Evaluating Everyday Prospective Memory in School-Age Children through Parent- and Self-Reports: Validating Questionnaires and Examining Relations to Executive Functions and Autistic Traits

2.1. Translation and Validation of the PRMQC The validation process followed the procedure for adapting the PRMQ [22]. First, permission to adapt the parent-reported PRMQC from English to Chinese was obtained from the original author of the previous study [23]. Second, two experienced researchers evaluated the items’ cultural relevance and content validity and found most items were suitable for the Chinese context except for item 9. This item, “Does your child repeat the same story to the same person on different occasions?” was designed to detect memory errors, but the behavior described in the item might also occur because children enjoy telling the same story repeatedly. Therefore, a new item with less ambiguity (“Does your child ask someone a question and ask it again when seeing them later?”) was developed to form a 17-item version. Third, a researcher with a psychological background and proficiency in both English and Chinese translated the questionnaire from English into Chinese. The other researchers with similar backgrounds translated it back into English and discussed and solved inconsistencies. For the self-reported PRMQC, the parent-reported version (17-item) was adapted by replacing the phrase “your child” with “you”. In order to evaluate the contextual relevance and linguistic clarity of the items, a pilot test on the parent- and self-reported PRMQC was conducted in a sample of 12 children aged from 6 to 11 years and their parents. According to the parents, most items were common in children’s daily lives and were clearly expressed except for item 13. Parents found it difficult to judge this item (“Does your child look at something without realizing he/she has seen it moments before?”). Therefore, a new item was developed, “Does your child ever do something but cannot remember if he or she has done it after a while?”. Because children might watch entertainment on other devices (e.g., tablet, mobile phone), the word “television” in item 15 (“Does your child forget what he/she watched on television?”) was changed to “cartoon”. Parents of younger children from 6 to 8 years old also pointed out their difficulty of understanding frequency words such as “rarely” and “occasionally”, and thus, a frequency diagram with dots representing response options was added. As some children reported difficulty in understanding some items, the expressions of these items were slightly adjusted, with the content remaining the same. After these revisions, both the parent-reported and self-reported PRMQC comprised of 18 items. Each item described a PM or RM error of children in their daily lives. Parents rated the frequency of their children committing each memory error using a 5-point Likert scale (1 = never, 2 = rarely, 3 = occasionally, 4 = sometimes, 5 = often). Children rated the frequency of themselves committing each memory error, using the same 5-point Likert scale as parents, but with the aid of frequency diagrams. The finalized questionnaires are presented in Appendices A and B. 2.2. Participants and Procedure The study was approved by the Ethics Committee of the Institute of Psychology, Chinese Academy of Sciences (H20033). Written consent was obtained from all participating parents or caregivers. The children completed the printed self-reported version of the PRMQC (18 items) in class, with two experimenters giving instructions and providing assistance. The children took the parent-reported PRMQC home for their parents/caregivers to complete and brought them back to their teachers. A total of 1127 pairs of parents and children from age 6 to 12 years were recruited from a primary school in China. The common exclusion criteria for the parent-reported and the self-reported PRMQC included: (1) the parent reported his/her child had a history of neurological diseases, developmental disorders, or mental illness; (2) two or more items of the PRMQC were missing. In addition, the parent-reported questionnaire was also considered invalid if it was not completed by parents (e.g., grandparents, uncles, aunts) or if parents reported they were not familiar with child’s daily life. Based on these criteria, the final valid sample included 938 parent-reported questionnaires and 941 self-reported questionnaires. Their demographic information is presented in Table 1. Two weeks later, a total of 80 pairs of children and their parents were invited again to complete the PRMQC questionnaires again. Given that the research was conducted during the COVID-19 period, the retest sample was a convenient sample. The child participants were from two classes and were between 8 and 11 years old. The inclusion and exclusion criteria remained unchanged, with an additional requirement: the same parent had to complete the questionnaire in both the initial test and the retest sessions. 