Dentist and practice characteristics associated with restorative treatment of enamel caries in permanent teeth: multiple-regression modeling of observational clinical data from The National Dental PBRN

Abstract

Purpose

Current evidence in dentistry recommends non-surgical treatment to manage enamel caries lesions. However, surveyed practitioners report they would restore enamel lesions that are confined to the enamel. We used actual clinical data to evaluate patient, dentist, and practice characteristics associated with restoration of enamel caries, while accounting for other factors.

Methods

We combined data from a National Dental Practice-Based Research Network observational study of consecutive restorations placed in previously unrestored permanent tooth surfaces and practice/demographic data from 229 participating network dentists. Analysis of variance and logistic regression, using generalized estimating equations (GEE) and variable selection within blocks, were used to test the hypothesis that patient, dentist, and practice characteristics were associated with variations in enamel restorations of occlusal and proximal caries compared to dentin lesions, accounting for dentist and patient clustering.

Results

Network dentists from 5 regions placed 6,891 restorations involving occlusal and/or proximal caries lesions. Enamel restorations accounted for 16% of enrolled occlusal caries lesions and 6% of enrolled proximal caries lesions. Enamel occlusal restorations varied significantly (p<0.05) by patient age and race/ethnicity, dentist use of caries risk assessment, network region, and practice type. Enamel proximal restorations varied significantly (p<0.05) by dentist race/ethnicity, network region, and practice type.

CLINICAL SIGNIFICANCE

Identifying patient, dentist, and practice characteristics associated with enamel caries restorations can guide strategies to improve provider adherence to evidence-based clinical recommendations.

Introduction

Current evidence supports remineralization as a treatment choice over restoration for carious lesions confined to the enamel.^1,2 Non-surgical strategies such as dental sealants and fluoride varnishes preserve tooth structure and can arrest caries progression.^3–6 However, substantial percentages of dentists routinely report that they restore teeth that have enamel caries for both low and high caries risk patients.^7–12 The data about dentists’ self-reported decisions to restore enamel lesions are based on examinations of extracted teeth^12–14 or radiographs for hypothetical patients.^8,11,15–17 There are little actual clinical data examining determinants of enamel lesion restorations.

The decision about whether to allow for remineralization or to restore dental caries is multifactorial and can vary significantly among dentists, particularly for lesions that do not extend beyond the dentin-enamel junction (i.e., early caries or lesions confined to enamel only).^18–20 It can be difficult in clinic settings to determine true lesion depth and whether caries are active or inactive.^{2, 20,21} The impact of improved diagnostic tools has been limited.^14,22–24 Other factors contributing to the decision include the patient’s history of caries,^25,26 poor oral health and expected compliance,²⁷ and preferences for surgical or medical treatment.²⁸ Preferences are influenced by treatment costs and patients’ insurance coverage,^25,29 including reimbursement rates for preventive treatments (e.g., fluoride varnish).^10,30 Additionally, the use of quality-based performance measures, clinical practice guidelines, or application of evidence-based practice concepts may reduce the likelihood that a dentist with restore enamel caries.^{2,11,15, 31}

This paper examines clinical data from a National Dental Practice-Based Research Network study of restorations on previously unrestored permanent tooth surfaces.³² The original restoration study included observational data on restorations for primary caries or non-carious defects, the tooth number and surfaces involved, and the restorative materials used. For restorations due to primary caries, data were collected on pre- and post-operative lesion depths and diagnostic techniques used. We use these data to test the hypothesis that specific patient, dentist, and practice characteristics are associated with variations in restorative treatment of enamel lesions. In particular, we test the hypothesis that network region, practice type, practice busyness, and use of caries risk assessment are significant predictors of variations in the percentage of enamel caries on restored occlusal and proximal surfaces, while controlling for other dentist, practice, and patient characteristics. Ideally, a clinical evaluation of this type would include unrestored carious lesions, which would have enabled us to directly estimate the percentage of enamel lesions that were restored. Nonetheless, given the paucity of clinical data examining the determinants of enamel lesion restorations, this study provides new evidence of key determinants of general dentists’ restorative treatment practices for enamel caries.

