Parental–Caregivers Perceptions Questionnaire (P-CPQ): translation and evaluation of psychometric properties of the French version of the questionnaire

The P-CPQ began with two questions asking parents for a global rating of their children’s oral health: “How do you evaluate the health of your child’s teeth, lips, jaws and mouth?” (Choice of Responses: Excellent, Very good, Good, Fair, Poor) and “To what extent is your child’s well-being affected by the state of their teeth, lips, mouth and jaws?” (Possible responses: Not at all, Very little, Some, A lot, Very much). The P-CPQ is composed of 31 items divided into four sub-scales including: Oral Symptoms (OS) (6 items), Functional Limitations (FL) (7 items), Emotional Well-Being (EWB) (8 items) and 10 items for Social Well-Being (SWB). The questions referred to the frequency of events occurring during the previous 3 months. A-five point Likert-like scale was used with the following options of response: “Never” (score 0), “Once or twice” (1), “Sometimes” (2), “Often” (3) and “Nearly every day” (4).

This cross-sectional study was performed in the framework of the Odontology Centre of the Hospital Rothschild in Paris (Assistance publique Hôpitaux de Paris), France.

The psychometric properties were evaluated on a sample of 142 parents/caregivers of children between 8 to 10 years old. The study focused on parents accompanying their children. Inclusion criteria were: participants must be able to read, write and self-administer the questionnaire. Parents or caregivers coming from different locations in Paris and even from farer regions were approached with a consent form and the questionnaire through their children who underwent clinical examinations at the clinical centre. Children were divided into three groups before the beginning of the data collection according to their oral health conditions. The groups were:

Parents were informed on the objectives of the study and had given their written consent to answering the questionnaire.

A convenience sample was selected for this cross-sectional study. Sample comprised parents/caregivers accompanying their children at the Odontology Centre of the Hospital Rothschild in Paris. To determine the minimum sample size required we referred to the Everitt et al.’s recommendation for a sample size ranging from 100 to 200 [20] and to a “ 4 subjects to 1 item ratio”. One hundred forty two parents/caregivers of children aged 8 (n = 52), 9 (n = 33) and 10 years old (n = 57) agreed to participate in the survey.

Two interviewers approached the parents by way of an informative letter, a consent form that they were asked to sign and the French P-CPQ questionnaire. Parents were invited by the two interviewers to self-complete the questionnaire in the waiting room of the odontology centre while their children underwent clinical examination.

The study was approved by the Research Project Ethics Evaluation Committee of the Hospital Robert Debré (AP-HP), under approval N° 2016/250–2.

The original version of the P-CPQ was obtained using an article available with open access, shared by the terms of “Creative Commons License” which allowed its use without restriction, distribution and reproduction, with the necessary condition of citing the authors’ works.

The steps of the translation and the evaluation process of psychometric properties of the French-P-CPQ are presented in a flow chart. (Fig. 1) The sociocultural and conceptual adaptation of the P-CPQ was carried out according to the specific recommendations for the self-administered measures [19, 21].

First, two independent translators translated the P-CPQ questionnaire from English to French. A committee met to decide on a single translation, which was then back translated into English by two other translators including a dentist and a linguist with perfect knowledge of English. Both were totally unaware of the existence of the original version. The committee of experts composed of all the translators and the researchers compared the two back-translations with the original in English, and then revised and agreed on the penultimate French version.

The penultimate version was subjected to a pre-test on a group of fourteen parents/caregivers of children 8 to 10 years old who did not belong to the final sample, in order to discuss the pertinence of items and to identify problems of comprehensibility. The main purpose of the pre-test was to detect misunderstandings, ambiguities, or other problems participants may encounter with the questions and response options. The respondents were briefed about the purpose of the pre-test. They were invited to highlight questions they think difficult to understand. They were also asked to give comments on the clarty of the questionnaire. This led to the revision of the questionnaire by the committee and the final version of the French-P-CPQ was obtained. (Additional file 1).

The P-CPQ scores for each participant was calculated by summing the valid responses scores to all items, whereas the domain’ scores were obtained by summing the items scores within each of the four domains. As there were 31 items, the final score could vary from 0 (no impact on the children’s Oral Health Quality of Life) to 124 (maximum impact). The highest scores indicated the highest levels of negative impact of oral conditions on the child’s quality of life perceived by the parents. “I don’t know” was permitted as response to an item and was given a score of 0(zero) as a “never” option. The choice of zero was based on data from Jokovic et al. [9].

