Development and external validation of a risk prediction score (DASHI) for cardiovascular events following acute respiratory infections: derivation and validation retrospective cohort study

Background Acute respiratory infections increase the short-term risk of myocardial infarction (MI) and stroke in primary care patients. Clinical guidelines for acute respiratory infections in primary care do not consider the risk of cardiovascular events, and CVD risk prediction tools target long-term risk. We aimed to develop and validate a prediction tool for the risk of cardiovascular disease events within 28-days of acute respiratory infection. Methods The design was a retrospective cohort study using two different databases of routinely collected data from electronic health records from January 1999 to December 2019. We used Clinical Practice Research Datalink (CPRD) Aurum data to derive models, and CPRD GOLD data from a different population for external validation. This data is from UK primary care, with data linkage to Hospital Episode Statistics, Office of National Statistics mortality data, and Index of Multiple Deprivation data. Participants were patients aged 40 years or older with no history of cardiovascular events, and a first diagnosis with acute respiratory infection. The outcome was a composite of new diagnoses of myocardial ischaemia (myocardial infarction, angina, acute coronary syndromes, or ischaemic cardiomyopathy), stroke or transient ischaemic attack, or deaths with these diagnoses, within 28 days of presentation with an acute respiratory infection. We derived a list of 57 potential predictors based on prior studies and asked clinical experts to rank them. We derived two logistic regression models, one with the top ranked variables, and another including additional lower ranked variables. We derived a clinical prediction score from the most parsimonious logistic regression model. We validated each model and the score in the external dataset using C statistics, calibration plots, and expected to observed ratios. We examined clinical utility using decision curve analysis. Findings The derivation cohort comprised 3.8 million patients with an acute respiratory infection (mean age 56.5 years, (SD 13.7); 57.7% female), of whom 11,996 had a subsequent cardiovascular outcome (0.3%). The validation cohort included 2.6 million patients (mean age 56.7 years, SD 13.6, 58.0% female), of whom 6868 (0.3%) had a subsequent cardiovascular outcome. The DASHI score comprised five clinical variables: Diabetes (1 point, yes/no), Age (40-59, 0 points; 60-79, 2 points; 80+, 4 points), current Smoking (1 point, yes/no), Heart failure (1 point, yes/ no), and Infection diagnosis (Upper Respiratory Tract Infection-0 points. Lower Respiratory Tract Infection (LRTI)-1 point, or LRTI with a pneumonia diagnosis-4 points). Upon external validation, each model and the score showed similar performance. The score showed good discrimination (AUC 0.85, IQR 0.848-0.849) and calibration with an expected to observed ratio of 0.85 (IQR 0.85-0.85). Interpretation The DASHI score allows primary care clinicians to estimate the risk of cardiovascular complications within 28 days in patients with acute respiratory infections. Funding This research was funded in part by the Wellcome Trust [211182/Z/18/Z] and NIHR [NIHR300738]. For the purpose of open access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. Copyright (c) 2025 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons. org/licenses/by/4.0/).