Materials and Methods

Study population

This study involved 229 National Dental Practice-Based Research Network practitioners who reported providing restorative dentistry in their practices. At the time of data collection, the network mainly comprised five regions: AL/MS (Alabama/Mississippi); FL/GA (Florida/Georgia); MN (dentists employed by HealthPartners and private practitioners in Minnesota); PDA (Permanente Dental Associates, Portland, OR, in cooperation with Kaiser Permanente’s Center for Health Research); and SK (Denmark, Norway, and Sweden).^{33, 34} To enroll in the network, all practitioners completed an enrollment questionnaire about their practice characteristics and about themselves. To be eligible for this study, dentists also completed a caries diagnosis and treatment survey, and training in the study protocol implementation and human subjects protection. Descriptions of questionnaires and protocol training procedures are detailed at the network’s website (nationaldentalpbrn.org).³⁵

Each participating network dentist enrolled up-to 50 consecutive restorations placed in permanent teeth. Staggered enrollment periods began in April, 2006 and were completed in December, 2008. Each dentist recruited patients ages ≥6 years who had at least one restoration placed on a previously unrestored surface of a permanent tooth. Eligible restorations had to have been placed during the first office visit within the enrollment period. The study included up-to four separate restorations per patient but excluded restorations placed at subsequent visits during the enrollment period. Informed consent and child assent were obtained from all study participants.

Study design

We used a consecutive recruitment design to enroll restorations among eligible patients. Once a practice began enrollment, dentists approached every patient scheduled to have a restoration on a previously un-restored permanent tooth surface. Enrollment continued until 50 separate restorations were enrolled, with up-to four restorations allowed per participant. Enrollment targets were based on a priori precision of estimation estimates for 95% confidence intervals (CIs) around 50% to provide conservative precision estimates, adjusting for cluster sizes of 50, 75, and 100. A sample size of 50 restorations for each of 100 clinicians will yield CI widths from 6.8% to 14.1% for interclass correlations (ICCs) of 0.10 to 0.50, respectively. Regional coordinators in each network region trained practitioners to assess eligibility, obtain informed consent and child assent, complete data collection, and transmit data forms to study staff. Each region’s Institutional Review Board reviewed and approved the study.

Dentists were informed that the study sought to examine the reasons restorations were placed, the diagnostic techniques dentists used to determine primary caries, and the types of restorative materials used for each restoration placed. The number of data elements collected for each restoration was limited to prevent disruption of patient flow. Demographic and practice data for each dentist were obtained from a questionnaire completed at network enrollment and from a caries diagnosis and treatment survey.

Measures

For each restoration placed, dentists recorded the tooth number, involved surfaces, why they placed the restoration (primary caries or selected non-carious defects), diagnostic techniques used (clinical assessments including probing, radiographs, transillumination or optical technique), pre- and post-operative lesion depth, and restorative material used (e.g., amalgam, composite resin, etc.) including base, lining, or bonding material. All study questions were curated using the caDSR™ (formerly caBIG™)³⁶ development tool under the guidance of external content experts. The survey questionnaire included a test-retest process to check internal validity, and the clinical study included pilot testing of the data collection form.

We collected data on patient age, gender, race, ethnicity, insurance coverage, and the number of enrolled restorations at the patient’s first visit during the enrollment period. Practitioner age, gender, race, ethnicity, practice region and type (private practice, large prepaid group, public health), years since dental school graduation, and practice busyness, were obtained from the enrollment questionnaire (see Makhija et al.³⁷ for a more detailed description of the data). Caries risk assessment, including whether a form was used, was obtained from the caries diagnosis and treatment survey. Restoration data included tooth number, primary reason for placing the restoration (primary caries or non-carious defect), diagnostic techniques used, pre- and post-operative lesion depth, carious surfaces, and restorative materials used. Other than the presence of primary caries and patient demographics, no other information was available regarding the dentist’s decision to place the restoration.

For lesion depth, dentists were asked “how deep did you estimate that the deepest part of the primary caries lesion was, preoperatively?” Dentists checked either a) E1: outer ½ of enamel, b) E2: inner ½ of enamel, c) D1: outer 1/3 of dentin, d) D2: middle 1/3 of dentin, e) D3: inner 1/3 of dentin, or f) uncertain. Lesions were classified into five non-mutually-exclusive categories based on the dentists’ pre-operative clinical assessments of surfaces involved in the restoration: posterior proximal, anterior proximal, posterior occlusal, posterior smooth surface, anterior smooth surface. Each dentist made diagnostic decisions using visual, tactile, and radiographic methods typical of his or her “routine care” approach.³⁸ An objective of the network is to investigate dental care as it occurs in routine clinical practice, involving the use of diagnostic and treatment methods as they are used in real-world community-based, nonacademic clinical practice settings, which is the type of setting in which nearly all patients receive dental care. Therefore, although the network makes a point of standardizing data collection processes, it seldom performs studies that require standardization or calibration of diagnostic and treatment methods across practices. Consequently, we intentionally did not conduct calibration and inter-examiner reliability training for this study. Analyses included restorations involving primary caries on at least one occlusal or proximal tooth surface. We excluded a small number of restorations involving only lingual or buccal surfaces, unknown lesion depth, as well as restorations that had been placed because of a non-carious defect.

