Predicting the risk of severe COVID-19 outcomes in primary care: development and validation of a vulnerability index for equitable allocation of effective vaccines

General practitioners (GPs) need a valid, user-friendly tool to identify patients most vulnerable to COVID-19, especially in the hypothesis of a booster vaccine dose. The aim of this study was to develop and validate a GP-friendly prognostic index able to forecast severe COVID-19 outcomes in primary care. Indeed, no such prognostic score is as yet available in Italy. In this retrospective cohort study, a representative sample of 47,868 Italian adults were followed up for 129,000 person–months. The study outcome was COVID-19-related hospitalization and/or death. Candidate predictors were chosen on the basis of systematic evidence and current recommendations. The model was calibrated by using Cox regression. Both internal and external validations were performed. Age, sex and several clinical characteristics were significantly associated with severe outcomes. The final multivariable model explained 60% (95%CI 58–63%) of variance for COVID-19-related hospitalizations and/or deaths. The area under the receiver-operator curve (AUC) was 84% (95% CI: 83–85%). On applying the index to an external cohort, the AUC was 94% (95% CI: 93–95%). This index is a reliable prognostic tool that can help GPs to prioritize their patients for preventive and therapeutic interventions.

全科医师（General practitioners, GPs）亟需一款有效且易用的工具，以识别出最易罹患新冠重症的患者，尤其在需开展加强针疫苗接种的场景中。本研究旨在开发并验证一款便于全科医师使用的预后指数，能够在基层医疗环境中预测新冠重症转归。目前意大利尚无此类预后评分工具。
本项回顾性队列研究（retrospective cohort study）纳入了47868名意大利成年患者的代表性样本，总随访时长达129000人月。研究结局为新冠相关住院及/或死亡。候选预测因子基于系统性证据与现行临床指南遴选。研究采用Cox回归（Cox regression）模型进行校准，并完成了内部验证与外部验证。
年龄、性别及多项临床特征与新冠重症转归显著相关。最终的多变量模型可解释60%（95%置信区间58%~63%）的新冠相关住院及/或死亡病例的变异度。受试者工作特征曲线下面积（area under the receiver-operator curve, AUC）为84%（95%置信区间83%~85%）。将该指数应用于外部队列时，其受试者工作特征曲线下面积达94%（95%置信区间93%~95%）。
该预后指数是一款可靠的临床预测工具，可帮助全科医师优先为高危患者开展预防与治疗干预。