Extracting Dental School Periodontal Records to Measure Treatment Outcomes and Risk Modification

Objectives: The rationale for creating new personalized dental therapies is based upon the observation that not all patients respond favorably to a standardized care. We hypothesized that Latent (Hidden) Class Analysis (LCA) methods using clinical measures to better refine clinical phenotypes (at a 7-level tooth class and 7-level subject class) derived from the Dental Atherosclerosis Risk in Communities Study (DARIC) and applied to harmonize data extracted from dental school periodontal databases might provide insight into creating effective measures of treatment outcomes.
Methods: We accessed electronic health records (EHR over the past 5 years) to identify 1219 patients from the periodontal clinics at UNC. LCA was used to classify patients into 7 patient groups (Periodontal Profile Classes, PPC) and each of their teeth into 7 teeth categories (Tooth Profile Classes, TPC). The 7 TPC classes are Health, Recession, Crowned, Hi GI, Interproximal Disease, Diminished Periodontium, and Severe Disease. All patients had baseline data and follow-up exam data within 270 days, with probing depths (PD) for >80% teeth. A positive tooth-based, treatment effect was considered a mean reduction in PD of 1+mm.
Results: Among patients who received scaling and root planning [ADA D4341/D4342], 35.7% of the teeth with Severe Disease TPC demonstrated a mean improvement of 1 or more mm, vs 66.7% of similar teeth following surgery [ADA D4240-41, D4260-61], p=0.006. By comparison using the ADA/CDC definition of severe disease, 16.7% of the teeth showing a positive response following scaling and root planing vs 17.4% following surgery, p=0.85.
Conclusions: These data suggest that LCA classifications, which consider 7PPCs and 7TPCs, may be more sensitive for detecting certain types of treatment responses than other methods. Furthermore, it suggests that measuring treatment outcomes might be streamlined based upon LCA classifications.