Adolescent reserve capacity, socioeconomic status and school achievement as predictors of mortality in Finland - a longitudinal study

Study design

A longitudinal study was designed linking two data sources by means of unique national personal identification numbers. Baseline data were obtained from the Adolescent Health and Lifestyle Surveys (AHLS) of 1985, 1987, 1991, 1993 and 1995. Nationally representative samples of 12-, 14-, 16-, and 18-year-old Finns born on certain days in June, July and August were drawn each study year from the Population Register Centre. Overall response rate was 79% (N = 41,833), with 72% (N = 19,509) for boys and 86% (N = 22,324) for girls, respectively. A self-administered questionnaire was sent in February, followed by two re-inquiries to non-respondents. The variables used in our study were comparable across all survey rounds.

Follow-up data and information on family SES were respectively obtained from the Finnish Official Cause-of-Death Register and from the Register of Completed Education and Degrees, containing statistics on every resident in Finland. The follow-up started on 30 April, each survey year, and ended 31 December, 2009, or when the participant died. Average follow-up time was 18.4 years. It ranged from 1 to 25 years and had a total of 770,161 person-years. At the end of the follow-up, the participants were aged 27 to 43 years.

Statistics Finland performed the data linkage of the national registries and the AHLS data according to a contract specifying the rights and duties of both parties. The study protocol was approved by its Institutional Review Board and the Data Protection Ombudsman. The Joint Commission on Ethics of the University of Turku and the Turku University Hospital stated that no human rights were violated in the research protocol and approved it. Identification of the study participants was withheld from the investigators at all stages of the study. The first review boards at the universities were established in Finland in the 1980s. AHLS was reviewed by the Ethical Review Board of the University of Helsinki, Department of Public Health in 1986. Parental consent was not considered by the ethics review board at that time. In later surveys, the latest in 2017, the relevant review boards have waived the parental consent.

Outcome and predictor variables

Family SES was based on parents’ education from Statistics Finland categorized into basic, secondary and high. Data was obtained nearest to the year when the adolescent was aged 15 years. If parents belonged to different categories, the highest was selected. If one parent was missing (2%), the available parent’s data was used.

For school achievement, adolescents were categorized as having: highest, 2nd highest, 2nd lowest or lowest academic achievement. The respondents were asked to assess whether their end-of-term school report was much better, slightly better, average, slightly poorer or much poorer than the class average. For 12–14-year-olds (all in comprehensive schools), the last two were combined. For 16–18-year-olds, the first two were further combined and school status (high school/vocational school/not attending school) was additionally used. Respective categories included: high school, better than class average; vocational school, better or high school, average; vocational school, poor to average or high school, poor; and, not at school.

Statistical analysis

Cox proportional hazards models, stratified by sex, were fitted to determine the relationship of predictor variables with mortality and calculate hazard ratios (HR) with 95% confidence intervals (CIs). Adherence to the proportional hazards assumption was checked using log-log survival curves and a formal significance test based on the unscaled and scaled Schoenfeld residuals [30]. First, a crude model, which considered family SES, each reserve capacity dimension and school achievement, was fitted to analyse each predictor’s unadjusted effect on mortality risk (Model 1). Then, to study whether the reserve capacity variables modified the relationship between SES and mortality, all statistically significant (p < 0.05) reserve capacity variables together with SES were included in a backward selection procedure until none could be deleted from the model (Model 2). Finally, school achievement was added (Model 3). An interaction term between family SES and school achievement was also tested. Model fit was assessed using likelihood ratio tests and Akaike Information Criteria (AIC) [31]. Post-estimation tests were done (checking of residuals and other plots) to ensure that the final model had the best fit. Respondents with missing data (5%) in one or more main variables studied were dropped from analysis. All analyses were performed using STATA version 12.1.

Characteristics of the study population

Table 1 lists the detailed characteristics of the total population according to family SES, dimensions of reserve capacity, school achievement and outcome status. Less than one-fourth of boys and girls had low family SES. Generally, most adolescents had positive reserve capacity characteristics but boys and girls differed in terms of perceived stress symptoms and tooth brushing frequency, which were more common among the girls, and ease of talking about issues to father, which was more common among the boys. High achievement in school was also more common among the girls compared to boys.

Among boys, with 358,787 person-years of follow-up time (mean 18.4 years), mortality rate was 10.1 per 10,000 population. In contrast, mortality rate among the girls with 411,373 person-years of follow-up time (mean 18.4 years), was lower (p < 0.001) at 3.3 per 10,000 population.

Predictors of mortality in adolescent boys

Accounting for reserve capacity significantly reduced the effect of low SES on the risk of death, more so when school achievement was controlled for. Among boys whose parents had secondary education, HR estimates decreased by almost 19%, from 1.6 (95% CI 1.1–2.4) to 1.3 (95% CI 0.9–1.9). Total reduction in HR estimates was greater (27%) among those whose parents had basic/lower education, from 2.2 (95% CI 1.5–3.3) to 1.6 (95% CI 1.1–2.4). Interestingly, HR estimates for reserve capacity did not change markedly even with adjustment for the effect of school achievement.

Predictors of mortality in adolescent girls

In the crude model (Model 1), the lowest SES category, although not statistically significant, showed an inverse relationship with mortality (HR 1.4, 95% 0.8–2.4). However, this effect was diluted and HR estimates became null when reserve capacity and school achievement were taken into account. Similar to results seen in boys, HR estimates for reserve capacity did not change markedly even when school achievement was added into the model.

