Child Perceptions Questionnaire in Croatia: Two Domains for Measuring Oral Health.

OBJECTIVE: To perform cross-cultural adaptation and to test psychometric properties of the 8-item CPQ in 11-14 year-olds: stepwise-regression (RSF:8) and item-impact (ISF:8) short-forms.
MATERIALS AND METHODS: The sample included 237 orthodontic patients aged 11-14 at two University Dental Clinics in Croatia. Structural and construct validity, reliability and responsiveness were assessed. Intraoral examination included an assessment of dental caries and malocclusion severity.
RESULTS: Two domains instead of originally suggested four are more appropriate for the assessment of oral health-related quality of life (OHRQoL) in both ISF:8 and RSF:8 (60.05% and 52.24% variance; α=0.56-0.85). Oral symptoms and functional limitations from the original instrument were grouped in one dimension that defines oral function, while emotional and social well-being was grouped in a dimension of psychosocial well-being. Instruments are able to detect differences between subjects with low and high caries and malocclusion severity. They were stable when there were no changes in oral conditions, while it was possible to detect differences induced by correction of malocclusion following orthodontic treatment (p<0.05).
CONCLUSION: A 8-item CPQ demonstrated good psychometric properties but points to the fact that two domains instead of four are more appropriate for the assessment of OHRQoL in 11-14 year-olds.

Introduction

Specific health-related questionnaires are being made in dentistry to evaluate patients' perception of their appearance, emotions and daily functioning in order to determine adequate treatment procedures and provide more information about patients' needs and demands. Recently, a new definition of oral health has been introduced by the World Dental Federation: “Oral health is multi-faceted and includes the ability to speak, smile, smell, taste, touch, chew, swallow and convey a range of emotions through facial expressions with confidence and without pain, discomfort and disease of the craniofacial complex” (). The Child Perceptions Questionnaire (CPQ) is in line with this definition, providing a simple method of evaluating children's well-being related to orofacial condition through dimensions of oral symptoms (OS), functional limitations (FL), emotional (EW) and social well-being (SW) (). The length of the questionnaire and time needed for answering it limits and its usage in clinical practice. Therefore, short-forms consisting of eight and 16 items were developed, by using item-impact (ISF:8, ISF:16) and stepwise-regression (RSF:8, RSF:16) methods ().

Our aim was to evaluate psychometric properties and to compare the cross-cultural validity and reliability of the RSF: 8 and ISF:8. The hypotheses were that translated questionnaires are equivalent to originals in groupings of items, and that capture concepts are similar to previously defined dimensions of validated oral health-related quality of life (OHRQoL) instruments. It was expected that the instruments are able to distinguish the children with higher and lower oro-facial problems, as well as to detect changes induced by dental treatment. It was assumed that the instruments are stable and will not detect changes in well-being and function unless there are changes in oral conditions.

Materials and Methods

Cross-cultural adaptation

Two experts (a dentist and an English language teacher) independently performed forward translation of the English CPQ. Both translations were then independently back-translated in English by other two experts (a dentist and an English language teacher). Expert committee (three dentists and developmental psychologist with a good knowledge of both languages) compared the original, translations and back-translations and designed first versions of questionnaires. Content validity was evaluated using a probe technique (n=20). Based on their comments, a few linguistic modifications were made. The Committee agreed on the final versions of the questionnaires.

Participants

The sample included 237 subjects (45% males) aged 11-14 years, consecutive patients referred for consultation or treatment at two University Dental Clinics in Croatia (Rijeka and Zagreb) during 2015-2017. A calculated minimum sample size was 220 participants considering type I error rate of 0.05, desired power of 0.80, anticipated model fit with root mean square error of approximation from 0.05-0.08 and with determined degrees of freedom 48 (). The local Ethics Committee approved the study (No. 2170-24-01-15-2) with written informed consent provided by each participant's parent. The study was performed according to Declaration of Helsinki 1964 and its later amendments.

