Using Propensity Score Subclassification to Estimate the Population-Average Causal Effect of Temporomandibular Dysfunction Experience on Oral Health-Related Quality of Life Among Australian Adults

Background: Temporomandibular dysfunction (TMD) experience might impair oral health-related quality of life (OHRQoL). Causal inference using population-based cross-sectional data is challenging given the potential for bias. Propensity Score Subclassification (PS-Subclassification) provides a tool to mitigate confounding bias. The aim of this study was to estimate the Population-Average Treatment Effect (PATE) of having TMD experience among Australian adults on OHRQoL using PS-Subclassification and statistically estimated Minimally Important Differences (MID-S). Method: Australia's National Survey of Adult Oral Health (NSAOH) 2004-06 data were used which included a Computer Assisted Telephone Interview, mailed questionnaire and oral epidemiological examination. Data included demographics, socioeconomics, caries experience (DMFT index), periodontitis, TMD experience using the TMD Diagnostic Criteria Question, the Oral Health Impact Profile (OHIP-14) and perceived stress. Analysis steps included: (1) generating propensity scores (PS) for TMD experience probability using causal model-derived confounders while incorporating survey design elements; (2) PS-Subclassification and weighting; (3) assessing common support and group balance and (4) estimating the PATE for TMD experience on OHIP-14 overall and domains scores using complex samples GLM. Results: Of the 4063 NSAOH participants, 397 with TMD and 3656 without TMD were included in PS-Subclassification (all data were used) and shared common support for their PS and established adequate covariate balance (SMD < 0.2). Experiencing TMD had higher OHIP-14 total scores (B = 3.498, 95% CI: 2.218-4.778) with a small MID-S (Cohen's F2 = 0.03). TMD experience impaired all OHIP-14 domains (p < 0.05) with physical pain and psychological domains among the highest impaired OHIP-14 domains with a small MID-S. Conclusion: TMD experience impaired the overall OHRQoL measured by the OHIP-14 among Australian adults with a small MID-S. Physical pain and psychological domains were among the highest impaired OHRQoL domains with a small MID-S. Clinicians and policymakers might consider these findings to support TMD screening and patient-centred management.