Overview

The 4C Mortality Score is a bedside prognostic index developed to stratify risk of in-hospital death among adults admitted to hospital with COVID-19. It was derived from a large UK cohort collected under the ISARIC WHO Clinical Characterisation Protocol, using only variables that are typically available early in the admission—demographics, a concise comorbidity count, basic vital signs, conscious level, and two widely measured blood tests (urea and C-reactive protein). The final model sums eight components into a total ranging from 0 to 21. Higher totals associate with higher observed mortality in the validation dataset; the published work provides a discrete lookup of observed in-hospital mortality percentage by integer score for scores 1 through 21.

This score is intended as a transparent, reproducible summary of early presentation severity and baseline vulnerability. It does not replace clinical judgment, multidisciplinary assessment, or institutional pathways, and it must be interpreted in light of how epidemiology, variants, vaccination, and therapeutics have evolved since the original cohorts were assembled.

Background and development

During the first phase of the COVID-19 pandemic, hospitals needed tools that could be applied consistently across sites using a common data dictionary, without requiring specialised imaging or experimental assays at the point of presentation. The ISARIC collaboration standardised case report forms across many hospitals, enabling multivariable modelling with rigorous internal validation.

The 4C score distils the prognostic information from that setting into a simple additive rule with prespecified cut-offs for age, respiratory rate, oxygen saturation on room air, Glasgow Coma Scale (GCS), urea, and CRP, plus sex and number of defined comorbidities. The derivation and validation process compared candidate predictors, selected a parsimonious set, and reported performance metrics appropriate for prognostic models, including discrimination and calibration in held-out data. External validation studies in other countries and time periods have since examined whether similar coefficients and risk tables remain stable; clinicians should treat any single historical risk table as context-specific unless reproduced in their own environment.

Who the score applies to

The score was designed for adult inpatients hospitalised with COVID-19, using measurements reflecting status at or near hospital presentation, as defined in the primary study. It is not validated for primary care triage of outpatients with mild illness, for paediatric populations, or for patients in whom the principal reason for admission is unrelated to acute SARS-CoV-2 disease. It is also not a pathophysiological severity score in the same sense as ICU organ-failure scores; rather, it estimates a population-level probability of the outcome of in-hospital mortality conditional on the patterns seen in the development cohort.

When applying the score, ensure that SpO₂ reflects peripheral oxygen saturation on room air (or the same oxygenation context used in the original protocol), that GCS is recorded with standard technique, and that laboratory values use the same units as the published thresholds (urea in mmol/L; CRP in mg/L).

Score components

Each variable below contributes points independently; the total 4C score is the sum. The categories and points follow the primary publication’s Table 2 and the public ISARIC 4C materials.

Age (years)

• 18–49: 0 points
• 50–59: 2 points
• 60–69: 4 points
• 70–79: 6 points
• 80 or older: 7 points

Age captures both biological reserve and the steep gradient of severe outcomes observed across adult decades in early pandemic hospitalised cohorts.

Sex recorded at birth

• Female: 0 points
• Male: 1 point

The association between male sex and worse outcomes in those cohorts likely reflects a mix of biological and social determinants; the score encodes the empirical association rather than implying a single causal mechanism.

Number of comorbidities

Count only conditions that meet the study definition (each qualifying condition counts as one comorbidity):

• Chronic cardiac disease
• Chronic respiratory disease excluding asthma
• Chronic renal disease with estimated glomerular filtration rate 30 mL/min/1.73 m² or lower
• Mild-to-severe liver disease
• Dementia
• Chronic neurological conditions
• Connective tissue disease
• Diabetes mellitus (diet-controlled, tablet-controlled, or insulin-treated)
• HIV/AIDS
• Malignancy
• Clinician-defined obesity

Points:

• 0 comorbidities: 0 points
• 1 comorbidity: 1 point
• 2 or more comorbidities: 2 points

Respiratory rate (breaths per minute)

• Below 20: 0 points
• 20–29: 1 point
• 30 or higher: 2 points

Tachypnoea integrates respiratory drive, lung involvement, metabolic stress, and sometimes anxiety or pain; in the model it functions as a rapid bedside marker of physiological stress.

Peripheral oxygen saturation on room air

• SpO₂ 92% or higher: 0 points
• SpO₂ below 92%: 2 points

Hypoxaemia on room air identifies pulmonary shunt or ventilation–perfusion imbalance due to pneumonia, oedema, or other pathology. If supplemental oxygen is already in use, the value entered must be consistent with how the original data were collected; mixing “on oxygen” saturations with cut-offs defined for room air can misclassify risk.

