Globally, the troubling burden of obesity continues to be evident among both children and adults(1) as the prevalence of elevated BMI and its associated morbidity and mortality continue to soar(1,Reference Friedrich2) . Since childhood is a critical period for shaping lifelong obesity risk(Reference Cunningham, Kramer and Narayan3), identifying modifiable risk factors at an early age is vital for the promotion and maintenance of a healthy weight status(Reference Deckelbaum and Williams4).
Characteristics of the mealtime environment and the behaviour of family members around food have been associated with child eating behaviours and weight status in several predominantly developed countries, including the USA, the UK and Australia(Reference Wardle, Guthrie and Sanderson5–Reference Nowicka, Sorjonen and Pietrobelli10). For example, overly demanding maternal feeding behaviours and restriction, such as use of punishments and rewards, were found to be inversely associated with children’s ability to self-regulate food intake, which can lead to overeating and weight gain(Reference Faith, Berkowitz and Stallings11–Reference Fisher and Birch14). Indulgent feeding behaviours and low involvement in child feeding have also been associated with higher weight status among children(Reference Faith, Scanlon and Birch6,Reference Hughes, Shewchuk and Baskin15) . Although evidence supporting causal associations between maternal feeding behaviours and child obesity is lacking(Reference Lumeng, Kaciroti and Retzloff16), results from several studies show that the association between maternal feeding and child eating behaviours is bidirectional; mothers may adjust and adapt the way they approach feeding in response to the child’s temperament, behaviour and weight status(Reference Byrne, Jansen and Daniels9,Reference Garcia, Power and Beck17,Reference Haszard, Russell and Byrne18) and the mother’s approach to feeding can in turn influence the child’s counter-response and behaviour around food(Reference Galindo, Power and Beck19). Therefore, maladaptive feeding and eating behaviours of mothers and their children may be described as a viscous cycle of obesogenic events that can ultimately promote an unhealthy weight status. Hence, validated instruments that aid in the evaluation of both maternal feeding and child eating behaviours are needed in order to efficiently identify maladaptive behaviours and subsequently promote a healthier mealtime environment.
As seen in many countries around the world, the prevalence of obesity and obesity-related diseases (e.g. type 2 diabetes) in the Kingdom of Saudi Arabia (KSA) has been increasing alarmingly(Reference Al-Othaimeen, Al-Nozha and Osman20). The prevalence of overweight and obesity among children is concerning, where about 32 % of children were found to be overweight or obese(Reference El-Mouzan, Foster and Al-Herbish21). Moreover, KSA is classified as a country experiencing advanced nutrition transition, such that dietary patterns are changing rapidly and drastically to mimic those seen in Western countries(Reference Sibai, Nasreddine and Mokdad22,Reference Ng, Zaghloul and Ali23) . It is hypothesized that these changes in dietary intake are secondary to changes in socio-economic status and social and familial norms(Reference Popkin24). Although these sociocultural adaptations are likely to be accompanied by changes in the family mealtime environment, family mealtimes and the behaviour of family members around food in Saudi society have not been efficiently explored. Although one study found that Saudi mothers of pre-school children may use restriction and pressure to eat in feeding(Reference Almarhoon, Ramsay and Fletcher25), we were unable to identify any other studies evaluating maternal feeding behaviours and practices as they relate to child weight status. Additionally, we were able to identify only one study that evaluated initial validity of a child feeding assessment tool in a convenience sample of highly educated Saudi mothers living in the USA(Reference Almarhoon, Ramsay and Fletcher25).
The Child Feeding Questionnaire (CFQ) is a feeding behaviour assessment tool that was first developed by Birch et al. in 2001 to assess parental beliefs, attitudes and practices relating to child feeding in a US cohort of parents of 2–11-year-old children(Reference Birch, Fisher and Grimm-Thomas26). The questionnaire consists of thirty-one items with 5-point Likert response scales and generates seven factors: Perceived Responsibility; Perceived Parent Weight; Perceived Child Weight; Concern about Child Weight; Restriction; Pressure to Eat; and Monitoring. Although some studies involving CFQ subscales have shown poor or mediocre model fit(Reference Boles, Nelson and Chamberlin27) and problematic structure of some factors (mainly Restriction)(Reference Boles, Nelson and Chamberlin27–Reference Corsini, Danthiir and Kettler30), several studies have reported evidence of validity and reliability in US(Reference Birch, Fisher and Grimm-Thomas26,Reference Musher-Eizenman and Holub31,Reference Kaur, Li and Nazir32) and non-US samples(Reference Nowicka, Sorjonen and Pietrobelli10,Reference Geng, Zhu and Suzuki29,Reference Corsini, Danthiir and Kettler30,Reference Liu, Mallan and Mihrshahi33,Reference Camcı, Bas and Buyukkaragoz34) .
