Background
Mental wellbeing has emerged as an important construct in population health, described as a fundamental human right and essential for a sustainable and functional society [
1]. Mental wellbeing has been defined as ‘a state of wellbeing in which every individual realises his or her own potential, can cope with the normal stresses of life, can work productively and fruitfully, and is able to make a contribution to her or his community’ [
2], encompassing concepts such as resilience, self-efficacy and optimism [
3]. As opposed to mental illness, which is either prevented or treated, mental wellbeing can only be promoted [
4], and promotion has been shown to relate to improved health and longevity in adults [
5]. Despite this, little information is available on the prevalence or social patterning of mental wellbeing in young people [
3], particularly compared to the extensive data on mental illness [
6].
In the United Kingdom (UK), local authority care includes the provision of accommodation for children and young people who are unable to live with their parents. There are a variety of reasons for young people to enter care, with approximately two thirds entering care due to abuse and neglect [
7]. As of March 2019, 6846 young people in Wales were in the care of their local authority, the majority (71%) accommodated in foster care placements [
7]. While most young people in care in the UK report their experiences to be good [
8], and report satisfaction with their life [
9], there is clear evidence that those who have experienced care do not fare as well as the general population in relation to their physical health, cognitive and language skills [
10], and mental health [
11‐
13], which in turn can impact their development and journey to adulthood [
14‐
16]. Studies have begun to investigate subjective wellbeing of children and young people in foster care in the UK [
9], and foster and residential care internationally [
17‐
19]. These studies have consistently identified lower levels of subjective wellbeing of those in care compared to their peers not in care, with those in residential care demonstrating the lowest levels of wellbeing.
Studying mental wellbeing requires the use of reliable measurement tools. The Warwick–Edinburgh Mental Wellbeing Scale (WEMWBS) was developed in 2007 [
20] and is one of the most widely used measures of mental wellbeing [
21]. It contains 14-items covering both psychological functioning and subjective wellbeing facets of mental wellbeing. A brief seven-item version (SWEMWBS) was subsequently developed using the Rasch measurement model which had preferable psychometric properties to the full version, though it is focussed more on functioning than subjective aspects of mental wellbeing [
22]. While measures of subjective wellbeing have been developed for young people in foster care, including specific care-related aspects such as birth parent contact [
23], a brief measure of mental wellbeing may be of particular use in population research where practical constraints often restrict the scope for detailed surveys.
It is often assumed that scores represent the same level of the construct for members of different groups. However, the nature and magnitude of relationships between items and a latent construct may differ across groups, meaning that comparisons between groups cannot be meaningfully made unless the measure is capturing the same thing in each sub-group [
24‐
26]. Thus, if we want to know if policies and interventions are working as well for children in care as for the rest of the population, we need to be able to measure this equally well in both groups. Testing for invariance of measures makes it possible to verify whether the members of different groups or cultures ascribe the same meanings to the items of a questionnaire [
27], which is critical for informing both practice and research [
28]. Most studies examining the psychometric properties of SWEMWBS have been undertaken in adults [
22,
29,
30]; the few studies conducted with adolescents have found acceptable measurement invariance properties by age and gender [
31,
32], and demonstrate good external construct validity [
32].
The present study
The aims of the present study were to: (1) Confirm the unidimensionality of SWEMWBS; (2) assess measure invariance of SWEMWBS across children and young people in care compared to their peers not in care; and (3) undertake comparison of mean differences in mental wellbeing across those groups. While other specific mental health measures, such as the Strengths and Difficulties Questionnaire [
33], have been examined across groups of care-experienced children and young people, to the best of our knowledge no study has yet examined the equivalence of SWEMWBS across these groups.
Results
Descriptive statistics for the sample can be found in Table
1. The response category proportions as well as item means, and standard deviations (data treated as continuous) can be found in Table
2. Across all items, it is worth noting that most children and young people positively rate items assessing their mental wellbeing. Visual examination of polychoric correlation matrices showed significant interitems correlation coefficients, ranging from 0.30 to 0.56 for the full sample (see Online Table S1). The matrices indicated increasing intercorrelation in the different care-experienced groups, with the strongest intercorrelations within the residential care group (ranging from 0.61 to 0.77). The scale demonstrated good internal consistency reliability across all groups (
α = 0.84 in full sample,
α = 0.82 ‘not in care’ group;
α = 0.86 in Foster Care,
α = 0.90 in Residential Care and
α = 0.81 ‘kinship care group).
