ABSTRACT
Background: To reduce the risk of hospital-acquired venous thrombosis (HA-VTE) in medical patients, guidelines recommend assessing HA-VTE risk and providing prophylaxis for those at high risk. Risk assessment models (RAMS) including objective risk factors available at admission remain an unmet clinical need. Aim(s): To develop and validate a RAM for HA-VTE in medical inpatients using data available to providers within 24-h of hospital admission. Method(s): We developed a HA-VTE RAM at the University of Vermont Medical Center (Burlington, Vermont, USA, Table 1) and validated this RAM at Michigan Medicine (Ann Arbor, Michigan, USA, Table 2). HA-VTE and the risk factors were identified using previously validated computable phenotypes. The RAM was developed using a Bayesian LASSO approach with model performance assessed using area under the receiver operating curves (AUC) and the slope of observed versus expected plot. People admitted with VTE were excluded. The research was approved by the Institutional Review Board funded by the National Institutes of Health and the Centers for Disease Control and Prevention, USA. Result(s): Table 1 presents the risk factors, odds ratios (OR) and 95% credible intervals (CI) for the HA-VTE RAM, which included 11 risk factors. For the development cohort, based on 219 events among 62,468 admissions, the AUC of the model was 0.75 and the observed versus expected slope was 1.11 (Table 2). In the validation cohort there were 48,265 admissions and 363 HA-VTE events with a younger population and a higher incidence of HA-VTE. The AUC and the observed versus expected slope were 0.69 and 0.89 (Table 2). Conclusion(s): We developed and validated a HA-VTE RAM in populations. The model fit and calibration are promising especially given these are two geographically diverse institutions. Further validation is in progress at additional hospitals as well as in people hospitalized with COVID-19.