Given the immediate need for valid and reliable, culturally sensitive instruments to assess feeding in KSA, the objective of the present study was to examine the psychometric properties of an Arabic version of the CFQ (CFQ-A) in a sample of Saudi pre-schoolers and their mothers by: (i) testing the original seven-factor CFQ-A model in addition to a modified nine-factor CFQ-A model and examining model fit and internal reliability of subscales; (ii) examining test–retest reliability of CFQ-A factors; and (iii) examining intercorrelations among CFQ-A factors as well as correlations with child BMI Z-score (BMIZ).
In developing the modified nine-factor CFQ-A model, we hypothesized that Saudi mothers are more invested and indulgent in feeding and that they like to ensure that their children are ‘eating enough food’ and that they are at the heavier end of the normal weight range(Reference Almarhoon, Ramsay and Fletcher25,Reference Al-Hussyeen and Al-Sadhan35,Reference Mosli, Bakhsh and Madani36) . We therefore added the two additional factors, Use of Food as a Reward and Concern about Child’s Diet. Moreover, we hypothesized that Perceived Child Weight, Concern about Child Weight and Concern about Child’s Diet are correlated with feeding practices including Restriction, Pressure to Eat, Monitoring, Perceived Responsibility and Use of Food as a Reward, and that these maternal perceptions and practices are also correlated with child BMIZ(Reference Gregory, Paxton and Brozovic37). Results from the present study can help provide evidence regarding the appropriateness of use of an Arabic version of the CFQ in Saudi Arabia. Findings can also help identify and inform future work needed for further development of efficient feeding assessment tools among Saudi/Arab populations.
Methods
Sample and procedures
A total of 209 mothers and children were included in the present study. Participants were recruited from eight different pre-schools around the city of Jeddah, KSA. Schools were randomly selected based on location; two schools were located in the northern area, two in the southern area, two in the eastern area and two in the western area. Of the eight pre-schools, four were public (government-subsidized) and four were private pre-schools. A description of the study and consent forms were placed in the backpacks of all students enrolled in the pre-schools. Research assistants contacted mothers who returned signed consent forms and completed the study questionnaire with them over the telephone. The study questionnaire included the CFQ-A, questions that assessed use of food as a reward and concern about child’s diet, and questions that assessed demographic characteristics.
Within 1 week of completing the questionnaire, the study team performed school visits in order to obtain weight and height measurements of participating children. Study inclusion criteria were as follows: child is Saudi or a permanent resident of KSA; between 3 and 5 years old; resides with his/her mother and is healthy with no serious medical problems or history of food allergies; and mother is an Arabic speaker. In order to examine test–retest reliability, forty randomly selected mothers (19 %) completed the questionnaire for a second time approximately 2 weeks after the initial telephone call.
Measures
Demographic characteristics
Mothers were asked questions that assessed demographic characteristics including the child’s sex, birthdate and nationality; the mother’s birthdate, educational level, employment status, marital status and nationality; as well as the family’s total monthly income. Birthdates and dates of interviews were used to calculate child and maternal age.
The Arabic version of the Child Feeding Questionnaire
Mothers completed the CFQ-A in Arabic language, following a forward–backward translation process conducted by bilingual professionals(Reference Nowicka, Sorjonen and Pietrobelli10,Reference Cramer, Perrine and Devinsky38) . The final version of the CFQ-A was pretested using a different sample consisting of sixty mothers of pre-school children and was reviewed by an expert committee of four health-care professionals. Clarifications were made to wording of questions and response options based on comments from mothers and expert committee members.
