cancer patients were followed for up to 39 months after the beginning of Medicare eligibility. A sham date was inserted after 36 months to separate the prevalence and incidence periods. Using 36 months as the gold standard, the prevalence period was then shortened in 6-month increments to examine the impact on the number of conditions first detected during the incidence period. RESULTS: In the breast cancer cohort, shortening the prevalence period from 36 to 6 months increased the incidence rates (per 1,000 patients) of all conditions; for example: hypertension 196 to 243; diabetes 34 to 76; chronic obstructive pulmonary disease 29 to 46; osteoarthritis 27 to 36; congestive heart failure 20 to 36; osteoporosis 22 to 29; and cerebrovascular disease 13 to 21. Shortening the prevalence period has less impact on those without cancer. CONCLUSIONS: Selecting a short prevalence period to rule out pre-existing conditions can, through misclassification, substantially inflate estimates of incident conditions. In incidence studies based on Medicare claims, selecting a prevalence period of ≥24 months balances the need to exclude pre-existing conditions with retaining the largest possible cohort.