Developing Children's Oral Health Assessment Toolkits Using Machine Learning Algorithm

OBJECTIVES

Evaluating children's oral health status and treatment needs is challenging. We aim to build oral health assessment toolkits to predict Children's Oral Health Status Index (COHSI) score and referral for treatment needs (RFTN) of oral health. Parent and Child toolkits consist of short-form survey items (12 for children and 8 for parents) with and without children's demographic information (7 questions) to predict the child's oral health status and need for treatment.

METHODS

Data were collected from 12 dental practices in Los Angeles County from 2015 to 2016. We predicted COHSI score and RFTN using random Bootstrap samples with manually introduced Gaussian noise together with machine learning algorithms, such as Extreme Gradient Boosting and Naive Bayesian algorithms (using R). The toolkits predicted the probability of treatment needs and the COHSI score with percentile (ranking). The performance of the toolkits was evaluated internally and externally by residual mean square error (RMSE), correlation, sensitivity and specificity.

RESULTS

The toolkits were developed based on survey responses from 545 families with children aged 2 to 17 y. The sensitivity and specificity for predicting RFTN were 93% and 49% respectively with the external data. The correlation(s) between predicted and clinically determined COHSI was 0.88 (and 0.91 for its percentile). The RMSEs of the COHSI toolkit were 4.2 for COHSI (and 1.3 for its percentile).

CONCLUSIONS

Survey responses from children and their parents/guardians are predictive for clinical outcomes. The toolkits can be used by oral health programs at baseline among school populations. The toolkits can also be used to quantify differences between pre- and post-dental care program implementation. The toolkits' predicted oral health scores can be used to stratify samples in oral health research.

KNOWLEDGE TRANSFER STATEMENT

This study creates the oral health toolkits that combine self- and proxy- reported short forms with children's demographic characteristics to predict children's oral health and treatment needs using Machine Learning algorithms. The toolkits can be used by oral health programs at baseline among school populations to quantify differences between pre and post dental care program implementation. The toolkits can also be used to stratify samples according to the treatment needs and oral health status.

Child and Parent Demographic Characteristics and Oral Health Perceptions Associated with Clinically Measured Oral Health

OBJECTIVE

To examine child and parent reports about the child's oral health and assess the associations of these reports with clinical assessments of oral health status by dental examiners.

METHODS

Surveys with 139 items for children and 133 items for parents were administered by Audio Computer-Assisted Self-Interview Software. In addition, the Children's Oral Health Status Index (COHSI) was computed from a dental examination.

RESULTS

A total of 334 families with children ages 8 to 17 y participated at 12 dental practices in Los Angeles County. Ordinary least squares regression models were estimated separately for child and parent surveys to identify items uniquely associated with the COHSI. Ten of 139 items the children reported regarding their oral health were associated with the COHSI. The strongest associations were found for child's age, aesthetic factors (straight teeth and pleased with teeth), and cognitive factors related to perception of dental appearance (pleased/happy with the look of the child's mouth, teeth, and jaws). Nine of 133 parent items about the child's oral health were associated with the COHSI in the parent model, notably being a single parent, parent's gender, parent born in the United States, pleased or happy with the look of their child's teeth, and accessing the Internet.

CONCLUSION

These child and parent survey items have potential to be used to assess oral health status for groups of children in programs and practices in lieu of dental screenings.

KNOWLEDGE TRANSLATION STATEMENT

The paper's results inform the development of a toolkit that can be used by schools, public health agencies, and dental programs to identify children with low oral health status based on parents' and children's responses to survey items across demographic, physical, mental, and social domains. These survey items can be used to inform parents of the desirability of proactively addressing inadequacies in their child's oral health status, enabling them to more rationally address dental needs.

An empirical test of the validity of the Oral Health Status Index (OHSI) on a minority population

Disease varies in different populations based on sociodemographic variables, and there is limited understanding of this interaction. The purpose of this methodological study was to determine the validity of the Oral Health Status Index, a disease-based index, on a Hispanic population by comparing it with the NIDCR epidemiological measures of disease, with the addition of demographic and behavioral variables. The epidemiologic data were collected according to the criteria defined by the NIDCR, including: a modified Decayed Missing Filled Surfaces Index, gingival inflammation, calculus, and destructive periodontal disease measures. The demographic and behavioral variables were gathered from 240 interviews with Hispanic adults in two community clinics. Bivariate analysis was used to determine relationships between the descriptive epidemiologic, demographic, and behavioral variables and the Oral Health Status Index (OHSI). There were statistically significant differences (p < 0.05) in mean OHSI scores among the demographic variables age, education, income, and place of birth; and the behavioral variables alcohol consumption, flossing, and acculturation. Multiple regression analysis with the OHSI as the dependent variable showed that the statistically significant (p < 0.001) epidemiologic predictors were: percentage of Decayed Teeth/Decayed, Filled Teeth; Number of Replaced Teeth/Missing Teeth; and millimeters of mesial attachment loss. These collectively explained 47.49% of the variance in the regression. The addition of demographic variables to the epidemiologic regression identified age (p < 0.05), gender (p < 0.01), and place of birth (p < 0.01) as significant predictors that explained an additional 4.12% of the variance, collectively bringing the total explained variance to 51.61%. The behavioral variables did not contribute significantly to predicting the OHSI regression score. The Oral Health Status Index in this study is validated by its correlation with both the epidemiologic measures and the demographic variables. This combination of variables separated the Hispanics into Mexicans and Central/South Americans.