Methods

Self-administrated questionnaires included CPQ items, self-assessed oral health and well-being on a 5-point scale (0-4; excellent-bad), satisfaction with esthetic appearance, self-perceived orthodontic treatment need and demand for treatment (0-4; not at all-a lot), Oral Impacts on Daily Performances (OIDP), Oral Health Impact Profile (OHIP) and Psychosocial Impact of Dental Esthetics Questionnaire (PIDAQ) (-). The intraoral examination included detection of dental caries according to WHO criteria by three calibrated experts, with weighted Kappa for intra-rater reliability 0.892-0.923 and for inter-rater 0.812-0.840 (95% CI 0.710-0.990; p<0.001). Malocclusion severity was assessed by using the Index of Orthodontic Treatment Need - Dental Health Component (IOTN DHC), by two orthodontists with weighted Kappa for intra-rater reliability 0.838-0.958 and for inter-rater 0.671 (95% CI 0.469-0.873; p<0.005). Stability over time was assessed on 50 children from two public schools in one-week interval without any dental intervention. The ability of the instrument to detect changes in oral conditions was evaluated on 33 subjects presenting class II/1 malocclusion subjected to orthodontic treatment by functional appliance during the period of approximately one year.

Statistical analysis

In statistical analysis, the explanatory and confirmatory factor analyses were used to test structural validity. Internal consistency was checked by calculating the Cronbach's alpha and inter-item correlations. Stability over time was assessed by paired t-test and correlations. The Mann-Whitney test was performed to test discriminant validity. CPQ domains were compared between two clinical groups based on caries severity and malocclusion severity. Cut-off points were determined by calculating the highest specificity and sensitivity values using Receiver Operating Characteristic curves. Convergent validity was assessed by the Spearman correlation coefficient r. Responsiveness was assessed using a Wilcoxon test. Changes in the domains were compared with the amount of reduction of OJ, and collated between groups based on global transition judgements collected post-treatment (Kruskal-Wallis Test). Commercial software SPSS v.22.0 (IBM Corp., Armonk, NY, USA) was used.

Results

Structural validity

Exploratory factor analysis with principal component analysis, Varimax rotation and extraction method based on eigenvalues greater than 1 revealed two-factor structure for both ISF:8 and RSF:8, which explained 60.05% and 52.24% of the variance, respectively. The items under OS and FL and under EW and SW were grouped together (Table 1, p. 11). Confirmatory factor analysis revealed a good model fit (standardized root mean square residual = 0.05 for ISF: 8, 0.06 for RSF:8), and moderate correlations between OS+FL and EW+SW domains (in ISF:8 r=0.60, in RSF:8 r=0.70).

Table 1

Factor loadings and items' grouping

	F1	F2
ISF:8
Food caught between teeth (OS)*	0.742
Bad breath (OS)	0.825
Difficulty eating/drinking hot/cold foods (FL)	0.598
Difficulty chewing firm foods (FL)	0.564	0.367
Upset (EW)		0.819
Felt irritable/dissatisfied (EW)		0.832
Avoided smiling/laughing (SW)		0.830
Asked questions (SW)		0.736
RSF:8
Mouth sores (OS)	0.719
Bad breath (OS)	0.553
Unable to sleep (FL)	0.559
Difficulty saying words (FL)	0.680
Upset (EW)	0.477	0.633
Concerned what people think about teeth/mouth (EW)	0.388	0.673
Teased/called names (SW)		0.776
Argued with children/family (SW)		0.808

*original items' grouping

Reliability

Internal consistency was higher for ISF:8 and for EW+SW domain. No significant differences were observed in absolute reliability assessment, while for relative reliability, higher level of agreement was observed in RSF domains (Table 2, p. 12).

