Demystifying DMFT/DMFS: Pragmatic nationwide assessment of dental caries outcomes from NHANES

Project Summary. Dental caries continues to remain a major public health burden in the US. Future dental
treatment plans are expected to benefit from development of advanced statistical methods for efficient caries
experience (CE) assessment by integrating various statistical challenges encountered in discrete caries data.
Publicly available nation-wide survey databases, such as the NHANES, are important, but somewhat under-
utilized resources for such evaluations and practical interpretations, mainly due to some unique statistical
complexities posed by these large databases, including unequal sampling weights, data granularity, complex
hierarchical structure, and possible spatial association (between tooth surfaces) determining disease
progression. The popular DMFT/DMFS summary indices for assessing CE are plagued with various
inconsistencies. Furthermore, these databases contain a large number of CE predictors, and confounders (such
as socio-demographics, smoking status, etc), leading to further complexities in conducting prudent cross-
sectional evaluation. Effective analysis and pragmatic summarization of the cross-sectional association between
a variety of CE factors and caries outcomes, incorporating the aforementioned complexities within a unified
paradigm are often beyond the capabilities of existing statistical tools and software packages, available for
complex surveys. Furthermore, those currently available statistical methods for handling spatial discrete
responses might be computationally prohibitive for analyzing large observational data, such as NHANES. In this
project, we address these challenges, and propose a single-number, nationally-representative summary of CE
index, from about 24,500 adult dentate subjects, who are part of the nationwide NHANES 2011- Mar 2020
complex survey. We further refine this index, and present a pragmatic validation using the NHANES database.
We have the following 3 high-impact aims: (1) construct and validate a pragmatic, interpretable, single index
model using DMFS/T counts, (2) refine and validate the index for spatially-referenced tooth-surface level binary
D/M/F/S outcomes, and finally (3) produce free software (R package) for the estimation and computation of this
personalized CE index for any future patient. This proposal will generate new knowledge on national-level CE
evaluation via the development of this comprehensive and unique index, which will also unravel the complex
covariate-response relationship to assess tooth-surface-level CE by moving away from the traditional DMFT/S
summary indices. Our fully generalizable methods are also expected to stimulate future research in developing
principled data-analytic tools for understanding burden of other diseases, and related comorbidities.

Grant Number: 1R21DE034124-01A1
NIH Institute/Center: NIH
Principal Investigator: Dipankar Bandyopadhyay