Related Experiment Video
Updated: Jun 2, 2026
Inverse Probability of Treatment Weighting (Propensity Score) using the Military Health System Data Repository and National Death Index
Published on: January 8, 2020
Using administrative databases to calculate Framingham scores within a large health care organization
Olaniyi James Ekundayo1, Stefanie D Vassar, Linda S Williams
1University of California, Los Angeles, CA, USA.
Insights
Administrative data can identify individuals with a worse Framingham risk profile, indicating a higher likelihood of stroke. Further research is needed to refine stroke prediction tools for specific populations.
Area of Science:
- Cardiovascular epidemiology
- Health services research
Background:
- Framingham risk calculators are typically used in one-on-one clinical settings.
- Increasing clinical data in administrative datasets allows for novel risk assessment applications.
Purpose of the Study:
- To evaluate the utility of administrative data-derived Framingham scores in identifying individuals at high risk for stroke within one year.
- To assess the predictive accuracy of Framingham risk scores using administrative data.
Main Methods:
- A nested case-control study design was employed.
- Compared 313 first-time stroke patients with 25,361 controls using administrative data from 2007.
- Framingham risk scores for generalized cardiovascular disease and stroke were calculated.
Main Results:
- Stroke cases exhibited significantly higher Framingham risk scores compared to controls (P<0.0001).
- The c-statistic for the generalized cardiovascular disease score was 0.68, and for the stroke score was 0.64.
- Stroke patients had a worse risk profile, including older age, higher blood pressure, and higher cholesterol.
Conclusions:
- Administrative data-derived Framingham risk profiles correlate with future stroke development.
- There is a need to enhance stroke predictive tools and determine optimal applications and populations for their use.
Background And Purpose:
Framingham calculators are typically implemented in 1-on-1 settings to determine if a patient is at high risk for development of cardiovascular disease in the next 10 years. Because health care administrative datasets are including more clinical information, we explored how well administrative data-derived Framingham scores could identify persons who would have stroke develop in the next year.
Methods:
Using a nested case-control design, we compared all 313 persons who had a first-time stroke at 5 Veterans Administration Medical Centers with a random sample of 25,361 persons who did not have a first-time stroke in 2008. We compared Framingham scores and risk using administrative data available at the end of 2007.
Results:
Stroke patients had higher risk profile than controls: older age, higher systolic blood pressure and total cholesterol, more likely to have diabetes, cardiovascular disease, left ventricular hypertrophy, and more likely to use treatment for blood pressure (P<0.05). The mean Framingham generalized cardiovascular disease score (18.0 versus 14.5) as well as the mean Framingham stroke-specific score (13.2 versus 10.2) was higher for stroke cases than controls (both P<0.0001). The c-statistic for the generalized cardiovascular disease score was 0.68 (95% CI, 0.65-0.70) and for the stroke score was 0.64 (95% CI, 0.62-0.67).
Conclusions:
Persons who had a stroke develop in the next year had a worse Framingham risk profile, as determined by administrative data. Future studies should examine how to improve the stroke predictive tools and to identify the appropriate populations and uses for applying stroke risk predictive tools.
