Key result
Machine learning with dietary data predicts CV mortality with a ~0.93 C-statistic over standard models.
Why the study?
The study tested whether adding nutrition predictor variables and/or using machine learning models improves cardiovascular mortality risk prediction compared to standard Cox models lacking nutrition variables.
Does the addition of nutrition predictor variables and machine learning models improve cardiovascular mortality risk prediction compared to standard Cox models?
Observational (n=41,990)
Yes
Does the addition of nutrition predictor variables and machine learning models improve cardiovascular mortality risk prediction compared to standard Cox models?
Incorporating nutrition data into machine learning models significantly improves the accuracy and discrimination of cardiovascular mortality risk prediction.
Answers are grounded in this paper and cite their sources.
Nutrition-ML integration may refine CV mortality prediction in cohorts; hypothesis-generating and requires prospective validation before clinical use.
OBJECTIVES: We aimed to test whether or not adding (1) nutrition predictor variables and/or (2) using machine learning models improves cardiovascular death prediction versus standard Cox models without nutrition predictor variables. DESIGN: Retrospective study. SETTING: Six waves of Survey (NHANES) data collected from 1999 to 2011 linked to the National Death Index (NDI). PARTICIPANTS: 29 390 participants were included in the training set for model derivation and 12 600 were included in the test set for model evaluation. Our study sample was approximately 20% black race and 25% Hispanic ethnicity. PRIMARY AND SECONDARY OUTCOME MEASURES: Time from NHANES interview until the minimum of time of cardiovascular death or censoring. RESULTS: A standard risk model excluding nutrition data overestimated risk nearly two-fold (calibration slope of predicted vs true risk: 0.53 (95% CI: 0.50 to 0.55)) with moderate discrimination (C-statistic: 0.87 (0.86 to 0.89)). Nutrition data alone failed to improve performance while machine learning alone improved calibration to 1.18 (0.92 to 1.44) and discrimination to 0.91 (0.90 to 0.92). Both together substantially improved calibration (slope: 1.01 (0.76 to 1.27)) and discrimination (C-statistic: 0.93 (0.92 to 0.94)). CONCLUSION: Our results indicate that the inclusion of nutrition data with available machine learning algorithms can substantially improve cardiovascular risk prediction.
No takes yet. Share an insight, caveat, or question.
Rigdon et al. (2019) conducted an observational in Cardiovascular mortality risk prediction (n=41,990). Machine learning algorithms with 24-hour dietary recall nutrition data vs. Standard Cox proportional hazards model excluding nutrition data was evaluated on Cardiovascular death (model evaluated via discrimination/C-statistic and calibration) (C-statistic 0.93, 95% CI 0.92-0.94). Combining comprehensive 24-hour dietary recall nutrition data with machine learning algorithms improved cardiovascular mortality risk prediction compared to standard clinical models, achieving a C-statistic of 0.93.
Synapse has enriched 10 related papers. Consider them for comparative context: