OVERESTIMATION OF EARLY CHILDHOOD CARIES USING THE dmfs INDICES

Abstract

Purpose:

The preferred epidemiological caries assessment method is the decayed, missing, filled surfaces (dmfs) score, which records all crowned/missing primary teeth’s surfaces as carious. This study evaluated the dmfs score’s accuracy in capturing “caries-affected” (vs. treated) surfaces of crowned/extracted teeth.

Methods:

A high-caries risk cohort of children, eight to 18 months at baseline, were recruited from a non-fluoridated, rural, minority, and low-income community. Oral examinations occurred every 12 months for 5 years, identifying children with at least one caries-related crown/extraction (N = 45). Observed scoring counted all crowned/extracted surfaces as carious. Private dentists’ clinical records were also reviewed to determine how many surfaces were actually carious at crown/extraction appointments (53 actual scores for N = 19). Differences in actual and observed scoring were evaluated (sign test, α = 0.05 with two-tailed P-values).

Results:

Most children in the study group had more than one crown/extraction. Actual scoring revealed two to three fewer carious surfaces per tooth than observed scoring; cumulatively, observed scoring added two to 27 more surface counts per participant (P < 0.001).

Conclusions:

Observed scoring exaggerated early childhood caries burdens when crowns/extractions were prevalent. Modified dmfs scoring, individualized or population-corrected crown/extraction counts, could more accurately estimate disease.

INTRODUCTION

Socioeconomically disadvantaged children experience most of the early childhood caries (EEC), resulting in higher decayed, missing, filled surfaces (dmfs) scores. They often require crowns and/or extractions due to the extent of disease^1–4. Positively speaking, recent United States data demonstrates more disease is currently treated than untreated⁵. Unlike clinicians, annual field examinations often do not record carious surfaces just prior to crown/extraction treatment. Therefore, caries activity can be missed until post-restoration or extraction at research follow-up. Children/parents are unreliable data sources since they are unlikely to be aware of the number of carious surfaces a tooth had at a crown/extraction visit⁶. Pediatric epidemiological studies rarely utilize radiographs because of ethics (e.g., frequent x-ray exposure), practicality (e.g., lack of child cooperation), and finances. Therefore, most observational ECC studies (i.e., investigators are not directly involved with children’s dental care) must rely on dmfs scoring to measure primary teeth’s caries^6,7. The dmfs score counts all surfaces of crowned/extracted carious teeth as carious, regardless of surface status at treatment time^3,8,9. However, at the inception of the DMFS score (caries scoring for permanent dentitions), an evidence-based recommendation suggested three surface caries counts be used for crowned/extracted carious teeth⁶. Additional research shows that fewer than all surfaces of crowned/extracted permanent teeth are carious^3,6,8,9. ECC overestimates likely exist, but researchers need the supporting data to demonstrate it^8,9.

Counting all crowned and extracted surfaces as carious therefore results in the dmfs index measuring treatment on these teeth, not disease, since clinically all tooth surfaces do not have to be carious to warrant such treatments^3,4,9. The dmfs index, minus the m- and f-measurements, could be used to quantify ECC and avoid crown/extraction assumptions, but it would miss disease documentation by ignoring all missing, crowned, and traditionally-filled teeth, only including untreated carious surfaces^9,10. The dmft is another alternative for overcoming the dmfs index’s crown/extraction dilemma since it only measures if a tooth was carious. The dmft index loses a significant amount of disease discrimination by looking at the tooth generally and not surface by surface¹⁰. These instances of caries data loss become compounded since researchers use disease data to measure disease burdens and identify risk factors, eventually creating risk factor assessments; clinicians use the risk assessments to determine patient risk and appropriate courses of action; and public officials use disease burden data to evaluate the effectiveness of current (and the need for additional/new) caries prevention/risk-reduction public health programs^9,11. Therefore, caries epidemiological studies must rely on accurate disease measurements^5,11,12.

