• Component Description
• Eligible Sample
• Protocol and Procedure
• Quality Assurance & Quality Control
• Data Processing and Editing
• Analytic Notes
• References
• Codebook
  • SEQN - Respondent sequence number
  • FCX02DI - Fluorosis DI: #2
  • FCX03DI - Fluorosis DI: #3
  • FCX04DI - Fluorosis DI: #4
  • FCX05DI - Fluorosis DI: #5
  • FCX06DI - Fluorosis DI: #6
  • FCX07DI - Fluorosis DI: #7
  • FCX08DI - Fluorosis DI: #8
  • FCX09DI - Fluorosis DI: #9
  • FCX10DI - Fluorosis DI: #10
  • FCX11DI - Fluorosis DI: #11
  • FCX12DI - Fluorosis DI: #12
  • FCX13DI - Fluorosis DI: #13
  • FCX14DI - Fluorosis DI: #14
  • FCX15DI - Fluorosis DI: #15
  • FCX18DI - Fluorosis DI: #18
  • FCX19DI - Fluorosis DI: #19
  • FCX20DI - Fluorosis DI: #20
  • FCX21DI - Fluorosis DI: #21
  • FCX22DI - Fluorosis DI: #22
  • FCX23DI - Fluorosis DI: #23
  • FCX24DI - Fluorosis DI: #24
  • FCX25DI - Fluorosis DI: #25
  • FCX26DI - Fluorosis DI: #26
  • FCX27DI - Fluorosis DI: #27
  • FCX28DI - Fluorosis DI: #28
  • FCX29DI - Fluorosis DI: #29
  • FCX30DI - Fluorosis DI: #30
  • FCX31DI - Fluorosis DI: #31

National Health and Nutrition Examination Survey

2011-2012 Data Documentation, Codebook, and Frequencies

Fluorosis - Clinical (FLXCLN_G)

Data File: FLXCLN_G.xpt

First Published: April 2019

Last Revised: NA

Component Description

Dental fluorosis clinical assessment was included in the NHANES oral health examination to monitor fluorosis in the population for selected age groups. Fluoride exposure, from any source, during the period of tooth development is associated with increased risk of developing dental (or enamel) fluorosis. Dental fluorosis is characterized by an increasing porosity or hypomineralization of the tooth enamel that leads to visual changes of the enamel that appear once a tooth erupts (Dean, 1934; Fejerskov et al., 1990). The severity of dental fluorosis depends on the dose and duration of fluoride ingestion during tooth development (Dean, 1942; Fejerskov et al., 1990).

In the United States, dental fluorosis is generally considered a cosmetic effect with no negative functional effect (Kaminsky et al., 1990; Fluoride Recommendation Work Group, 2001; US Department of Health and Human Services, 2015). The severe form of dental fluorosis, however, may have adverse dental effects because the pitting can compromise the protective function of the enamel and the affected area can break away (Clark and Slayton, 2014; Fejerskov, et al., 1990; National Research Council, 2006; US Department of Health and Human Services, 2015). But the severe form is rare in the U.S. (Beltran 2010; National Research Council, 2006).

Eligible Sample

A dental fluorosis clinical assessment was conducted on survey participants aged 6-19 years who received the Oral Health Examination (exam status code, OHDEXSTS, is complete or partial).

Protocol and Procedure

The 2011-2012 dental fluorosis clinical assessment followed the same protocols as conducted in 1999-2004.

The dental fluorosis clinical assessment was conducted at the NHANES mobile examination center (MEC) by dental examiners, who were dentists (D.D.S. or D.M.D.) licensed in at least one U.S. state. A health technician assisted in entering all examiner observations directly into a computerized data collection system at the MEC. Examiners used a surface reflecting mirror for the assessment. Teeth were not dried with air before assessment.

All fully erupted, permanent teeth (excluding third molars) were evaluated. The dental fluorosis clinical assessment proceeded tooth-by-tooth in a similar manner as the dental caries assessment, beginning with the maxillary right central incisor and proceeding posteriorly to the upper second molar. Then, the same sequence was repeated for the upper left, lower left, and lower right quadrants of the mouth.

