Exploring the methodological challenges of investigating comparison groups with different underlying characteristics: a case study

To examine challenges in forming accurate conclusions from claims data when comparing different patient groups, and to stress the need to adjust for confounding variables through application of statistical methods.Patients who were 60 or older and began amlodipine or felodipine in 1997 or 1998 were identified from a proprietary managed care claims database. Patients had 6 months of medical and outpatient pharmacy benefits coverage before and after initiating therapy. They were stratified on varying degrees of hypertension severity and on all-condition clinical complexity, and quantified with a proprietary Burden of Illness (BOI) score, defined as the expected resource use for the patient.s combined conditions during the 6 months before therapy.Of the 17,667 patients, 10,469 (59.3%) were female. The 12,389 amlodipine users were similar in age to felodipine users (n=5,278), 73 versus 74, respectively. More amlodipine users were assigned to the 2 most clinically complex quintiles of the BOI distribution (45.5% to 26.9%), with 24.3% of patients in the top quintile (9.7% for felodipine) and twice the prevalence of ischemic heart disease/angina (42.4% versus 19.4%). High-severity hypertension was found among 33.6% of amlodipine users, contrasted with 17.9% for felodipine. A significant difference in the mean average daily dose (ADD) between the amlodipine (5.93 mg) and felodipine (5.8 mg) groups (P=0.0007) was noted.These results stress the importance of assessing the impact of potentially confounding variables and the need to address differing patient characteristics through either risk adjustment or stratification in making head-to-head comparisons.