Self-reported social class in adolescents: validity and relationship with gradients in self-reported health

Measurement of social difference in health of adolescents presents challenges not present in the study of social gradients in health among adults. Reports of socio-economic position by adults are assumed to be accurate but the accuracy of child reports is unknown [1, 2]. Indicators based on occupation, education or income have been used to assess socio-economic position [3, 4]. With the mass entry of women into the labor market and with the emergence of new family structures, determining the socio-economic status of children and adolescents on the basis of the father's occupation may no longer be an adequate representation of family socioeconomic position. In self-administered adolescent questionnaires, up to 40% or responses on paternal occupation are invalid or missing, which has prompted the development of other measures of material well-being that are easier to elicit [5‐7].

The accuracy of responses has been positively associated with age and the nearness in time and space with the parental characteristics being investigated [8]. The level of agreement between adolescents and proxy-respondents reporting education and occupation has been shown as moderate [9, 10] improving with age, and worsening when the adolescent does not live with the parent about whom information is being asked. No gender differences were found.

In terms of analyzing social differences in health, the measurement of adolescent health is a challenge. In the last decade a number of questionnaires have been developed which attempt to measure self-perceived health and health-related quality of life, which have proved to be sensitive to social inequalities. The Child Health and Illness Profile – Adolescent Edition (CHIP-AE) [11] is a generic self-administered questionnaire developed in the United States and which has been adapted in Spain, with acceptable reliability and validity [12, 13].

The objectives of this study were to examine the validity of measuring occupational social class (OSC) and family level of education (LE), reported by adolescents aged 12 to 18, and the relationship between social position and self-reported health.

A total of 1453 adolescents was included in the final analysis (response rate = 82%, average age 14.9). 259 adolescents were absent on the day of questionnaire administration and in 62 cases, parents refused permission for their children to participate in the study. 60% were aged less than 16; 51.2% were boys; 60.1% came from the urban sample of which 17.3% lived in a neighborhood with a low ICEF (table 1). Eighty percent lived in a standard family with both parents. In 50.7% the highest level of education in the family was primary, and 54.5% belonged to the least privileged social class (IV–V).

Responses were received from 91 of the 160 proxy-respondents (response rate = 56%). Mothers were the principal informants (83/91). Distribution of socio-demographic variables for the sub-sample of adolescents whose proxy-respondents were interviewed and the rest of the adolescents were similar, except that there were a higher proportion of families with a secondary level of education and with paternal OSC IV–V in the sub-sample of adolescents whose parents participated in the study. No differences were found in adolescents' health status between participants and non-participants.

The present study has attempted to help researchers to decide on which indicator of socioeconomic position could be used in adolescents, and to analyze the influence of socioeconomic position on self-perceived health. Occupational social class is an indicator widely used in adults while adolescents are in general excluded from this analysis.

Although one in seven adolescents didn't allow to assign the family's occupational social class, overall, adolescents' reports provide valid information about their family socioeconomic status. The response to the question about the LE of both parents was more exhaustive, but with a low level of agreement between adolescents and their mothers, when compared with the OSC.

The percentage of missing responses about parental OSC and LE are similar to those of other studies [9, 10]. Girls answered more exhaustively than boys, although for those over 15 this only happened in relation to questions covering LE. As a consequence, younger adolescents could answer questions about OSC in an acceptable manner. Questions about OSC presented more missing responses, although the calculation of the variable on family OSC could still be determined in 86% of cases. Non-standard families were associated with a higher frequency of missing response about paternal OSC and LE, which indicates the importance of living with the parent about whom information is being collected.

In this study, contrary to the finding of Wardle [7], there was no relation between the economic level of the urban area where the school was located, as measured by the ICEF, and missing answers to the questions on OSC and LE. The kappa values found in this study were above 0.50 except for the question about paternal and highest education in the family. These values were similar to those found by Lien [9] and somewhat lower than those of Ensminger [10]. In the present study the agreement between adolescents and mothers was not influenced by the age or sex of the adolescent, although the type of family influenced the agreement on the paternal OSC. Overall, the percentage of agreement about occupation were higher than that about other family characteristics such as the presence of smokers in the family, the presence of physical or emotional problems, or the declaration about lifestyle and diet (data not shown). From this it may be concluded that adolescents are better informants about OSC than about other characteristics of family life.