Use of food as a reward and concern about child’s diet
Saudi mothers may be more indulgent in feeding, traditionally use food as a reward to encourage desirable behaviours and may consciously or subconsciously follow the traditional belief that heavier children are healthier(Reference Almarhoon, Ramsay and Fletcher25,Reference Al-Hussyeen and Al-Sadhan35,Reference Mosli, Bakhsh and Madani36) . We therefore elected to evaluate a modified CFQ-A model that assessed Use of Food as a Reward as a separate construct, rather than including these items with the Restriction subscale, as employed in the original questionnaire(Reference Birch, Fisher and Grimm-Thomas26). This approach yielded satisfactory results among a sample of Chinese mothers(Reference Liu, Mallan and Mihrshahi33). Furthermore, we included a separate subscale for assessing Concern about Child’s Diet.
Mothers completed questions that were adapted from the Meals in Our Household Questionnaire, a questionnaire developed to assess family meals in the households of 3–11-year-old children(Reference Mosli, Miller and Peterson40). Questions included six items that assessed Use of Food as a Reward (FR; response options ranging from 1 = ‘never’ to 5 = ‘always’)(Reference Anderson, Must and Curtin39). These items were: FR1, ‘I give my child food to keep him/her quiet when shopping or travelling’; FR2, ‘I give my child food to reward him/her for good behaviour’; FR3, ‘I withhold a food my child likes as a consequence for bad behaviour’; FR4, ‘My child expects to be given a favourite food as a reward’; FR5, ‘I give my child a special food to celebrate an achievement’; and FR6, ‘I give my child food to persuade him/her to do something he/she does not really want to do’. In addition, mothers completed seven out of seventeen items that assessed Concern about Child’s Diet (CD; response options ranging from 1 = ‘not at all concerned’ to 6 = ‘extremely concerned’)(Reference Anderson, Must and Curtin39). We selected items that reflected concern about child’s diet in general, rather than those that measured concern about child’s consumption of specific food types (e.g. CD14 and CD17 concerning eating vegetables and meat). These items were: CD1, ‘Child is not eating enough’; CD2, ‘Child is eating too much’; CD3, ‘Child eats a lot of junk food’; CD4, ‘Child eats only a few types of food’; CD5, ‘Child is not getting good nutrition’; CD6, ‘Child has poor eating habits’; and CD7, ‘Child will not try new foods’.
The same translation and pretesting process was utilized with these items, which were used to construct a modified nine-factor CFQ-A model as described in following sections.
Child BMI Z-score
Trained staff members followed standardized procedures to measure children’s weights and heights. Calibration of the instruments was checked periodically. Shoes and heavy clothing were removed and the child was asked to stand still on a digital scale, with his/her weight equally distributed on both feet. Each child was weighed twice and if the two readings were inconsistent by more than 0·1 kg, the child was weighed two more times and the average of the two measurements was calculated. Similarly, height was measured twice and if the measurements differed by more than 0·5 cm, two more measurements were taken and the average of the two measurements was calculated. BMI was calculated for each child by dividing weight (in kilograms) by the square of height (in metres)(Reference Mosli, Miller and Peterson40). BMIZ were calculated based on the age- and sex-specific WHO growth standards for children aged ≤5 years and the WHO growth reference data for children who were between 5 and 6 years old(Reference Onis, Onyango and Borghi41,42) . Two children had missing weight and height data due to transferring to other schools and were therefore excluded from analyses involving child BMI data.
Statistical analysis
Analyses were conducted using the statistical software package IBM SPSS Statistics version 21.0. Descriptive statistics were used to examine sample characteristics, as well as the distribution of questionnaire factors, maternal BMI and child BMIZ.
To examine internal reliability, Cronbach’s α was calculated for each factor in the seven- and nine-factor CFQ-A models. A Cronbach’s α value of 0·70 or higher is considered favourable(Reference DeVellis43,Reference Kline44) . Spearman correlation coefficients were calculated to examine the 2-week test–retest reliability of the CFQ-A factors.
To assess validity of the seven- and nine-factor models, confirmatory factor analysis was performed using lavaan version 0.5-23(Reference Rosseel45) in R version 3.3.1. Diagonally weighted least squares was used to estimate model parameters. Diagonally weighted least squares was specifically designed for ordinal data and is more appropriate/recommended for use with Likert-type scales(Reference Li46,Reference Lin, Pakpour and Broström47) .
Fit indices were estimated for each of the models by calculating the root-mean-square error of approximation, Tucker–Lewis index, comparative fit index, standardized root-mean-square residual, and the likelihood ratio test or the minimum sample discrepancy function/df(Reference Cole48,Reference Hoelter49) . The suitability of confirmatory factor analysis was assessed prior to conducting the analysis, and inspection of the correlation matrix showed that all variables in both the seven-factor and nine-factor models had correlation coefficients greater than 0·30.