2.3. Data Analysis The data were entered using EpiData software, and 10% of the questionnaires were randomly selected for verification. As the questionnaires with less than 2 items missing were considered valid, the missing values of the items in these questionnaires were replaced by series means. As the original items 9 and 13 were found to be confusing or difficult for parents to judge, they were replaced by items 17 and 18, resulting in a revised 16-item version. All subsequent statistical analyses were conducted based on this revised 16-item version. Three typical structural models of PRMQC were tested using CFA with Mplus 7.0. The single-factor model assumed a single factor reflecting the overall memory function of an individual. The PM-RM correlated factor model assumed the questionnaire comprised of two related factors (i.e., PM and RM). The bi-factor model contained a general memory factor comprising all items and two specific and orthogonal factors (i.e., PM and RM, no correlation). We used the WLSMV (weighted least squares mean and variance adjusted estimation) method to estimate the parameters of the models. The fitting index included χ², CFI (comparative fit index), TLI (Tucker-Lewis index) and RMSEA (root mean square error of approximation). χ² reflects the fit between the tested model and the theoretical hypothesis model, but this index needs to take account the sample size and normality of the data. The criteria for a reasonable model were CFI ≥ 0.90, RMSEA ≤ 0.08, and TLI ≥ 0.90 [35,36]. Based on the results of structural models of the PRMQC, the preliminary norms for the PRMQC were generated using the method described in a previous study [18]. The internal consistency was assessed by calculating Cronbach’s α coefficients for the questionnaire. The test-retest reliability was assessed using the Pearson correlation analysis. The relationships between children’s age and PRMQC scores (both parent-reported and self-reported versions) were examined using the Pearson correlation analyses. Gender differences in the PRMQC scores were assessed using the independent samples t-tests. Additionally, we evaluated the association between the parent-reported and self-reported PRMQC scores using the Pearson correlation. To further explore developmental patterns, we divided the sample into a younger group (7–9 years) and an older group (10–12 years) and repeated the correlation analyses within each subgroup. The Fisher’s Z-tests were used to compare the strength of correlations between the two age groups. All statistical tests (except for Fisher’s Z-test) employed Bonferroni correction for multiple comparisons (α = 0.025, k = 2). 2.4. Results 2.4.1. Confirmatory Factor Analysis CFA results of both versions are presented in Table 2. In both versions, the bi-factor model showed negative loads for specific items, suggesting anomalies of the model and that it should not be adopted. For the parent-reported PRMQC, the Chi-square test revealed no significant difference between the single factor model and the PM-RM correlated factor model (Δχ² = 3.62, p = 0.057). For the self-reported PRMQC, the Chi-square test revealed significantly smaller χ² the PM-RM correlated factor model than the single factor model (Δχ² = 12.93, p < 0.001). Compared with the single factor model, the PM-RM correlated factor model allows the assessment of both PM and RM factors. Therefore, the PM-RM correlated factor model was selected as the best model for both parent- and self-reported PRMQC, and the factor loadings are presented in the supplementary materials (Figure S1). In both versions, the PM and RM scores were calculated by summing up the items associated with PM and RM. The means (and standard deviations) of parent- and self-reported PM and RM scores are presented in Table 1. The preliminary norms for both versions are developed, and the tables for converting the PM and RM raw scores to T scores are presented in the supplementary materials (Table S1). 2.4.2. Reliability In the parent-reported PRMQC, the Cronbach’s α for the PM scale (8 items) and the RM scale (8 items) were 0.77 (CI: 0.75–0.79) and 0.75 (CI: 0.72–0.77), respectively. In the self-reported PRMQC, the Cronbach’s α for the PM scale (8 items) and the RM scale (8 items) were 0.73 (CI: 0.70–0.75) and 0.74 (CI: 0.71–0.76). The valid test-retest sample included 62 pairs of children and their parents. For the parent-reported PRMQC, the correlations were moderate for both the PM scale (r = 0.43, p < 0.001) and the RM scale (r = 0.49, p < 0.001). The self-reported PRMQC demonstrated good consistency for both the PM scale (r = 0.63, p < 0.001) and the RM scale (r = 0.63, p < 0.001). 2.4.3. The Effects of Children’s Age and Sex on the PRMQC Scores In the parent-reported PRMQC, children’s age was not significantly correlated with either the PM score (r = 0.05, p = 0.119) and RM score (r = 0.02, p = 0.549). The parent-reported PM scores for boys and girls were 15.81 (SD = 4.96) and 15.98 (SD = 5.21), respectively, and there was no significant sex difference (p = 0.604, Cohen’s d = 0.03). The parent-reported RM scores for boys and girls were 14.12 (SD = 4.48) and 14.76 (SD = 4.90), with lower scores in boys than in girls (p = 0.038, Cohen’s d = 0.14), but this difference was no longer significant after Bonferroni correction. In the self-reported PRMQC, children’s age significantly correlated with the PM score (r = 0.14, p < 0.001), but not with the RM score (r = 0.03, p = 0.442). The PM scores for boys and girls were 14.58 (SD = 4.55) and 15.22 (SD = 5.23), respectively, with higher scores in girls than in boys (p = 0.046, Cohen’s d = 0.13), but this difference was no longer significant after Bonferroni correction. The RM scores for boys and girls were 14.38 (SD = 4.81) and 14.86 (SD = 4.84), with no significant sex difference (p = 0.125, Cohen’s d = 0.10). 