Table 1
Sample descriptive statistics (N/%)
Gender |
Male | 1273 (46) | 506 (47) | 234 (46) | 59 (47) | 474 (44) |
Female | 1464 (52) | 571 (53) | 263 (51) | 53 (42) | 577 (54) |
Prefer not to say | 58 (2) | 9 (1) | 16 (3) | 14 (11) | 19 (2) |
School year |
Year 7 | 513 (18) | 199 (18) | 106 (21) | 25 (20) | 183 (17) |
Year 8 | 550 (20) | 213 (20) | 103 (20) | 25 (20) | 209 (20) |
Year 9 | 617 (22) | 242 (22) | 110 (21) | 28 (22) | 237 (22) |
Year 10 | 570 (20) | 210 (19) | 108 (21) | 33 (26) | 219 (20) |
Year 11 | 545 (20) | 222 (20) | 86 (17) | 15 (12) | 222 (21) |
Family affluence |
Low | 1022 (39) | 359 (34) | 150 (32) | 38 (35) | 475 (46) |
Medium | 822 (31) | 325 (31) | 143 (30) | 36 (33) | 318 (31) |
High | 810 (31) | 365 (35) | 179 (38) | 34 (31) | 232 (23) |
Ethnicity |
White British | 2307 (85) | 915 (86) | 412 (83) | 72 (60) | 908 (87) |
White non-British | 158 (6) | 46 (4) | 45 (9) | 16 (13) | 51 (5) |
Black and Minority Ethnic | 256 (9) | 98 (9) | 42 (8) | 33 (27) | 83 (8) |
Language |
English | 2691 (96) | 1029 (95) | 499 (97) | 117 (93) | 1046 (98) |
Welsh | 104 (4) | 57 (5) | 14 (3) | 9 (7) | 24 (2) |
Table 2
Item descriptive statistics
Item 1 “I’ve been feeling optimistic about the future” | 379 (13.56) | 562 (20.11) | 830 (29.70) | 678 (24.26) | 346 (12.38) | 3.02 ± 1.22 |
Item 2 “I’ve been feeling useful” | 320 (11.45) | 579 (20.72) | 964 (34.49) | 667 (23.86) | 265 (9.48) | 2.99 ± 1.13 |
Item 3 “I’ve been feeling relaxed” | 249 (8.91) | 510 (18.25) | 803 (28.73) | 806 (28.84) | 427 (15.28) | 3.23 ± 1.18 |
Item 4 “I’ve been dealing with problems well” | 323 (11.56) | 534 (19.11) | 804 (28.77) | 751 (26.87) | 383 (13.70) | 3.12 ± 1.21 |
Item 5 “I’ve been thinking clearly” | 247 (8.84) | 492 (17.60) | 832 (29.77) | 769 (27.51) | 455 (16.28) | 3.25 ± 1.18 |
Item 6 “I’ve been feeling close to other people” | 236 (8.44) | 419 (14.99) | 667 (23.86) | 783 (28.01) | 690 (24.69) | 3.46 ± 1.24 |
Item 7 “I’ve been able to make up my own mind about things” | 171 (6.12) | 300 (10.73) | 582 (20.82) | 841 (30.09) | 901 (32.24) | 3.72 ± 1.20 |
Factorial structure
Categorical confirmatory factorial analysis was used to test a unidimensional model in which the scores obtained for the 7 items of the scale all contribute to the evaluation of children and young people’s mental wellbeing. Considering the sensitivity of the chi‐square statistic to sample size [
54], we assessed a number of additional indices. Model fit was assessed to be adequate as despite the significant Chi-square all other indices showed excellent fit (
χ2 (df = 14) = 190.75,
p < 0.001; CFI = 0.988; TLI = 0.982; RMSEA = 0.067 [0.059-0.0.076]; SRMR = 0.024). Furthermore, all standardised factor loadings were statistically significant (
p < 0.001) and ranged from 0.548 to 0.814 (item 1 = 0.548; item 2 = 0.703; item 3 = 0.701; item 4 = 0.743; item 5 = 0.814; item 6 = 0.657; item 7 = 0.718), higher than the threshold of 0.5 [
43]. See Online Table S2 for item factor loadings.
