J D Bader1, D A Shugars. 1. Sheps Center for Health Services Research, University of North Carolina, Chapel Hill 27599-7590, USA. jim_bader@unc.edu
Abstract
OBJECTIVES: This study developed descriptive models of dentists' restorative treatment decisions for individual teeth. Such models could be useful in personnel planning, in assessing the effects of dental treatment programs, and in furthering understanding of dentists' decision-making processes. METHODS: Logistic regression was used to construct models of the probability of individual teeth receiving a recommendation for restorative treatment. Independent variables for the models were data from epidemiologic oral examinations and self-administered questionnaires of subjects who were seeking treatment at a dental school. Data for the dependent variable, the probability of treatment, were collected from multiple dentists' treatment plans of these subjects. Separate models were constructed for molar, premolar, and anterior teeth. An assessment of the models' utilities in a different population consisted of comparing the treatment probabilities estimated by the models with those actually experienced by a community sample of 317 individuals who visited dentists in the 18 months following our examination. RESULTS: Constructed models for molar, premolar, and anterior teeth returned kappa values of 0.60, 0.62, and 0.65, respectively, for the original data set. The models were less accurate in identifying which teeth received treatment among subjects in the community sample, with kappas of 0.10, 0.18, and 0.20, respectively. CONCLUSIONS: Models of dentists' restorative treatment decision making based on clinical and nonclinical data can determine the probability of treatment for individual teeth with reasonable accuracy. Hence, the approach holds promise for developing measures of normative treatment need. However, the models are not accurate predictors of dichotomous decisions by individual dentists regarding treatment interventions. Both differences in the subject samples used to develop and assess the models and individual dentist idiosyncrasies may contribute to this inaccuracy.
OBJECTIVES: This study developed descriptive models of dentists' restorative treatment decisions for individual teeth. Such models could be useful in personnel planning, in assessing the effects of dental treatment programs, and in furthering understanding of dentists' decision-making processes. METHODS: Logistic regression was used to construct models of the probability of individual teeth receiving a recommendation for restorative treatment. Independent variables for the models were data from epidemiologic oral examinations and self-administered questionnaires of subjects who were seeking treatment at a dental school. Data for the dependent variable, the probability of treatment, were collected from multiple dentists' treatment plans of these subjects. Separate models were constructed for molar, premolar, and anterior teeth. An assessment of the models' utilities in a different population consisted of comparing the treatment probabilities estimated by the models with those actually experienced by a community sample of 317 individuals who visited dentists in the 18 months following our examination. RESULTS: Constructed models for molar, premolar, and anterior teeth returned kappa values of 0.60, 0.62, and 0.65, respectively, for the original data set. The models were less accurate in identifying which teeth received treatment among subjects in the community sample, with kappas of 0.10, 0.18, and 0.20, respectively. CONCLUSIONS: Models of dentists' restorative treatment decision making based on clinical and nonclinical data can determine the probability of treatment for individual teeth with reasonable accuracy. Hence, the approach holds promise for developing measures of normative treatment need. However, the models are not accurate predictors of dichotomous decisions by individual dentists regarding treatment interventions. Both differences in the subject samples used to develop and assess the models and individual dentist idiosyncrasies may contribute to this inaccuracy.
Authors: Michael S McCracken; David R Louis; Mark S Litaker; Helena M Minyé; Rahma Mungia; Valeria V Gordan; Don G Marshall; Gregg H Gilbert Journal: J Am Dent Assoc Date: 2016-08-01 Impact factor: 3.634
Authors: Theodorus G Mettes; Wil J M van der Sanden; Michel Wensing; Richard P T M Grol; Alphons J M Plasschaert Journal: Implement Sci Date: 2007-04-20 Impact factor: 7.327