Adaptation and validation of the PEDSQL™ oral health scale for toddlers in Chilean population

The PedsQL™ Oral Health Scale is a questionnaire containing five items with five response alternatives: 0 = never a problem, 1 = almost never a problem, 2 = sometimes a problem, 3 = often a problem, 4 = almost always a problem. The score for each item is reversed and transformed to a linear scale of 0–100 (0 = 100; 1 = 75; 2 = 50; 3 = 25; 4 = 0). The scale scores are calculated as the sum of the items divided by the number of items answered, such that a higher score indicates better OHRQoL. The version for toddlers is completed by parents or guardians [11].

The cross-cultural adaptation of the PedsQL™ Oral Health Scale into Chilean Spanish followed international standards for the development of new linguistic versions equivalent to the original instrument [16].

The English version of the PedsQL™ Oral Health Scale was translated into Spanish independently by two professional translators whose native language was Spanish. The translators were asked to maintain the conceptual equivalence of the original version, rather than a literal translation (to obtain natural expressions which expressed the same concept), and to score the difficulty in finding the conceptual equivalence for each item from 1 (minimum difficulty) to 10 (maximum difficulty).

The two Spanish translations of the PedsQL™ Oral Health Scale were compared by a panel consisting of: two experts in OHRQoL assessment, two pediatric dentists, and the two translators. The discrepancies were discussed until a first unified version was obtained. This unified Spanish translation was reviewed by a group of parents of pre-school children (3 fathers and 4 mothers), in order to check its applicability and comprehensibility. This preliminary version was translated back into English, separately, by two native English speakers who also evaluated the difficulty in finding equivalent expressions. Finally, the equivalence between the original version and these back-translations was evaluated by the expert panel who rated the items as: A = conceptually and linguistically equivalent to the original item; B = functionally equivalent, but with grammatical differences; or C = equivalence is not obvious. Expressions evaluated with category C were reviewed by the expert panel to improve the Spanish wording to achieve equivalence. The report on equivalence between the original and back-translated versions was sent to the author of the original PedsQL™ Oral Health Scale for evaluation.

A cognitive debriefing interview was applied to 15 parents (2 fathers and 13 mothers, aged between 24 and 37 years) of children aged between 2 and 5 years, to assess the clarity and comprehensibility of the Spanish version. The parents first self-completed the whole questionnaire and afterwards, to assess what the parents had understood, they were asked open questions about their responses. A set of questions was used during the interview in order to obtain standardized information.

To assess the metric properties of the Chilean version of the PedsQL™ Oral Health Scale, a cross-sectional study was carried out in public preschools in the district of Carahue (Southern Chile). Eleven preschools funded by the Chilean government participated in the study, which was carried out between April and October 2016.

Pre-school children aged 2 to 5 years were included, together with their parents. The inclusion criteria were: children without any systemic diseases, long-term medication or special health needs, and that both children and parents agreed to take part in the study. Authorization was obtained through a written informed consent in the case of the parents and the verbal consent of the children. The parents were invited to a meeting in the preschool, during which the participating children had their mouths examined and parents were asked to complete three questionnaires regarding their child: the PedsQL™ 4.0 Generic Core Scale, the PedsQL™ Oral Health Scale and the Early Childhood Oral Scale (ECOHIS). The parents also completed a structured questionnaire to collect information on the child’s age, gender, socio-economic status and history of oral hygiene habits, as well as their general and dental health status. These questionnaires were sent by mail to the parents who did not attend the meeting.

Four previously trained and standardized researchers carried out the dental examinations in rooms specially equipped and adapted for the study, under artificial light, involving visual inspection of the oral cavity after prophylactic brushing with a toothbrush. The examiners were blinded to the questionnaire answers. The caries diagnosis was based on the criteria proposed by the World Health Organization (WHO) for Oral Health Surveys [17]. The types of traumatic dental injuries were classified according to Andreasen and Andreasen [18], and malocclusion was assessed by the presence or absence of at least one of the following alterations: anterior open bite, overjet> 4 mm and anterior cross-over bite [19]. The researcher standardization process consisted first of a theoretical calibration on the study protocol, diagnostic criteria, dental examination system, and filling in a clinical record. Practical calibration was then carried out on 15 children selected at random from a school located in the same district, but not included in the study. The inter-examiner agreement was high with kappa coefficients of 0.83 and 0.70 for caries and malocclusion traits. A series of 20 images was used to evaluate the reliability of traumatic dental injuries (kappa = 0.79). The global intra-examiner agreement was kappa = 0.81.