Statistical analysis

We assessed factors associated with restoration of enamel caries lesions involving a) occlusal and b) proximal surfaces, using generalized linear models (GLM) in SAS^® 9.2 PROC GENMOD (Cary, NC). We conducted analysis of variance (ANOVA) and logistic regression using generalized estimating equations (GEE) to assess patient-, dentist-, and practice-level predictors of restoration of enamel lesions compared to lesions extending into the dentin, controlling for clustering of restorations within patients (up to four restorations) and dentists (up to 50 patients per dentist). We used a three-step approach to modeling. First, univariate ANOVA was used to separately identify individual patient and dentist predictors of enamel restorations for occlusal and proximal lesions. Second, we conducted separate blocked analyses for patient and dentist variables using predictors that were significant (p≤0.10) in step 1, for each surface type. Blocked analyses included main effects and two-way interactions. For step 3, we fitted separate final models for occlusal and proximal restorations using patient and dentist predictors that were significant (p≤0.10) during the blocked analyses from step 2. Since network region and practice type (private, large prepaid group, public health) were highly correlated, we conducted the analyses in step 3 using region as a covariate, then re-ran the models after substituting practice type for region. Variables with p-values ≤ 0.05 were considered statistically significant predictors and were retained in the final models. All dentists in the PDA region were large-group practitioners, and thus were used as the referent group for the final models.

Results

A total of 229 network dentists recruited 4,397 patients with diagnoses of caries who had a total of 6,891 restorations placed on previously unrestored occlusal and proximal tooth surfaces. Participating dentists were predominantly male, non-Hispanic, White, and had graduated from dental school a mean of 19 years earlier (Table 1). Overall, most dentists reported they were able to treat all patients with no burden, while about one-third indicated that their practices were busy (too busy to treat all patients or all treated, burdened). There were significant regional variations (p-value of at least <0.05) among dentists by gender, ethnicity, practice busyness, mean years since dental school graduation, caries risk assessment, and practice type. Differences in mean practitioner age and race were not significant.

Table 1.

Demographic and practice characteristics of participating National Dental PBRN dentists

	AL/MS N=63	FL/GA N=37	MN N=31	PDA N=51	SK N=47	P
Male (%)	89	89	67	80	53	<0.0001
Age (mean)	45	42	44	34	42	0.5999*
Hispanic (%)	0	16	0	4	0	0.0002
Race (%)**						0.1127†
White	92	97	89	80	98
African American	3	3	7	4	0
Asian	5	0	4	13	2
Other	0	0	0	1	0
Years since dental school graduation (mean)	21	21	17	15	20	0.0156*
Practice busyness (%)						0.0078
Too busy to treat all	10	3	18	9	24
All treated, burdened	18	11	29	16	33
All treated, no burden	53	63	50	66	38
Not busy enough	18	23	4	9	5
Assess caries risk (%)	76	49	77	86	85	0.0004
Practice type (%)						<0.0001†
Public health	2	3	0	0	36
Large prepaid group	0	0	90	100	0
Private practice	98	97	10	0	64

AL = Alabama; GA = Georgia; FL = Florida; GA = Georgia; MN = Minnesota; PDA = Permanente Dental Associates; SK = Scandinavia. P-values (P) based on chi-square test unless noted.

^*One-way ANOVA.

^†Fisher’s exact test.

^**Totals may not equal sums due to rounding.

Most participating patients were aged 18-44, with a mean age of 33 years across all regions (Table 2). Patients were mostly female, non-Hispanic, White, and had some form of dental insurance (insurance type was not collected). Overall, 62% of participants had a single restoration. Patient enrollment logs indicated that more than 95% of eligible restorations were included in the study. Patients differed significantly across regions (p-value of at least <0.05) by age group, gender, ethnicity, race, dental insurance, and number of enrolled restorations.

Table 2.