Glasgow Coma Scale (GCS)

• GCS 15: 0 points
• GCS below 15: 2 points

Any reduction in consciousness raises concern for encephalopathy, severe hypoxia, hypercapnia, sepsis, sedating drugs, or intracranial processes; the score treats reduced GCS as a high-risk feature regardless of aetiology.

Urea (mmol/L)

• Below 7: 0 points
• 7–14 inclusive: 1 point
• Above 14: 3 points

Elevated urea may reflect dehydration, reduced renal perfusion, intrinsic kidney injury, or catabolic stress. The model uses urea rather than creatinine alone in its final form; laboratories reporting blood urea nitrogen (BUN) in mg/dL require conversion before applying the published bands (a commonly used approximation is urea (mmol/L) ≈ BUN (mg/dL) × 0.357, but local laboratory standards should prevail).

C-reactive protein (CRP, mg/L)

• Below 50: 0 points
• 50–99: 1 point
• 100 or higher: 2 points

CRP rises with systemic inflammation; very high values in COVID-19 often parallel cytokine-mediated lung injury and broader physiological derangement. If CRP is reported in mg/dL, multiply by 10 to obtain mg/L.

Interpreting the total score

The total ranges from 0 (lowest points from every domain) to 21 (maximum points). The published validation materials provide observed in-hospital mortality percentages for each integer total from 1 through 21 in the validation subset. Score 0 lies below the tabulated minimum in the public lookup (which begins at 1); clinically it still represents the lowest-risk pattern within the model’s structure and should be interpreted with the same caveats as all prognostic estimates.

These percentages describe what happened to groups of patients with that score in a specific UK hospitalised cohort at a specific time. They are not guaranteed calibration for:

• Different countries, ethnic mixes, or baseline health systems
• Later viral variants with altered virulence or vaccine escape patterns
• Widespread vaccination or outpatient antiviral use changing the case mix at admission
• Modern inpatient treatment protocols (steroids, anticoagulation strategies, respiratory support algorithms) that alter the trajectory after admission

Therefore, the numeric mortality associated with each score should be used for context, communication, and proportionate planning, not as a fixed personal probability for an individual patient in a different era of the pandemic.

Observed in-hospital mortality by score (validation cohort)

The following table reproduces the discrete mapping disseminated with the ISARIC 4C materials for scores 1–21. Values are expressed as percentages of patients with that score who died in hospital in the validation data.

Practical use alongside pathways

In many institutions, prognostic scores inform—but do not dictate—decisions about level of care, frequency of monitoring, ceiling-of-care discussions, and research stratification. The 4C score’s strength is simplicity and transparency: clinicians can see which domains drive the total and explain the estimate to patients and families. Its weakness is temporal drift: absolute risk estimates from 2020-era UK inpatients may overstate or understate current risk depending on local case mix and treatment effectiveness.

Best practice is to combine the score with dynamic assessment (oxygen requirements over hours to days, need for respiratory support, development of secondary infection, thrombotic events, and renal or cardiac complications), patient values, and bed capacity. Any single admission snapshot misses deterioration after presentation; repeat scoring was explored in some follow-on work but is outside the scope of the original 4C definition.

Strengths and limitations

Strengths include wide availability of inputs, fast calculation without imaging, clear cut-offs, and an explicit mapping from score to observed outcome frequency in the validation cohort. The score aligns with how front-line teams already conceptualise risk: age, sex, comorbidity burden, respiratory distress, hypoxaemia, conscious level, renal stress, and systemic inflammation.

Limitations include dependence on accurate and contemporaneous data entry, sensitivity to unit errors for urea and CRP, potential misclassification if SpO₂ is measured on supplemental oxygen but interpreted as room air, and reduced transportability across health systems and pandemic phases. Patients with missing variables were handled in specific ways in the research pipeline; in real time, missing data should prompt direct reassessment rather than silent imputation. The score should not be used in isolation to label a patient “unsuitable” for intensive treatment; such decisions require ethical, legal, and professional frameworks that no calculator can replace.

Educational disclaimer

This article supports education and shared understanding of the 4C Mortality Score. It is not medical advice. Clinicians remain responsible for assessment, diagnosis, treatment, and documentation according to licence, local regulation, and institutional policy.