These implications indicate the need for investigations into epidemiological ECC measurement through dmfs scoring of crowns/extractions. Diagnostic notes and pretreatment radiographs, collectively, should provide optimal documentation of carious primary tooth surfaces prior to restoration/extraction such that diseased surfaces can be quantified with certainty¹³. To understand if ECC experiences are being evaluated properly, this study quantified the dmfs score’s accuracy in capturing the amount of “caries-affected” (vs. treated) crowned/extracted primary teeth surfaces. It was hypothesized that, for a cohort of high-caries risk children, the m- and f-components of the observed dmfs score recorded more carious surfaces for crowns/extractions than were actually diseased, overestimating caries incidence.

METHODS

This epidemiologically methods-based investigation was ancillary to an ongoing cohort study of high-caries risk children to understand childhood dental caries epidemiology better. The community is high-caries risk because of their low socioeconomic status, minority (African American) population, and minimally fluoridated community water sources. Their rural residency and limited dental care access also increase their disease risks^14–16.

Pre-recruitment calculations determined the appropriate number of participants that would power the parent study sufficiently to identify caries risk factors, accounting for lengthier follow-ups and consequential participant attrition. Healthy (i.e., no baseline systemic diseases) children, eight to 18 months were recruited through convenience sampling (e.g., fliers, community center events, community advocates, etc.) from Uniontown, Alabama, between July 2008 and December 2009. After baseline, annual study visits continued for five years through early 2015. Study staff implemented appointment scheduling and reminding efforts to reduce loss to follow-up.

Children with at least one study-observed crown/extraction (for caries) and complete follow-up (i.e., attended baseline and all follow-up visits) for study re-consenting were included in this methodological investigation. Study staff repeated the parent study’s informed consent and waiver of assent procedure to include identification of study participants’ treating, non-study related community dentists and clinical treatment record review. Designated dentists provided clinical treatment records, including chart notes and radiographs, to study staff. All studies complied with Declaration of Helsinki guidelines; University of Alabama at Birmingham’s Institutional Review Board (Birmingham, AL) approved all research investigations.

Three non-treating, study dentists, periodically re-trained and calibrated (κ = 0.90+) for caries recording quality assurance, documented cavitated lesions on each tooth surface during annual visual oral examinations¹⁴. The valid and reliable World Health Organization’s (observed) dmfs score counted all crowned and extracted tooth surfaces as carious (i.e., score of five, including anterior teeth’s incisal surfaces)¹⁰. Resulting from one pilot-tested, study-trained dentist (i.e., principal investigator, NKC) reviewing clinical treatment notes and radiographs, actual dmfs scores incorporated the exact number of carious surfaces at crown/extraction treatment appointments. If exact surfaces could not be identified, children had “no actual score”. Crowned and extracted tooth scores were summed to produce a cumulative year-five score per individual for each scoring method. Untreated and filled caries scores were excluded due to equivalent counting across the two scoring approaches.

Children’s demographics, dietary and oral hygiene histories, as reported by legal guardians in baseline questionnaires, described their ECC crown/extraction-related qualities. Baseline difference tests (i.e., Chi-square/Fisher’s exact, Kruskal-Wallis, non-parametric equivalents) evaluated any selection bias between the actual scoring (i.e., carious surfaces documented) and no actual scoring (i.e., carious surfaces undocumented) groups. A sign test measured paired differences in observed and actual cumulative ECC dmfs scores. The sign test would be sufficiently powered (i.e., P ≥ 0.75) if there was a minimum of seventeen data pairs¹⁷. SAS 9.3 (SAS Institute Inc., Cary, NC, USA) conducted all statistical analyses at significance levels of α = 0.05 with two-tailed P-values reported.