The socio-economic status of the family was estimated through the healthcare provision level determined by the state health insurance. Families were classified in the low socio-economic status if they had no resources and a taxable monthly income lower than or equal to USD 367.73.

The sample size was estimated following the recommended standard of 2 to 20 participants per item with a minimum of 100 to 250 subjects [20, 21]. Considering this last number of participants recommended, and assuming a 20% of potential missing answers, the sample size required was of 300 children.

The descriptive analysis of the socio-demographic and clinical characteristics of the sample was carried out calculating frequencies and percentages. The feasibility and acceptability of the PedsQL™ Oral Health Scale was examined by calculating the percentage of parents who did not respond to some items. If more than 50% of the items on the scale were missing, the scale score was not calculated. When 50% or fewer items were missing, we imputed them by the mean of the completed items [11].

Distribution of the PedsQL™ Oral Health Scale scores was assessed by obtaining the range, mean, standard deviation, and floor and ceiling effects (percentage of patients with minimum and maximum theoretical scores, respectively). Small floor or ceiling effects (< 15%) are considered acceptable [21].

Two approximations were applied to estimate reliability: a) internal consistency assessed by Cronbach’s alpha coefficient [22]; and b) test-retest reproducibility evaluated by the intraclass correlation coefficient (ICC) [23]. Cronbach’s alpha coefficient ranges from 0 to 1 with values > 0.70 being considered acceptable [21]. An ICC of < 0.40 indicates poor to fair agreement, 0.41–0.60 moderate agreement, 0.61–0.80 good agreement and > 0.80 excellent agreement [24]. The PedsQL™ Oral Health Scale was re-administered between 2 and 4 weeks after the first administration in a sub-sample of 50% of the participants in each preschool, which were selected at random. Test-retest analysis was performed with this sub-sample, after excluding participants who reported change in oral health.

Confirmatory factorial analysis (CFA) was carried out to confirm the single dimension of the Chilean version proposed by developers of the PedsQL™ Oral Health Scale. For non-standardized solutions, the pattern of fixed and free factorial loads remained constant. Various fit indices were calculated to evaluate the model, including the Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI) and the Root Mean Square Error Approach (RMSEA). CFI and TLI values greater than or equal to 0.95 indicate an excellent fit of the model, while values between 0.90 and 0.95 suggest only an acceptable fit of the model. RMSEA values below or equal to 0.06 indicate a good fit of the model, while values between 0.06 and 0.08 suggest only acceptable fit of the model [25].

Construct validity was assessed by following a known groups approach, comparing groups with presence or absence of oral conditions (caries, traumatic dental injuries and/or malocclusion) and groups defined by responses to the following questions: “In general, how would you rate the general health of your child?”, and “In general, how would you rate the dental health of your child?”. The possible answers to these two questions were: 1 = Excellent; 2 = Very good; 3 = Good; 4 = Fair; 5 = Poor. We hypothesized lower scores (poorer OHRQoL) in the PedsQL™ Oral Health Scale for the children whose parents reported regular or poor oral health; also, lower scores in the PedsQL™ Oral Health Scale among the groups of children with diagnosis of dental health problems in the oral examination. Comparisons between groups were performed with non-parametric Mann Whitney tests due to the score distributions of the PedsQL™ scales.

The convergent validity was assessed by examining the correlation between the PedsQL™ Oral Health Scale and ECOHIS, using Spearman’s correlation coefficients interpreted as follows: insignificant relation when r < 0.20; weak between 0.20 and 0.40; moderate between 0.40 and 0.60; moderate-to-strong between 0.60 and 0.80; and strong relation when r > 0.80 [26]. The convergent validity implies showing that different instruments measuring a similar concept present at least moderate correlations. We hypothesized that the correlation coefficients between ECOHIS and PedsQL™ Oral Health would be moderate-to-strong, since both were designed to measure OHRQoL. The data were analyzed using Stata 15 (Stata Corp, College Station, TX, USA).

