NHANES 2007–2018: Cardiovascular Disease Risk in Kidney Stone Patients

A 47% increased risk of cardiovascular disease was found in participants with kidney stones compared to those without. This dataset, derived from 34,770 participants in the NHANES 2007–2018 cycles, was used to develop and validate an interpretable machine learning model for identifying CVD in kidney stone patients. The model development and validation involved 1,491 participants from 2007–2016 and an independent temporal cohort of 296 participants from 2017–2018.

Use Cases

• Training a logistic regression model for CVD risk prediction based on 15 identified predictors.
• Conducting SHAP analysis for global and local model interpretability in a clinical context.
• Validating the association between kidney stone disease and cardiovascular disease using weighted multivariable logistic regression.
• Performing temporal validation of a clinical prediction model using an independent cohort.

Strengths

• Data is sourced from the large, nationally representative NHANES survey, involving 34,770 participants.
• The developed logistic regression model achieved an area under the ROC curve of 0.801 in the internal test set.
• Includes an independent temporal validation cohort of 296 participants from the 2017–2018 NHANES cycle.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the primary modeling dataset is unknown, which may limit suitability assessment.
• The cross-sectional study design means causality between kidney stones and CVD cannot be inferred.

Provenance

Source

National Health and Nutrition Examination Survey (NHANES)

Collection Method

Survey data analysis and machine learning model development.

Time Range

2007–2018

Freshness

Last updated 2026-05-29 05:31:41; freshness should be verified.

Geography

United States