RESULTS

Investigators initially recruited ninety-six infants; twenty-seven children were lost to follow-up throughout the study. Sixty-six children had complete follow-up and were available for further consenting to contact their community/private dentist. Forty-five children, mostly males, had at least one crown/extraction, however only 19 records (42.2%) had the information needed (i.e., surface-level details and/or radiographs for calculating actual caries scores, totaling fifty-three crowns/extractions) for chart review (Figure 1). Actual scored and no actual chart-scored groups were similar across baseline ECC-related characteristics (data not shown).

Figure 1.

Flowchart of the cohort with actual scores available to assess dmfs indices for crown- and/or extraction-treated early childhood caries.

*specific carious surfaces charted and/or radiographed in clinical dental treatment records.

More than two-thirds of actual scored children had more than one crown/extraction. Most (79%) of the reviewed teeth were crowns. Each tooth’s actual score was single-sourced (i.e., determined from a chart note or radiograph). Radiographs were the main (75%) actual score data source.

The mean actual caries count per crown/extraction was 2.2 surfaces (crowns: 2.1; extractions: 2.3). Observed year-five ECC dmfs counts suggested a range of two to 27 additional “caries-affected” surfaces per participant. The pairwise difference between the cumulative ECC dmfs scoring methods was significant (P < 0.001). Observed scores overestimated actual counts, producing a median score difference of nearly one whole tooth or four tooth surfaces (Table 1).

Table 1.

Summary of children’s (N = 19) cumulative early childhood caries (ECC) decayed, missing, filled (carious) surfaces (dmfs) scores and differences by disease counting approach.

Cumulative ECC dmfs Score*	Range	Quartile1**	Median	Quartile3†	p-value
Observed‡	35	5	10	20	--------
Actual§	12	2	6	8	--------
Score Differences§§	25	3	4	12	< 0.001

^*ECC surfaces summed for crowns and extractions only

^**25^th percentile: one quarter of data points are numerically oriented below this data point

^†75^th percentile: three quarters of data points are numerically oriented below this data point

^‡all surfaces assigned as carious per treated tooth

^§specific surfaces identified as carious prior to treatment in clinical dental treatment records

^§§pairwise (observed – actual) score count difference per participant

Bold p-value: statistical significance (α ≤ 0.05)

DISCUSSION

Actual scoring demonstrated that not all crown/extracted tooth surfaces were carious in the group of charts reviewed, countering observed dmfs score assumptions and supporting our study’s hypothesis. Observed scores produced values consistently exceeding actual ECC counts, differing by as much as five teeth (i.e., 25 surfaces). Here, where crowns/extractions were abundant, dmfs assumptions inflated true ECC burdens.

Extractions are appropriate for un-restorable teeth, including deep, one surface lesions^3,9 just as crowns are indicated for multi-surface caries, even when only two surfaces are involved⁴. Extensively damaged crowned/extracted teeth where higher caries scores are appropriate when a minimal number of surfaces are actually affected can exist; these are rare and exceptions to our findings. All surfaces of three treated teeth (six percent) were carious at treatment time in our study. Actual scoring finding an average of two carious surfaces per treated tooth agrees with permanent tooth studies which found, on average, no more than two and three carious permanent surfaces at each crown and extraction site, respectively^3,6,9. Investigators use modified DMFS scores to address the measure’s limitations, using M-component correction factors that range from arbitrary surface count assignment to population-specific values^3,6,9. Observed dmfs scoring likely warrants similar adjustments to overcome its inaccuracies.

Valid epidemiological measurements must be as accurate as possible to facilitate statistical analyses; mismeasurement creates biased results¹¹. By specifying all surfaces of crowns/extractions as carious, the observed dmfs measure is estimating ECC treatment (on all surfaces) and not disease. With higher amounts of treated (vs. untreated) ECC and crowns being a cornerstone in pediatric dental care, these exaggerations will likely perpetuate in future epidemiological studies, obscuring true disease burdens^1,4,5. Further, more crowns/extractions should suggest better dental care access (i.e., improved dental public health). Instead by demonstrating more disease than present, observed crown/extraction dmfs counts can inaccurately portray dental public health initiatives as ineffective. Clinician’s caries risk assessments could wrongly include risk factors for treated versus actual caries disease, resulting in less optimal ECC prevention^5,12.