Comparing measures of reliability for indices of gingivitis and plaque

The purpose of this methodological study was to compare methods used to assess reliability for gingival inflammation and plaque. Duplicate examinations were conducted by one examiner on 17 subjects (506 scoring sites), using the gingival index (GI), bleeding points index (BPI), and plaque index (PI). The percentage of agreement, the weighted and unweighted kappa coefficients, and Pearson correlation coefficients were calculated as statistics of reliability for mesial buccal site scores and whole mouth mean scores when appropriate. For mesial buccal sites the respective values of the GI, BPI, and PI for weighted kappas were: 0.47, 0.49, and 0.75; for the correlation coefficients: 0.47, 0.49, and 0.76; for unweighted kappas, 0.39, 0.49, and 0.39; and for percentage of agreement 66.2%, 76.1%, and 51.2%. For whole mouth means the correlation coefficients for the GI, BPI, and PI were 0.87, 0.59, and 0.87, respectively. In conclusion, the most useful statistics in assessing the intraexaminer reliability of a solo examiner in descending order were the weighted kappa coefficient, Pearson correlation coefficient, the unweighted kappa coefficient, and percentage of agreement.

Managed care and dentistry: promises and problems

At its inception in the mid-1950s, managed care held a number of promises for dental care providers and dental patients. Sometime during the development of managed care, however, many programs lost sight of the importance of provider equity. Using data from current programs, the authors contrast the original promises with the realities of managed care in dentistry.

Evaluating the efficacy of a new flossing aid

The purpose of this clinical trial was to compare the efficacy of a new flossing aid (Flosser) with finger flossing on preventing plaque and gingival inflammation. 35 adults who did not use dental floss routinely were assigned randomly to one of 2 treatment groups (Flosser or finger flossing) in a 2-period, single-blind crossover study. After prophylaxis, subjects were instructed to use the flossing aid or finger floss 1 x per day and to continue brushing for 30 days. Gingival inflammation (GI & BPI) and plaque (PI) were assessed prior to the prophylaxis and at 30 days. After a 30 day "washout" period, subjects were again reassessed for gingival inflammation and plaque, given a prophylaxis, assigned the opposite treatment (2nd treatment period) that they received the first treatment period, and assessed (GI, BPI & PI) after 30 days. Comparing the mean difference of the 30-day buccal interproximal scores between the treatment groups (flossing aid scores minus finger flossing scores) showed that the mean differences with 95% CI were: -0.013 +/- 0.067 [GI], -0.017 +/- 0.044 [BP] and 0.019 +/- 0.014 [PI]. No statistically significant differences from zero (0.05 alpha) were observed using the t-test. There was a high level of compliance (90%) with the prescribed regiment, and subjects preferred (56%) the flossing aid slightly over finger flossing. Even though there were no statistically or clinically significant differences in gingivitis and plaque scores between the 2 flossing groups, the positive inclination for the flossing aid makes it a desirable addition to the armamentarium of preventive dentistry.

Effects of alexidine.2HCL mouthwash on plaque and gingivitis after six months

Two-hundred and fourteen subjects ranging in age from 22 to 63 years used either a mouthwash that contained 0.035% w/v alexidine.2HCL (2-ethylhexyl bisbiguanidine dihydrochloride) or a placebo twice daily on a doubleblind basis for six months in conjunction with routine oral hygiene procedures. The results indicated that alexidine.2HCL was safe and was statistically more effective in producing a clinically significant reduction in gingivitis than was the placebo.

The effect of an antimicrobial mouthwash on dental plaque and gingivitis in yound adults

The effectiveness of an antimicrobial mouthwash (alexidine, 2-ethylhexyl bisbiguanidine dihydrochloride) in preventing plaque and gingivitis was evaluated in 45 young male adults. In a study using a double-blind cross-over design, subjects received a thorough prophylaxis and refrained from all routine oral hygiene for 2 weeks. During this period, subjects rinsed twice daily with either a treatment (alexidine) or placebo mouthwash. Plaque and gingivitis were assessed on selected teeth on days 2, 7 and 14. The characteristics of plaque that were recorded include: surface area score, dry weight and differential counts of microorganisms. After the 14th day, subjects resumed their regular oral hygiene practices for 3 weeks before starting the second experimental period. Aftter a second prophylaxis, the treatment and placebo groups from the first 2 weeks were crossed over and the procedures followed previously were repeated during the sixth and seventh weeks. The alexidine mouthwash was statistically and clinically effective in decreasing plaque scores and plaque weight, and statistically but not clinically effective in decreasing gingivitis scores. No systemic side effects were observed, but an asymptomatic brown tongue stain was observed in about half of the subjects. There was no evidence to suggest that alexidine disturbed the relative composition of the microorganisms in plague, but there was a suggestion that it decreased the number of microorganisms almost 2-fold when compared with the placebo group.