Talk to the Veterans Crisis Line now
U.S. flag
An official website of the United States government

Health Services Research & Development

Veterans Crisis Line Badge
Go to the ORD website
Go to the QUERI website
HSRD Conference Logo



2019 HSR&D/QUERI National Conference Abstract

Printable View

4084 — Secondary ASCVD Risk Prediction using Electronic Health Record Data

Lead/Presenter: Wyndy Wiitala,  COIN - Ann Arbor
All Authors: Wiitala WL (VA Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI), Ratz DP (VA Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI), Hayward RA (VA Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI; Dept of Internal Medicine, University of Michigan, Ann Arbor, MI) Burns JA (VA Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI) Keusch J (VA Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI) Visnic S (VA Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI) Sussman JB (VA Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI; Dept of Internal Medicine, University of Michigan, Ann Arbor, MI)

Objectives:
Risk prediction plays a key role in the primary prevention of cardiovascular disease, but it is less well-developed in secondary prevention, which refers to people who have already had cardiovascular events. With the introduction of new, expensive, understudied drugs, predicting who is likely to have future events becomes even more important. We examined if the VA electronic health record could effectively be used to predict atherosclerotic cardiovascular disease (ASCVD) for secondary prevention.

Methods:
We defined a cohort of 742,787 VA patients from 2009 who had a heart attack or stroke in the prior 5 years. Outcome events included fatal and non-fatal heart attack or stroke during 5-years of follow-up. Logistic regression models were applied separately for men and women using 80% training and 20% testing data. Predictors included traditional cardiovascular disease risk factors, including longitudinal evaluations, history of heart attack vs. history of stroke, time since last event, number of total prior events, and medication use.

Results:
The sample included 742,787 patients, over 20,000 of whom were women. 52% had a history of heart attack, 35% stroke, and 13% both. Many patients also had a history of peripheral arterial disease (41%), coronary artery bypass surgery (6%), primary coronary intervention (15%). Of male patients, 30% had another event during the 5 years of follow-up, compared to 23% of women. The c-statistic for predicting any future events was 0.69 for men and 0.76 for women. Although calibration was good for both models, the model for women was slightly better calibrated. The average predicted risk for men was 0.30 (SD = 0.14; IQR: 0.19, 0.38) and for women was 0.23 (SD = 0.16; IQR: 0.11, 0.31).

Implications:
Using data entirely within the electronic health record, we were able to predict secondary events as successfully as prior primary risk scores.

Impacts:
These findings could guide population health management or use of new cardiovascular medicines, among other possibilities.