Table 1 summarizes factors and subscales of the original seven-factor CFQ-A, as well as the modified nine-factor CFQ-A. The modified nine-factor CFQ-A model includes two additional factors: Use of Food as a Reward and Concern about Child’s Diet(Reference Anderson, Must and Curtin39). As described earlier, items relating to using food as a reward were removed from the Restriction subscale and were included only in the Use of Food as a Reward subscale in the modified nine-factor model. These items were RST3A and RST3B. Scores for all factors were calculated as the mean of contributing items, with higher scores reflecting more of the given behaviour(Reference Birch, Fisher and Grimm-Thomas26).
Summary of factors and subscales of the original seven-factor and modified nine-factor Arabic version of the Child Feeding Questionnaire (CFQ-A)
Intercorrelations among the modified CFQ-A factors as well as correlations with child BMIZ were examined. Interpretation of the correlation coefficients were based on Cohen’s guidelines(Reference Cohen50); correlations between 0·5 and 1·0 were considered large, correlations between 0·3 and 0·5 as medium and correlations between 0·1 and 0·3 as small. Only children with complete BMIZ data (n 207) were included in the correlation analysis between CFQ-A factors and child BMIZ. For all statistical analyses, the significance level was set at 0·05.
Results
Sample characteristics
Mean child age was 4·79 years (sd = 0·79 years), and about half of the sample (51·7 %) was male. The majority (approximately 70 %) of children and mothers were Saudi. Of the sixty-four non-Saudi children, about half (n 29, 45·3 %) were Egyptian, ten (15·6 %) were Syrian, eight were Yemeni (12·5 %), six were Jordanian (9·4 %), five (7·8 %) were Lebanese, five (7·8 %) were Palestinian and one (1·6 %) was Sudanese. In addition, the majority of mothers (80 %) had a college education or higher, and about half (51·2 %) reported that they were housewives (Table 2).
Characteristics of the sample of pre-schoolers and their mothers (n 209) recruited from eight different pre-schools in Jeddah city, Kingdom of Saudi Arabia, October 2017–April 2018
Data are presented as mean and sd for continuous variables or as n and % for categorical variables.
Table 3 shows the distribution (including mean, sd and range) of questionnaire factors.
Distribution and internal consistencies of factors of the Arabic version of the Child Feeding Questionnaire (CFQ-A) in the sample of pre-schoolers and their mothers (n 209) recruited from eight different pre-schools in Jeddah city, Kingdom of Saudi Arabia, October 2017–April 2018
Internal and test–retest reliability
Table 3 shows the respective Cronbach’s α for each factor of the seven- and nine-factor CFQ-A models. For the seven-factor CFQ-A model, three factors fell below the 0·70 threshold: Perceived Parent Weight (Cronbach’s α=0·67), Restriction (Cronbach’s α = 0·60) and Pressure to Eat (Cronbach’s α=0·49). These same factors were also below the 0·70 threshold for the nine-factor CFQ-A model: Perceived Parent Weight (Cronbach’s α = 0·67), Restriction (Cronbach’s α = 0·62) and Pressure to Eat (Cronbach’s α = 0·49). All factors showed excellent 2-week test–retest reliability, with Spearman correlations > 0·70 (all P < 0·05).
Confirmatory factor analysis
The seven-factor CFQ-A model
Forced extraction of seven factors explained 19·53, 13·71, 10·17, 9·12, 7·10, 6·07 and 4·84 % of the total variance, respectively. The seven-factor solution explained 70·5 % of the total variance. The root-mean-square error of approximation of 0·02 indicates good fit(Reference MacCallum, Browne and Sugawara51). The Tucker–Lewis index and comparative fit index were calculated to be 0·96 and 0·97, respectively. The seven-factor model is illustrated in Fig. 1.