The SPSS 24.0 software for Windows was used to process and analyse the data. Initially, demographic data was presented in numbers and percentages. Descriptive analysis were also performed to assess the frequency distribution of oral impacts and measures of central tendency (means and standard deviation) of the total scale and subscales scores of the French-P-CPQ. Ceiling or floor effects was evaluated on the basis of the percentage of respondents with the maximum or minimum P-CPQ score and was considered present if this was in 15% or more of the respondents [22]. The second step of the analysis included tests to confirm the reliability and validity of the French version of the P-CPQ. Reliability was assessed using two methods: test-retest reliability and internal consistency reliability. The reliability of the French version of the P-CPQ was tested on the parents (n = 85) who, two weeks later, filled the questionnaire once again for the retest. Test-retest reliability was evaluated using Intra-Class Correlation (ICC) calculated by two-way analysis of variance [23]. Reliability was considered acceptable if the ICC was higher than 0.7 and if the 95% confident interval of an ICC estimate is 0.83–0.94, the level of reliability can be regarded as “good” to “excellent” [24]. Internal consistency was evaluated using Cronbach’s alpha with values α > 0.7 considered acceptable [25].

To analyse construct validity, the associations between French-P-CPQ scores and the two global indicators were determined using Spearman’s rho correlation coefficient. The two questions presented separately at the beginning of the text were considered for the construct validity and Spearman rho correlation was used for this purpose. Therefore, associations were analysed between total scale score and subscales scores with the global ratings of oral health and overall wellbeing. The hypothesis tested was that parents reporting negative impact of children’s oral health on their every day life and poor oral health status have higher P-CPQ scores than parents reporting positive impact or good oral health. Factor analysis was used to evaluate the structure of the French-P-CPQ version to confirm whether the questionnaire contains appropriate groups of constructs and items. In the present study, the Exploratory Factor Analysis (EFA) with the extraction of factors by component principal analysis (CPA) and varimax rotation with Kaiser normalisation was carried out to evaluate the factor structure of the French-P-CPQ scale. Before performing factor analysis, Kaiser-Meyer-Olkin (KMO) criteria factors and Bartlett’s sphericity test were assessed to evaluate the justifiability of performing factor analysis on the French-P-CPQ. The KMO value of ≥0.80 indicated sample size accuracy and justifiability of factor analysis. The eigenvalue> 1 and scree test, were used to determine how many factors to retain. Kaiser’s criterion suggests retaining all factors that are above the eigenvalue of 1. A KMO > 0.6 and variance explained by loaded factors> 60% were considered. A variable loads strongly if its loading is ≥.40 in a particular factor and is considering clean if its absolute difference between loadings exceeds 0.20 [26].

EFA was followed by a partial confirmatory factor analysis (PCFA) to verify the goodness of fit of the model that was built according to the results of EFA. PCFA was performed to validate the factorial validity of the model derived from the results of the EFA. For this, we have performed PCFA with Maximum Likelihood Estimation method, and by calculating the implied model chi-square (Goodness-of-fit test) and the degree of freedom associated with the residual correlation matrix. The residual correlation resulted was used in SPSS to estimate the value of the Standardized Roots Mean Residual (SRMR) value that would get from a CFA [27]. Thus, with SPSS, to calculate an incremental close fit index, two chi-square values were required (the chi-square obtained by Barlett’s test of sphericity and the implied chi-square) with the respective degree of freedom. Model fit verification was done using normed chi square test and several different indices of fit including the Normed Fit Index (NFI), Comparative Fit Index (CFI), Tucker Lewis Index (TLI) and Root Mean Squared Error for Approximation (RMSEA). A normed chi² < 3 is considered a good model. An acceptable scores of CFI, NFI was >.90 respectively; TLI = .90 and RMSEA value of 0.06 and lower was indicative of a good model fit [28]. A cut-off value of .05 was used for SRMR and RMSEA. Cronbach’s α was also determined for reliability of the extracted 7-factors.

The expression “sores in mouth” in English, which means “plaie dans la bouche” in French, seemed ambiguous and was finally translated into “douleurs dans la bouche” (pain in mouth). Since chewing “corn on the cob” is not suitable for the nutritional habits of children in France, it was excluded. The pre-test results showed that the French-P-CPQ was well understood by the respondents.

The mean (standard deviation) age of children was 9.04(0.88), of which 74(52.1%) were boys and 68(47.9%) girls. A total of 142 parents of children aged from 8 to10 years old answered the P-CPQ questionnaire among whom 84 (59.1%), 23 (16.2%) and 35 (24.6%) were respectively parents of healthy children (with no oral-facial malformation) = health group, parents of children with oral-facial clefts = cleft group and parents of children with anomalies linked to a rare disease other than clefts = other group. 105(73.3%) of the respondents were mothers, 26(18.3%) fathers and 11(7.7%) were others caregivers. (Table 1).

The number of “I don’t know” (DK) responses per parent was very limited. Only three parents had ticked DK as responses to one item of emotional wellbeing or social wellbeing. Overall, 94.4% of the parents reported oral impact. In the study population, the scores for the total scale from 0 to 81 with a mean (standard deviation) of 18.73(18.79), 90.8% of them reported experiencing Oral Symptoms in the previous three months, 75.4% reported Functional Limitations, 72.5% reported Emotional Well-Being, and 64.1% reported Social Well-Being. The frequency of a total score of zero was 5.6% not exceeding 15% (no floor effect). No respondent achieved the maximum possible score (no ceiling effect) of the P-CPQ. Thus, no floor or ceiling effects in total scores. In subscales, FL, EWB and SWB, > 15% of respondents scored “never” (Table 2). The proportion of individuals who achieve the lowest numeric value of a score was found in the “healthy” group.

