https://orcid.org/0000-0002-3552-6534
Figures
Abstract
Introduction
Some caregivers are hesitant about topical fluoride for their children despite evidence that fluoride prevents caries and is safe. Recent work described a five domain model of caregivers’ topical fluoride hesitancy. We developed the Fluoride Hesitancy Identification Tool (FHIT) item pool based on the model. This study sought to evaluate the FHIT’s psychometric properties in an effort to generate a short, simple to score, reliable, and valid tool that measures caregivers’ topical fluoride hesitancy.
Methods
In 2021 and 2022, we conducted an observational, cross-sectional study of caregivers, collecting data from two independent caregiver samples (n1 = 523; n2 = 612). The FHIT item pool included 33 items. We used confirmatory factor analyses (CFA) to examine whether the FHIT items measured five separate domains as hypothesized and to reduce the number of items. We then fit item response theory (IRT) models and computed Cronbach’s alpha for each domain. Last, we examined the construct validity of the FHIT and evaluated scoring approaches.
Results
After dropping 8 items, CFA supported a five factor model of topical fluoride hesitancy, with no cross-loadings (RMSEA = 0.079; SRMR = 0.057; CFI = 0.98; TLI = 0.98). We further reduced the items to four per domain (20 items total). Marginal alphas showed that the item sets provided reliability of ≥0.90 at hesitancy levels at and above average. The domains correlated more strongly with each other and topical fluoride refusal than with other questions on the survey.
Citation: Carle AC, Pallotto I, Edwards TC, Carpiano R, Kerr DC, Chi DL (2024) Psychometric properties of the Fluoride Hesitancy Identification Tool (FHIT). PLoS ONE 19(1): e0297188. https://doi.org/10.1371/journal.pone.0297188
Editor: Hadi Ghasemi, Shahid Beheshti University of Medical Sciences School of Dentistry, ISLAMIC REPUBLIC OF IRAN
Received: September 14, 2023; Accepted: December 31, 2023; Published: January 17, 2024
Copyright: © 2024 Carle et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data cannot be shared publicly because it contains potentially identifying or sensitive patient information. The Unviersity IRB imposed these restrictions. Data are available by contacting the University of Washington School of Dentistry Office of Research (dentres@uw.edu) for researchers who meet the IRB's criteria for access to confidential data.
Funding: DC: This study was funded in part by the U.S. National Institute of Dental and Craniofacial Research (NIDCR) grant nos. R01DE026741 and T90DE021984, the William T. Grant Foundation Scholars Program, and the Center for Advanced Study in the Behavioral Sciences (CASBS) at Stanford University. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Despite evidence that topical fluoride prevents caries and is safe [1–6], caregivers are increasingly hesitant about their children receiving it [7, 8]. Hesitancy may lead to topical fluoride refusal. In turn, insufficient exposure to topical fluoride increases the likelihood of dental caries. Hesitancy may lead to topical fluoride refusal. In turn, insufficient exposure to topical fluoride increases one’s risk for dental caries (tooth decay). Untreated tooth decay may lead to pain, infection, and hospitalization, as well as social and economic consequences, including missed school, poor grades, teasing, and bullying among children and underemployment and lower earnings among adults [7, 9–12]. Despite this, little work has addressed the concept of topical fluoride hesitancy and no published measure of topical fluoride hesitancy currently exists.
Recent work developed a six domain conceptual model of caregivers’ topical fluoride hesitancy [13]. Five domains specifically related to reasons for topical fluoride hesitancy: thinking it is 1) unnecessary, 2) a chemical that should be kept out of their child’s body, and 3) harmful, and feeling 4) there is too much uncertainty about topical fluoride, and 5) distrust about and pressured by dentists and their motives for topical fluoride. We used this model to develop the first tool to measure topical fluoride hesitancy, the Fluoride Hesitancy Identification Tool (FHIT), a pool of 33 items intended to measure the five domains specifically related to hesitancy [14].
The current study sought to evaluate the FHIT item pool’s psychometric properties and create a short, simple to score, reliable, and valid tool to measure caregivers’ topical fluoride hesitancy. First, we used confirmatory factor analyses (CFA) to examine whether the FHIT items measured five separate domains. Second, we used psychometric information to reduce the number of questions and minimize respondent burden while maintaining reliability. Third, we fit item response theory (IRT) models for each domain, delivering a contemporary test theory evaluation of the FHIT’s psychometric properties and a scoring method. Fourth, we computed Cronbach’s alphas to evaluate reliability using classical test theory (CTT). Fifth, we examined the FHIT’s construct validity. And, sixth, we compared different scoring methods.
