1070. Stroke Incidence Among VA Patients with Diabetes
Thomas W Findley, MD, PhD, VA New Jersey Healthcare System and University of Medicine and Dentistry of New Jersey, School of Health Related Professions, M Safford, VA New Jersey Healthcare System and University of Alabama at Birmingham School of Medicine, P Findley,
VA New Jersey Healthcare System and Rutgers University, D Miler,
CHQOER, Bedford VAMC, M Maney,
VA New Jersey Healthcare System, M Rajan,
VA New Jersey Healthcare System, L Pogach,
VA New Jersey Healthcare System and University of Medicine and Dentistry of New Jersey, New Jersey Medical School
Objectives: Estimating hospitalized stroke incidence in at-risk populations would assist primary and secondary prevention efforts and could be accomplished economically using claims data. Although several ICD-9 code-based strategies for identifying stroke have been reported, their operation in different health systems has not been described.
Methods: We obtained VA and Medicare utilization data for the 552,128 veteran VA users identified as having diabetes between October 1, 1997 and September 30, 1998. We calculated hospitalized stroke rates in this population, using four strategies: two developed in the VA, one developed in the private sector, and the approach reported annually by the National Hospital Discharge Survey. To create a ‘best estimate’ of the population hospitalized stroke rate, we used a combination of the VA and private sector strategies.
Results: The four approaches yielded widely different hospitalized stroke estimates (8.41-22.61/1000). The algorithms developed in the VA detected relatively more VA strokes, have unclear operating characteristics, and have not been studied in Medicare data. The algorithm developed in private sector data detected relatively more private sector strokes, does not assist population managers in identify at-risk populations, and has not been studied in the VA. The Health, US, estimates detected 1.4-2.7 times the number of strokes than the other algorithms. The ‘best estimate’ of hospitalized strokes using a hybrid approach was 12.3/1000 diabetes patients.
Conclusions: Currently available strategies for estimating stroke incidence using claims data have serious shortcomings. Further validation and standardization of claims data based approaches to estimating stroke is warranted.