Cronbach’s alpha of the total scale score was 0.85, indicating excellent internal consistency and for the subscales, values ranged from 0.77 for Social wellbeing to 0.81 for Oral Symptoms. The reliability of the French-P-CPQ according to the test-retest measured by the intra-class correlation coefficient (ICC) provided a global value of 0.88, CI 95% [0.82–0.92]. ICC values varies from 0.76 (Functional Limitations) to 0.86 (Social Wellbeing) (Table 3).

The correlations between global ratings (Oral health and overall wellbeing) and the total score (r = 0.54 and r = 0.46), the subscales: oral symptoms (r = 0.41 and r = 0.38), functional limitations (r = 0.38 and r = 0.34), emotional wellbeing (r = 43 and r = 43), social wellbeing (r = 45 and r = 41) were weak but highly significant (p < 0.001) (Table 4).

We carried out an exploratory factor analysis EFA on the answers of the respondents to the 31 items. Before EFA, the Kaiser-Meyer-Olkin (KMO) test and Bartlett’s test of sphericity were conducted to evaluate the factorability and ensure that EFA was adequate for principal component analysis (PCA). The Kaiser-Meyer-Olkin test indicated good adjustment to latent factors (KMO = 0.885) and Barlett’s test of sphericity was significant at p < 0.001 meaning that EFA can be applied to the dataset. An exploratory factor analysis was conducted to determine the factor structure of the French-P-CPQ and to explore item placement within factors. Varimax rotation was used. This procedure led to a different grouping from the initial framework and the 31 items of the French-P-CPQ were loaded into 7 different factors with Eigenvalues> 1 and the total variance explained was 65.23%. The first factor taken alone explains 36.33% of the total variance of the 31 items in the analysis which included: 1-First factor: consists of the totality of the original emotional wellbeing subscale (7/7 items) [14‐20], and social wellbeing subscale (5/10 items)(items: 2, 25, 27, 28, 29) related to “relationship”; 2-Second Factor. “Eating difficulties”: represents 4/8 items of functional limitations domain (items 7, 11, 13, 14); 3-Third factor: “oral dental/pain” (items 1, 3, 4, 12); 4-Forth factor: “Oral Symptom and discomfort ”(items 5, 9, 24, 26); 5-Fith factor: “functional limitation” (items 8, 10); 6- Sixth factor: “discomfort”: (items 22, 30, 31) and 7-Seventh factor: “oral hygiene” (items 2, 6).. (Table 5) The maximum likelihood estimation analysis values associated with the PCFA solution was equal to 391.315 with 269 degree of freedom (df) and p < 0.001 which is smaller than the corresponding null model chi square Bartlett’s test of sphericity of 2443.025 and p < 0.001. Using the Kaiser criterion of retaining all components with eigenvalues above 1, 12 items loaded strongly (0.4+) on the first component; 4 items loaded strongly (0.4+), or on the second, third and fourth component respectively; 2, 3 and 2 items loaded strongly (0.4+), on the fifth, sixth and seventh component respectively. Item 13 (difficulty eating foods he/she would to eat) was identified to have the highest factor loadings of 0.856, while item 23 (hard time paying attention in school) had the lowest factor loadings of 0.443. The reliability of the 7-factors model was tested and the results showed a scale Cronbach’s alpha of> 0.80.

Based on khi² values and their degrees of freedom, the close fit indices were calculated to be NFI = 0.837; CFI = 0.939; TLI = 0.889; RMSEA = 0.049 and SRMR = 0.036 (Table 6).

The Visual Summaries of Fit of distributions of the residuals showed that the frequency distributions of the correlation residuals or covariance residuals had a normal shape.

When only the first 4-factors, were retained, (eigenvalue> 1), these explained 47, 241% of total variance. Factor 1: contained 15 items from EWB (5/7items) and SWB (9/10 items); Factor 2: were found 7/8items of FL; Factor 3: was composed by 6 items: 2 items moved from EWB ( “been upset” and “irritable and frustrated” and 3 items moved form SO (bleeding gum; bad breath and food caught); and 1 item from SWB “Not wanted/unable to take part in activities (sport, drama, clubs)” . Factor 4 with only one item can’t be considered as a factor. It’s about “Food stuck to roof of mouth” . The values of absolute fit indices showed a poor level of fit: khi²/df = 1.69; NFI = 0.759; TLI = 0.837; CFI = 0.879; RMSEA = 0.073.

Discriminant validity confirmed the hypothesis. As expected, the mean of total scale score was higher for parents/caregivers with children in the cleft and other rare disease groups and lower in the paediatric healthy group. The difference between groups reached statistical significance (p < 0.001 and p < 0.05) (Table 7).