Demographic characteristics of patient with at least one occlusal or proximal carious lesion

	AL/MS N=1162	FL/GA N=699	MN N=848	PDA N=1003	SK N=685	P
Age group (%)						<0.0001
<18 years	26.3	19.2	22.3	16.7	21.3
18–44 years	51.5	50.1	55.3	62.2	59.8
45–64 years	17.8	21.5	16.6	18.3	15.6
≥65 years	4.4	9.2	5.8	2.8	3.4
Male (%)	45.7	45.6	41.4	49.8	50.2	0.0016
Hispanic (%)	1.7	11.8	6.4	7.8	0.6	<0.0001
Race (%)						<0.0001
White	78.2	85.1	75.9	84.1	97.6
African American	19.2	12.9	16.6	4.4	0.3
Asian	1.0	1.3	6.0	5.7	2.0
American Indian/Alaska Native	1.6	0.4	0.4	0.5	0.0
Native Hawaiian/Pacific Islander	0.1	0.3	0.0	0.6	0.2
Other	0	0	1.2	4.7	0.0
Dental insurance (%)	83.1	68.5	85.3	93.4	65.3	<0.0001
Enrolled restorations (%)						<0.0001
One	55.1	64.5	63.8	53.8	79.1
Two	27.5	23.8	22.9	27.1	17.2
Three or more	17.4	11.7	13.3	19.0	3.7

P-values (P) based on chi-square test unless noted.

^*One-way ANOVA.

Out of the 6,891 enrolled restorations, we excluded 115 with missing or uncertain preoperative depth, leaving 6,776 restorations for analysis. A total of 4,064 restorations involved an occlusal surface, 4,149 involved a proximal surface, and 1,437 involved both occlusal and proximal surfaces (Table 3). Data collection forms indicated that 60% of restorations were diagnosed using multiple techniques. The most common combination was clinical assessments, including probing, plus radiographs (47%), while no other combination accounted for more than 6% diagnoses. Overall, diagnoses were made using clinical assessment for 77% of cases, radiographs for 66% of cases, and optical techniques for 8% of case. See Rindal et al.³⁹ for an extensive examination of caries diagnostic techniques used in this study population. Dentists restored 718 occlusal surfaces that extended into the enamel only, which accounted for 16% of all occlusal restorations. Dentists in AL/MS (24%) were much more likely than dentists from other regions to have enrolled an enamel occlusal restoration, while MN (6%) and SK dentists (4%) were the least likely to enroll enamel lesions. Network dentists enrolled 246 proximal lesions that extended into the enamel only (6% of all proximal lesions). Regional variations in the percentages of restorations involving proximal surfaces limited to the enamel were similar to regional variations for occlusal lesions.

Table 3.

Treatment thresholds for occlusal and proximal caries by pre-operative depth

	AL/MS N (%)	FL/GA N (%)	MN N (%)	PDA N (%)	SK N (%)
Occlusal surfaces*	1483 (100.0)	773 (100.0)	556 (100.0)	916 (100.0)	336 (100.0)
E1	104 (7.0)	30 (3.9)	3 (0.5)	22 (2.4)	4 (1.2)
E2	258 (17.4)	102 (13.2)	30 (5.4)	84 (9.2)	10 (3.0)
D1	667 (45.0)	405(52.4)	319 (57.4)	517 (56.4)	145 (43.2)
D2	327 (22.1)	188 (24.3)	155 (27.9)	218 (23.8)	147 (43.8)
D3	127 (8.6)	48 (6.2)	49 (8.8)	75 (8.2)	30 (8.9)
Proximal surfaces**	986 (100.0)	561 (100.0)	819 (100.0)	1184 (100.0)	599 (100.0)
E1	18 (1.8)	9 (1.6)	8 (1.0)	4 (0.3)	1 (0.2)
E2	102 (10.3)	28 (5.0)	30 (3.7)	43 (3.6)	3 (0.5)
D1	440 (44.6)	278 (49.6)	445 (54.3)	749 (63.3)	262 (43.7)
D2	275 (27.9)	197 (35.1)	253 (30.9)	289 (24.4)	227 (37.9)
D3	151 (15.3)	49 (8.7)	83 (10.1)	99 (8.4)	106 (17.7)

^*Includes any restoration involving an occlusal surface.

^**Includes any restoration involving a mesial or distal surface. Lesion depths defined as extending into the outer half of the enamel (E1), inner half of the enamel (E2), outer third of the dentin (D1), middle third of the dentin (D2), and inner third of the dentin (D3).

Table 4 presents GEE parameter estimates and standard errors for the step-wise modeling of early occlusal and proximal caries. Characteristics with univariate p-values ≤0.10 were included in separate blocked analyses of patient and dentist characteristics for each enamel lesion category. Patient age group (p=0.002), race-ethnicity (p=0.019), and dental insurance (p=0.039) met the significance test criterion for inclusion in the patient characteristic blocked analysis for enamel occlusal restorations, while patient gender (p=0.040) met the requirement for enamel proximal lesions. Dental insurance was not significant (p=0.189) in the blocked analysis and was excluded from the final models. Dentist gender (p=0.103) and caries risk assessment (p=0.011) met the criterion for inclusion in the dentist characteristic blocked analysis for enamel occlusal restorations. Dentist gender (p=0.040) and race-ethnicity (p=0.001) met the requirement for enamel proximal lesions. All dentist characteristics included in the blocked analyses met the inclusion criterion for the final modeling. Univariate significance tests for practice region and practice type met the inclusion criterion for the final modeling.