Table 2

Reliability assessment

	ISF:8		RSF:8
	OS+FL	EW+SW	OS+FL	EW+SW
Internal consistency (n=237)
Mean ±SD	4.6 ±2.9	2.4 ±3.1	2.5 ±2.2	2.1 ±2.7
Range	0-15	0-15	0-12	0-14
α	0.67	0.85	0.56	0.75
α if item deleted	0.55-0.65	0.78-0.85	0.40-0.54	0.63-0.74
Inter-item correlation (range)	0.34(0.22-0.50)	0.56(0.47-0.73)	0.25(0.17-0.35)	0.44(0.32-0.65)
Absolute reliability (n=50)
ME*	0.83	0.61	0.74	0.68
SDC*	2.29	1.70	2.06	1.89
Paired differences mean(95% CI)	0.04(-0.10-0.17)	0.12(-0.06-0.29)	0.06(-0.15-0.03)	0.01(-0.09-0.12)
p*	0.569	0.182	0.182	0.709
LOA*	-0.92-1.10 (90.19%)	-1.10-1.33 (98.04%)	-0.55-0.67 (91.18%)	-0.72-0.74 (93.14%)
Relative reliability (n=50)
Spearman r	0.626	0.646	0.878	0.785
Weighted kappa	0.535	0.453	0.832	0.732
ICC*(95% CI)*	0.98(0.96-0.99)	0.88(0.79-0.93)	0.98(0.97-0.99)	0.98(0.97-0.99)

* Measurement error ME was calculated as square root of the residual variance. Smallest detectable change SDC was calculated as 1.96 * √2 * ME. P-value of the Paired t-test. Limits of agreement LOA were calculated as paired differences mean ± 1.96 * standard deviation of the differences between two measurements (percentage of test and retest that are within limits of agreement). The intraclass correlation coefficient ICC (95% confidence interval CI)

Convergent validity

Convergent validity was better in EW+SW, compared to OS+FL (r=0.34-0.54, compared to 0.19-0.37). Additionally, all domains correlated with PIDAQ PI (r=0.32-0.39, p<0.05). In ISF: 8 EW+SW correlated with PIDAQ SI and OHIP summary score (r=0.37 and 0.55, p<0.05), and OS+FL with PIDAQ SI (r=0.41, p=0.006). In RSF: 8, EW+SW correlated with OIDP and OHIP summary scores (r=0.41 and 0.53, p< 0.05). No correlations were observed for dental caries and IOTN DHC.

Discriminant validity

The areas under the Receiver Operating Characteristic curve of 0.616 for dental caries, and of 0.746 for IOTN DHC were detected. For IOTN DHC the cut-off point of three and for dental caries of four were determined. ISF: 8 OS+FL was able to detect differences between subjects based on caries and malocclusion severity, while RSF: 8 EW+SW for malocclusion severity (Figure 1, p.15).

Figure 1

Discriminant validity between clinical groups based on severity of dental caries, and based on malocclusion severity (asterisks mark differences significant at p<0.05)

Responsiveness

Following orthodontic treatment, 13 children reported no change, little improvement was reported by 12 and great improvement by eight (global transition judgements). All domains had a significant difference between clinical groups based on global transition judgement, except for EW+SW in RSF: 8 (p=0.068). All domains were able to detect differences in OHRQoL provided by orthodontic treatment (p<0.05). Changes in RSF: 8 EW+SW demonstrated a linear correlation with the amount of OJ reduction (Table 3, p. 13).

Table 3

Responsivenes

	RSF:8		ISF:8
	OS+FL	EW+SW	OS+FL	EW+SW
No change (n=13)a	0.00 (1.29)	0.46 (1.33)	0.23 (2.05)	0.31 (0.75)
Little improvement (n=12)a	0.33 (1.23)	0.67 (1.97)	0.58 (2.02)	0.67 (1.67)
Great improvement (n=8)a	3.25 (1.67)	3.88 (3.64)	5.00 (2.67)	5.75 (2.92)
pb	0.001	0.068	0.001	˂0.001
All (n=33)a	0.91 (1.89)	1.36 (2.64)	1.52 (2.93)	1.76 (2.88)
pc	0.010	0.006	0.008	0.001
ESd	0.45	0.48	0.46	0.56
SRMe	0.48	0.52	0.52	0.61
rf	0.187	0.463	0.179	0.176
pf	0.322	0.010	0.343	0.352

aMean baseline score - mean follow-up score (standard deviation). bKruskal-Wallis Test for between-subjects comparisons by global transition judgement collected post-treatment. cWilcoxon Signed Ranks Test for within-subjects comparisons for total score before and after treatment. dEffect size ES was calculated as Z/√n. eStandardized response mean SRM was calculated as mean change/standard deviation of the change scores. fPearson correlation coefficient r with corresponding p value for correlation between changes in CPQ domains and changes in overjet (mm).