Seven-factor model of the Arabic version of the Child Feeding Questionnaire (CFQ-A), with the factors Perceived Responsibility, Perceived Parent Weight, Perceived Child Weight, Concern about Child Weight, Restriction, Monitoring and Pressure to Eat (see Table 1 for a description of the factors and subscales). Fit indices: root-mean-square error of approximation = 0·02 (95 % CI 0·00, 0·04), Tucker–Lewis index = 0·96, comparative fit index = 0·97, standardized root-mean-square residual = 0·07, χ 2 = 267·7 (P < 0·001), minimum sample discrepancy function/df = 1·10 (ε 1–ε 29 are error terms)
The nine-factor CFQ-A model
Forced extraction of nine factors had explained 15·42, 13·43, 11·84, 7·57, 6·79, 5·01, 4·16, 3·62 and 3·59 % of the total variance, respectively. The nine-factor solution explained 71·4 % of the total variance. The root-mean-square error of approximation of 0·04 indicates good fit(Reference MacCallum, Browne and Sugawara51). The Tucker–Lewis index and comparative fit index were estimated to be 0·90 and 0·88, respectively. The nine-factor model is illustrated in Fig. 2. A comparison of fit indices of the seven- and nine-factor models is shown in Table 4.
Nine-factor model of the Arabic version of the Child Feeding Questionnaire (CFQ-A), with the factors Perceived Responsibility, Perceived Parent Weight, Perceived Child Weight, Concern about Child Weight, Restriction, Monitoring, Pressure to Eat, Use of Food as Reward and Concern about Child’s Diet (see Table 1 for a description of the factors and subscales). Fit indices: root-mean-square error of approximation = 0·04 (95 % CI 0·03, 0·05), Tucker–Lewis index = 0·88, comparative fit index = 0·90, standardized root-mean-square residual = 0·07, χ 2 = 694·0 (P < 0·001), minimum sample discrepancy function/df = 1·41 (ε 1–ε 18 and ε 21–ε 39 are error terms)
Comparison of fit indices between the seven-factor and nine-factor models of the Arabic version of the Child Feeding Questionnaire (CFQ-A) in the sample of pre-schoolers and their mothers (n 209) recruited from eight different pre-schools in Jeddah city, Kingdom of Saudi Arabia, October 2017–April 2018
CMIN, minimum sample discrepancy function; χ 2, chi-square statistic; TLI, Tucker–Lewis index; CFI, comparative fit index; SRMR, standardized-root-mean square residual; RMSEA, root-mean-square error of approximation.
Factor–factor and factor–child BMI Z-score correlations
As shown in Table 5, the highest correlations between factors were between Restriction and Monitoring (r = 0·50, P < 0·001), Perceived Responsibility and Monitoring (r = 0·40, P < 0·001) and Perceived Responsibility and Restriction (r = 0·30, P < 0·001). There were small positive correlations between Perceived Child Weight and Restriction (r = 0·20, P < 0·001) and Perceived Child Weight and Monitoring (r = 0·17, P = 0·01), as well as between Concern about Child’s Diet and Concern about Child Weight (r = 0·18, P < 0·001) and Concern about Child’s Diet and Use of Food as a Reward (r = 0·16, P = 0·01). Although marginally significant, Perceived Responsibility was also positively correlated with Perceived Child Weight (r = 0·13, P = 0·06) and Pressure to Eat (r = 0·12, P = 0·07).
Factor–factor and factor–child BMI Z-score (BMIZ) correlations of the modified nine-factor model of the Arabic version of the Child Feeding Questionnaire (CFQ-A), October 2017–April 2018
n 209 unless otherwise specified.
*P < 0·10, **P < 0·05, ***P < 0·01.
We detected negative correlations between Perceived Responsibility and Concern about Child’s Diet (r = −0·24, P < 0·001) and between Perceived Responsibility and Use of Food as a Reward (r = −0·19, P < 0·001). Furthermore, Concern about Child’s Diet was negatively correlated with Perceived Parent Weight (r = −0·15, P = 0·03) and Perceived Child Weight (r = −0·23, P = 0·001). Concern about Child Weight was negatively correlated with Pressure to Eat (r = −0·14, P = 0·04).
There were significant correlations between child BMIZ and five out of the nine factors; Perceived Child Weight (r = 0·49, P < 0·001), Perceived Parent Weight (r = 0·30, P < 0·001), Restriction (r = 0·21, P < 0·001) and Monitoring (r = 0·16, P = 0·02) were each positively correlated with child BMIZ, while Concern about Child’s Diet was negatively correlated with child BMIZ (r = −0·18, P = 0·01; Table 5).
