A Clinical Risk Prediction Tool for Cardiovascular Events Following Respiratory Infections

Abstract

Objective: The risk of cardiovascular (CVD) events increases for four weeks following a respiratory infection. This short-term increase in risk may be an opportunity for clinical intervention, providing one could predict the risk of infection-related CVD. We therefore sought to develop and validate a risk prediction model for use in patients with respiratory infections in primary care. Design and method: We used routinely collected data from two UK general practice databases, provided by Clinical Practice Research Datalink (CPRD). We used cohorts of patients without pre-existing CVD, starting from their first respiratory infection after the age of 40. We followed patients for 28 days. Our outcome was a composite of cerebral and cardiac events. We used logistic regression to develop risk prediction models in CPRD Aurum, and validated it in CPRD Gold. We selected predictors by asking clinicians to prioritize a list of potential predictors. This model uses ten variables. Results: The development cohort comprised 3.8 million patients with respiratory infections, and 7,081 CVD events in the following 28 days (0.18%). The validation cohort comprised 2.6 million patients with infections, and 6,868 events (0.26%). In external validation, we estimated a C statistic of 0.86. Using a 1% threshold of risk as a cut off, the positive predictive value was 3% (3.2 to 3.4%), with a negative predictive value close to one. Conclusions: It is possible to predict risk of primary infection-related CVD. Identifying patients at greatest risk of infection-related CVD opens up possibilities for further research applications.