Table 4.

GEE univariate and blocked parameter estimates (PE) and standard errors (SE) for patient and dentist factors associated with placement of restorations for enamel carious lesions on occlusal and proximal surfaces

	Occlusal lesions*				Proximal lesions*
	Univariate		Blocked**		Univariate		Blocked**
	PE	SE	PE	SE	PE	SE	PE	SE
Patient characteristics
Age
<18 years	0.438	0.369	0.315	0.364	0.525	0.356	-	-
18–44 years	−0.099	0.351	−0.229	0.337	0.011	0.314	-	-
45–64 years	−0.214	0.341	−0.312	0.315	0.233	0.327	-	-
65+ years	0.000	0.000	0.000	0.000	0.000	0.000	-	-
Male	−0.097	0.103	-	-	−0.251	0.110	−0.251	0.110
Hispanic/racial minority	−0.288	0.129	−0.219	0.133	−0.204	0.237	-	-
No dental insurance	0.319	0.164	−0.207	0.167	0.101	0.197	-	-
1 enrolled restoration	0.036	0.109	-	-	−0.088	0.118	-	-
Dentist characteristics
Male	0.442	0.294	0.464	0.280	0.705	0.324	0.708	0.349
Hispanic/racial minority	0.180	0.426	-	-	−1.543	0.467	−1.448	0.447
Years since graduation†
0–5 years	0.111	0.352	-	-	0.280	0.376	-	-
6–15 years	0.060	0.263	-	-	−0.327	0.334	-	-
16–19 years	0.299	0.336	-	-	0.402	0.443	-	-
≥20 years	0.000	0.000	-	-	0.000	0.000	-	-
Assesses caries risk
No	1.006	0.334	1.027	0.333	0.819	0.440	-	-
Yes – no form	0.548	0.317	0.539	0.315	0.596	0.417	-	-
Yes – with form	0.000	0.000	0.000	0.000	0.000	0.000	-	-
Practice not busy‡	0.237	0.268	-	-	0.041	0.322	-	-

^*Includes restorations involving both occlusal and proximal surfaces.

^**Blocked analyses conducted separately for patient and dentist characteristics with univariate p-values ≤ 0.10.

^†Categories for years since dental school graduation selected for group balance.

^‡All treated, no burden/not busy enough compared to Too busy to treat all/all treated, burdened.

The GEE analysis strategy resulted in final main-effects models that included patient’s age group, gender, and race/ethnicity and accounted for clustering within patient and provider (Table 5). Dentist variables included gender, race/ethnicity, use of caries risk assessment, and network region and practice type. The average cluster size was approximately 18 for each restoration type. The observed values of ICC were 0.145 and 0.244 for proximal and occlusal restorations, respectively, yielding variance inflation factors of 3.5 for proximal and 5.1 for occlusal restorations. Thus, the effective sample sizes were 1,185 for proximal and 797 for occlusal restorations. Expected widths of 95% CIs, taking the observed percentages of 5.9% and 15.9% as the underlying true values, are 2.6% for proximal and 5.0% for occlusal restorations. None of the two-level interactions between covariates was significant.

Table 5.

Factors associated with placement of restorations for enamel carious lesions on occlusal and proximal surfaces

	Occlusal lesions*			Proximal lesions*
	OR	95% CI	P	OR	95% CI	P
Patient characteristics
Age			0.0024
<18 years vs. 18–44 years	1.74	1.33–2.29	<.0001	NA	NA	NA
45–64 years vs. 18–44 years	0.91	0.68–1.22	0.5393	NA	NA	NA
65+ years vs. 18–44 years	1.08	0.59–1.97	0.8096	NA	NA	NA
Male vs. female	NA	NA	NA	1.25	0.99–1.57	0.0831
Non-Hispanic white vs. non-white	1.36	1.07–1.73	0.0145	NA	NA	NA
Dentist characteristics
Male vs. female	0.89	0.53–1.51	0.6566	1.09	0.53–2.23	0.8133
Non-Hispanic white vs. non-white	NA	NA	NA	5.35	1.72–16.64	0.0057
Assesses caries risk			0.0447
Yes (with form) vs. no	0.44	0.22–0.90	0.0248	NA	NA	NA
Yes (no form) vs. no	0.56	0.35–0.92	0.0206	NA	NA	NA
Practice region or type
Network region			<.0001			<.0001
AL/MS vs. PDA	4.90	2.07–11.61	0.0003	2.71	1.01–7.25	0.0479
FL/GA vs. PDA	2.90	1.15–7.30	0.0240	1.43	0.48–4.20	0.5187
MN vs. PDA	2.80	1.28–6.10	0.0097	0.76	0.24–2.34	0.6289
SK vs. PDA	0.71	0.23–2.21	0.5523	0.13	0.03–0.50	0.0032
Practice type**			0.0219			0.0008
Private vs. large prepaid group	1.71	0.97–3.02	0.0625	2.05	0.71–2.88	0.0407
Public vs. large prepaid group	0.49	0.14–1.76	0.2769	0.18	0.02–1.39	0.1007

