Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
March 6, 2019PLoS ONEOpen Access

Exploring the use of machine learning for risk adjustment: A comparison of standard and penalized linear regression models in predicting health care costs in older adults

View Full Paper
Ask AI
Bookmark
Share

Authors

HKHong KanEli Lilly (United States)HKHadi KharraziBloomberg (United States)HCHsien‐Yen ChangJohnson & Johnson (United States)

Discussion

Loading...

Member takes

Implication

Retrospective cohort study compares penalized and standard regression for health care cost prediction in older adults, indicating improved accuracy with penalized methods.

Key Points

  • The study aims to compare the prediction performance of standard and penalized linear regression models for estimating health care costs in older adults.
  • Conducted a retrospective cohort study of 81,106 Medicare Advantage patients with continuous insurance during 2009-2013.
  • Utilized OLS and penalized linear regression models to predict health care costs in 2013 based on predictors from 2009-2012.
  • Analyzed prediction performance metrics such as R2 and prediction ratios.
  • OLS regression showed an R2 of 16.3%, while penalized regression models had R2 values ranging from 16.8% to 16.9%.
  • When using 2009-2012 predictors, OLS predicted health care costs with an R2 of 15.0%, compared to 18.0-18.2% for penalized regression.
  • Lasso regression demonstrated superior prediction ratios across various predicted risk levels compared to standard models.

Cite This Study

Kan et al. (2019) studied this question.

synapsesocial.com/papers/6a1022a12badbc352aff6366https://doi.org/10.1371/journal.pone.0213258
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 4 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1The Elements of Statistical Learning: Data Mining, Inference, and Prediction2013 · 19,253 citations
  2. 2Regularization Paths for Generalized Linear Models via Coordinate Descent2010 · 21,048 citations
  3. 3Ridge Regression: Biased Estimation for Nonorthogonal Problems2000 · 6,552 citations
  4. 4A proposed national research and development agenda for population health informatics: summary recommendations from a national expert workshop2016 · 64 citations