Each tooth was scored according to the Dean’s Fluorosis Index (DFI) and assigned one of the DFI disease severity categories (Table 1), based on the area of the tooth surface with visible fluorosis and presence of pitting: normal (DFI=0), questionable (5), very mild (1), mild (2), moderate (3), or severe (4) (Dean, 1934; Dean, 1942). Missing teeth, deciduous (primary) teeth, permanent teeth not fully erupted, and teeth in which more than one-half of the visible surface area was obscured by a restoration, caries, or orthodontic appliance were not assessed. These teeth were coded as cannot be assessed (9). A tooth having a non-fluoride opacity was coded as 8. The NHANES Oral Health Examiners Manual provides detail on assessing a tooth for fluorosis.

 

Table 1. Dean’s Fluorosis Index (DFI) criteria and scoring on the NHANES dental fluorosis clinical assessment

NHANES DFI Value	Fluorosis Severity Level	Description
0	Normal	No fluorosis detected
1	Very mild	Opaque, paperwhite areas involving less than ¼ of the tooth surface
2	Mild	Opaque, paperwhite areas involving ¼ to less than ½ of the tooth surface
3	Moderate	Opaque paperwhite areas involving ½ or more of the tooth surface
4	Severe	Discrete or confluent pitting in involved areas
5	Questionable	Slight aberration of normal enamel appearance, including white flecks
8	Non-fluoride opacity	Coded if non-fluoride opacity
9	Cannot be assessed	Coded if the tooth was missing, not fully erupted, one-half or more of the tooth was replaced with a restoration, covered with orthodontic band, or destroyed by caries.

 

Quality Assurance & Quality Control

The specific QA practices for the dental fluorosis clinical assessment are documented in detail in the NHANES Oral Health Examiners Manual.

All dental examiners received an initial training which consisted of lecture, model review, practice simulations and standardization sessions. Following successful initial training, examiners received field training at the MEC consisting of more practice simulations, standardizations, and calibration sessions.

During data collection, the reference examiner visited each dental examiner 2-3 times a year to conduct a random number of replicate examinations during each visit. The reference examiner determined if retraining and future monitoring of the dental examiner was needed. The reference examiner also conducted an annual retraining session for all dental examiners to reinforce existing protocols and to introduce protocol updates as needed.

Since 1999, data for the oral health examination have been recorded directly onto a computerized data collection system at the MEC. The system is integrated centrally and allows for ongoing monitoring of much of the data. As part of the quality control practice, all data are reviewed systematically for logical inconsistencies. Before data release, the collected data are further reviewed.

Data Processing and Editing

While the dental examiners assess the teeth in quadrants, starting from the central incisor and moving to the molars, the data files are produced with the teeth numbered using the Universal or ADA Dental Numbering System. Note that there is no data for tooth numbers 1, 16, 17, and 32. Third molars were not assessed in this clinical examination.

Analytic Notes

1999-2004 and 2011-2016 DATA QUALITY ASSESSMENT CONCLUSIONS

NHANES is currently the only survey providing national estimates on dental fluorosis. Two-year estimates of fluorosis prevalence by DFI category demonstrated variability within and across the six year time periods. Inter-examiner reliability statistics found that agreement ranged from 0.51-0.98 for the nine primary dental examiners in 1999-2016. These values indicate moderate (0.41-0.60) to almost perfect (0.81-0.99) agreement based on Landis and Koch, and mostly adequate agreement (33 out 36 kappas >=0.6) based on McHugh. There was also high percent agreement in defining very mild or greater fluorosis. The proportion of the total number of examinations for which a gold standard examination was conducted was, however, low and the DFI scoring method has high examiner subjectivity. The observed increase in dental fluorosis prevalence with age between 2001-2004 and 2011-2014, based on the analysis of the synthetic cohort, is not biologically plausible. This suggests that there may have been some change in the way the examiners evaluated the level of fluorosis over time.

The quality assessment findings in this document should be strongly considered when determining whether these data are appropriate for the user’s analytic objectives, including studies of prevalence and trends. An NCHS Vital and Health Statistics Series 2 Report on the data quality for the NHANES 1999-2004 and 2011-2016 dental fluorosis clinical assessment data is available at NCHS website (National Center for Health Statistics and National Center for Chronic Disease Prevention and Health Promotion, 2019).

 

DATA EVALUATION OF 1999-2004 AND 2011-2016 FLUOROSIS CLINICAL ASSESSMENT DATA

As described previously, several QA processes were implemented during the dental fluorosis clinical assessment. In addition, evaluation of rater variability and reliability was assessed on the final data. The summary of the data evaluation of the dental fluorosis clinical assessment data from 1999-2004 and 2011-2016 for participants aged 6-19 years (the common age range across all survey years) is provided below.

For these analyses, a person was assigned a dental fluorosis severity value based on the lesser of the two most affected tooth-level DFI values. For all analyses, tooth-level DFI value of cannot be assessed (code 9) and non-fluoride opacity (code 8) were recoded to missing and questionable recoded to 0.5. Only youth with at least two teeth with a non-missing DFI value were assigned a person-level DFI and included in further analyses. Qualitative interpretation of numeric kappa statistic value ranges are: less than chance agreement (<0), slight agreement (0.00-0.20), fair agreement (0.21-0.40), moderate agreement (0.41-0.60), substantial agreement (0.61-0.80), and almost perfect (0.81-0.99) (Landis and Koch, 1977). Kappa <0.60 may also be used as a general indicator of inadequate agreement among raters (McHugh 2012).

Intra-Examiner Reliability: Evaluation of Replicate Examinations from 1999–2001

During 1999–2001, approximately 10% of examined participants aged 6-49 had a repeat second fluorosis examination. Details on these replicate examinations were previously published (Dye et al. 2007) and are summarized here. The weighted kappa statistics comparing the DFI values for the same participant by the same examiner ranged from 0.56 – 0.72, across the three main dental examiners who conducted examinations from 1999 through 2001. These kappa values are considered moderate to substantial agreement. The difference in DFI values for the same teeth assessed in the same way days apart, however, indicates the subjective nature of the DFI scoring method.

Inter-Examiner Reliability: Evaluation of Gold Standard Examinations from 1999–2004 and 2011–2016

Gold standard examinations were conducted by a reference examiner on 3.6% (n=356) and 2.8% (n=210) of participants aged 6-19 years in 1999–2004 and 2011–2016, respectively. Complete data for analysis (i.e., at least two teeth with valid DFI values for both examiner and reference) were available for 339 participants aged 6-19 years in 1999-2004 and 198 in 2011-2016. These analyses did not use the survey examination weights.

Weighted kappa statistics were computed using different weighting schemes (explained below) which assigned specific values to the various possible levels of disagreement between dental examiner and reference examiner on the DFI scale (i.e., difference in one category, two categories, etc.).

Unweighted kappa (Cohen et al. 1960) assigns a weight of 1 for perfect agreement (e.g., examiner DFI = 1 and reference DFI =1) and 0 for no agreement (e.g., examiner DFI = 1 and reference DFI =2); Custom #1 weights assigned 1 for perfect agreement, 0.667 for 1 category difference, 0.333 for a 2 category difference, and 0 for more than 2 categories; Custom #2 (based on Kumar et al. 2000) assigned 1 for perfect agreement, 0.5 for 1 category difference, and 0 for all others; the other two weighting schemes were based on Cicchetti and Allison (1971) and Fleiss and Cohen (1973).

Unweighted kappa values for examiners ranged from 0.35 to 0.78, with five of nine examiners having values below 0.60 during the two 6-year time periods (Table 2). When considering the results across all four weighting schemes, the agreement between the dental examiners and reference examiner ranged from 0.51 through 0.87 for the five primary dental examiners who conducted examinations in 1999-2004 and from 0.60 through 0.98 for the four primary examiners during 2011-2016. For both time periods, these reliability statistics would be considered moderate to almost perfect agreement based on the Landis and Koch interpretation of kappa values. When McHugh’s requirement of a kappa value of at least 0.60 for adequate agreement is applied, the majority of the kappa values indicate adequate agreement (i.e., 33 out of 36 kappa values (computed for 9 examiners for 4 different weighting schemes) were at or above 0.60).

 

Table 2. Percent agreement and kappa statistics on person-level Dean’s Fluorosis Index values assigned by dental examiner and reference examiner for participants aged 6-19 years with available gold standard observations by 6-year survey period: National Health and Nutrition Examination Survey, 1999-2004 and 2011-2016

Survey Years	Examiner	n	Agreement (%)	Unweighted kappa	Weighted Cicchetti-Allison	Weighted Fleiss-Cohen	Weighted Custom #1	Weighted Custom #2
									1999-2004	A	62	58.1	0.45	0.62	0.77	0.60	0.54
									1999-2004	B	53	73.6	0.64	0.77	0.87	0.78	0.76
1999-2004	C	48	72.9	0.61	0.70	0.81	0.70	0.69
1999-2004	D	97	62.9	0.51	0.71	0.86	0.68	0.64
1999-2004	E	38	47.4	0.35	0.62	0.80	0.57	0.51
2011-2016	F	66	63.6	0.53	0.70	0.82	0.69	0.65
2011-2016	G	26	84.6	0.78	0.93	0.98	0.87	0.84
2011-2016	H	60	66.7	0.48	0.64	0.77	0.62	0.60
2011-2016	I	25	80.0	0.69	0.81	0.90	0.73	0.71

NOTES: All kappa statistics were computed using the following Dean’s Fluorosis Index values: 0 = Normal, 0.5 = Questionable (recoded from collected value of 5), 1 = Very mild; 2 = Mild, 3 = Moderate, 4 = Severe. All values of 8 (non-fluoride opacity) and 9 (could not assess) were set to missing.

 

Percent agreement between dental and reference examiners in categorizing fluorosis

When DFI categories are collapsed into broader categories and fluorosis is defined as mild or greater severity, the examiner and reference agreed on 88.8% of cases in 1999-2004; agreement was 89.4% in 2011-2016. When fluorosis is defined as moderate or severe, agreement is 97.1% in 1999-2004 and 94.4% in 2011-2016. It is expected that agreement would improve when broader categories are assessed since agreement is achieved as long as both examiners rate the same DFI category or higher.

As a means of detecting systematic scoring biases, differences between the dental examiner and reference examiner were further explored by examining the direction of the disagreement between the dental examiner’s and the reference examiner’s person-level DFI values (Table 3). Among the cases where there was disagreement between the examiner and reference, in 1999–2004, the examiner scored higher than the reference examiner on 48.8% of these cases, while the reference examiner scored higher than the dental examiner on 51.2%. In 2011–2016, the examiner scored higher than the reference examiner on 54.8% of cases, whereas the reference examiner scored higher than the examiner on 45.2%. The majority of the disagreement was by one DFI level: 82.7% in 1999–2004 and 88.7% in 2011–2016.

 

Table 3. Disagreement on person-level Dean’s Fluorosis Index values assigned by dental examiner and reference examiner for participants aged 6-19 years with available gold standard observations, by 6-year survey period: National Health and Nutrition Examination Survey, 1999-2004 and 2011-2016

Survey Years	# Gold Standard Exams	Disagree 1		Examiner > Reference	Reference > Examiner	Differ by 1 level	Differ by 2 levels	Differ by 3 levels	Differ by 4 levels	Differ by 5 levels
		n	%	%	%	%	%	%	%	%
1999-2004	339	127	37.5	48.8	51.2	82.7	15.0	2.4	0	0
2011-2016	198	62	31.3	54.8	45.2	88.7	11.3	0	0	0

¹Disagreements are based on differences in person-level Dean’s Fluorosis Index values: 0 = Normal, 0.5 = Questionable (recoded from collected value of 5), 1 = Very mild; 2 = Mild, 3 = Moderate, 4 = Severe. For example, examiner score = questionable and reference score = very mild is reported in this table as differing by 1 level.

 

Overall, for the two 6-year survey periods, there was similarity between the dental examiner and reference examiners in the person-level DFI values (Table 4). In 1999-2004, 48.7% were classified by the dental examiner and 49.8% by the reference examiner with DFI values of very mild or greater. In 2011-2016, the percentages were 74.2% and 73.7%, respectively.

 

Table 4. Percent of survey participants aged 6-19 years with gold standard observations classified at each person-level Dean’s Fluorosis Index value: National Health and Nutrition Examination Surveys, 1999-2004 and 2011-2016

	1999-2004 (n=339)		2011-2016 (n=198)
Dean’s Fluorosis Index level	Dental Examiners (%)	Reference Examiner (%)	Dental Examiners (%)	Reference Examiner (%)
Normal	26.0	30.1	13.1	11.6
Questionable	25.4	20.1	12.6	14.6
Very mild	29.8	26.8	34.9	39.9
Mild	13.0	15.3	23.7	17.7
Moderate	5.0	5.6	13.1	13.1
Severe	0.9	2.1	2.5	3.0

 

Prevalence of Dental Fluorosis Severity Among Youth

Weighted 2- and 6-year estimates of dental fluorosis severity were computed on children 6-19 years (the common age range across all years) who had at least 2 teeth with a non-missing DFI value (1999-2004, n=9395; 2011-2016, n=7158). These analyses used the examination sample weights.

During 1999-2004, 25.3% (SE=1.5) of youth 6–19 years were estimated to have very mild fluorosis, 7.7% (SE=0.5) mild, 3.2% (SE=0.4) moderate, and 0.4% (SE=0.1) severe (Table 5, Figure 1). In 2011–2016, 35.6% (SE=2.5) of youth 6–19 years were estimated to have very mild fluorosis, 21.5% (SE=2.2) mild, 13.4% (SE=1.5) moderate, and 1.0% (SE=0.2) severe (Table 5, Figure 2). In 1999–2004, 3.6% of youth 6–19 years had moderate or severe fluorosis; whereas, in 2011–2016 the percent increased to 14.4%.

Variability in the prevalence of different dental fluorosis severity levels was seen within each 6-year time period, as well as across the two time periods (Table 5, Figures 1 and 2). For example, during 1999-2004, questionable fluorosis ranged from 0.4% to 34.6%. During 2011-2016, mild fluorosis prevalence ranged from 9.1% to 40.4% and moderate fluorosis prevalence ranged from 1.3% to 20.6%.

 

Table 5. Percentage (weighted) dental fluorosis severity levels, based on person-level Dean’s Fluorosis Index, among youth 6-19 years: National Health and Nutrition Examination Surveys 1999-2004 and 2011-2016

Survey Years	n	Normal Percent (SE)	Questionable Percent (SE)	Very Mild Percent (SE)	Mild Percent (SE)	Moderate Percent (SE)	Severe Percent (SE)
1999-2004	9,395	44.3 (2.8)	19.2 (1.6)	25.3 (1.5)	7.7 (0.5)	3.2 (0.4)	0.4 (0.1)
1999-2000	3,103	65.9 (4.6)	0.4 (0.2)	22.3 (3.1)	7.0 (0.9)	3.9 (1.1)	0.5 (0.2)
2001-2002	3,326	48.8 (5.4)	20.6 (3.6)	21.8 (2.1)	6.4 (0.8)	2.1 (0.3)	0.2 (0.1)
2003-2004	2,966	20.0 (2.8)	34.6 (1.8)	31.9 (2.6)	9.5 (1.0)	3.5 (0.7)	0.5 (0.2)
2011-2016	7,158	18.1 (1.6)	10.4 (1.1)	35.6 (2.5)	21.5 (2.2)	13.4 (1.5)	1.0 (0.2)
2011-2012	2,304	34.1 (4.0)	8.8 (0.9)	19.5 (2.1)	14.9 (1.6)	20.6 (3.7)	2.0 (0.5)
2013-2014	2,502	6.1 (1.2)	5.7 (1.3)	28.7 (4.2)	40.4 (5.2)	18.4 (2.1)	0.8 (0.3)
2015-2016	2,352	14.3 (2.9)	16.8 (3)	58.4 (4.9)	9.1 (2.2)	1.3 (0.4)	0.1 (0.1)

SE=standard error

 

Figure 1. Percent (weighted) dental fluorosis severity levels, based on person-level Dean’s Fluorosis Index (DFI), among youth 6-19 years, National Health and Nutrition Examination Surveys 1999-2004

 

Figure 2. Percent (weighted) dental fluorosis severity levels, based on person-level Dean’s Fluorosis Index (DFI), among youth 6-19 years, National Health and Nutrition Examination Surveys 2011-2016

 

Assessing Biological Plausibility of Prevalence Estimates

Further evaluation was conducted on the NHANES 1999-2004 and 2011-2016 dental fluorosis clinical assessment data to determine whether the prevalence estimates presented above could be consistent with the known etiology of fluorosis.

Fluoride ingestion prior to tooth eruption is the only known cause of fluorosis (Fejerskov et al., 1990); therefore, there should be no change in fluorosis prevalence among erupted permanent teeth of the same tooth type in the same birth cohort over time. For this analysis, a synthetic birth cohort that included youth aged 6-9 years in 2001-2004 (born 1992–1998; n=1097) and youth aged 16-19 years in 2011-2014 (born 1992–1998; n=1193) was constructed to determine if the percent with fluorosis, for a given tooth type (specifically first permanent molars, which usually appear between 6-7 years; ADA, 2006) was constant with age, as would be expected given that fluorosis develops before teeth erupt. Logistic regression and computed predictive marginals were used to estimate the prevalence of mild or greater and moderate or severe dental fluorosis in the first permanent molars (identified in the data file as tooth numbers 3, 14, 19, 30) among youth aged 6-9 years in 2001-2004 and 16-19 years in 2011-2014. Adjusted models, which included sex and race and Hispanic origin, were run to control for possible population changes over the time period.

Prevalence (adjusted for age and race and Hispanic origin) of mild or greater fluorosis in the first permanent molars of youth aged 6-9 years in 2001–2004 was 9.5% (standard error (SE) = 0.01) and among youth aged 16-19 years in 2011–2014 was 46.9% (SE = 0.04, p<.001). This increase between surveys was unexpected, because the prevalence of fluorosis in the first permanent molar should not change over time among a similar birth cohort given that tooth eruption has already occurred by 6-9 years of age. Similarly, adjusted prevalence of moderate and severe fluorosis in the first permanent molars of the birth cohort increased from 2.8% (SE = 0.01) in 2001–2004 to 17.7% (SE = 0.3) in 2011–2014 (p<.001). Estimates were similar in unadjusted models.

Impact of Oversampling in NHANES

During 1999–2016, there were changes in the racial and ethnic composition of the U.S. population. Also, in 2011–2016, NHANES started oversampling Asian American persons and continued the oversampling of Hispanic persons, which began in 2007. The oversampling of adolescents was also discontinued in 2007. It may be possible that underlying changes in the population composition of the U.S. or NHANES sample design changes affected the overall prevalence estimates of dental fluorosis and contributed to some of the observed differences in estimates between 1999–2004 and 2011–2016.

Closer examination of the race and Hispanic origin estimates from 2011–2016, however, showed no differences in the prevalence of dental fluorosis for the specific severity categories across the different race and Hispanic origin groups. In 1999–2004, the race and Hispanic origin groups (specifically, non-Hispanic white, non-Hispanic black, and Mexican-American) follow a similar pattern to 2011–2016 (data not shown).

Impact of Geographic Variation in Water Fluoridation Levels 

Determining the water fluoridation level in a given NHANES location is not straightforward given that a county may include multiple water systems. For example, in 2013, there were 345 total public water systems (PWSs) for seven of the NHANES locations visited that year. Nine of the 345 PWSs had levels higher than 1.2 mg/L. These nine were all from a location with a substantial number of PWSs. 

As stated earlier, dental fluorosis typically occurs from fluoride exposure during early tooth development before eruption, typically. Therefore, water fluoride levels many years before the NHANES examination would need to be assessed to determine exposure levels rather than levels for participants at the time of examination. For example, fluoride exposure for a 19-year-old participant examined in 1999 would have occurred during 1980–1986. Determining actual fluoride exposure during childhood, however, is further complicated by the uncertainty of migration patterns among families (i.e., if they ever moved, taking them from one PWS to another).

1999-2004 and 2011-2016 DATA QUALITY ASSESSMENT SUMMARY

There are potential sources of error in the measurement of fluorosis in general, and specifically in sample surveys, such as NHANES. For time-varying factors, these may include intra-subject (within subject) variation and intra- or inter-examiner (within or between examiner) variations in measurement. Within-subject variation could occur if there were changes in the underlying disease process between assessments. However, dental fluorosis is the result of exposure to fluoride from when a child is born up to about 8 years of age. Therefore, intra-subject variability due to physiological or biological reasons is unlikely. Intra- or inter-examiner error, however, may occur due to lack of adherence to the protocol by examiners and changes in quality control procedures over time. During 1999–2004 and 2011–2016, there were only minor changes to the dental fluorosis assessment protocol (no dental explorer used for assessment in 2011–2016) and the QA procedures (repeat examinations by the same examiner only occurred in 1999–2001). There was no change to the measurement scale (DFI). This index is, however, known for its subjective nature, which leads to potential variability, especially at the low end of the index where distinguishing between very mild (i.e., less than 25% of the tooth) and mild (25% to less than 50%) can be difficult. Evaluation of gold standard examinations found that agreement (across 4 different weighting schemes) ranged from 0.51 to 0.98 for the 9 primary dental examiners in 1999-2016. These values indicate moderate (0.41-0.60) to almost perfect (0.81-0.99) agreement based on Landis and Koch, and mostly adequate agreement (33 out 36 kappas >=0.6) based on McHugh. It is possible, however, that the quantity of repeats as a percentage of the total examined persons (2.8%–3.6%) was inadequate for making conclusions on the reliability of a tool with such subjective variability. Additionally, the assessment of intra-examiner reliability from 1999–2001, as published by Dye et al. (2007), did not show perfect agreement in scoring of the same person only a few days later (weighted kappa statistics comparing person-level scores for the same participant by the same examiner ranged from 0.56 to 0.72). This demonstrates the substantial subjectivity and variability of this scoring method. There was no assessment of intra-examiner reliability in 2011–2016, and procedures to ensure calibration of the reference examiner over time are not available. Therefore, the possibility of a shift in how the examiners assessed dental fluorosis over time cannot be ruled out.

The impact of the complex sampling scheme of NHANES and geographic variability in the presence of fluoride in drinking water on the results were also evaluated. While there were significant sample design changes between 1999–2004 and 2011–2016, and specifically with the oversampled groups, when proper weighting procedures are used, the final estimates produced should still be reflective of the civilian noninstitutionalized U.S. population of youth aged 6–19 years, and estimates from the different survey periods should be comparable. There were no changes to the inclusion and exclusion criteria for the examination for youth aged 6–19 years specifically and no differences in the percentage of the sample who completed the fluorosis assessment.

Variability in the prevalence of different dental fluorosis severity levels was seen within each 6-year time period, as well as across the two time periods. For example, during 1999-2004, questionable fluorosis ranged from 0.4% to 34.6% and during 2011–2016, mild fluorosis prevalence ranged from 9.1% to 40.4%. The reasons for this variability cannot be determined but likely include random error, true change in prevalence, or changes in the application of measurement processes. There may be other factors outside of these that could also contribute to the variability in 2-year prevalence estimates. 

