Providers Against Cavities in Children's Teeth

Study Design:

This study will utilize a cluster randomized clinical trial design (Phase III) in primary care settings. Eighteen practices will be randomized to 2 arms: A) provider-level CSM theory-based didactic and skills training to deliver oral health facts to parents, a prescription and a list of dentists accepting Medicaid + practice-level EMR changes for documenting oral health; B) AAP based didactic training with no provision of resources or changes to the practice EMR. Arm A parents/caregivers will receive oral health facts and prescription to take their child to the dentist and improve oral health behaviors in the home, while Arm B parent/caregivers will receive usual AAP-based care for oral health. Each arm will consist of 9 practices (n= 18), 33 to 34 providers (n= 67), and 512 parent/caregiver and child dyads (n= 1024). Each parent/caregiver and child dyads will be recruited at the first WCV and then followed for two consecutive WCVs (for a 24 month duration). Each provider will complete training prior to enrolling any parent/caregiver and will participate in the study for a total of 24 months duration. Immediately after randomization of practices, recruitment will be rolled-out, i.e. parent/caregivers will be recruited during a 3-month period in 6 practices at a time, with recruitment at all 18 practices expected to be completed in 9 months.

The primary outcome is receipt of dental care assessed through data abstracted from Medicaid claims files, clinical dental screenings and parent/caregiver Dental Attendance Questionnaire responses. The secondary outcomes are development of new caries, changes in oral health behaviors and oral health quality of life, dental care costs, and implementation of the interventions.

Participants:

Subjects will be pediatric providers (Pediatricians/Nurse Practitioners) and parent/caregivers and their 3-6 year old Medicaid-enrolled children from 18 primary care practices located in 6 counties in NE Ohio. The study is offered to all pediatricians/nurse practitioners in the recruited practices and will be offered to all eligible caregivers and their children excluding those with serious medical or behavioral conditions which would preclude them from participating in the dental screening. All provider and parent participants meeting the eligibility criteria will be enrolled in the study upon signing the consent form.

Procedures:

WCV #1 (Baseline): Before WCV #1, Providers (pediatrician/nurse practitioners) will receive oral health didactic education and skills training (based on study arm) to communicate core OH facts to parents/caregivers. They will complete pre- and post-tests before and after the OH didactic education session. During the well-child visit, caregivers will complete the following self-administered Baseline questionnaires: Illness Perception Questionnaire-Revised for Dental (IPQ-RD) and Parent Questionnaire. A dental hygienist will perform the child dental screening examination and study staff will record results on the ICDAS Form. During the WCV, the provider will deliver oral health facts, give a prescription to take the child to the dentist + list of local Medicaid-accepting dentists, and document oral health in EMR, based on study arm. Following the provider encounter, caregivers will provide feedback about the OH information given to them during the visit with a short self-administered Exit Questionnaire. At the end of the WCV, caregivers will be given the Follow-up IPQ-RD to be completed and returned within 2 weeks (in postage paid envelope). At six months, caregivers will report whether the child visited the dentist and also complete an annotated cost questionnaire.

WCV #2 (12 month follow-up): Before WCV #2 Providers will receive an OH didactic education booster session.

During and after the well child visit, providers and parent/child dyads will complete the same assessments and procedures done in WCV #1.

WCV #3 (24 month follow-up): There is no provider education booster session before WCV #3.

During the well child visit, providers and parent/child dyads will complete the same assessments and procedures done in WCV #1 and #2, except the IPQ-RD follow-up questionnaire. The 6 month assessments will not be completed during the third well child visit.

Analysis Plan:

Primary Statistical Analysis: For the primary outcome, the investigators will use as an overall dental attendance score the number of years (over the 24 months of follow-up) in which the child visited the dentist. This will be an ordinal outcome with possible scores of 0, 1 or 2. To assess the intervention effect, the investigators will use a generalized estimating equations (GEE) approach, with practices as clusters, based on a proportional odds marginal model. The model covariates will include an intervention indicator variable (equal to 1 for bundled intervention, 0 for enhanced usual care) and a set of baseline variables representing potential confounders. A standard error correction (for example, the method by Morel et al. 2003) will be used to adjust for a small number of clusters and 95% confidence intervals will be computed. This will be an intent-to treat analysis as all randomized participants providing the necessary measurements - regardless of any lack of compliance - will be included in the analysis.

