Providers’ preferences for pediatric oral health information in the electronic health record: a cross-sectional survey

In an effort to increase the number of young children in North Carolina (NC) Medicaid who have a dental home, we disseminated a decision tool to improve oral health screening, risk assessment and referrals in medical offices. As part of the evaluation of this initiative [17], we developed a survey to assess primary care providers’ oral health promotion and disease prevention activities for infants and toddlers (children under the age of 4 years). Additionally, the survey examined the availability of EHRs for well-child visits, participation in meaningful use incentive programs, and provider information needs and preferences for oral health and other preventive services for well-child visits. We received Institutional Review Board approval from the University of North Carolina at Chapel Hill (IRB study #07–1942).

We surveyed primary care physicians in NC who provided care for Medicaid-enrolled children younger than 4 years of age from August – December 2013. Physicians who did not conduct well-child visits for this aged child, practiced in a tertiary academic health center or community clinic, or were not involved in any patient care were excluded from the study.

We developed the sampling frame using multiple sources of information including the National Plan and Provider Enumeration System [18], the NC Health Professions Data System [19], and NC Medicaid well-child visit data and Into the Mouths of Babes program participation records [20]. We verified the data and identified additional primary care practices and physicians by conducting online searches and making phone calls to practices. The final sampling frame included 1364 primary care physicians in 435 practices. We received a response from 50.3% or 219 of the 435 practices. We randomly selected one physician per practice to respond to the survey. If the selected physician did not respond, we randomly selected another physician from the same practice. We ensured that physicians who worked at multiple practices were surveyed only once.

We piloted the questionnaire with providers in 11 primary care practices participating in another study [21]. Sampled physicians were mailed up to three requests for participation via U.S. mail. To potentially reduce the non-response rate, we provided physicians with two options for completing the survey–a paper survey using a pre-paid envelope or an online survey developed using Qualtrics Survey Software (Provo, UT). Respondents were entered into a drawing for one of five Kindle Fire HD e-readers (a value of $200 at the time of survey administration).

Prior studies have shown that practice characteristics, such as practice ownership, size, and urban location, affect primary care providers’ oral health activity for children [22‐24]; therefore, we collected these data for our sample of providers. Practice ownership was coded as a categorical variable that included physician or physician group owned, academic medical center, non-academic affiliated hospital, and other. Practice size was measured as the number of physicians within the practice and was treated as a continuous variable. We transformed the zip code of the practice into a rural-urban commuting area code [25] and categorized the zip codes into urban and rural. Additionally, we included two binary variables including whether the practices used EHRs to conduct well-child visits and whether practices exclusively used an electronic system.

We collected information on provider characteristics, including proportion of pediatric patients seen per week, oral health activities performed, and years since graduation from medical school. We hypothesized that the proportion of pediatric patients seen per week and the amount of oral health screening and dental referral activity would be positively associated with providers’ information preferences. We measured the proportion of pediatric patients as the ratio of pediatric patients (under age 4) to the total number of patients seen per week. We measured the amount of screening activity and oral health referral activity by asking physicians to estimate the percentage of all well-child visits (0%, 1–10%, 11–25%, 26–50%, 51–100%) in which they perform these activities. We also included years since graduation from medical school as a proxy for age because age is negatively associated with EHR adoption [26].

To assess providers’ oral health information preferences, we developed survey items based on the American Academy of Pediatrics’ clinical guidelines for infant and toddler oral health and recommendations from the U.S Preventive Services Task Force [27, 28]. Ten items assessed the importance (i.e., not important, somewhat important, or very important) providers place on an EHR containing oral health information for (1) risk assessment, such as listing risk factors for tooth decay; (2) intervention, such as listing prescriptions for fluoride supplements; and (3) referrals to a dentist.

To determine appropriateness of reducing any of the oral health information preferences survey items into a composite measure, we conducted a principal component analysis of the 10 items. We applied two decision rules to determine whether there was sufficient evidence for combining survey items into a composite index including a Kaiser-Meyer-Olkin Measure and a Bartlett’s Test of Sphericity [29]. We conducted a parallel analysis test to determine the number of factors to retain by comparing the observed eigenvalues extracted from the correlation matrix analyzed with those obtained from uncorrelated normal variables [30]. Based on the results, we retained one factor. We used factor scores from the principal components as weights, and a final oral health-information-preference composite index, ranging from 0 to 10, was constructed from the 10 items. The mean score was 7.13 (SD 2.19).

We asked providers about the importance of EHR information about other preventive aspects of well-child visits using the same 3-level response options as used for the oral health items. We developed these items based on recommendations from the American Academy of Pediatrics clinical guidelines for well-child visits [27],—specifically, how important it is for the EHR to plot growth charts and calculate height, weight, and body mass index (BMI); track adherence to well-child visits; track immunizations; calculate weight-based dosing; and calculate catch-up immunizations.

