Interpretable Machine Learning for Predicting Early Onset Hypertension in the US

Li-Wei Wu

doi:10.5888/pcd23.260109

Interpretable Machine Learning for Predicting Early Onset Hypertension in the US

ORIGINAL RESEARCH — Volume 23 — October 22, 2026

Print

PEER REVIEWED

Fairness assessment across racial and ethnic groups, machine learning model analysis for development of early onset hypertension, Medical Expenditure Panel Survey, 2022. White race is the reference group.
Rate difference	Race and ethnicity, metric value
Rate difference	Other	Hispanic	African American	Asian
True positive	0.03	0.11	−0.05	0.08
False positive	−0.01	0.05	−0.03	0.11
False negative	−0.03	−0.11	0.05	−0.08
Average of true positive and false positive	0.01	0.08	−0.04	0.09

Figure 1. Fairness assessment across racial and ethnic groups, machine learning model analysis for development of early onset hypertension, Medical Expenditure Panel Survey, 2022. White race is the reference group.

Return to Article

This SHAP summary plot shows the ten features with the greatest influence on the XGBoost model’s predicted probability of early-onset hypertension. In summary, older age, elevated BMI, high cholesterol diagnosis, residing in the Southern region, a higher number of emergency department visits, diabetes diagnosis, and fair perceived health were associated with an increased predicted probability. Conversely, excellent perceived health, female sex, and Hispanic ethnicity were associated with a decreased predicted probability.

Importance and summary plots of the top 10 SHAP features from the XGBoost model, machine learning model analysis for development of early onset hypertension, Medical Expenditure Panel Survey, 2022. A, SHAP features importance plot; B, SHAP summary plot. Abbreviations: BMI, body mass index; ED, emergency department; SHAP, Shapley Additive Explanations.
SHAP feature	Mean SHAP value
Age	+0.69
BMI	+0.29
High cholesterol	+0.22
Excellent perceived health	+0.22
Sex	+0.16
Hispanic ethnicity	+0.11
South region	+0.09
ED visit	+0.09
Diabetes	+0.08
Sum of 36 other features	+0.61

Figure 2. Importance and summary plots of the top 10 SHAP features from the XGBoost model, machine learning model analysis for development of early onset hypertension, Medical Expenditure Panel Survey, 2022. A, SHAP features importance plot; B, SHAP summary plot. Abbreviations: BMI, body mass index; ED, emergency department; SHAP, Shapley Additive Explanations.

Return to Article

Importance of SHAP features in machine learning model analysis for development of early onset hypertension, Medical Expenditure Panel Survey, 2022. A, age; B, race and ethnicity; C, CHD; and D, heart attack. Abbreviations: BMI, body mass index; CHD, coronary heart disease; ED, emergency department; SHAP, Shapley Additive Explanations.
SHAP feature	Age (graph A)
SHAP feature	≤40 years	>40 years
Age	0.69	0.70
BMI	0.31	0.27
High cholesterol	0.16	0.31
Excellent perceived health	0.22	0.21
Sex	0.18	0.14
Hispanic ethnicity	0.11	0.12
South region	0.09	0.10
ED visit	0.10	0.08
Diabetes	0.06	0.11
Fair perceived health	0.06	0.06

Importance of SHAP features in machine learning model analysis for development of early onset hypertension, Medical Expenditure Panel Survey, 2022. A, age; B, race and ethnicity; C, CHD; and D, heart attack. Abbreviations: BMI, body mass index; CHD, coronary heart disease; ED, emergency department; SHAP, Shapley Additive Explanations.
SHAP feature	Race and ethnicity (graph B)
SHAP feature	Other	White	Hispanic	African American	Asian
Age	0.74	0.67	0.72	0.73	0.65
BMI	0.31	0.32	0.25	0.28	0.30
Excellent perceived health	0.20	0.21	0.21	0.24	0.23
High cholesterol	0.21	0.22	0.21	0.22	0.23
Sex	0.18	0.18	0.14	0.15	0.17
Hispanic ethnicity	0.09	0.09	0.17	0.10	0.09
South region	0.09	0.09	0.09	0.11	0.09
ED visit	0.10	0.08	0.09	0.13	0.07
Diabetes	0.09	0.07	0.08	0.10	0.07
African American race	0.03	0.03	0.03	0.21	0.03

Importance of SHAP features in machine learning model analysis for development of early onset hypertension, Medical Expenditure Panel Survey, 2022. A, age; B, race and ethnicity; C, CHD; and D, heart attack. Abbreviations: BMI, body mass index; CHD, coronary heart disease; ED, emergency department; SHAP, Shapley Additive Explanations.
SHAP feature	CHD (graph C)
SHAP feature	No	Yes
Age	0.69	0.74
High cholesterol	0.22	0.53
BMI	0.29	0.26
Excellent perceived health	0.22	0.16
Diabetes	0.08	0.25
Sex	0.16	0.16
CHD	0.00	0.27
Hispanic ethnicity	0.11	0.10
ED visit	0.09	0.11
South region	0.09	0.10

Importance of SHAP features in machine learning model analysis for development of early onset hypertension, Medical Expenditure Panel Survey, 2022. A, age; B, race and ethnicity; C, CHD; and D, heart attack. Abbreviations: BMI, body mass index; CHD, coronary heart disease; ED, emergency department; SHAP, Shapley Additive Explanations.
SHAP feature	Heart attack (graph D)
SHAP feature	No	Yes
Age	0.69	0.65
High cholesterol	0.22	0.46
BMI	0.29	0.31
Excellent perceived health	0.22	0.16
Sex	0.16	0.17
Diabetes	0.08	0.25
Hispanic ethnicity	0.11	0.10
ED visit	0.09	0.11
South region	0.09	0.10
Fair perceived health	0.06	0.11

Figure 3. Importance of SHAP features in machine learning model analysis for development of early onset hypertension, Medical Expenditure Panel Survey, 2022. A, age; B, race and ethnicity; C, CHD; and D, heart attack. Abbreviations: BMI, body mass index; CHD, coronary heart disease; ED, emergency department; SHAP, Shapley Additive Explanations.

Return to Article

Top

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.