Discussion
Our study including 209 Saudi mothers and their pre-school children provides initial evidence of the validity and reliability of a modified nine-factor CFQ-A model. Our results suggest that modifying the restriction subscale and including a separate subscale to assess using food as a reward may be suitable for Saudi mothers. This approach has previously provided satisfactory results among a sample of Chinese mothers of pre-schoolers(Reference Liu, Mallan and Mihrshahi33), suggesting that parenting/feeding strategies among Middle-Eastern and Asian families might be comparable. Indeed, when comparing our restriction factor mean with that of the study involving Chinese mothers(Reference Liu, Mallan and Mihrshahi33) (total mean scores out of 5), we observed that they are almost similar (mean=4·40, sd = 0·94 and mean=4·18, sd = 0·80, respectively), whereas a larger difference is seen when comparing with Western samples of pre-school (mean = 2·72, sd=0·99)(Reference Nowicka, Sorjonen and Pietrobelli10) and school-age children (mean = 2·50, sd=;0·95)(Reference Birch, Fisher and Grimm-Thomas26). Likewise, perceived responsibility was higher among our Saudi mothers (mean = 4·23, sd = 0·93), as well as Chinese (mean = 4·26, sd = 0·71) and Turkish mothers (mean = 4·14, sd = 0·67), compared with US (mean = 3·40, sd = 0·95) and Swedish (mean = 4·02, sd = 0·69) mothers. Interestingly, monitoring among our sample of Saudi mothers (mean = 4·44, sd = 0·80) was noticeably higher than that of other samples of Chinese (mean = 3·68, sd = 1·12), Turkish (mean = 2·96, sd = 0·70), US (mean = 3·60, sd = 0·90) and Swedish (mean = 3·87, sd = 0·88) mothers. These cross-cultural similarities/differences might be attributed to societal norms regarding the role and responsibility of mothers in feeding their children, as well as availability and access to quality early childhood care centres (e.g. daycare centres and pre-schools), which affects the amount of time mothers spend being in charge of feeding their children.
Both the original seven-factor model as well as the modified nine-factor model showed good fit to data (root-mean-square error of approximation < 0·05)(Reference Browne, Cudeck, Bollen and Long52). Furthermore, our findings suggest adequate internal reliability for six out of the nine factors in the modified nine-factor CFQ-A model. The three factors with Cronbach’s α that fell below the 0·70 cut-off were: Perceived Parent Weight, Restriction, and Pressure to Eat. This is consistent with results from other validation studies that also found lower internal reliability for the Restriction and Pressure to Eat factors(Reference Nowicka, Sorjonen and Pietrobelli10,Reference Boles, Nelson and Chamberlin27) . One possible explanation for the less satisfactory estimates for the Restriction and Pressure to Eat subscales could be the variation between mothers in defining some phrases used in questionnaire items, such as ‘I have to be sure that my child does not eat too many high-fat foods’. For example, while one mother might define ‘being sure’ as actively or physically removing food items from the child’s reach, another might define this as more subtle behaviours such as stating the harms of high-fat foods in front of the child. In line with previous studies, ours showed evidence of adequate test–retest reliability(Reference Nowicka, Sorjonen and Pietrobelli10).
Overall, intercorrelations between factors in the modified model were in the expected direction and give support to our hypothesis that perceived child weight and concern about child weight and diet are each associated with feeding practices among Saudi mothers. Results suggest that mothers who apply higher restriction may monitor their children’s eating to a greater extent. Mother who perceive that they are highly responsible for their children’s eating may apply greater restriction. Although the original CFQ study by Birch et al. did not find a correlation between perceived responsibility and restriction(Reference Birch, Fisher and Grimm-Thomas26), our findings are consistent with those of a study involving Turkish mothers(Reference Polat and Erci53). In line with our findings, both the original CFQ study and the Turkish study reported a positive correlation between perceived responsibility and monitoring(Reference Birch, Fisher and Grimm-Thomas26,Reference Polat and Erci53) . Additionally, our findings suggest that mothers who perceive their children to have a higher weight status may apply higher restriction and monitoring, and mothers who are more concerned about their children’s weight may be more concerned about their diet.