Given this small study, future ECC research should determine if crowns/extractions result in significant overestimations of disease with dmfs scoring in epidemiological studies. In this regard, studies should incorporate chart review or (validated) population specific correction factors into dmfs scoring’s m- and f-components directly to improve accuracy of ECC burden. Reporting disease burden and/or risk factor analyses with observed and modified (or actual) dmfs scores would allow results to reassess existing literature to demonstrate differences in risk factor identification and facilitate improved clinical caries risk assessments and public health.

This introductory investigation used a high quality longitudinal dataset to demonstrate a potential epidemiological methods inaccuracy in ECC observations. A homogenous (i.e., age, race/ethnicity, & socioeconomic status) cohort of high-caries risk children emphasized disease measurements without having to adjust for population differences. In this regard, existing National Institute of Health-funded prospective, strongly reliable (κ = 0.90+)¹⁴ data with five years of follow-up maximized crown/extraction observations.

Study design elements are also acknowledged to be study limitations. Participant availability for re-consent, guardian-provided extraction reasons, and non-adjustable radiographic equipment for very young children impeded data collection. In addition, clinicians do not create patient records for research purposes; charts were incomplete and/or inconsistent across participants—an inherent limitation of chart review studies. Dentistry reimburses for treatment (not diagnostics), resulting in practicing dentists often denoting diseased tooth number and treatment proposed crown/extraction rather than diagnostic details, such as tooth surfaces involved. Twenty-six clinical treatment charts did not contain surface details adequate for surface diagnosis and therefore were not useful for chart review. This missing data potentially biased (underestimated) study results and restricted data representativeness. Similarities in potential baseline ECC risk factors between actual and no actual scoring groups suggests the groups had similar disease amounts, minimizing any sample size, biased results, and data representativeness concerns. However, implementation of detailed diagnostic coding by dental professionals could alleviate this issue in future dental epidemiology studies. Potential selection bias with reliance on participants completing the study (i.e., 5-year dmfs data) also likely underestimated disease burden. The distinct population limits the study’s generalizability to demographically similar groups. Lower-caries risk groups will have fewer crowns/extractions, likely resulting in less pronounced inflations of caries scoring data.

Regardless, this study demonstrated instances where observed ECC dmfs scoring reported additional crown/extraction surfaces as “caries-affected” because they were treated, but were not actually diseased. Accepted epidemiological methodology (i.e., observed dmfs) may overstate disease extents for crowns/extractions in the primary dentition. Future studies are needed to confirm the overestimation findings and establish acceptable population-specific correction factors, if chart review is not feasible, for scoring teeth. These findings could improve disease burden and risk factor descriptions over the observed dmfs measure. Better ECC measurement accuracy and subsequent risk factor identification can create more effective ECC prevention/control efforts (e.g., clinical caries risk assessment tools, dentist-facilitated patient dental education materials, etc.) that when implemented will improve pediatric oral health.

CONCLUSIONS

1. Clinically, high-caries risk children’s crowned/extracted carious teeth may not have all tooth surfaces diseased at treatment time.

2. When crowns/extractions are prevalent in epidemiological studies, observed dmfs scoring can significantly overestimate true disease incidence (and prevalence), misconstruing ECC epidemiology.

3. Practicing dentists’ incorporation of detailed diagnostic caries descriptions, especially when crown and extraction treatments are included, can aid in the accuracy of disease measurement in epidemiological studies.

4. Future more extensive studies should investigate potential dmfs scoring adjustments to improve the measure’s validity, ECC epidemiological research, and, indirectly, clinicians’ caries risk assessments which, in turn, will collectively better inform clinician and public health decisions promoting individual and population ECC prevention/control efforts, respectively.