{
  "abstract": "Edentulism is an end-stage oral condition that severely impairs the quality of life and overall health of older adults. Using cross-sectional data from the National Health and Nutrition Examination Survey (NHANES), this study integrated multidimensional information – including sociodemographic factors, chronic disease history, and dietary characteristics – to develop and validate an interpretable machine learning model for accurate prediction of edentulism risk. A retrospective cross-sectional design was adopted, including 9230 NHANES participants aged ≥ 20 years from 2005 to 2018, among whom the prevalence of edentulism was 7.86%. Participants were split into training and testing sets using a 70%:30% stratified random allocation. LASSO regression was first applied to select 11 core variables from 44 candidate features: age, years of education, smoking status, poverty-to-income ratio (PIR), magnesium, dietary fiber, total sugar, caffeine, diabetes history, and heart disease history. Five machine learning models (Random Forest, XGBoost, LightGBM, Decision Tree, and KNN) were then constructed, and their performance in the testing set was comprehensively evaluated based on discrimination (AUC, AUPRC), calibration (ECE), and clinical net benefit (decision curve analysis [DCA]). SHAP (SHapley Additive exPlanations) analysis was subsequently performed to interpret the optimal model. Baseline characteristics indicated that individuals with edentulism exhibited a pattern of “advanced age – low socioeconomic status – insufficient dietary antioxidant intake – accumulation of chronic conditions.” Restricted cubic splines (RCS) showed significant nonlinear protective associations of dietary fiber and magnesium intake with edentulism risk, whereas total sugar intake demonstrated a nonlinear positive association. Among all models, gradient boosting algorithms (XGBoost, LightGBM) achieved the highest discrimination (AUC = 0.855). Considering discrimination, reliability of predicted probabilities (ECE = 0.0097), and clinical net benefit, Random Forest was ultimately selected as the optimal model. SHAP analysis revealed that age, years of education, smoking status, and PIR were the most influential predictors of edentulism risk. This study successfully developed a Random Forest–based prediction model for edentulism risk, demonstrating strong discriminative performance, reliable calibration, and substantial clinical net benefit. Age, socioeconomic factors, and smoking status were identified as core risk determinants, while inadequate magnesium and dietary fiber intake emerged as key nutritional intervention targets. The model provides a useful decision-support tool for early screening and precision nutrition strategies among high-risk populations.",
  "abstract_shown": true,
  "affiliations": [
    "Department of Blood Transfusion, Affiliated Hospital of Youjiang Medical University for Nationalities, Baise, China",
    "School of Stomatology, Xiamen Medical College, Xiamen, Fujian, China",
    "School of Basic Medical Science, Fujian Medical University, Fuzhou, Fujian, China",
    "School of Clinical Medicine, Xiamen Medical College, Xiamen, Fujian, China",
    "School of Basic Medical Science, Xiamen Medical College, Xiamen, Fujian, China",
    "Guangxi Key Laboratory of Artificial Intelligence for Genetic Diseases of Long-dwelling Nationalities, Baise, Guangxi, China."
  ],
  "article_type": "research article",
  "arxiv": null,
  "author_total": 5,
  "authors": [
    {
      "affs": [
        0,
        1
      ],
      "name": "Shiyuan Wu"
    },
    {
      "affs": [
        2
      ],
      "name": "Muze Huang"
    },
    {
      "affs": [
        3
      ],
      "name": "Zi-han Shen"
    },
    {
      "affs": [
        4
      ],
      "name": "Yaw-Syan Fu",
      "orcid": "0000-0001-8715-9644"
    },
    {
      "affs": [
        0,
        5
      ],
      "name": "Rong Wang",
      "orcid": "0000-0002-0225-1191"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/5se9ezwcjtqqx.bib",
  "cited_by": 0,
  "doi": "10.1097/md.0000000000051182",
  "fulltext_url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13727241/?tool=EBI",
  "id": "5se9ezwcjtqqx",
  "issns": [
    "0304-5412",
    "1536-5964",
    "0025-7974"
  ],
  "landing_url": "https://europepmc.org/article/PMC/PMC13727241",
  "language": "eng",
  "licence": "CC-BY",
  "licence_label": "CC BY",
  "licence_source": "article",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [
    [
      "heading",
      "4700"
    ],
    [
      "heading",
      "Observational Study"
    ],
    [
      "heading",
      "Research Article"
    ]
  ],
  "openalex": null,
  "pdf_url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13727241/pdf/?tool=EBI",
  "pmcid": "PMC13727241",
  "pmid": null,
  "published": "2026-10-30",
  "publisher": "Wolters Kluwer Health",
  "ris": "https://research.draconitic.com/paper/5se9ezwcjtqqx.ris",
  "sources": [
    {
      "first_seen": "2026-11-03T15:20:51Z",
      "native_id": "PMC:PMC13727241",
      "source": "europepmc",
      "url": "https://europepmc.org/article/PMC/PMC13727241"
    },
    {
      "first_seen": "2026-11-05T12:23:21Z",
      "native_id": "PMC13727241",
      "source": "pmc",
      "url": "https://doi.org/10.1097/md.0000000000051182"
    }
  ],
  "subjects": [
    {
      "field": "medicine",
      "origin": "journal-openalex:F27",
      "subfield": "general"
    }
  ],
  "title": "Development and interpretable validation of a machine-learning model for edentulism risk in US adults : A cross-sectional study integrating five algorithms and dietary data from NHANES 2009–2014",
  "title_text": "Development and interpretable validation of a machine-learning model for edentulism risk in US adults : A cross-sectional study integrating five algorithms and dietary data from NHANES 2009–2014",
  "type": "peer-reviewed",
  "updated_at": "2026-11-06T12:22:57Z",
  "url": "https://research.draconitic.com/paper/5se9ezwcjtqqx",
  "venue": "Medicine",
  "versions": []
}