We used descriptive statistics to assess information preferences for oral health and other preventive aspects of well-child visits. Furthermore, we analyzed associations between the oral health-information-preference composite index and key provider- and practice-level characteristics using an ordinary least squares regression model with bootstrapped standard errors. Since only one physician per practice was sampled, we assumed observations were independent and did not control for potential clustering effects. We ran three specifications of the model–one with a linear version of the dependent variable, one with a logarithmic version of the dependent variable, and one with the logarithmic version of the independent and dependent variables as a sensitivity analysis. We compared the results across the three models to ensure that estimates were robust and not sensitive to model specification. Since all three models produced similar estimates with the same level of statistical significance, we report the findings of the linear model for ease of interpretation. To assess whether missing values were missing at random, we compared the characteristics of individuals with and without missing data for the main variables of interest and did not find significant differences in characteristics. Therefore, we dropped missing cases from the model, reducing the sample size from 221 to 211. For these analyses, we used the statistical software Stata, version 13.0.

The analytical sample included 211 providers, 95.9% of sampled physicians. The majority of physicians worked in a practice owned by a physician or physician group (73.5%), and a practice located in an urban area (87.7%) (Table 1). Nearly 80% of physicians reported exclusively using an electronic EHR system for conducting well-child visits. On average, physicians worked in practices with 3.2 (SD 2.4) other physicians. Most physicians reported screening for oral health problems (89.6%) during at least half of well-child visits with infants and toddlers, and 51.2% reported making an oral health referral in at least half of well-child visits. The mean percentage of all patients seen per week who were infant or toddler was 48.0%.

Table 2 summarizes results about preferences for oral health information in the EHR. The largest percentage of physicians indicated that tracking topical fluoride applications was very important (69.2%). The smallest percentage of physicians indicated that providing test results for fluoride content of drinking water (31.3%) was very important.

Table 2 also summarizes preferences for having non-dental preventive well-child information in the EHR. The majority of physicians identified each of these elements as being very important, with the largest percentage of physicians indicating that tracking immunizations (94.3%) was very important and the lowest percentage indicating that calculating weight-based dosing (76.8) was very important. By comparison, this measure was rated as very important by more respondents than the highest-rated type of oral health information (tracking topical fluoride applications, 69.2%).

Table 3 provides results for the regression model examining the association between the composite index scores and provider and practice characteristics. Among provider characteristics, percentage of pediatric patients, oral health referral activity, and oral health screening activity were significantly associated with oral health information preferences. Specifically, holding all else constant, a one percentage point increase in the percentage of toddler and infant patients was associated with an approximately 13.3 percentage point increase in the reported importance of oral health information in the EHR (p = 0.017). Compared to physicians who conducted oral health referrals in less than 25% of well-child visits, physicians who conducted oral health referrals in more than 51% were associated with a higher reported importance for oral health information in the EHR (p = 0.014). Similarly, physicians who conducted oral health screenings in more than 51% of well-child visits reported significantly higher importance for oral health information as compared to physicians who conducted oral health screenings in less than 25% of well-child visit (p = 0.013). We found that other provider characteristics, such as years since graduation from medical school and exclusive use of an EHR system for well-child visits were not significantly associated with oral health information preferences. Also, we did not find significant associations between oral health information preferences and practice characteristics, such as size, rural location, and ownership.

This study was limited to Medicaid providers of services for children younger than 4 years of age in NC. Because NC was an early adopter of Medicaid reimbursement policies for preventive oral health services [37], NC-based physicians may have greater experience with oral health service delivery than physicians in other states, hindering the generalizability of our results. However, physicians with experience providing pediatric oral health services are better positioned to judge which elements of oral health information would be useful to support oral health screening and dental referral activity. Similarly, most of the practices in the sample were located in an urban area (87.7%), owned by a physician or physician groups (73.5%), and exclusively used EHRs for conducting well-child visits (80.6%). As a result, practice patterns and information preferences may not be generalizable to all primary care practices. Additionally, the survey did not collect information on availability of pediatric-specific information within the practice’s current EHR system, whichmay be an omitted variable from the OLS model. It is possible that preferences for oral health information could be a function of a providers’ current access to oral health information. In other words, the study could not identify whether the practices in the sample had protocols for oral health screenings, services, or referrals, and if documenting these activities was part of usual care. Omitting this variable could explain, at least in part, why our model did not account for more than 18% of the variation in practitioner responses. Nonetheless, this study makes a contribution to the literature by identifying primary care providers’ oral health information preferences in the EHR and provides evidence for future researchers to build upon.