Measurement invariance
Having verified unidimensionality of the SWEMWBS, we estimated a one-factor configural invariance model. Results from each of the successively stricter invariance tests are reported in Table
3. Configural invariance (baseline model) provided an acceptable fit to the data, (CFI 0.984, TLI 0.958, RMSEA 0.073,
p < 0.001), meaning the constructs had similar patterns of free and fixed loadings across groups [
46]. Metric invariance was subsequently tested whereby factor variances remained freely estimated but factor loadings were held invariant. Findings showed that the model fit the data well and there was no change in CFI (0.000) and a reduction in RMSEA (− 0.017), thus the items therefore loaded onto factors similarly across groups [
46]. Scalar invariance was then tested whereby indicator thresholds were now also held invariant, fit indices suggested acceptable fit with this constraint as indicated by little change to CFI (0.003) and a reduction in RMSEA (− 0.010).
Table 3
Measurement invariance tests of SWEMWBS across care status
Configural | 0.984 | 0.073 (0.068–0.082) | – | – |
Loadings | 0.984 | 0.056 (0.049–0.063) | No change | − 0.017 |
Loadings, thresholds | 0.987 | 0.046 (0.040– 0.053) | 0.003 | − 0.010 |
Additivea | | | 0.003 | − 0.027 |
Latent mean differences in mental wellbeing
Based on the establishment of scalar invariance across care status groups, latent mean comparisons can be made between care status groups. The NIC (not in care) group served as the reference group. Findings showed that young people currently in all types of care reported significantly lower mental wellbeing scores than those not in care (kinship care: Est = − 0.364 ± 0.051, p < 0.001; foster care: Est = − 0.319 ± 0.075, p < 0.001), and those in residential care reported the lowest levels of mental wellbeing; (Est = − 0.882 ± 0.192, p < 0.001).
Discussion
In this study, we analysed the short version of the WEMWBS in a population-based sample of school-aged students to explore measurement invariance and latent mean differences between young people currently in care (foster, kinship and residential settings) and those not currently in the care of local authority. The 1-factor CFA test showed that the SWEMWBS exhibited satisfactory model fit and demonstrated unidimensionality, thus this short single-factor instrument may be useful in reducing respondent burden in future studies. The current study established configural, metric and scalar invariances across care status groups, suggesting that differences in SWEMWBS scores between care status groups can be attributed to differences in the underlying latent trait rather than to the measure itself. Researchers employing SWEMWBS in future studies can compare the mental wellbeing scores meaningfully across those in different types of local authority care compared to their peers not in care.
Our findings revealed that young people in all types of care reported significantly lower mental wellbeing scores than their peers of the same age not in care. This result is consistent with findings from previous studies testing wellbeing scores using traditional methods [
17‐
19]. Research shows that developmentally specific factors including parents' availability and wellbeing, family relationships and interactions, quality of care, and supportive learning environments are critical for children’s wellbeing [
55]. Thus, given that the majority of children entered care due to abuse or neglect [
7] and the strong evidence base showing the long-lasting impact of early trauma and adversity [
56‐
59] we suggest that the lower mental wellbeing of those in care may be due to early negative experiences prior to or during care; however, we do not have the data available to test this assumption. Given that the purpose of the care system is to address these factors by: protecting children from further harm, addressing a child’s need for good parenting, and enabling them to recover from traumatic experiences [
9], further work to promote mental wellbeing of young people in care is needed. A scoping review [
6] highlighted a number of interventions which may be beneficial in improving children and young people’s wellbeing. The review also highlighted a decreasing emphasis on wellbeing as children grew into teens or young adults, with more interventions available for this age group which intervene in the development of mental illness rather than promoting wellbeing.
Our analysis has several strengths, but also several limitations. Our large-scale nationally representative sample provides evidence of the utility of SWEMWBS for measuring mental wellbeing among young people in care in the UK. A limitation of this study is the lack of testing for invariance across other categories, such as gender, family affluence and ethnicity. Previous studies have shown that SWEMWBS demonstrates strong measurement invariance across sex and age differences in adults [
60] and a further study showed measurement invariance across the full age range of secondary school students [
32]; however, as previous research has shown that these factors are all connected to strong structural inequities [
61], future work should address this limitation by testing for invariance across family affluence and ethnicity. Furthermore, it is possible that results from Wales may not generalise internationally, though evidence of the psychometric properties of SWEMWBS in adults is consistent across multiple cultures [
62]. Self-reported data may have been biased by standard limitations (e.g. memory recall biases, social desirability, etc.). While the living situation question enabled us to identify that the ‘kinship care’ group are living with family other than their parents we cannot be sure if they are subject to a formal care order. As the SHW survey is only completed by young people in mainstream schooling, the views of children not in mainstream school are not included, this is particularly significant given that approximately 40% of children in care attend non-mainstream schools such as special schools, pupil referral units and alternative provisions [
63].
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.