Furthermore, we found that mothers who perceive that they are highly responsible for their children’s eating also reported lower concern about their child’s diet. This is probably due to these mothers exerting higher restriction and monitoring, leading them to feel more in control of what their children eat. Additionally, some of the Concern about Child’s Diet items were related to the child receiving ‘good nutrition’ and him/her having ‘poor eating habits’, which can be relevant to both underweight and overweight children. This might explain the negative correlation between Concern about Child’s Diet and Perceived Child Weight. Our results showed a negative correlation between Concern about Child Weight and Pressure to Eat, which is consistent with results from previous studies in Western samples(Reference Nowicka, Sorjonen and Pietrobelli10,Reference Faith, Berkowitz and Stallings11,Reference Birch, Fisher and Grimm-Thomas26) .
Correlations detected between the modified CFQ-A factors and child BMIZ also give support to our hypothesis that maternal perceptions and practices around feeding are associated with child BMIZ in a sample of Saudi mothers and pre-schoolers. While some studies did not detect an association between restriction and child BMI(Reference Liu, Mallan and Mihrshahi33,Reference Gregory, Paxton and Brozovic37) consistent with our results, others have reported a link with overeating and obesity risk(Reference Fisher and Birch13,Reference Fisher and Birch14,Reference Farrow, Haycraft and Blissett54) . Furthermore, although only marginally significant, our finding of a negative correlation between pressure to eat and child BMIZ is consistent with those of previous studies(Reference Liu, Mallan and Mihrshahi33,Reference de Souza Rezende, Bellotto de Moraes and Mais55) suggesting that mothers may pressure their children to eat more when they are thinner.
Since the present analyses did not take potential confounders into account, further studies that test adjusted statistical models are needed to further establish the association between maternal feeding behaviours and child BMI in a Saudi sample. Our findings, in addition to evidence from previous studies(Reference Almarhoon, Ramsay and Fletcher25,Reference Mosli, Bakhsh and Madani36) , suggest that Saudi mothers may feel great responsibility towards their role in feeding their families. Future studies that characterize feeding behaviours and examine associations with weight status are needed to help inform intervention studies aiming to enhance adaptive feeding practices and promote healthy eating behaviours among children in Saudi Arabia(Reference Patrick, Nicklas and Hughes56).
Our study was the first to evaluate validity and reliability of an Arabic version of the CFQ in a cohort residing in KSA. We were able to slightly modify the CFQ in order to better capture traditional feeding practices among Saudi mothers. Child anthropometry was objectively measured, and the sample was recruited from eight different pre-schools with varying degrees of socio-economic status. Limitations of the study included that our sample size was relatively small. Mothers were recruited through an opt-in approach by placing consent forms in children’s backpacks. This might have increased the likelihood of self-selection bias, such that mothers who were more interested in the topic of feeding might have been more inclined to participate. Administration of the questionnaire over the telephone might have increased the likelihood that mothers will give socially desirable responses, although this approach can help minimize missing data. Furthermore, since the majority of mothers in our study had a college degree or higher, our results may not be generalizable to other population groups with lower educational levels.
Future studies with larger sample sizes that include mothers with different educational backgrounds and studies that employ different recruitment strategies are warranted. Longitudinal designs are also needed in order to evaluate the effect of feeding behaviours over time and draw inferences regarding causal associations. Additionally, qualitative studies and those that include direct observation of mother–child interactions during mealtimes may be useful in identifying unique feeding behaviours among Saudi and Arab families.
Conclusion
Identifying characteristics of the mealtime environment and how they relate to child eating behaviours and weight status can help inform effective prevention and intervention strategies among Saudi families. Culturally sensitive assessment tools are needed in order to evaluate feeding behaviours among Saudi mothers. Our study provides evidence that supports the validity and reliability of the original seven-factor CFQ-A as well as a modified nine-factor CFQ-A model. Future larger studies are needed to further establish the psychometric properties of the CFQ-A as well as other feeding assessment tools.
Acknowledgements
Financial support: This project was funded by the Deanship of Scientific Research (DSR) at King Abdulaziz University, Jeddah (grant number J-734-142-38). The author acknowledges and thanks DSR for technical and financial support. Conflict of interest: The author has no financial relationships relevant to this article to disclose. Authorship: R.H.M. designed the study, oversaw data collection, analysed the data, wrote the manuscript and approved final the version as submitted. Ethics of human subject participation: This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving research study participants were approved by the Unit of Biomedical Ethics at King Abdulaziz University. Written informed consent was obtained from all participants.