Methods
Design and procedures
We conducted a multisite, observational study. Two independent samples of English-speaking caregivers with at least one child <18 years (nSample 1 = 523 and nSample 2 = 612) were recruited from November 9th 2020 t0 September 30th 2021 from four urban, pediatric dentistry clinics and social media sites. The majority were recruited from the University of Washington’s Center for Pediatric Dentistry (65% and 93%, samples 1 and 2 respectively).
We administered the FHIT’s 33-item pool within an 85-item REDCap survey [15]. For participants with multiple children, the youngest child was the referent child. Participants provided online consent. Participants provided online consent. Before seeing the survey questions, participants were presented with an informed consent page that contained informed consent information. This page ended with "Click ’Next Page’ to confirm informed consent and begin the survey." All questions were optional and participants could discontinue at any time. All participants who completed a survey were entered into a raffle to win an Apple iPad, 2 Philips Sonicare toothbrushes, or gift cards. The study was approved by the UW Institutional Review Board (STUDY00005790) and conforms to Strobe Guidelines.
Measures
Fluoride Hesitancy Identification Tool.
As described in detail elsewhere [14], our team used elicitation interviews, team writing and editing, and cognitive interviews to develop and refine a 33 item pool to measure five domains of topical fluoride hesitancy (Table 1). This item pool demonstrated content validity [14]. For item stems beginning with “I am concerned,” participants indicated the extent to which they were: extremely, somewhat, slightly, or not at all concerned. For all other items, participants indicated the extent to which they strongly agreed, agreed, disagreed, or strongly disagreed with the item stem. All responses were scored 0–3. Zero corresponding to the option indicating the least hesitancy.
Survey questions
We used the following survey questions (analytical values in parentheses) to establish construct validity. A fluoride refusal item asked: “Regarding topical fluoride in general for your child/children, which statement below best describes you?” Caregivers responded: I always say yes (0), I say yes, but I have thought about saying no (1), sometimes I say no (2), most of the time I say no (3), or I always say no (4). A fluoride opposition item asked, “On a scale of 0 to 10 with ‘0’ being ‘not at all opposed’ and ‘10’ being ‘totally opposed’, how opposed are you to topical fluoride for your child/children?” For a dislike item (“I say no or have thought about saying no to topical fluoride because my child/children doesn’t/don’t like getting it”) and a cost item (“I say no or have thought about saying no to topical fluoride because of cost”), caregivers responded: disagree (0) or agree (1). A dental caries likelihood item asked, “How likely is your child to get a cavity?” Caregivers responded: extremely unlikely (0), unlikely (1), likely (2), or extremely unlikely (3). A dental caries severity item asked, “How bad would it be for your child to get a cavity?” Caregivers responded: extremely bad (0), somewhat bad (1), or not that bad (2). Caregivers indicated their child’s: gender (male (0) or female (1)), dental insurance type (private (0) or public (1)), race (White (0) and non-White (1)), and ethnicity (non-Hispanic (0) and Hispanic (1)). And, caregivers indicated their own: gender (male (0) or female (1)), age in years, and annual household income (<$15,000 (0), $15,000 to $25,000 (1), $25,000 to $50,000 (2), $50,000 to $75,000 (3), $75,000 to $100,000 (4), $100,000 to $150,000 (5), and >$150,000 (6)). Although caregivers could select multiple races, we created a white/non-white variable due to small frequencies (Table 2).
Analytical plan
Measurement model.
We used CFA [16, 17] to evaluate whether the items measured the five domains as hypothesized (Fig 1). This also allowed us to examine whether the item sets met some IRT model assumptions [18, 19]. In Sample 1, we used CFA to determine whether the model fit well and, if necessary, modify it. In Sample 2, we evaluated the fit of the model that resulted from the Sample 1 CFA using independent data. We also tested whether a single factor, two factor, or bifactor model fit the data well. We did this to rule out alternative explanations for multidimensionality. We treated root mean square error of approximation (RMSEA) and standardized root mean square residual (SRMR) values <0.05 as demonstrating ideal fit and <0.08 as demonstrating acceptable fit. We considered close fit index (CFI) and Tucker-Lewis index (TLI) values ≥0.95 as indicating ideal fit and values >0.90 and <0.95 as demonstrating acceptable fit. We evaluated fit based on the majority of indices [20, 21]. We used modification indices (MIs), which provide the expected change in fit if a constraint is relaxed [17], to evaluate sources of misfit, like local dependence (LD) and cross-loadings. LD means that the items share a correlation even after accounting for the correlation caused by the domain they measure [16, 17]. Because it can bias parameters and mask the extent to which a response reflects an underlying domain, we dropped one item from each item pair if we found LD. If an item measures more than one domain (cross-loads), it is unclear how much a participant’s response is driven by one domain or another. Thus, we dropped cross-loading items [16, 17].
Note: Circles represent each of the five domains measured by the FHIT. The large square label the domains and are not part of the measurement model. Small squares represent the items. Table 1 gives the item content corresponding to each number. The arrows from the circles to the small squares depict loadings and indicate which items measure a domain. The arrows among the circles depict the correlations among the domains. For simplicity, we do not show all possible measurement parameters.
Item reduction.
We sought to reduce the items for each domain to minimize respondent burden [22, 23], We used the CFA results and selected the four items that resulted in the greatest information (a measure of reliability) across each domain’s levels [18, 24], We dropped items with the lowest loadings. When two items had similar loadings, we kept the item with thresholds that differed the most from other items. This insured that we selected items that provided reliability across as much of each domain as possible.
Item response theory.
After establishing the measurement model and reducing items, we fit five separate graded response model (GRM) IRT models for each domain [25] using Sample 2 data and output scores for each domain. We then created marginal reliability curves [26], which translate IRT test information into a Cronbach alpha metric [18, 24]. Unlike CTT, IRT is more realistic and does not assume that reliability is constant [18, 24]. We used a Wald χ2 and plots comparing observed and expected scores to evaluate fit [27].
Classical test theory.
We calculated Cronbach’s alphas using each domain’s final four item set and Sample 2 data [28]. We considered values ≥0.90 excellent, <0.90 and ≥0.80 good, and values <0.80 and ≥0.70 adequate.
Construct validity.
Using Sample 2, we evaluated construct validity by examining the correlations between and across mean levels of each domain and responses to several survey items [16]. We did this three ways. We used 1) structural equation modeling (SEM) [17] to examine the polyserial correlations between and across mean levels of the latent domains and survey items. We also examined polyserial correlations between and across mean levels of 2) domain scores output from IRT and 3) domains scores created by averaging the item responses within a domain and survey items. Last, we examined the extent to which those expressing no hesitancy on each domain reported refusing topical fluoride.
Scoring evaluation.
Finally, we evaluated whether using a simple scoring approach based on the average of item responses within a domain resulted in substantive conclusions similar to SEM and IRT. SEM is expected to give the best estimate of the relationship between each domain and other variables. However, SEM is analytically complicated, does not provide a score, and can only be used with a relatively large number of individuals. Outputting individuals’ scores using IRT should provide the most accurate method of generating individuals’ scores. However, while IRT provides a score, IRT is still relatively complicated and difficult to implement in clinical settings, especially in real time. An item average-based approach is the least accurate and treats the ordinal responses as if they are interval level measures, yet it is simple to compute in real time in clinical settings and may be more intuitively interpretable because the resulting score is in the metric of the items. While theoretically more accurate than an item average, using IRT scores or item averages when evaluating relationships rather than SEM may lead to biased results [29]. Thus, to evaluate the validity of using the theoretically less accurate but more feasible item average-based approach or alternatively IRT, we compared probability values and standardized parameters and effect sizes across SEM, IRT-based scores, and item average-based scores. We used an average vs. a sum score because this allows calculation even if a caregiver skips an item [17].
We conducted all analyses in Stata [30] and Mplus [31]. All analyses appropriately modeled ordinal data. CFA and SEM used weighted least square mean and variance adjusted estimation and IRT used full information maximum likelihood. Both types allow missing item level data [31].