OR: Odds ratio; CI: Confidence Interval; PBRN: Practice-Based Research Network; PDA: Permanente Dental Associates.

^*Includes restorations involving both occlusal and proximal surfaces. ORs (CIs) for occlusal and proximal models are adjusted for covariates and patient and dentist clustering.

^**Substituting practice type into the model led to trivial changes in the ORs (CIs) for other variables.

The final model for occlusal lesions shows that patient age (p=0.0024), race/ethnicity (p=0.0145), caries risk assessment (p=0.0447), and network region (p<0.0001) were significantly related to variations in enamel restoration enrollment, adjusting for other covariates and for within-patient/dentist clustering. Patient gender and dentist race-ethnicity were not significant in blocked analysis and we excluded them from the final model. Contrast tests indicated patients ages <18 years were significantly more likely (p<0.0001) than 18–44 year olds to have an enamel restoration. Results for other age groups did not differ from 18–44 year olds. Non-Hispanic white patients were more likely to have an enamel occlusal restoration compared to patients who identified themselves as Hispanic or African American, Asian, or other race. Patients treated by dentists who conducted a caries risk assessment were about half as likely to have an enamel occlusal restoration compared to patients with dentists who did not assess caries risk, whether the dentist used a form or not. Patients in the MN, FL/GA, and AL/MS regions were 2.80 to 4.90 times more likely to have an enamel occlusal restoration compared to PDA region patients. Differences between SK and PDA patients were not statistically significant. When we substituted practice type for network region, patients receiving care from private practice dentists were 71% more likely to have an enamel occlusal restoration compared to patients from large prepaid group practices; however, this result was not statistically significant.

The final model for enamel restorations shows that practitioner race/ethnicity (p=0.0057) and network region (p<0.0001) were significantly related to variations in enamel lesion restoration, adjusting for other covariates and for within patient/dentist clustering. Patient age and race/ethnicity, and dentist gender and caries risk assessment, were not significant during blocked analysis and we excluded these factors from the final model. Contrast tests indicated that enamel proximal restorations were more than five times more likely to be enrolled by non-Hispanic white dentists compared to dentists who identified themselves as Hispanic or African American, Asian, or other race. Compared to PDA region patients, AL/MS patients were nearly three times more likely, and SK patients were about 13% as likely, to have an enamel lesion restoration. Patient gender was included in the final model, but was not significantly related to enamel restorations at p<0.05. When practice type was substituted for network region, we found that patients from private practices were twice as likely as patients from large prepaid group practices to have an enamel proximal restoration; we found no difference between public health and large prepaid group practice.

Discussion

We found significant variations in the percentage of enrolled enamel caries restorations associated with network region, practice type, and select practitioner and patient characteristics. Network region and practice type were highly correlated, and both were significantly associated with enamel restorations for both occlusal and proximal caries, after accounting for other factors. Overall, enamel lesions accounted for 16% of occlusal surface and 6% of proximal surface restorations among network dentists. These percentages are somewhat higher than the Domejean-Orliaguet et al.²⁸ study of general dentists in France, which found that enamel caries represented 12% of enrolled pit and fissure surface restorations and 2% of proximal surface restorations. Compared to French dentists, the percentages of enamel restorations for occlusal surfaces were higher than 12% for AL/MS (24%) and FL/GA (17%) patients and lower for MN (6%) and SK (4%) patients. In all regions except SK, the percentages of enamel proximal restorations were higher than 2%.

Network region and practice type were significant predictors of variations in enamel restorations, controlling for other factors. Region and practice type were highly correlated and not able to be included in a single regression model. AL/MS and FL/GA dentists were predominantly independent or small-group private practitioners. PDA and MN dentists were mostly in large, multi-specialty, pre-paid group practices (the latter from HealthPartners). Scandinavian dentists were either in independent, small group or public-health practices.

