Development and internal validation of a multivariable logistic model for pulp disease in children from the central Peruvian Amazon: a secondary analysis of a case–control study
Abstract Background Pulp disease is a frequent and largely preventable consequence of untreated dental caries in children, yet its determinants are seldom modelled jointly in socioeconomically vulnerable settings, and probability models that could support screening in primary care are rarely develop...
Guardado en:
| Autores principales: | , |
|---|---|
| Formato: | Preprint |
| Lenguaje: | inglés |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | Enlace permanente (DOI o Handle) Descargar el archivo |
| Etiquetas: |
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | Abstract Background Pulp disease is a frequent and largely preventable consequence of untreated dental caries in children, yet its determinants are seldom modelled jointly in socioeconomically vulnerable settings, and probability models that could support screening in primary care are rarely developed or validated. We aimed to identify factors independently associated with pulp disease and to develop and internally validate a multivariable logistic model estimating its probability in children from the central Peruvian Amazon. Methods This was a secondary analysis of the individual-level dataset of a case–control study conducted in 2023 at the primary health care centre of Pichanaki (Chanchamayo, Junín, Peru), comprising 270 children aged 5 to 11 years: 90 cases with pulpal or periapical pathology diagnosed on clinical and radiographic grounds and 180 controls without pulpal involvement. Biological (anemia, hemoglobin concentration, nutritional status), behavioural (fluoride toothpaste, mouthwash and dental floss use, ferrous sulfate intake) and socioeconomic variables (maternal educational level and occupation, maternal age, household income) were retrieved from a previously collected and validated database. Variables with p < 0.20 in bivariable analysis, together with clinically relevant covariates, were entered into a multivariable binary logistic regression model; multicollinearity was assessed with the variance inflation factor. Discrimination was quantified with the area under the receiver operating characteristic curve (AUC), calibration with the Hosmer–Lemeshow test and a calibration curve, and internal validity with a 70/30 training–test split and k-fold cross-validation. The optimal classification threshold was derived from the Youden index, and performance was re-estimated across alternative thresholds. Results Dental caries (adjusted odds ratio [aOR] 12.08, 95% confidence interval [CI] 3.85–44.62), ferrous sulfate intake (aOR 11.63, 95% CI 3.85–40.76), anemia (aOR 5.73, 95% CI 1.99–18.14) and non-professional maternal educational level (aOR 3.83, 95% CI 1.36–11.82) remained independently associated with higher odds of pulp disease, whereas fluoride toothpaste use (aOR 0.09, 95% CI 0.03–0.25), dental floss use (aOR 0.13, 95% CI 0.04–0.36) and mouthwash use (aOR 0.30, 95% CI 0.10–0.83) were inversely associated. Maternal age below 20 years was no longer significant after adjustment (aOR 2.34, 95% CI 0.73–7.81). The final model was adequately calibrated (Hosmer–Lemeshow χ² = 6.11; p = 0.635; Nagelkerke R² = 0.817) and discriminated well (AUC 0.974, 95% CI 0.958–0.990), with an accuracy of 93.3%, sensitivity of 90.0% and specificity of 95.0% at the 0.50 threshold. Internal validation yielded an AUC of 0.975 in the 70/30 split and 0.966 in cross-validation. At the Youden-derived threshold of 0.31, sensitivity reached 96.7% and specificity 92.2%. Conclusions In this vulnerable paediatric population, pulp disease was explained by a combination of caries experience, anemia-related indicators, maternal educational level and everyday preventive behaviours, and these routinely available variables separated affected from unaffected children with high accuracy. Because predictors were recorded concurrently with the outcome within a case–control sample, the model should be regarded as a screening and prioritisation aid rather than a causal or population-calibrated risk tool; external and prospective validation is required before clinical deployment. |
|---|