Results
Measurement model
We first fit a five factor model based on the qualitative research. This model did not fit well (RMSEA = 0.10; SRMR = 0.09; CFI = 0.94; TLI = 0.94). MIs indicated LD for eight different item pairs and we dropped one item from each. MIs also revealed that 3 items measured multiple domains and we dropped these items [16, 17]. Finally, “I trust that my child’s dentist will give me a choice to say no to topical fluoride” had a low loading (<0.2), indicating it did not measure the underlying domain well. So, we dropped this item [16, 17].
After dropping these 8 items, the resulting 25-item model, with all remaining items loading on their expected domains, demonstrated sufficient fit (RMSEA = 0.08; SRMR = 0.059; CFI = 0.98; TLI = 0.98). We then tested the fit of the 25 item model in the independent sample. This model fit acceptably: RMSEA = 0.079; SRMR = 0.057; CFI = 0.98; TLI = 0.98. Thus, we considered this the final model. Table 1 presents its standardized loadings and factor correlations. Given that the factor correlations were large (though not problematically large for SEM) [17], we also examined single, two, and a bifactor models. None demonstrated acceptable fit. Therefore, we considered the 25-item, five factor model, the final model.
Item reduction
As Table 1 shows, for the Concerns Topical Fluoride is Unnecessary and Concerns Topical Fluoride is Harmful domains, we dropped the items with the lowest loadings. For the Feeling Uncertainty about Topical Fluoride domain, we kept the item whose loading equaled 0.81 (vs. 0.85) because the thresholds indicated that this item measured lower hesitancy levels than the other items in the set and thus increased the range of reliability relative to the others. The remaining 2 domains were already at four items each.
Item response theory
Next, we fit GRM for each domain based on each domain’s final four item set. Each fit well. Table 1 lists the parameters. Fig 2 shows the marginal reliability curves. For all domains, reliability is greater than 0.80 at average levels of the latent variable (statistically identified as 0 in IRT) and rapidly and consistently 0.90 and greater for average to high levels of hesitancy. For all domains except Concerns Topical Fluoride is Unnecessary, reliability relatively rapidly decreases at less than average levels.
Classical test theory
Cronbach’s alpha values equaled 0.92, 0.90, 0.77, 0.88, and 0.89 for Concerns Topical Fluoride is Unnecessary, Concerns Topical Fluoride is Harmful, Feeling Distrust and Pressured by Dentists and their Motives, Feeling Uncertainty about Topical Fluoride, and Feeling that Topical Fluoride Should be Kept Out of Child’s Body, respectively.
Construct validity
Table 3 presents the validity analyses across all three scoring approaches. Given agreement, we present only the item average findings here. Each of the domains correlated significantly with topical fluoride refusal: ρUnnecessary = 0.64, (p < 0.01), ρHarmful = 0.48, (p < 0.01), ρDistrust = 0.55, (p < 0.01), ρUncertainty = 0.54, (p < 0.01), and ρKept Out = 0.55, (p < 0.01) and opposition to topical fluoride: ρUnnecessary = 0.55, (p < 0.01), ρHarmful = 0.43, (p < 0.01), ρDistrust = 0.49, (p < 0.01), ρUncertainty = 0.50, (p < 0.01), and ρKept Out = 0.52, (p < 0.01). Among caregivers who reported refusing topical fluoride, hesitancy levels did not differ significantly across caregivers that refused due to reasons of cost or a child’s dislike of topical fluoride. Except for Feeling that Topical Fluoride Should be Kept Out of Child’s Body (ρKept Out = −0.12, p = 0.01), the extent to which a caregiver considered it likely that a child would get a cavity was not significantly related to hesitancy levels. However, Concerns Topical Fluoride is Harmful, Feeling Uncertainty about Topical Fluoride, and Feeling that Topical Fluoride Should be Kept Out of Child’s Body were all significantly related to the extent to which a caregiver would consider it bad if a child developed a cavity: ρHarmful = 0.55, ρUncertainty = 0.55, and ρKept Out = 0.55 respectively. Decreasing household income was significantly related to increasing hesitancy on each domain: ρUnnecessary = −0.23, (p < 0.01), ρHarmful = −0.18, (p < 0.01), ρDistrust = −0.17, (p < 0.01), ρUncertainty = −0.22, (p < 0.01), and ρKept Out = −0.19, (p < 0.01). Caregiver age was not. Further, as Table 3 shows, while hesitancy levels did not differ significantly across child gender, female caregivers generally reported greater hesitancy levels: dUnnecessary = 0.34, (p < 0.01), dHarmful = 0.31, (p < 0.01), dDistrust = 032., (p < 0.01), and dKept Out = 0.26, (p = 0.01). Caregivers of non-White children reported less hesitancy on each domain [dUnnecessary = −0.41, (p < 0.01), dHarmful = −0.39, (p < 0.01), dDistrust = −034., (p < 0.01), dUncertainty = −0.39, (p < 0.01), and dKept Out = −0.39, (p < 0.01)] than caregivers of White children. However, caregivers of Hispanic children did not report significantly different levels of hesitancy in any domain relative to caregivers of non-Hispanic children. Finally, caregivers of children with public dental insurance reported significantly lower levels of hesitancy due to Concerns Topical Fluoride is Harmful (dHarmful = -0.20, p = 0.03) and Feeling Uncertainty about Topical Fluoride (dUncertainty = −0.28, p = < 0.01) than caregivers of children with private insurance but no other domains.