In the event of missing data (for either year) for dental attendance, the investigators will conduct sensitivity analyses by imputing responses under conservative assumptions (favoring the null hypothesis) and re-running the analysis described above on the completed data.

Analysis of Secondary Outcomes: Summary statistics (including means and standard errors) for secondary outcomes will be calculated by intervention group. The same approach as above will be used for binary or ordinal secondary outcomes (oral hygiene, frequency of sweet snacks and beverages). Namely, ordinal outcomes for each variable will be defined that summarize outcomes over time. For continuous outcomes (e.g., OH quality of life, cost), the above method will be modified by using a linear model (identify link) for GEE, modeling the mean response over time as a summary measure. These outcomes will each be tested for normality using the Shapiro-Wilk statistic; outcomes for which normality is violated will be transformed where appropriate or an alternative model used. For count outcomes (e.g., dft accumulated over time), the investigators will use a loglinear model (log link) assuming a negative binomial or other appropriate (e.g., zero-inflated negative binomial) distribution. For proportion outcomes (e.g., dt/dft), the investigators will use GEE with a logit link, assuming the proportion follows a beta binomial or zero-inflated beta binomial distribution. Similarly, the investigators will fit appropriate GEE models to test for the effect of each implementation strategy on the corresponding outcome (e.g., % prescriptions given as a measure of adoption). As in the dental attendance analysis, the intervention indicator as well as pertinent baseline variables will be included in the model. In addition, interaction terms (baseline variables by intervention) will be included to test for possible effect modification.

For secondary analyses, a GLIMMIX model approach will be considered as an alternative, which may more easily allow for more than one cluster level if needed. Another alternative approach is to model the repeated measurements (again using GEE or GLIMMIX) which will add an additional cluster level - namely, for individuals). Goodness of fit of alternative models will be compared using QIC for GEE (or AIC for GLIMMIX).

Missing Data: In the likely event of missing responses, the investigators will first assess (In the context of repeated measures analyses) whether the data are missing completely at random (MCAR), that is, whether missingness of the given outcome is dependent only on participant baseline characteristics and not further on the observed outcome at an earlier time. This will be done by modeling missing data indicators for the repeated measurements of each outcome using a GEE (or GLIMMIX) approach with a logit link and including appropriate baseline (control) variables and the outcome at the previous time if available. The MCAR null hypothesis will be rejected if the previous outcome has a statistically significant effect on the probability of missing. A nonsignificant effect would support the use of GEE (which assumes MCAR). In addition to assessing the overall effects of the interventions, the investigators will investigate the mechanisms (or paths) through which interventions impact dental attendance.

Data Management:

The study staff will collaborate and interact with the NIH-appointed Coordinating Center (CC) to perform data management and quality control activities. Study data will be collected and stored using the REDCap platform hosted by University of California-San Francisco, the home institution of the CC. REDCap is a secure, web-based application designed to support remote data capture for research studies.

Study forms will be completed by participants on paper, and subsequently entered into REDCap by study staff, or on a tablet directly into REDCap. Paper forms will be securely stored in a locked file cabinet. Recorded audio will be deleted from the digital recording device immediately after being stored on a secure CWRU School of Dental Medicine network drive.

Data for this study will include: (1) dental screening data, (2) study questionnaires, (3) abstracted medical data, (4) abstracted Medicaid dental claims data, (5) cost data (6) data from observation/audiotaping of providers, and (7) EMR audit data. Additionally, audio recordings will be used for fidelity monitoring. Form revisions should be minimal; however, should they occur, changes will be submitted to the CC for updating and dissemination to study staff.

Quality control is primarily conducted at the study team level through internal processes of data review/data monitoring using periodic custom reports generated by the CC. The CC will assist with the design of project-specific custom reports. The CC will run regular validation reports to detect data anomalies and will work with the local project staff to resolve any data anomalies that arise during data entry. REDCap's native data resolution workflow will be used to document and fix any data anomalies. The Data Manager will also respond to data queries generated by the PI, Study Coordinator, or other study staff.

The CC will generate regular reports showing enrollment and potential data anomalies, which will be sent to PIs, Project Coordinators, and other relevant study staff.