Related Experiment Video
Updated: Apr 27, 2026
Computerized Adaptive Testing System of Functional Assessment of Stroke
Published on: January 7, 2019
Validity of claims-based stroke algorithms in contemporary Medicare data: reasons for geographic and racial
Hiraku Kumamaru1, Suzanne E Judd1, Jeffrey R Curtis1
1From the Department of Epidemiology, Harvard School of Public Health, Boston, MA (H.K.); Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women's Hospital and Harvard Medical School, Boston, MA (H.K., J.J.J.); Department of Biostatistics (S.E.J., J.D.R., G.H.) and Department of Epidemiology (J.R.C.), University of Alabama at Birmingham School of Public Health; Department of Medicine, University of Alabama at Birmingham School of Medicine (J.R.C., R.R., M.M.S.); Duke Clinical Research Institute, Department of Medicine, Duke University School of Medicine, Durham, NC (N.C.H., S.S.); Department of Neurology, University of Cincinnati, OH (B.M.K.); and Department of Neurology, Mayo Clinic, Jacksonville, FL (T.G.B.).
Insights
Administrative claims accurately identify stroke in Medicare patients, with high positive predictive value and specificity. However, these algorithms have low sensitivity, making them unsuitable for estimating stroke incidence.
Area of Science:
- Epidemiology
- Health Services Research
- Medical Informatics
Background:
- Accuracy of stroke diagnosis in administrative claims for Medicare enrollees is not well-studied.
- This study assesses the validity of diagnostic coding algorithms for stroke identification in the Medicare population.
Purpose of the Study:
- To evaluate the accuracy of claims-based algorithms for identifying stroke in contemporary Medicare enrollees.
- To determine the positive predictive value, specificity, and sensitivity of these algorithms using adjudicated strokes as the gold standard.
Main Methods:
- Linked data from the REasons for Geographic And Racial Differences in Stroke (REGARDS) Study to Medicare claims for fee-for-service enrollees.
- Used adjudicated strokes from medical records as the gold standard to calculate accuracy measures for diagnostic coding algorithms.
- Evaluated algorithms using International Classification of Diseases, Ninth Revision, Clinical Modification codes for acute ischemic and hemorrhagic stroke.
Main Results:
- An algorithm using primary discharge diagnosis codes for stroke had a positive predictive value of 92.6% and specificity of 99.8%.
- An algorithm using only acute ischemic stroke codes showed similar performance, with a positive predictive value of 91.1% and specificity of 99.8%.
- Both algorithms demonstrated low sensitivity (around 59%), indicating they miss a significant proportion of stroke cases.
Conclusions:
- Claims-based algorithms possess high positive predictive value and specificity for identifying stroke in Medicare populations.
- These algorithms are suitable for etiologic and comparative effectiveness studies but not for estimating overall stroke incidence due to low sensitivity.
Background:
The accuracy of stroke diagnosis in administrative claims for a contemporary population of Medicare enrollees has not been studied. We assessed the validity of diagnostic coding algorithms for identifying stroke in the Medicare population by linking data from the REasons for Geographic And Racial Differences in Stroke (REGARDS) Study to Medicare claims.
Methods And Results:
The REGARDS Study enrolled 30 239 participants ≥45 years in the United States between 2003 and 2007. Stroke experts adjudicated suspected strokes, using retrieved medical records. We linked data for participants enrolled in fee-for-service Medicare to claims files from 2003 through 2009. Using adjudicated strokes as the gold standard, we calculated accuracy measures for algorithms to identify incident and recurrent strokes. We linked data for 15 089 participants, among whom 422 participants had adjudicated strokes during follow-up. An algorithm using primary discharge diagnosis codes for acute ischemic or hemorrhagic stroke (International Classification of Diseases, Ninth Revision, Clinical Modification codes: 430, 431, 433.x1, 434.x1, 436) had a positive predictive value of 92.6% (95% confidence interval, 88.8%-96.4%), a specificity of 99.8% (99.6%-99.9%), and a sensitivity of 59.5% (53.8%-65.1%). An algorithm using only acute ischemic stroke codes (433.x1, 434.x1, 436) had a positive predictive value of 91.1% (95% confidence interval, 86.6%-95.5%), a specificity of 99.8% (99.7%-99.9%), and a sensitivity of 58.6% (52.4%-64.7%).
Conclusions:
Claims-based algorithms to identify stroke in a contemporary Medicare cohort had high positive predictive value and specificity, supporting their use as outcomes for etiologic and comparative effectiveness studies in similar populations. These inpatient algorithms are unsuitable for estimating stroke incidence because of low sensitivity.
