Birth cohort confounds effect estimates of guideline changes on prevalence of statin use in diabetic patients

Background: In earlier research we found that birth cohort effects potentially confound effect estimates of guideline changes on statin use. Conventional methods do not take the birth cohort dimension into account. Objectives: We investigated whether including the birth cohort dimension in time series analysis leads to a more accurate estimation of the effect of a guideline change on the trend of statin use in diabetic patients. Methods: Outpatient pharmacy data from a drug prescription database in the Netherlands (IADB.nl) were used to obtain the age- (30-85) and sex-specific number of users of statin per 1,000 diabetic patients (prevalence) per year from 1998 to 2011. The intervention took place in 2002 and targeted persons above age 70. Using likelihood ratio tests we compared an age-specified interrupted time series model which included the birth cohort dimension with one in which birth cohort was excluded. Results: The model which included the birth cohort dimension had a significantly better fit to the data than the model which excluded birth cohort (p <0.001). Both models found significant effects of the guideline change at age-group 70+ years and below 70 years (p <0.05) but effect estimates differ. For ages below 70, the model estimated an increase in statin use among diabetics, but when cohort was included the estimated effect became stronger: 29.5% increase vs. 24% (p <0.05). For ages above 70 an increase was also estimated, but now the model including birth cohort predicted smaller effect: 24.5% vs. 29.5% (p = 0.09). Both models found that, for both age groups, the increase in prevalence of statin use levels off from 2006 onwards (p