Key result
Disease-specific neural networks showed variable accuracy versus a population model for predicting high costs, performing better for diabetes (C=0.786 vs 0.767) but worse for cardiac conditions.
Why the study?
Does a disease-specific neural network model improve the prediction of high medical costs compared to a total population model in patients with chronic conditions?
Population
55,777 health plan members with 24 months of continuous enrollment, including 33,908 with diabetes, 19,264…
Comparison
Disease-specific neural network predictive models vs Total population neural network predictive model
Design
Cohort
Follow-up
1 year
Authors
Loading...
Variable performance cautions against adopting disease-specific models for cardiac cost prediction; leaves open validation in larger cohorts.
Observational (n=55,777)
Does a disease-specific neural network model improve the prediction of high medical costs compared to a total population model in patients with chronic conditions?
The predictive power of disease-specific versus population-based neural network models for medical costs varies by disease, with larger cohorts generally favoring disease-specific models.
Crawford et al. (2005) conducted an observational in Diabetes, asthma, and cardiac conditions (n=55,777). Disease-specific neural network models vs. Total population neural network model was evaluated on Accuracy in predicting probability of high medical costs (top 15% of distribution). Disease-specific neural networks showed variable accuracy versus a population model for predicting high costs, performing better for diabetes (C=0.786 vs 0.767) but worse for cardiac conditions.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: