Validation of maternal report of early childhood caries status in Ile-Ife, Nigeria

Background: To determine the validity of maternal reports of the presence of early childhood caries (ECC), and to identify maternal variables that increase the accuracy of the reports. Methods: This secondary data analysis included 1155 mother–child dyads, recruited through a multi-stage sampling household approach in Ile-Ife Nigeria. Survey data included maternal characteristics (age, monthly income, decision-making ability) and maternal perception about whether or not her child (age 6 months to 5 years old) had ECC. Presence of ECC was clinically determined using the dmft index. Maternally reported and clinically determined ECC presence were compared using a chi-squared test. McNemar’s test was used to assess the similarity of maternal and clinical reports of ECC. Sensitivity, specificity, positive and negative predictive values, absolute bias, relative bias and inflation factor were calculated. Statistical significance was determined at p < 0.05. Results: The clinically-determined ECC prevalence was 4.6% (95% Confidence interval [CI]: 3.5–5.0) while the maternal-reported ECC prevalence was 3.4% (CI 2.4–4.6). Maternal reports underestimated the prevalence of ECC by 26.1% in comparison to the clinical evaluation. The results indicate low sensitivity (9.43%; CI 3.13–20.70) but high specificity (96.9%; CI 95.7–97.9). The positive predictive value was 12.8% (CI 4.3–27.4) while the negative predictive value was 95.7% (CI 94.3–96.8). The inflation factor for maternally reported ECC was 1.4. Sensitivity (50.0%; CI 6.8–93.2) and positive predictive value were highest (33.3%; CI 4.3–77.7) when the child had a history of visiting the dental clinic. Conclusions: Mothers under-reported the presence of ECC in their children in this study population. The low sensitivity and positive predictive values of maternal report of ECC indicates that maternal reporting of presence of ECC may not be used as a valid tool to measure ECC in public health surveys. The high specificity and negative predictive values indicate that their report is a good measure of the absence of ECC in the study population. Child’s history of dental service utilization may be a proxy measure of presence of ECC. This was a secondary analysis of a dataset of 1549 maternal-child dyads primarily collected to determine the prevalence of ECC and examine maternal psychosocial factors that were risk indicators for ECC. The data were collected from residents in Ife Central local government area, one of the 774 local government areas in Nigeria. The sample size required for this secondary analysis was determined based on an assumption that 50% of mothers with autonomous decision-making ability were able to accurately report the presence of ECC in their children in the absence of any accessible data. We estimated, based on a margin of error of 5% and 95% confidence level, that the required sample size for this study was 768. We had access to data from 1155 mother–child dyads who reported the presence of ECC in their children. A multi-stage sampling technique was used to collect the primary data. First, 70 of the 700 enumeration areas in the local government area was selected by simple random method (balloting), followed by selection of every other household on each street in the enumeration areas. The sample of the enumeration areas was limited to 10% in line with prior suggestions that a 10% sample was an adequate sample size for household surveys [28]. The final stage was the selection of an eligible respondent in each household for interview and clinical examination. Only one child and mother dyad in each household was eligible for participation in the study. A child was eligible for study participation if (s)he was below the age of 6 years, living with a caregiver, present at the time of the survey and for whom parental consent for study participation was obtained. Children with chronic medical conditions that required prolonged use of sweetened medications and those with medical conditions that increased their risk for ECC were excluded from the primary study. Trained and experienced field workers collected the data electronically using an interviewer-administered structured questionnaire on Open Data Kit—an online/offline platform for the collection and management of data. The questionnaire was administered to the mothers. Folayan et al. [29, 30] provided detailed descriptions of the sampling process and study procedure. Socio-demographics: Information on the socio-demographic profile of the mother (age, income, educational status) were extracted from the primary study dataset. The mother’s age at her last birthday was categorized into three groups: ≤ 29 years, 30–39 years, 40 years and over. Mother’s income was defined as monthly salary for persons in paid employment, and an estimate of monthly income for self-employed persons. Income was categorized using the national Nigerian currency and wage into three categories: ≤ N18,000 ($49)/month, N18,001–N60,000/month, and > N60,000 (168)/month [31]. Mothers’ educational status was defined as no formal education, primary school only, secondary school only, or tertiary (post-secondary) education. Data about women’s participation in making decisions concerning (1) their own health care, (2) major household purchases, and (3) visits to family or relatives without having to get permission were extracted from the primary study dataset. The questions exploring women’s decision-making ability were adopted from the Nigeria Demographic and Health Survey [32]. When others make any of these decisions on behalf of the mother, the mother was regarded as having no decision-making ability for the item scored. Data on children’s dental visit history were extracted from the primary study dataset. Mothers were asked if their children had ever visited the dentist (yes or no), and if the child had a hole in their teeth (yes or no). The prevalence of ECC was determined as the proportion of children reported by their mothers to have caries. Data on the early childhood caries profile of the 1155 children generated in the primary study by five calibrated dentists who conducted the oral examination for each child were extracted for this study. Calibration of the five dentists was conducted by first training them on caries assessment using a colored picture chart with varying presentations of decayed, missing and filled teeth, followed by examining a group of five children with caries and making a diagnosis using the World Health Organization scoring criteria. The scoring for each of the five children was repeated three times with an interval of one week between each visit. Intra-examiner agreement for each of the dentists was calculated using Cohen’s Kappa and the inter-examiner agreement (between the dentists and the trainer) was calculated using the Cohen’s kappa coefficient. The intra- and inter-examiner reliability tests were all greater than 0.80. ECC was determined in the primary study using the decayed-missing-filled teeth (dmft) index as recommended by the World Health Organization [33]. Radiographic assessment was not conducted. The dmft score was an aggregated score of the d, m and f scores for each child. ECC was considered present when the dmft score was > 0 and absent when the dmft was 0. The study had access to the aggregated dmft score for each child and not the respective d, m and f scores. The final analytic sample included only children who had maternally reported and clinically determined presence of ECC (N = 1155). Descriptive analyses were performed, including calculation of mean values and 95% confidence intervals (CI) for maternal reported and clinically determined presence of ECC. The bivariate association was tested between clinical and maternal reported ECC, and selected maternal characteristics separately using a chi-squared test. McNemar’s test was used to assess differences between paired data (i.e. clinical versus maternal reporting of ECC). In addition, sensitivity, specificity, positive and negative predictive values, absolute bias, relative bias and inflation factor (gold standard prevalence/self-reported prevalence) were also calculated [34]. Estimates of sensitivity, specificity, and positive and negative predictive values were stratified by the socio-demographic profile of the mother (age, income, educational status) and maternal decision-making status. Statistical analyses were conducted with Intercooled STATA (release 15) for windows. Statistical significance was inferred at p ≤ 0.05. Ethical approval for the study was obtained from the Obafemi Awolowo University Teaching Hospitals Complex Health Research Ethics Committee (NHREC/27/01/2009a and IRB/EC/0004553). Study participants for the primary study were recruited after receiving written consent from the mothers for their own study participation, and written consent for their child’s participation in the study.