2.4.4. The Relationships between the Self- and Parent-Reported PRMQC There were 844 pairs of parents and children who completed PRMQC. The correlations between the child and parent reports were significant, both for the PM score (r = 0.40, p < 0.001) and for the RM score (r = 0.36, p < 0.001). Children were further divided into a younger group (7 to 9 years old, N = 416) and an older group (10 to 12 years old, N = 416). The self-parent correlations were significantly higher in the older group compared with the younger group; this was evident in both the PM score (younger: r = 0.28, older: r = 0.48; Z = 3.38, p < 0.001) and the RM score (younger: r = 0.25, older: r = 0.46; Z = 3.47, p < 0.001). These results remained significant after Bonferroni corrections. 2.5. Discussion Study 1 validated the previously developed parent-reported and newly developed self-reported PRMQC in a large sample of school-age children and their parents. Results suggest that the PM-RM correlated factor model is the best model for both versions, which is consistent with the optimal model of the Chinese PRMQ for adults [22]. Moreover, it had acceptable reliability (Cronbach’s α ≥ 0.7), indicating acceptable reliability of the model. However, only children older than 7 years were able to complete the questionnaire, suggesting that parent-reports are more suitable for children under age 7. In addition, the child and parent reports of the PRMQC showed moderate correlations (r = 0.40 for PM, r = 0.36 for RM). A recent study [37] recruited children and adolescents aged 5–18 years, along with their parents, to report on the daily memory functioning of the children and adolescents. The results revealed two common factors of the parent- and child-reported versions, and the parent-child correlations on these factors varied in size (r = 0.68 for “forgetting”, and r = 0.19 for “remembering”). These findings suggest that the questionnaire taps a similar construct of the parent and child version but also indicates that children and parents may have different access to children’s internal memory processes. Indeed, for memory questionnaires with both child- and parent-reported versions, parents and children share a common awareness of children’s daily memory errors, but each informant also offers unique perspectives. For example, parents and children may demonstrate stronger agreement regarding memory failures at home (e.g., forgetting household chores) compared to those at school (e.g., failing to bring required materials), reflecting differences in observation opportunities and situational awareness. Interestingly, we noticed the child-parent correlations in PM and RM scores were larger in the older children group (age 10–12) than the younger children group (age 7–9), which may reflect the development of metacognitive ability in school-age children [38]. As children grow older, they become familiar with memory tasks and have more experience with memory failures, which raises their awareness and retrieval of daily memory errors. In the parent-reported PRMQC, children’s age did not correlate with daily PM error, which is consistent with a previous study using a different parent-reported questionnaire (i.e., CEMQ) to measure school-age children’s daily PM [39]. In the self-reported PRMQC, age significantly correlated with self-reported PM, showing that older children reported more PM errors than younger children. On the one hand, the finding may reflect children’s increased awareness of daily PM errors as their metacognitive ability increases [38], or the demand for daily PM tasks is more likely to exceed children’s PM abilities as they grow older. On the other hand, the correlation was relatively small (r = 0.14). Thus, the relationship between age and self-reported PM errors in children requires replication and should be interpreted cautiously. In terms of the sex effect on PM in typically developing children, previous studies using laboratory-based PM tasks indicated mixed results [7,40,41,42,43]. In the present study, both parents and self-reported ratings suggest that boys and girls committed similar amounts of PM errors in everyday life. This finding also suggests that both versions of the PRMQC were not biased in evaluating the daily PM functions of males and females. In a large adult sample between 18 and 59 years old, there was a trend that females reported more PM errors than males [22], which is probably related to females maintaining more PM intentions for family members [44]. Future research may consider investigating sex effects on daily PM across the lifespan.