COVID-19 Patient Clinical and Mortality Data for Severity Score Development

4,711 patient records from a single healthcare system contain demographics, comorbidities, admission lab values, and mortality outcomes for COVID-19. The data were derived from healthcare surveillance software and primary medical record review. The dataset was used to derive and validate a novel severity score for predicting disease severity and mortality.

Use Cases

• Predicting patient mortality based on demographics, comorbidities, and admission lab values mentioned in the description.
• Developing risk stratification models using variables like length of stay, D-dimer, and interleukin-6.
• Analyzing temporal trends in COVID-19 severity by comparing patient outcomes between the first and second three-week periods of the pandemic.
• Validating clinical severity scores against real-world patient data including oxygen saturation and mean arterial pressure.

Strengths

• Contains 4,711 patient records, providing a substantial sample size for analysis.
• Includes a wide range of clinical variables such as lab values (D-dimer, WBC, IL-6), comorbidities, and admission medications.
• Data is structured for model development, with a clear derivation (n=2,355) and validation (n=2,356) cohort split.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the specific data file is unknown, which may limit suitability assessment.
• Data may reflect geographic and temporal bias inherent to a single healthcare system over a limited pandemic period.

Provenance

Source

Derived from healthcare surveillance software (Clinical Looking Glass) and medical record review at Montefiore Medical Center/Albert Einstein College of Medicine.

Collection Method

Retrospective data collection from electronic health records.

Time Range

Covers a specific period during the COVID-19 pandemic, separated into two three-week phases.

Geography

Patients admitted to a single healthcare system; specific location not stated.