Loading forecasting engine

Connecting to forecasting server...

0s, cold boot

About the claims forecaster

This tool forecasts paid surgical malpractice claims in the United States: the expected payment year, the likely amount with a 90% range, and the premium paid when a claim goes to trial instead of settling. A national outlook projects payments still to come by incident year.

Models are trained on the National Practitioner Data Bank (NPDB) Public Use Data File and scored on years they never saw.

Inputs: Incident year, state, injury severity, primary and secondary allegation, patient age band and sex, care setting, practitioner license and graduation decade, who would pay, and years already unpaid.
Models: Timing: LightGBM reverse-time hazard. Amount: LightGBM with conformalized quantile intervals and a band classifier (AUROC 0.78 to 0.81). Trial: causal forest. Outlook: chain ladder averaged with a machine-learning development model.
Backtests: 90% amount range covered 91% of 2020 to 2025 payments. Outlook backtested on 2011 to 2024.
Data: NPDB Public Use Data File, paid claims only. Amounts in 2024 U.S. dollars.

Key terms

Research demonstration only. Not legal, actuarial or financial advice. Forecasts are conditional on the claim ending in a payment.