Researchers use population health data to determine cardiovascular disease risk

A recent study in the Journal of the American Medical Informatics Association found that using population health data could help identify patients with a high risk of developing cardiovascular disease who have not been previously identified as high risk due to lack of relevant cholesterol data. The study found that it is possible to determine the risk level for patients without individual cholesterol data by using calculations derived from other data within electronic health records (EHR). Primary care doctors then used this information to proactively engage patients newly identified as high risk.

Researchers from the University of North Carolina (UNC) School of Medicine and the UNC Gillings School of Global Public Health formed a network of 219 primary care clinics across the state of North Carolina from urban, suburban, and rural areas to examine 345,440 patients age 40–79. Of that pool, 108,515 did not have cholesterol scores (total cholesterol and HDL levels) on their EHRs, which is one of the factors used to determine precise cardiovascular risk.

The researchers performed two separate analyses for determining risk for patients lacking cholesterol data by using the data from the 236,925 patients who had cholesterol numbers, along with other factors for determining cardiovascular disease risk, such as smoking status, weight, diabetes, gender, etc.

Researchers then created an equation to estimate cholesterol levels for patients with missing data and used it to calculate atherosclerotic cardiovascular disease (ASCVD) risk scores. Using this estimation method, the researchers determined that 43,205 patients were at a ≥ 10%, 10-year ASCVD risk and were eligible for risk reduction interventions, such as aspirin or statins. These patients had not been previously identified as being high-risk.

The researchers then ran a simpler analysis by plugging in conservative cholesterol numbers (170 mg/dl total cholesterol and 50 mg for HDL) for the patients lacking data. Even with this conservative estimate, 40,565 patients (37% of those lacking cholesterol data) were found to have risk scores ≥ 10%.

The researchers concluded that the estimation method had a higher false-positive rate than using the conservative estimate, but using either analysis strategy still accurately identified high-risk patients who could benefit from intervention.