Existing research suggests that enamel caries restorations are reduced if dentists use quality-based performance measures, clinical practice guidelines, or apply evidence-based practice concepts.^2,11,15,16 Consequently, when we re-ran the models using practice type instead of network region we expected to find lower rates of enamel restorations among patients in the MN, PDA and SK regions, and with public and large prepaid group practices. PDA and HealthPartners have strong commitments to evidence-based practice and partially compensate dentists based on quality measure achievement. HealthPartners uses established practice guidelines promoting non-surgical treatment of enamel lesions,⁴⁰ while PDA reviews and disseminates treatment recommendations through a Clinical Effectiveness Committee. Yet, our results indicated that enamel occlusal restorations were nearly three times more likely among patients from the MN region compared to PDA patients. Compared to the PDA region, MN and FL/GA regions were similar for both occlusal and proximal enamel restorations, even though practice types were very different. Enamel restorations were extremely rare among SK dentists regardless of practice type, likely reflecting the widespread adoption of treatment recommendations promoting preventive and non-invasive treatment of enamel lesions in Denmark, Norway, and Sweden.^30,41 Educational differences between American and Scandinavian dentists^25,42,43 also may explain the stark divergence in treatment thresholds.

There were significant regional differences in participating dentists’ demographic characteristics, years since graduation, and practice busyness, but only dentist’s race-ethnicity was associated with differences in enamel restorations for proximal lesions only. Non-Hispanic White dentists were five times more likely than minority dentists to place an enamel restoration. This relationship warrants further examination, but is beyond the scope of this paper. Network dentists have fewer years since dental school graduation compared to dentists as a whole, but are similar across other demographic and practice characteristics.³⁷ We did not assess which dental school participants attended.

We expected to find a significant association between enamel restorations and dentists’ years of experience. Bonetti and colleagues found dentist’s attitudes about a treatment and its expected outcome are consistent predictors of treatment intentions.^44–46 Past positive outcomes from a particular treatment strategy can reinforce the use of this strategy.⁴⁵ These results are consistent with a recent qualitative study of dentists’ use of evidence that found that the most influential data were based on “concrete evidence seen in their patients’ mouths” and accumulated over years of practice.⁴⁷ In this study, we found no relationship between years since dental school graduation and enamel restorations. We doubt that experience has no role in restorative treatment decisions, but that its effect on decision-making may be more complex (even counter-veiling) than the number of years since graduation. For instance, we might expect older dentists would be more likely to restore enamel lesions if they established treatment patterns prior to publication of caries treatment recommendations. Conversely, perhaps network dentists with more practice experience are more familiar with difficult diagnosis and treatment decisions, and with the slow progression of enamel caries, and thus are less likely to restore enamel caries compared to more-recent graduates. Practical experience suggests that a large proportion of enamel caries (assessed preoperatively) actually extend into the dentin, which is a rationale for surgically treating enamel lesions.⁴⁸ Data from the parent clinical study found that 45% of dentist’s pre-operative depth assessments differed when compared to post-operative depths.⁴⁹ There was no relationship between diagnostic method used and variations in enamel restorations.

We expected that time constraints among busy practices might lead dentists to de-prioritize enamel restorations among enrolled patients, while dentists with available time might be more likely to do enamel restorations. However, we found no relationship between practice busyness and enamel restorations. There was significant variation in practice busyness across network regions; dentists from the FL/GA and AL/MS regions reported the highest rates (23% and 18%, respectively) of being “not busy enough.” Even though patients from these two regions also had the highest rates of enamel restorations, practice busyness appeared to not account for any of the association between network region and enamel restorations.

Treatment recommendations acknowledge the importance of assessing patients’ caries risk as part of the decision-making process for enamel caries.^1,2,40 Patients at high risk for caries may need restorative treatment of enamel caries, based on individual assessments of past and current caries activity, dental hygiene, and dietary habits.² Patient-level caries risk data were not available for this study. We did, however, evaluate the predictive value of providers’ self-reported use of caries risk assessment and the number of currently enrolled caries lesions for each patient. We expected that enamel restorations would be more common among dentists who did not assess caries risk, given that dentists assessing risk may be more likely to choose non-surgical treatments for their patients.^26,42 We found that patients of dentists who assess caries risk were less likely to have enamel restorations enrolled for occlusal lesions, while there was no association for proximal lesions. While the number of enrolled lesions per patient varied by region and practice type, we found no evidence that patients with more than one enrolled lesion were more likely to have enrolled enamel lesions. Thus, it appears that dentists’ general approach to caries risk assessment has some impact on treatment thresholds, but patient-level risk data may not.

