Model Performance Statistics for Predicting 1-Year Mortality in Ontario
by Steven Habbous·Updated 2mo ago
13.5 KB1files
Available on 1 platform
Sign in to view source links and access this dataset
Description
A 2026 study by Steven Habbous compared logistic regression and seven tree-based ensemble methods to predict 1-year mortality in Ontario, Canada. The dataset contains performance statistics for these models, evaluated on a 30% test set of 12,080,801 adults. Metrics include AUROC, PR-AUC, Brier score, and Integrated Calibration Index.
Use Cases
Benchmarking machine learning models for mortality prediction based on reported AUROC and PR-AUC scores.
Comparing calibration performance of different algorithms based on the Brier score and Integrated Calibration Index.
Assessing feature importance for mortality risk factors based on the description of age, long-term care, and palliative care as influential variables.
Strengths
Performance metrics are reported for eight distinct modeling methods, allowing for direct comparison.
Evaluation is based on a large population cohort of over 12 million individuals.
Uses multiple validation metrics including AUROC, PR-AUC, Brier score, and a calibration index.
Limitations
Row count is unknown, which may limit suitability assessment.
Column-level documentation is absent; field semantics must be inferred after download.
The dataset is very small (13.5 KB), suggesting it contains only summary statistics, not the underlying patient-level data.
Provenance
Source
Steven Habbous via figshare.
Collection Method
Model performance derived from analysis of Ontario administrative healthcare databases.
Time Range
Study population alive as of January 1, 2022, with predictors captured from a preceding 3-year window.
Freshness
Last updated 2026-04 23 17:25:46; freshness should be verified.