severityscore

Severity score, or severityscore, is a general term for a numerical metric intended to quantify how severely a patient is affected by a disease, injury, or syndrome. Such scores are used to stratify risk, guide treatment decisions, and compare outcomes across patients or groups. There is no universal severity score; instead, many domain-specific scales have been developed for particular conditions or settings.

Most severity scores combine multiple data elements, such as vital signs, laboratory results, imaging findings, and patient characteristics, into a weighted calculation. The resulting score is a single numeric value where higher scores typically indicate greater severity or worse prognosis. Scores may be derived from statistical models, simple rule-based rules, or modern machine learning approaches. Because patient status can change, some scores are designed to be recalculated over time.

Examples include APACHE II and APACHE III in critical care, SOFA for organ failure assessment, the Injury Severity Score (ISS) for trauma, MELD and Child-Pugh for liver disease, CURB-65 or qSOFA for pneumonia and sepsis risk, and SCORTEN for toxic epidermal necrolysis. Each system has its own inputs, interpretation rules, and validation data.

Validation typically assesses discrimination, such as the area under the receiver operating characteristic curve (AUROC), and calibration against observed outcomes in external cohorts. Limitations include variability in data quality, differences in patient populations, and potential drift in performance over time. Scores should support, not replace, clinical judgment.

In research datasets and electronic health records, a field named severityscore may be used to standardize reporting across studies, or as a feature in predictive models learned from data.