1. Caries Risk Prediction Models in a Medical Health Care Setting
- Author
-
Rahul Nair, Pui Ling Chay, S-M Saw, Y.S. Chong, Tosha Ashish Kalhan, C. Un Lam, Yung Seng Lee, Jonathan Y Huang, L. Shek, Kenneth Kwek, Fabian Yap, Bindu Karunakaran, Chin-Ying Stephen Hsu, Kok Hian Tan, M.C.F. Fong, C.K. Chng, and Keith M. Godfrey
- Subjects
Pediatrics ,medicine.medical_specialty ,Dental Caries ,Logistic regression ,Article ,Healthcare improvement science Radboud Institute for Health Sciences [Radboudumc 18] ,Cohort Studies ,03 medical and health sciences ,0302 clinical medicine ,Risk Factors ,Health care ,Medicine ,Humans ,030212 general & internal medicine ,Imputation (statistics) ,Tooth, Deciduous ,General Dentistry ,Receiver operating characteristic ,business.industry ,030206 dentistry ,Confidence interval ,Logistic Models ,Cohort ,business ,Risk assessment ,Delivery of Health Care ,Cohort study - Abstract
Contains fulltext : 225413.pdf (Publisher’s version ) (Closed access) Despite development of new technologies for caries control, tooth decay in primary teeth remains a major global health problem. Caries risk assessment (CRA) models for toddlers and preschoolers are rare. Among them, almost all models use dental factors (e.g., past caries experience) to predict future caries risk, with limited clinical/community applicability owing to relatively uncommon dental visits compared to frequent medical visits during the first year of life. The objective of this study was to construct and evaluate risk prediction models using information easily accessible to medical practitioners to forecast caries at 2 and 3 y of age. Data were obtained from the Growing Up in Singapore Towards Healthy Outcomes (GUSTO) mother-offspring cohort. Caries was diagnosed using modified International Caries Detection and Assessment System criteria. Risk prediction models were constructed using multivariable logistic regression coupled with receiver operating characteristic analyses. Imputation was performed using multiple imputation by chained equations to assess effect of missing data. Caries rates at ages 2 y (n = 535) and 3 y (n = 721) were 17.8% and 42.9%, respectively. Risk prediction models predicting overall caries risk at 2 and 3 y demonstrated area under the curve (AUC) (95% confidence interval) of 0.81 (0.75-0.87) and 0.79 (0.74-0.84), respectively, while those predicting moderate to extensive lesions showed 0.91 (0.85-0.97) and 0.79 (0.73-0.85), respectively. Postimputation results showed reduced AUC of 0.75 (0.74-0.81) and 0.71 (0.67-0.75) at years 2 and 3, respectively, for overall caries risk, while AUC was 0.84 (0.76-0.92) and 0.75 (0.70-0.80), respectively, for moderate to extensive caries. Addition of anterior caries significantly increased AUC in all year 3 models with or without imputation (all P < 0.05). Significant predictors/protectors were identified, including ethnicity, prenatal tobacco smoke exposure, history of allergies before 12 mo, history of chronic maternal illness, maternal brushing frequency, childbearing age, and so on. Integrating oral-general health care using medical CRA models may be promising in screening caries-susceptible infants/toddlers, especially when medical professionals are trained to "lift the lip" to identify anterior caries lesions.
- Published
- 2020