Modeling risk of gastrointestinal events among Medicaid NSAID users using propensity scores

The purpose of the study is to evaluate the gastrointestinal risk of nonsteroidal anti-inflammatory drugs compared with cyclooxygenase-2 inhibitors in a Medicaid managed care population. Medical and prescription claims were analyzed for all Medicaid-enrollees aged 18 years and older, who received a cyclooxygenase-2 or other prescription nonsteroidal anti-inflammatory drug between January 6, 2000 and January 6, 2002, and who did not use these drugs for at least 6 months prior. A logistic model was developed of the propensity for either treatment and stratified patients by quintiles of their propensity score, adjusting for demographics, indications for nonsteroidal anti-inflammatory drugs and gastrointestinal risks. The rates of gastrointestinal event (ICD-9 codes 531-534,578) among nonsteroidal anti-inflammatory drug and cyclooxygenase-2 users were compared. The model was adjusted for drug exposure which was calculated using the medication possession ratio. Of the total sample, 73% were female, 43% were Caucasian and 29% were older than 50 years. Both the direct and the propensity adjusted model, controlling for gastroprotective agents, and medication possession ratio showed significantly different rates of gastrointestinal events in nonsteroidal anti-inflammatory drug users as compared with cyclooxygenase-2 users (odds ratio = 1.874, 95% confidence interval 1.056, 3.326) and (odds ratio = 2.088, 95% confidence interval 1.061, 4.110) respectively. A significant difference in gastrointestinal event rates was found among patients in this Medicaid population on nonsteroidal anti-inflammatory drug versus cyclooxygenase-2 inhibitors, when the gastroprotective agent use and the medication possession ratio were controlled for. Misclassification bias was accounted for by adjusting for length of drug exposure.