Scholarlink

<p>Anemia remains a major public health problem among Peruvian women, yet the combined influence of altitude and nutritional status on hemoglobin is not fully understood. We analyzed 197,586 nonpregnant women aged 15–49 years from the Peruvian Demographic and Health Survey (ENDES), 2005–2024, to compare body mass index (BMI), waist circumference, and waist-to-height ratio as predictors of hemoglobin using interpretable machine learning. Random Forest models and SHAP values were used to evaluate predictive performance, identify the most influential predictors, and describe nonlinear associations, while adjusted linear regression examined interaction with altitude. All three anthropometric indicators performed similarly (R² = 0.451–0.462; RMSE = 1.20&#xa0;g/dL). In the main BMI model, altitude was the strongest predictor and showed a marked nonlinear association with hemoglobin, becoming steeper above about 2000&#xa0;m above sea level. BMI showed a modest nonlinear positive association, with visual attenuation at higher BMI values but without a clearly identifiable threshold. Although BMI showed a statistically significant interaction with altitude in linear models, the effect was minimal and not clinically meaningful in the interpretable analysis. Survey year showed a nonlinear temporal pattern, with lower model contributions in recent survey rounds. These findings identify altitude as the main determinant of hemoglobin and support continued monitoring of temporal patterns.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Interpretable machine learning maps nonlinear associations of altitude and body mass index with hemoglobin in Peruvian women

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月4日

查看全文
  • Víctor Juan Vera-Ponce,
  • Jhosmer Ballena-Caicedo,
  • Holly Estrella Delgado-Toro,
  • Fiorella E. Zuzunaga-Montoya,
  • Renzo Acosta-Porzoliz,
  • Félix García-Ahumada,
  • Oriana Rivera-Lozada,
  • Mario J. Valladares-Garrido

摘要

Anemia remains a major public health problem among Peruvian women, yet the combined influence of altitude and nutritional status on hemoglobin is not fully understood. We analyzed 197,586 nonpregnant women aged 15–49 years from the Peruvian Demographic and Health Survey (ENDES), 2005–2024, to compare body mass index (BMI), waist circumference, and waist-to-height ratio as predictors of hemoglobin using interpretable machine learning. Random Forest models and SHAP values were used to evaluate predictive performance, identify the most influential predictors, and describe nonlinear associations, while adjusted linear regression examined interaction with altitude. All three anthropometric indicators performed similarly (R² = 0.451–0.462; RMSE = 1.20 g/dL). In the main BMI model, altitude was the strongest predictor and showed a marked nonlinear association with hemoglobin, becoming steeper above about 2000 m above sea level. BMI showed a modest nonlinear positive association, with visual attenuation at higher BMI values but without a clearly identifiable threshold. Although BMI showed a statistically significant interaction with altitude in linear models, the effect was minimal and not clinically meaningful in the interpretable analysis. Survey year showed a nonlinear temporal pattern, with lower model contributions in recent survey rounds. These findings identify altitude as the main determinant of hemoglobin and support continued monitoring of temporal patterns.