Summary and interpretation of results

Our study found that family SES in adolescence significantly predicted risk of death only in boys. Among reserve capacity dimensions, poor perceived health (presence of chronic disease and weekly stress symptoms in boys; poor self-rated health in girls) as well as reduced social support (difficulty in talking to father in both groups; not living in a nuclear family in boys) generally increased the mortality risk of adolescents. Poor health-promoting behaviour (poor oral hygiene) increased the risk only in boys. Adolescents with low school achievement had 1.6–2.3 times higher risk of dying compared to the highest achievers. Reserve capacity and school achievement independently mitigated the effects of low SES on mortality risk among boys.

Family SES was related with boys’ mortality risk in adolescence and early adulthood in our study. In Finland, previous research also revealed that health inequalities in adolescence and early adulthood persisted in boys from low SES environments possibly due to risky living standards and lifestyle-related factors [32, 33]. Likewise, studies on adult SES measures and outcomes presented stronger effects of SES on mortality for men relative to women because of underlying gender roles and other social characteristics [6, 10]. Typically, though, socioeconomic differentials in morbidity and mortality were recognised as less salient in the adolescent population compared to adults due to a certain level of “equalisation” of risk exposures [32, 34].

Our findings showed that all reserve capacity dimensions significantly predicted mortality risk in boys. Among girls, similar results were observed, except for health-promoting behaviour. A particular study which found difference in psychosocial resources between teenage boys and girls used a different dimension from those analysed in our study [28]. Thus, we cannot conclusively say that there are gender differentials in reserve capacity. Moreover, most epidemiological studies which dealt with reserve capacity’s role in SES-health inequalities controlled for the effect of sex and combined results for both groups [17, 34, 35].

Since poor health perceptions are usually influenced by the presence of co-morbid conditions and symptoms [36], we included these along with self-rated health in the perceived health dimension. Studies have shown that perceived health was strongly and independently associated with mortality, even after controlling for known risk factors [36, 37], and objective physician ratings [38]. Researchers have explained that this indicator may have a summative property of capturing health aspects relevant to survival which are not measured by other health indicators [37]. In adolescence, health perceptions also reflect one’s overall sense of psychosocial functioning aside from physical health [39]. Based on our results, changing self-perceptions of health and alleviating stress symptoms might improve both psychosocial and physical functioning in adolescence.

In our study, physical activity was not associated with the risk of death. Perhaps, this was because among those who died and regardless of their SES, both boys and girls were physically active in their adolescent years. Such health-promoting behaviour is usually adopted early in life [20] and further reinforced by school environments [40]. However, lack of health-promoting behaviour in terms of poor tooth brushing habits, was associated with boys’ mortality risk. The girls in our study generally had good dental behaviour, hence, there was little variation in the distribution of exposure, unlike in boys. Tooth brushing behaviour, also formed during childhood, probably reflected family conditions, such as how well parents provide care and monitor their children’s health behaviour, to some extent [22].

Research on the effect of social support on mortality was extensive. A meta-analytic review showed that overall effect size of being in social relationships provided up to a 50% increase in odds of survival [24]. In our study, important aspects of social support were related to family structure and communication with father. Researchers have recognized that a “risky” family environment early in life predisposed children to various emotional and physical disorders [9, 17, 34]. In a study among Hungarian adolescents, a non-intact family structure was a significant determinant of risky health behaviours such as use of cigarettes, alcohol and marijuana [34]. Our results showed that poor communication with one’s father increased the mortality risk of adolescents. However, the mechanisms by which communication with one’s father influences health during adolescence is beyond the scope of our study. Nonetheless, our results, comparable to earlier findings [41], underscore the importance of paternal relationship as a form of social support. This is congruent with evidence that showed children had less emotional and behavioral problems with father’s involvement during childhood and adolescence [42].

School achievement also significantly predicted the risk of death in both genders in our study. Previous studies showed that school achievement in adolescence empowered a person to make healthy choices and adopt healthy habits [25, 26]. It also ensured completion of high school education, often leading to a college degree, greatly improving one’s future SES [8, 26]. In our study, increasing mortality risk in boys was estimated with each category below the highest achievement. In girls, only the lowest category was significantly related to risk of death. The lack of interaction between family SES and school achievement implies that both education-related variables exhibit a similar and expected gradient with mortality.

As shown in literature [11, 17, 19], our results demonstrated that reserve capacity reduced the effect of low SES on mortality risk among boys. Interestingly, the addition of school achievement into the model further weakened the effect of low SES on boys’ risk of death. Yet, it did not modify the risk estimates obtained from the reserve capacity dimensions, suggesting that these factors are important predictors which independently affect mortality risks in adolescents. The results of our study lend further support for the life-course approach to the SES-health relationship.

Strengths and weaknesses

Most studies have utilized either childhood or adult markers of SES. Adolescent indicators are seldom used, even though adolescence is a critical period in developing sound psychological and behavioural patterns, which are carried forward into adulthood [28]. Our prospective study addressed this research gap using large, nationwide samples with a long follow-up period and reliable register-based data. Our study added support to the importance of the life-course approach in epidemiologic research on SES-health inequalities.

Studies which dealt with a reserve capacity framework among adolescents were limited. The opportunity to combine survey data with register-based data on death made it possible to build a longitudinal dataset and study potential psychosocial factors mediating the SES-health gradient. Since the survey data was collected in the 1980s and 1990s, it was not designed to measure dimensions of reserve capacity. Due to this, we needed to use proxy measures for each reserve capacity dimension. The selection of variables was based on a cluster of single-item indicators which correlated with each other. However, proxy measures may give unreliable results and further research is needed to validate these.

Despite issues in measurements, we tried to analyse a wide range of reserve capacity dimensions. This follows the methodological framework of the proponents of reserve capacity who emphasised that it is “a bank of resilient resources that contributes to the SES and health relationship” [9, 17–19] instead of a single psychosocial factor or dimension. Moreover, we presented results disaggregated by sex, providing evidence to the interconnections of SES, gender and health inequalities.