Further analyses of the synthetic birth cohort born in 1992-1998, revealed that the prevalence of mild or greater fluorosis in the first permanent molars was 9.5% in youth aged 6–9 years in 2001–2004, but was 46.9% in youth aged 16–19 years in 2011–2014. This increase does not seem biologically plausible since fluorosis develops before teeth erupt and therefore prevalence in erupted teeth for the same birth cohort should not change over time. The analysis of the synthetic cohort with adjustment for possible changes in the demographic characteristics of the population over time, suggests that the observed increase in fluorosis prevalence does not reflect the genuine amount of change within the U.S. population.

For general information on analyzing NHANES data and the use of examination sample weights refer to the NHANES Analytic Guidelines and the on-line NHANES Tutorial.

References

• American Dental Association. For the dental patient. Tooth eruption: The permanent teeth. J Am Dent Assoc 137(1):127. 2006.
• Beltran-Aguilar ED, Barker L, Dye BA. Prevalence and severity of dental fluorosis in the United States, 1999–2004. NCHS Data Brief, no 53. Hyattsville, MD: National Center for Health Statistics. 2010.
• Cicchetti DV, Allison T. A new procedure for assessing reliability of scoring EEG sleep recordings. American J EEG Technol 11(3):101–10. 1971.
• Clark MB, Slayton RL. Fluoride Use in Caries Prevention in the Primary Care Setting. Pediatrics 2014;134(3):626–33.
• Cohen JA. Coefficient of Agreement for Nominal Scales. Educational and Psychological Measurement, 20(1): 37–46. 1960.
• Dean HT. Classification of mottled enamel diagnosis. J Am Dent Assoc 21(8):1421–6. 1934.
• Dean HT. The investigation of psychological effects by the epidemiological method. Fluorine and Dental Health 19:23–33. 1942.
• Dye BA, Barker LK, Selwitz RH, Lewis BG, Wu T, Fryar CD, et al. Overview and quality assurance for the National Health and Nutrition Examination Survey (NHANES) oral health component, 1999–2002. Community Dent Oral Epidemiol 35(2):140–51. 2007.
• Fejerskov O, Manji F, Baelum V. The nature and mechanisms of dental fluorosis in man. J Dent Res 69 Spec No:692–700. 1990.
• Fleiss JL, Cohen J. The equivalence of weighted kappa and the intraclass correlation coefficient as measures of reliability. Educ Psychol Meas 33:613–9. 1973.
• Fluoride Recommendations Work Group. Recommendations for using fluoride to prevent and control dental caries in the United States. MMWR Recomm Rep 50(RR–14):1–42. 2001.
• Kaminsky LS, Mahoney MC, Leach J, Melius J, Miller MJ. Fluoride: Benefits and risks of exposure. Crit Rev Oral Biol Med 1(4):261–81. 1990.
• Kumar JV, Swango PA, Opima PN, Green EL. Dean's fluorosis index: An assessment of examiner reliability. J Public Health Dent 60(1):57–9. 2000.
• Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 33(1):159–74. 1977.
• McHugh ML. Interrater reliability: The kappa statistic. Biochem Med (Zagreb) 22(3):276–82. 2012.
• National Center for Health Statistics, National Center for Chronic Disease Prevention and Health Promotion. Data quality evaluation of the dental fluorosis clinical assessment data from the National Health and Nutrition Examination Survey: 1999-2004 and 2011-2016. National Center for Health Statistics. Vital Health Stat 2(183). 2019. https://www.cdc.gov/nchs/data/series/sr_02/sr02_183-508.pdf.
• National Research Council. Health effects of ingested fluoride. Washington, DC: National Academies Press. 2006.
• U.S. Department of Health and Human Services Federal Panel on Community Water Fluoridation. U.S. Public Health Service recommendation for fluoride concentration in drinking water for the prevention of dental caries. Public Health Rep 130(4):318–31. 2015.

Codebook and Frequencies