Factors associated with cost conversations in oral health care settings

 Setting, population, and sample

Our research team partnered with Qualtrics to recruit 500 participants to complete a self-administered online survey. Sample size determination is less straightforward for nonprobability samples, and there is no widely accepted guideline. A probability-based sample of 385 would provide sufficient power for a prevalence estimation precision rate of plus or minus 5% at a 95% confidence level. Using this as a reference point, a sample size of 500 was identified as adequate for balancing the power to detect medium differences with survey cost and providing some cushion for nonresponse.

 Recruitment

Qualtrics relies on actively managed opt-in market panels for study recruitment, including niche panels for specialized recruitment campaigns. Potential participants are invited to participate via email and are incentivized via Qualtrics. Internet protocol address tracking and digital fingerprinting technology are leveraged to prevent duplication.

 Data collection

Data were collected using an online self-administered survey developed by researchers on the basis of a review of the cost conversations literature. The survey was developed specifically for our study and, as such, had not been validated previously. To minimize survey fatigue and obtain trustworthy answers, the survey was designed to be short (facilitated using skip patterns where applicable); average completion time was 6 minutes.

To ensure data quality, Qualtrics’ internal processes facilitate checks for evidence of satisficing, including short completion times and straightlining in questions using matrix style. Such observations are excluded and do not count toward the contractual final sample determination. For our study, there was 1 such observation. However, this respondent also responded inaccurately to the attention checks and was already accounted for in the surveys that had been deleted for data quality issues, as described previously.

 Measures

Responses to these 2 statements were combined into a single binary measure (yes, no) of cost conversation. Participants responding in the affirmative to either of these statements were deemed to have had a cost conversation during their last dental visit. Participants responding in the negative for both statements were deemed not to have had a cost conversation.

Assessed covariates included patient demographic and provider information. Demographic information included gender (female, male, nonbinary), age (18-24 years, 25-34 years, 35-44 years, 45-54 years, 55-64 years, ≥ 65 years), race or ethnicity (non-Hispanic White, non-Hispanic Black, non-Hispanic other, Hispanic), personal income (< $25,000, $25,000 through $49,999, $50,000 through $74,999, $75,000 through $99,000, ≥ $100,000), location of residence (county with population of < 50,000, county with population of ≥ 50,000), dental insurance coverage (yes, no), experience of financial hardship (yes, no), and when the dental visit occurred (within 6 months, 7-12 months ago, 13-24 months ago). Provider characteristics included the type of practice in which the visit occurred (public, private), type of provider seen (general dentist, specialist dentist, dental hygienist or other), and whether the patient was screened for financial hardship during the visit (yes, no).

 Data analysis

Participants who reported obtaining care from a free clinic (n = 28) were excluded for the purposes of our study owing to its focus on cost. After the exclusion of observations with missing data on the study variables, the final analytical sample for our study was 370, slightly fewer than the anticipated 385. However, had this been a probability-based study, the sample size of 370 would provide sufficient power for a prevalence estimation precision rate of plus or minus 5.1% at a 95% confidence level, just a little over the 5% margin of error estimated for a sample of 385.

Variables were described using frequencies. Bivariable associations were assessed using χ² tests. Multivariable logistic regression was used to identify the patient and provider characteristics independently associated with the occurrence of cost conversations. A hierarchical model building approach was adopted in which patient characteristics, provider characteristics, and financial hardship screening information were added in blocks sequentially (models 1-3, respectively). Statistical significance was assessed at the P < .05 level using 2-tailed tests. Robust SEs were obtained and clustered at the state level. All analyses were performed using STATA, Version 16 (StataCorp).

 Characteristics of survey respondents

 Occurrence of cost conversations in the dental setting

 Factors associated with cost conversations in the dental setting

 Limitations

Our study was a cross-sectional study, and therefore we cannot assert causal associations between each of the assessed correlates of cost conversations and the occurrence of cost conversations. Furthermore, we recruited participants using a nonprobability sampling approach from an opt-in market panel, resulting in potential selection and voluntary response bias. The extent to which the findings may be generalized to the broader population, especially older adults, may be considerably limited.

There were also some limitations with the survey instrument used in our study, including lack of previous validation. Participants could have also responded to questions related to cost conversations with different frames of reference, leading to potential bias. In addition, the questions used in assessing cost conversations focused on discussions related solely to medical costs. Furthermore, the quality checks used in the survey were placed at the beginning rather than interspersed within the survey, which allowed us to immediately exclude those who did not have serious intentions about the survey from the start. However, given the brevity of the survey, we believe quality issues due to survey fatigue were of limited concern. Collectively, these survey-related factors could have potentially influenced the validity of the responses and our resulting estimates.

In addition, we may not have accounted for all potential correlates of cost conversations, including patients’ education or employment and providers’ attitudes and beliefs toward cost conversations. In addition, detailed information on the type of dental visit the patient had (for example, diagnostic visit, preventive or hygiene visits, treatment visits, or reviews) or the site of care (including whether care was received in a hospital-based or dental education institution–based dental facility) were not obtained.

Despite these limitations, our study represents one of the only attempts to our knowledge to examine cost conversations within the dental setting. Because we took an empirical approach, our study results add to a cost conversation evidence base that has mostly been dominated by qualitative studies. In future studies, researchers can extend this exploratory study using nationally representative and probability-based sampling approaches to confirm the broad applicability of our study’s findings. Furthermore, additional studies are needed to develop valid measurement instruments for assessing the occurrence and quality of cost conversations.