Scholarlink

<p>Type 2 diabetes (T2D) poses a global health crisis, disproportionately affecting women through gestational and postmenopausal risks. Current screening overlooks multicollinearity among demographic and metabolic predictors. This study develops and validates a multivariate logistic regression model to quantify T2D risk in women, assessing clinical utility. Cross-sectional secondary analysis of the Pima Indian Diabetes Dataset (1980–1985; <i>n</i> = 751 women aged ≥ 21 years). Backward stepwise logistic regression modeled binary T2D status against eight predictors (pregnancies, glucose, blood pressure, skin thickness, insulin, BMI, diabetes pedigree function [DPF], age), with multicollinearity evaluated via variance inflation factors and correlations. Validation included ROC curves, bootstrap calibration, Hosmer-Lemeshow test, and decision curve analysis (DCA). Significant predictors post-adjustment: pregnancies (OR 1.13, 95% CI 1.06–1.21), glucose (OR 1.04, 95% CI 1.03–1.04), BMI (OR 1.09, 95% CI 1.06–1.13), DPF (OR 2.57, 95% CI 1.44–4.66); blood pressure showed inverse association (OR 0.99, 95% CI 0.98–1.00). Skin thickness, insulin, and age were non-significant. Model performance: AUC-ROC 0.84 (95% CI 0.82–0.86), Hosmer-Lemeshow p = 0.30, Nagelkerke R²=0.41. DCA indicated superior net benefit over treat-all/none strategies at thresholds 0.05–0.45. Marginal probabilities showed exponential risk escalation beyond BMI &gt; 20 kg/m² and glucose &gt; 50 mg/dL. The model reveals independent effects of key predictors, mitigating biases from univariate analyses. It aligns with evidence on metabolic clustering and genetic modulation in women, offering interpretable insights over machine learning approaches. Limitations include ethnic homogeneity and cross-sectional design. This validated framework supports targeted T2D screening, emphasizing glucose, BMI, pregnancies, and DPF, with potential for broader epidemiological use.</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

Multivariate risk stratification of type 2 diabetes in women: a clinically validated logistic regression model with decision curve analysis

  • 期刊论文

  • 发表日期：

    2026年7月24日

查看全文
  • Ruhiteswar Choudhury,
  • Tanusree Deb Roy

摘要

Type 2 diabetes (T2D) poses a global health crisis, disproportionately affecting women through gestational and postmenopausal risks. Current screening overlooks multicollinearity among demographic and metabolic predictors. This study develops and validates a multivariate logistic regression model to quantify T2D risk in women, assessing clinical utility. Cross-sectional secondary analysis of the Pima Indian Diabetes Dataset (1980–1985; n = 751 women aged ≥ 21 years). Backward stepwise logistic regression modeled binary T2D status against eight predictors (pregnancies, glucose, blood pressure, skin thickness, insulin, BMI, diabetes pedigree function [DPF], age), with multicollinearity evaluated via variance inflation factors and correlations. Validation included ROC curves, bootstrap calibration, Hosmer-Lemeshow test, and decision curve analysis (DCA). Significant predictors post-adjustment: pregnancies (OR 1.13, 95% CI 1.06–1.21), glucose (OR 1.04, 95% CI 1.03–1.04), BMI (OR 1.09, 95% CI 1.06–1.13), DPF (OR 2.57, 95% CI 1.44–4.66); blood pressure showed inverse association (OR 0.99, 95% CI 0.98–1.00). Skin thickness, insulin, and age were non-significant. Model performance: AUC-ROC 0.84 (95% CI 0.82–0.86), Hosmer-Lemeshow p = 0.30, Nagelkerke R²=0.41. DCA indicated superior net benefit over treat-all/none strategies at thresholds 0.05–0.45. Marginal probabilities showed exponential risk escalation beyond BMI > 20 kg/m² and glucose > 50 mg/dL. The model reveals independent effects of key predictors, mitigating biases from univariate analyses. It aligns with evidence on metabolic clustering and genetic modulation in women, offering interpretable insights over machine learning approaches. Limitations include ethnic homogeneity and cross-sectional design. This validated framework supports targeted T2D screening, emphasizing glucose, BMI, pregnancies, and DPF, with potential for broader epidemiological use.