We also examined the extent to which those expressing no hesitancy on a domain, reported refusing topical fluoride. Among those with an item average-based score of zero (no hesitancy) on the Concerns Topical Fluoride is Unnecessary, Concerns Topical Fluoride is Harmful, Feeling Distrust and Pressured by Dentists, Feeling Uncertainty about Topical Fluoride, and Feeling that Topical Fluoride Should be Kept Out of Child’s Body, 100%, 95%, 98%, 98%, and 97%, respectively, reported never refusing topical fluoride.
Scoring evaluation
As Table 3 shows, effect sizes (d) and standardized parameter estimates (ρ) across approaches were similar. The item average- and IRT score-based approaches occasionally underestimated the size of relationships to a small degree relative to SEM. With only two exceptions, the three approaches agreed perfectly with respect to the probability that a given effect occurred by chance. The IRT-based approach probability for the relation between the extent to which a caregiver considered it likely that a child would get a cavity and Concerns that Topical Fluoride is Harmful was less than 0.05 (p = 0.03) while the SEM probability value was greater than 0.05 (p = 0.06). Conversely, the item average- and IRT score-based approach probabilities for the relation between the extent to which a caregiver considered it likely that a child would get a cavity and Feeling Uncertainty about Topical Fluoride Science were greater than 0.05 (p = 0.08 and 0.10 respectively), while the SEM value was less than 0.05 (p = 0.04).
Discussion
This study sought to evaluate the FHIT item pool’s psychometric properties and establish a short, simple to score, reliable, and valid method of measuring domains of caregivers’ topical fluoride hesitancy for their children. Our results support the reliability and validity of a 20-item FHIT to measure five domains of topical fluoride hesitancy using the average of the four items measuring each domain.
CFA supported the FHIT’s internal validity. Caregivers’ fluoride hesitancy appears to be comprised of five domains: feeling that topical fluoride is 1) unnecessary, 2) a chemical that should be kept out of their child’s body, and 3) harmful, and feeling 4) there is too much uncertainty about topical fluoride, and 5) distrust about and pressured by dentists and their motives for topical fluoride. The FHIT items measure one and only one domain each. While the domains are correlated, a caregiver could conceptually be high on one and low on others. Efforts to understand, prevent, and address topical fluoride hesitancy should consider an individual’s reported values for each domain and their joint impact.
CTT and IRT supported the reliability of each FHIT domain. Cronbach’s alphas suggested the domains generally provided excellent reliability. IRT showed that the items provided high reliability at and above average hesitancy levels but that reliability tended to decrease relatively rapidly at lower than average levels of hesitancy. While a uniformly high level would be ideal, the decrease in reliability is unlikely to be problematic in contexts where the FHIT is likely to be used. Clinicians using the FHIT to identify individuals with topical fluoride hesitancy do not need to discriminate well among those with low hesitancy. Similarly, researchers evaluating efforts to address hesitancy will be most interested in the effect of these efforts among individuals with elevated hesitancy. The FHIT is well suited for these tasks.
Our results also support construct validity. Except for topical fluoride refusal, the domains correlated more strongly with each other than with other survey questions. And, as expected, the correlations between domain scores and refusal of and opposition to topical fluoride were large generally and relative to the size of other variables’ correlations with the domain scores. In addition, as expected, domain scores did not differ significantly across a child’s gender but did differ across caregiver’s gender. While we did not have specific hypotheses for other variables, validity was supported by the fact that the domains significantly correlated with (or levels significantly differed across) some variables (e.g., household income) but not others (e.g., caregiver’s age).
