Validating Medicare Claims Data For Stroke Interview with
Hiraku Kumamar, MD, MPH
Department of Epidemiology
Harvard School of Public Health, Boston, MA and
Soko Setoguchi-Iwata, M.D.
Duke Clinical Research Institute
Durham, NC 27715

Medical Research: What are the main findings of the study?

Answer: We evaluated the accuracy of discharge diagnosis of stroke in the Medicare claims database by linking it to a nationwide epidemiological study cohort with 30239 participants called REasons for Geographic And Racial Differences in Stroke (REGARDS). We found that among the 282 events captured using a strict claims definition of stroke, 91% were true events.  We also found that 12% of the overall strokes had been identified only by Medicare claims, strongly supporting the use of these readily available data for event follow-up in cohort studies.

Medical Research: Were any of the findings unexpected?

Answer: As more and more researches are being conducted using Medicare claims data, it was important to up-date the accuracy of stroke diagnosis as diagnostic modalities and coding practices change over time. The study was innovative in that it was conducted by linking the Medicare claims to national cohort of general subjects living in the community, not to patients with particular diseases like many of the previous studies.

Medical Research: What should clinicians and patients take away from your report?

Answer: Clinicians should know that when searching and critically appraising evidence from observational studies using stroke outcome captured by Medicare inpatient data, the results from studies comparing two or more therapies will be valid due to the high specificity of Medicare stroke diagnosis. However, studies that report incidence and prevalence rates of strokes using Medicare inpatient data will not be as reliable due to low sensitivity of inpatient Medicare diagnosis of stroke.

Medical Research: What recommendations do you have for future research as a result of this study?

Answer: More research is needed to explore the methods that maximize the accuracy of diagnoses recorded in these administrative databases which is especially useful for studies that looks for real-world evidence in the effectiveness and safety of medical products.  For stroke, future researches should try to identify definitions that preserve similar specificity to our definition but capture more cases (more sensitive), so that we can use them to accurately capture the frequencies of strokes in the Medicare population.

Citation :
Validity of Claims-Based Stroke Algorithms in Contemporary Medicare Data: Reasons for Geographic and Racial Differences in Stroke (REGARDS) Study Linked With Medicare Claims
Hiraku Kumamaru, Suzanne E. Judd, Jeffrey R. Curtis,Rekha Ramachandran, N. Chantelle Hardy, J. David Rhodes, Monika M. Safford, Brett M. Kissela, George Howard, Jessica J. Jalbert, Thomas G. Brott, and Soko Setoguchi

Circ Cardiovasc Qual Outcomes. 2014;CIRCOUTCOMES.113.000743published online before print June 24 2014, doi:10.1161/CIRCOUTCOMES.113.000743

Last Updated on July 4, 2014 by Marie Benz MD FAAD