several sources of real world data available to researchers. METHODS: We compare and contrast the pros and cons of data available from administrative (payment) databases, electronic medical record (EMR) databases, and surveys. RESULTS: Administrative claims databases provide fully-integrated, all-encounter patient data on diagnoses, procedures, and payments. However, data quality varies depending upon whether particular fields are required for provider payment. Data on lab and test values are typically lacking. Prescriptions that are written, but not filled by the patient, are usually not captured. Medical record data overlap, to a certain extent, with administrative data. While information on payments for services may not be included, detailed information on test results and lab values are usually captured in the EMR. Data are included on written prescriptions, but the researcher will not know whether the prescription was filled by the patient. Depending upon the clinical system covered, only some encounters (e.g., ambulatory care in the outpatient setting) may be available. Both administrative and EMR data hold the potential to provide longitudinal patient information that is not subject to recall or social desirability biases that often affect survey data. However, information on satisfaction with care, quality of life, activities of daily living, and many other metrics, may only be captured with survey data. CONCLUSIONS: Several sources of rich, longitudinal patient data are available to provide real world evidence on drug effectiveness and cost. In some cases, data may be combined to overcome limitations of a single source. With care, data may be found that will produce generalizable findings for the population of interest.

subgroup of patients (Nϭ473) eligible for evaluation under Method II was also evaluated under Method I, to assess differences in results compared with the overall population. Method I yielded higher dose estimates when compared to Method II. This was consistent for the total sample and subgroup. Ustekinumab cost per injection in Method I was estimated up to 68.2% higher than in Method II. Variability in cost estimates across doses was up to 18 times higher in Method I than Method II. CONCLUSIONS: Conducting ustekinumab drug utilization assessments with pharmacy claims requires a methodological adjustment to address multiple doses dispensed on the same claim. Unadjusted assessment may yield artificially high dose and cost estimates. Adjusted assessments offer a more realistic distribution of dose and less variability in cost estimates. The ability to conduct adjusted assessments requires the sample to have Ն 2 ustekinumab prescriptions and sufficient follow-up time.