UK BIOBANK–DERIVED MODELS

Multimorbidity Pattern Calculator

Assign UK Biobank-derived multimorbidity patterns and estimate incident sepsis risk using the study models.

Pattern and risk estimates

Awaiting inputs
Assigned pattern—Pattern with the highest posterior probability.
Posterior probability—Posterior probability of the assigned pattern.
10-year sepsis risk—Estimates at 1, 2, 5, 10 and 15 years.
10-year risk difference—Relative to No multimorbidity, in percentage points.

Research use only. These estimates are not intended to guide diagnosis or treatment.

Risk modelRisk estimation requires age, sex and race. Additional covariates are optional.

Pattern membership probabilities

Posterior probabilities of the five multimorbidity patterns.

Enter baseline diagnoses to estimate pattern membership.

Sepsis risk over time

Estimated risk at 1, 2, 5, 10 and 15 years.

Enter baseline characteristics to estimate sepsis risk.
Methods and interpretation

Condition definitions

ICD-10 codes are matched by prefix to 41 study-defined conditions. Conditions without evidence in the submitted data are treated as absent. When BMI is provided, obesity is defined as BMI ≥30 kg/m²; otherwise, ICD-10 obesity codes are used.

Pattern assignment

Participants with fewer than two conditions are classified as No multimorbidity. Otherwise, assignment uses the highest posterior probability from the five-class latent class model fitted in UK Biobank participants. Class priors and condition probabilities are held fixed.

Sepsis risk estimation

The Cox model was fitted in UK Biobank participants, including incident sepsis events. The full model includes pattern, age, sex, race, smoking status, alcohol use, Townsend deprivation index and university degree. When optional covariates are unavailable, a model refitted without those covariates is used.

Risk estimates are calculated as 1 − S(t). Deaths before sepsis are censored; competing mortality is not incorporated.

Interpretation

Risk estimation is restricted to the development cohort’s age range ( years). Absolute risk estimates have not been externally validated or recalibrated for other populations. Comparisons with No multimorbidity hold covariates constant and should not be interpreted as causal effects.