Discussion

Although reduced number of questions could affect the content validity of the questionnaire, the findings in the literature indicate that short-forms show good psychometric properties (), which is similar to our results. Nevertheless, two domains instead of four might be more appropriate for the assessment of OHRQoL. Oral symptoms and functional limitations are grouped in a dimension of oral function, while emotional and social well-being in a dimension of psychosocial well-being. Four-item structure allows for a more detailed evaluation of the instrument's properties by performing within domain analyses (). Although unidimensionality of the 8-item CPQ was originally suggested (), a recent study also pointed to a two-factor structure of the CPQ short-forms ().

Dimension of psychosocial well-being shows better consistency than oral function dimension. RSF:8 and ISF:8 are stable and do not detect changes in OHRQoL unless there is a change in the oral condition. This is important because they will not generate error due to imprecision and accuracy in reporting well-being. The results of the RSF applied on two separate occasions may be considered essentially equivalent since Bland and Altman limits of agreement for both ISF domains were, although narrow, 1.7 times wider compared to RSF.

Although ISF manages to detect the difference between patients with low and high caries and malocclusion severity, in dimension of oral function relationship between them is not linear. This means that an increased severity of malocclusion or number of cariogenic lesions does not indicate proportional increase in functional disability. Recently, in a systematic review, the authors highlighted the fact that the impact of malocclusion on OHRQoL is presumed to affect predominantly dimensions of emotional and social well-being (). Nevertheless, effects of malocclusion on OHRQoL should be evaluated and interpreted carefully and individually. Specific malocclusion could have greater impact on one's OHRQoL, while presence of any malocclusion per se may not (). These findings emphasize the importance of cross-cultural adaptation in the process of validation of the OHRQoL instruments.

Both CPQ short-forms are capable to detect changes after orthodontic treatment, with greater effect in psychosocial rather than oral function dimension. Emotional component of the OHRQoL has higher sensitivity to change in oral health and dental esthetics. This is important when assessing the need for, and planning of the orthodontic treatment. The relationship between the degree of malocclusion correction and QoL is not linear since- most of the children report increase in QoL regardless of the degree of correction. The reason could be in children's individual psychological traits.

In both RSF:8 and ISF:8, domain which defines psychosocial well-being performed better in within domain analyses, describing the latent construct, stability of the results over time, correlated to the more measurements of the similar construct and, in RSF model, correlated linearly to the amount of actual clinical improvement. Moreover, the context of the dimension of psychosocial well-being measures the concept of satisfaction with teeth appearance and their function to a greater extent than the oral function domain does. This implicates that emotional and social aspects are more important to children, rather than the symptomatology related to the oral condition. Furthermore, the dimension of psychosocial well-being can distinguish children with low and high severity of caries, although this is significant only in RSF version.

A possible limitation of this study could be a high variability in CPQ results. This may be due to children's individual psychological characteristics and different comprehension of esthetics (-), as well as their current emotional state when reporting the impacts of oro-facial condition on OHRQoL. The results of our study are based on a convenience sample of a clinical population. Future studies on general samples in various settings are needed in order to further evaluate RSF:8 and ISF:8 performance in our socio-cultural setting.

Conclusion

Two domains are most appropriate for the assessment of OHRQoL by 8-item CPQ in 11-14 year-olds. Oral symptoms and functional limitations are grouped under dimension that defines oral function, while the emotional and social well-beings are grouped in a dimension of psychosocial well-being. 8-item instruments could be more effective in detecting alterations of OHRQoL than the original long form since they reduce the time needed to complete the questionnaire, and therefore reduce the risk of item non-response.