With respect to scoring, our findings support the validity of an item average-based score. This score is created for each domain’s four item set by 1) assigning values of 0–3 to participants’ responses for each item, with 0 corresponding to the option indicating the least hesitancy, and 2) taking the average of the responses. The validity of this approach is important because users may not have access or ability to use SEM or IRT. Further, because this approach is in the metric of the item responses, an item average-based score allows interpretation relative to the item responses.
Finally, our study provides the first quantitative evidence that topical fluoride hesitancy is multifactorial. Previous qualitative work has suggested the complex nature of topical fluoride hesitancy [13, 32] but it was not clear to what extent that complexity represented a set of underlying domains. Our findings indicate that five correlated domains reflect parents’ hesitancy about topical fluoride. Thus, while dentists may assume that caregivers are hesitant due to lack of knowledge [33], our results suggest that education alone will likely not reduce hesitancy across these domains. Future efforts should evaluate the predictors of the different domains of topical fluoride hesitancy and develop and evaluate interventions to reduce topical fluoride hesitancy and refusal.
Limitations
Before concluding, we note some weaknesses. Participants were from mainly from a university-based dental school in Seattle and the FHIT is only available in English. Yet, the study’s strengths, which include independent samples, a confirmatory rather than exploratory approach, and evaluating a range of approaches to examine construct validity and support scoring, outweigh these weaknesses.
Conclusion
In conclusion, our study supports the reliability and validity of the 20-item FHIT as a measure of five domains of caregiver’s reasons for hesitance to topical fluoride.
References
- 1. Garcia RI, Gregorich SE, Ramos-Gomez F, Braun PA, Wilson A, Albino J, et al. Absence of Fluoride Varnish-Related Adverse Events in Caries Prevention Trials in Young Children, United States. Prev Chronic Dis. 2017;14:E17. Epub 20170216. pmid:28207379.
- 2. Iheozor-Ejiofor Z, Worthington HV, Walsh T, O’Malley L, Clarkson JE, Macey R, et al. Water fluoridation for the prevention of dental caries. Cochrane Database Syst Rev. 2015;2015(6):Cd010856. Epub 20150618. pmid:26092033.
- 3. Walsh T, Worthington HV, Glenny AM, Marinho VC, Jeroncic A. Fluoride toothpastes of different concentrations for preventing dental caries. Cochrane Database Syst Rev. 2019;3(3):Cd007868. Epub 20190304. pmid:30829399.
- 4. Marinho VC, Chong LY, Worthington HV, Walsh T. Fluoride mouthrinses for preventing dental caries in children and adolescents. Cochrane Database Syst Rev. 2016;7(7):Cd002284. Epub 20160729. pmid:27472005.
- 5. Marinho VC, Worthington HV, Walsh T, Chong LY. Fluoride gels for preventing dental caries in children and adolescents. Cochrane Database Syst Rev. 2015;2015(6):Cd002280. Epub 20150615. pmid:26075879.
- 6. Tubert-Jeannin S, Auclair C, Amsallem E, Tramini P, Gerbaud L, Ruffieux C, et al. Fluoride supplements (tablets, drops, lozenges or chewing gums) for preventing dental caries in children. Cochrane Database Syst Rev. 2011;2011(12):Cd007592. Epub 20111207. pmid:22161414.
- 7. Chi DL. Caregivers who refuse preventive care for their children: the relationship between immunization and topical fluoride refusal. American journal of public health. 2014;104(7):1327–33. pmid:24832428
- 8. Chi D, Basson A. Surveying dentists’ perceptions of caregiver refusal of topical fluoride. JDR Clinical & Translational Research. 2018;3(3):314–20. pmid:30938597
- 9. Jackson SL, Vann WF Jr., Kotch JB, Pahel BT, Lee JY. Impact of poor oral health on children’s school attendance and performance. Am J Public Health. 2011;101(10):1900–6. Epub 20110217. pmid:21330579.
- 10. Filstrup SL, Briskie D, da Fonseca M, Lawrence L, Wandera A, Inglehart MR. Early childhood caries and quality of life: child and parent perspectives. Pediatr Dent. 2003;25(5):431–40. pmid:14649606.