The translation of the 5 items was considered of low level difficulty (< 2.5). In the back-translation, the mean difficulty was 6.5 for item 3, and less than or equal to 3.5 for the rest. When the conceptual equivalence was compared between the back-translation and the original questionnaire, they were considered conceptually and linguistically equivalent in four of the five items. Only for item 3, “Having teeth that are dark in color”, the conceptual equivalence was rated B (conceptually equivalent, with grammatical differences): thus, after discussion by the group of researchers and translators, the expression in Spanish was modified to “Having some dark-colored teeth”. The cognitive interview showed that the instructions, items and response options were easily understood by the parents, and no problems were identified in distinguishing between the items or the different response options. All the parents agreed that the questions were intended to assess OHRQoL, and no modifications were needed as a result of the cognitive interviews.

In total, 435 children aged between 2 and 5 years were contacted for the study, two of whom were excluded because they had special health needs; 12 parents declined to sign the informed consent form and 93 children did not attend the clinical examination. Of the 328 children examined, 26 parents did not reply to the questionnaire (response rate = 92.1%), and one questionnaire contained answers to fewer than 50% of the items and was therefore excluded from the analysis. Finally, a total of 301 families (children and parents) took part in the study. The descriptive information details are shown in Table 1. The mean age of the children included was 47.8 months (4 years) (SD = 13.3), 54% were boys (n = 162), and 75.7% presented low socio-economic status. Of the parents, 83.3% reported that their children had good or very good general health and 70.4% reported good or very good oral health. Seventeen percent declared that their children brushed their teeth once a day or less often, and 83.3% that the children brushed their teeth twice a day or more. The prevalence of caries experienced (dmft≥1) in the population examined was 53.8%, with a mean dmft of 2.52 (SD 3.71); the prevalence of malocclusions was 39.4%, and the traumatic dental injuries were 14.6%.

All items of the PedsQL™ Oral Health Scale were completed by the participants, while 4.6% of them left some item without completion in the PedsQL™ 4.0 Generic Core Scale. The school functioning sub-scale reported the highest percentage of missing replies by the respondents (2.3%). The most reported problem was “having some dark-colored teeth” (32.6% almost never, sometimes, often or almost always) in the PedsQL™ Oral Health Scale; and for the PedsQL™ 4.0 Generic Core Scale, “[ …] has feeling angry been a problem for your child” (80.9% almost never, sometimes, often or almost always).

Table 2 shows distribution statistics and Cronbach’s alpha coefficient of the PedsQL™ scales. The mean scores in the PedsQL™ 4.0 Generic Core Scale and Oral Health Scale were 82.2 (SD = 12.2) and 89.1 (SD = 16.1) respectively. No floor effect was observed, but a substantial ceiling effect was observed in the PedsQL™ Oral Health Scale (46.5%). Cronbach’s alpha was 0.87 for the PedsQL™ 4.0 Generic Core Scale and 0.79 for the PedsQL™ Oral Health Scale. The ICC was 0.86 (95% CI = 0.78–0.91) for the PedsQL™ Oral Health Scale among the subsample of 84 parents who participated in the test-retest reliability study.

Figure 1 shows results of the confirmatory factor analysis (CFA) in which the five items of the PedsQL™ Oral Health Scale loaded on a single latent variable. However, the CFA model did not fit the data adequately: CFI = 0.927, TLI = 0.855 and RMSEA = 0.160. To improve the overall model fit, a Lagrange multiplier test was applied and covariances were incorporated between: item 1 (Having tooth pain) and 2 (Having tooth pain when eating or drinking something hot, cold, or sweet); and item 4 (Having gum pain) and 5 (Having blood on his or her toothbrush after brushing). The new model showed a better goodness of fit indices: CFI = 0.996, TLI = 0.987 and RMSEA = 0.048.

Table 3 shows the results for the construct validity based on known groups. The children whose parents reported poorer general and oral health obtained significantly lower scores in the PedsQL™ Oral Health Scale, meaning a lower quality of life (p < 0.05). Finally, although the PedsQL™ Oral Health Scale score showed significant differences between children with and without dental caries (p < 0.0001), differences were no statistically significant between those with and without malocclusion (p < 0.142) or traumatic dental injuries (p < 0.126). An inverse moderate-to-strong (r = − 0.64) correlation between the PedsQL™ Oral Health Scale and ECOHIS was found, while the correlation with the generic core scale was moderate (r = 0.25).