Adolescents under-estimated the parental LE (for the adolescents 70.3% of mothers had only primary education, vs. 52.7% for proxy-respondents; and 69.2% of fathers against 53.8%, respectively), which in part explains the low level of agreement compared with that of occupation. This finding would tend to support the theory of Looker that adolescents respond better about parental characteristics that are near in time, such as occupation, than about more distant ones. Changes in administrative structure, and academic reforms that have occurred in Spain in the past could also influence the lower validity of the answers about LE.

The most valid indicator of socio-economic position collected in this study seems to be paternal OSC, which presents high kappa values and is little influenced by socio-demographic factors which affect the adolescents' answers. However, due to the completeness of responses and a moderate kappa value, maternal LE could be a good indicator of those aspects of socio-economic position that also influence the self-perceived health of adolescents. For the purpose of collecting information about socio-economic position in adolescents it could be recommended to use father's occupational social class as a first choice and maternal level of education as the second.

The moderate social gradient of some aspects of self-perceived health among adolescents was found. This finding is consistent with the studies of others[10, 18] although the US studies found a clear gradient in the Satisfaction and Discomfort domains.

Family type was remarkably important in terms of the relation between socio-economic status and self-perceived health, in addition to the known differences by gender and age. Those from standard families presented better scores in all the CHIP-AE domains and sub-domains, similar results to the USA studies [10, 18]. The absence of gradient in Satisfaction and Discomfort domains might be due to the difference in the percentage of adolescents in the Spanish sample living with both parents (80% in the present study vs. 47% in the Ensminger study). Furthermore, single-parent families showed a significantly higher percentage of mothers with university studies, living in urban areas with a high ICEF, which differs from the American sample. The Resilience domain was the most sensitive to differences in OSC and LE, with a significant gradient in the majority of the sub-domains. In the Risks domain, differences were observed between adolescents from families with university qualifications (maternal, paternal or highest) and the rest, although the worst scores were for families with secondary education. Adolescents from families with a better socio-economic position, especially those with a high LE, generally presented more protective factors and lower level of risks than for other families. A social gradient was also noted in academic performance. Better academic achievement in high socio-economic position adolescents indicates the high likelihood of inter-generational transmission of social position. Paternal, maternal and highest LE showed a statistically significant relationship in more domains than the OSC gradient. These results support the theory that social and human capital are both important factors in developing a healthy life[10].

The main limitation of this study is the small sample size in the sub-sample of adolescents and proxies, although it has been possible to establish comparisons between the indicators reported on by adolescents and their mothers. No statistically significant differences were found comparing data of adolescents whose parents did not participate with that of those who took part in the study, except for a higher participation by families with maternal secondary LE. We were not able to analyze criterion validity, as it would have been necessary to collect information on self-reported OSC and LE of both parents to obtain a gold-standard with which to compare the answers from the adolescents. Furthermore, in the current study the kappa values may have been influenced by the high prevalence of the level of primary education in the sub-sample [19]. Most cross-sectional studies are conducted in school settings, and adolescents are the only source of information, while parents are rarely involved the opportunity to study. Therefore, socioeconomic gradients in health and health services would be lost in these studies unless we rely on self-reported data by adolescents. Nevertheless, it is worth noting that information from both adolescents and their parents should be collected whenever this would be possible. The present analysis does not allow us to know to what extent missing answers could be from adolescents who avoid answering what they might consider embarrassing replies [8]. Nevertheless, this factor may have little influence on the results, given that other sections of the questionnaire with less percentage of missing answers can be considered more embarrassing than the socioeconomic status (e.g. risky and sexual behaviours).

The absence of significant differences in the domains of Satisfaction and Discomfort could be related to specific characteristics of the sample, particularly the small number of individuals in some cells means. That is, gradients were observed but the number of observations could have been too small to achieve significance. Finally, it should be noted that ethnic and family composition was more homogenous when the questionnaire was administered than currently in Spain. This is due to the fact that in recent years the percentage of immigration in Europe in general and Spain in particular has increased, as the percentage of different types of family. It may be that more recent administration of the survey would result in different findings.

Adolescents seem to be acceptable informants on family OSC and LE. Social gradient in health also exists among adolescents. Health policy measures should take into account social class inequalities in health during adolescence.