- 11. Hyde S, Satariano WA, Weintraub JA. Welfare dental intervention improves employment and quality of life. J Dent Res. 2006;85(1):79–84. pmid:16373686.
- 12. Glied S, Neidell M. The economic value of teeth. Journal of Human Resources. 2010;45(2):468–96.
- 13. Chi DL, Kerr D, Patiño Nguyen D, Shands ME, Cruz S, Edwards T, et al. A conceptual model on caregivers’ hesitancy of topical fluoride for their children. PLoS One. 2023;18(3):e0282834. Epub 20230322. pmid:36947522.
- 14. Edwards TC, Carle AC, Kerr D, Carpiano RM, Nguyen DP, Orack JC, et al. Topical Fluoride Hesitancy Among Caregivers: Development of a Content-Valid Topical Fluoride Hesitancy Identification Item Pool. J Public Health Dent. 2023;83(1):116–22. pmid:36719013
- 15. Saini SJ, Carle AC, Forsyth AR, Chi DL. Association between caregiver opposition to topical fluoride and COVID-19 vaccines. Vaccine. 2023;41(5):1035–41. pmid:36567141
- 16.
McDonald RP. Test theory: A unified treatment. Mahwah, NJ: Erlbaum; 1999.
- 17.
Gunzler DD, Perzynski AT, Carle AC. Structural Equation Modeling for Health and Medicine: Chapman and Hall/CRC; 2021.
- 18.
Embretson S, Reise S. Item response theory for psychologists. Mahwah, N.J.: Lawrence Erlbaum; 2000.
- 19. Carle AC, Weech-Maldonado R. Validly interpreting patients’ reports: using bifactor and multidimensional models to determine whether surveys and scales measure one or more constructs. Medical Care. 2012;50:S42–S8. pmid:22895230
- 20. Hu L, Bentler PM. Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecification. Psychological methods. 1998;3(4):424–53.
- 21. Hu L, Bentler P. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal. 1999;6(1):1–55.
- 22. Reeve BB, Hays RD, Bjorner JB, Cook KF, Crane PK, Teresi JA, et al. Psychometric evaluation and calibration of health-related quality of life item banks: plans for the Patient-Reported Outcomes Measurement Information System (PROMIS). Medical care. 2007;45(5 Suppl 1):S22–31. pmid:17443115
- 23. Reeve BB. Item response theory modeling in health outcomes measurement. Expert Review of Pharmacoeconomics and Outcomes Research. 2003;3(2):131–45. pmid:19807361
- 24.
Embretson S, Hershberger S. The new rules of measurement: What every psychologist and educator should know: Lawrence Erlbaum; 1999.
- 25. Samejima F. Estimation of latent ability using a response pattern of graded scores. Psychometrika Monograph Supplement. 1969;34(4, Pt. 2):100.
- 26. Green BF, Bock D, Humphres RL, Linn MD. Technical Guidelines for Assessing Computerized Adaptive Tests. Journal of Educational Measurement. 1984;21(4):347–60.
- 27. Kondratek B. Item-fit statistic based on posterior probabilities of membership in ability groups. Applied Psychological Measurement. 2022;46(6):462–78. pmid:35991828
- 28. Cronbach L. Coefficient alpha and the internal structure of tests. Psychometrika. 1951;16(3):297–334.
- 29. Rabe-Hesketh S, Skrondal A, Pickles A. Maximum likelihood estimation of limited and discrete dependent variable models with nested random effects. Journal of Econometrics. 2005;128(Journal Article):301–23.
- 30.
StataCorp L. Stata statistical software: Release 17. College Station, TX: StataCorp LP; 2021.
- 31.
Muthén LK, Muthén BO. Mplus User’s Guide. Eighth Edition. ed. Los Angeles, CA: Muthén & Muthén; 1998–2017.
- 32. Leung E, Kerr D, Askelson N, Chi DL. Understanding topical fluoride hesitancy and refusal behaviors through the extended parallel process model and health belief model. J Public Health Dent. 2023;83(1):3–8. Epub 20220314. pmid:35288941.
- 33. Chi DL. Parent Refusal of Topical Fluoride for Their Children: Clinical Strategies and Future Research Priorities to Improve Evidence-Based Pediatric Dental Practice. Dent Clin North Am. 2017;61(3):607–17. Epub 20170504. pmid:28577640.