There were significant regional variations in patients’ age, gender, race-ethnicity, insurance coverage, and numbers of enrolled restorations. Patient age and race-ethnicity were associated with differences in enamel restorations for occlusal, but not proximal, surfaces. Two-way interactions between covariates were also not significant. We expected the percentage of enamel restorations would be lower for younger patients compared to older age groups, as dentists sought to preserve tooth structure and longevity through less-aggressive interventions.^1,28,50 Yet, we found that children were about 75% more likely than adults to have an enrolled occlusal restoration confined to the enamel, controlling for other factors. This counter-intuitive result warrants further study. There was no relationship between patient age and enamel proximal restorations.

We found no association between enamel restorations and patients having any insurance coverage. Previous studies have found mixed effects on treatment decisions associated with insurance coverage,^25,30 although reimbursement for preventive treatment (e.g., fluoride varnish) of caries may reduce the likelihood of an enamel restoration.^10,30 Time constraints for data collection precluded comparisons by insurance type, e.g., public or commercial, but we expect little impact as reimbursement rules are similar.

This study has several limitations. First, we did not enroll patients who had no restoration placed. Thus, we cannot assess the percentages of lesions restored for each enamel and dentin lesion depth, nor can we directly compare our clinical results to dentists’ self-reported treatment thresholds for hypothetical patients. A direct comparison would have been possible had we also collected clinical data for carious lesions treated non-surgically. Second, we did not provide caries diagnosis calibration training to participating dentists. Lesion depths were based on self-reported pre-operative clinical assessments using routine diagnostic techniques. Dentists used a combination of clinical assessment, optical techniques including transillumination, or radiographs (analog or digital).³⁹ While there could be considerable variation between dentists’ treatment thresholds, the large number (n=229) of widely dispersed practitioners participating in the study made calibration training impractical in terms of costs and time. Also, a key objective of this study was to quantify the percentage of restorations done in enamel using the method of lesion depth assessment that the dentist normally uses in routine practice; therefore, calibration is counter-productive and not desirable. We found that most dentists used a combination of visual and radiographic assessment, which was similar to other US practitioners.³⁷

Lastly, the original restoration study did not capture information about the behavioral determinants of dentists’ treatment decisions. There has been increasing interest in using psychological theories, such as the Theory of Planned Behavior (TPB) or Social Cognitive Theory, to understand healthcare provider behavior, including dentists.^{44–46,51–54} TPB posits that treatment intentions are predicated on a combination of provider’s a) attitudes about a condition, intervention, and expected outcome, b) existing subjective norms about the intervention, and c) perceived behavioral control over the intervention. Actual treatment decisions are subsequently determined by treatment intentions, along with an additional direct influence from perceived behavioral control. For this study, the available data did not support a comprehensive post-hoc application of the TPB to enamel caries restoration decisions. However, our findings suggest that dentists’ social norms about restorative treatment may reduce the likelihood of enamel lesion restorations. The wide-spread adoption of non-surgical treatment for enamel caries in Northern Europe^42,43,55 was reflected in our Scandinavia region. Compared to private office practice, enamel restoration rates were similar for large group and Scandinavian practices, except for occlusal surfaces, for MN patients. A recent review by Perkins and colleagues⁵⁴ found that subjective norms can be a strong predictor of healthcare provider behavior, including HMO providers. Dentists may also interact with each other and their patients in ways that affect regional practice norms, and are related to practice type and location but were beyond the scope of this study to detect.

Guidelines on the treatment and management of dental caries recommend that dental care providers use non-surgical interventions to treat caries lesions confined to the enamel unless caries risk is high.^1,2 This study found that network region and practice type were important predictors of variations in occlusal and proximal enamel restorations. Furthermore, restorative treatment thresholds for occlusal lesions appeared to contradict expectations that practitioners would limit surgical interventions for children. These issues require more detailed examination to understand dentists’ treatment thresholds in the context of routine clinical practice, particularly the roles of practice environment and caries risk assessment strategies. Prospective clinical studies are needed to evaluate factors associated with dentists’ decisions to remineralize, seal, or restore carious lesions. These studies should include assessments of lesion depth, and of the long-term effectiveness and cost-effectiveness of each treatment strategy. Furthermore, research is needed to understand the behavioral determinants of dentists’ caries treatment thresholds in the context of routine clinical care. Understanding what drives dentists’ decisions can better inform efforts to infuse evidence-based decision-making into routine practice.