Context

The International Mobility in Aging Study (IMIAS) is a prospective cohort study with the primary objective of examining how events across the life-course affect the health, and particularly, the mobility, of people age 65–74. By recruiting equal numbers of men and women across four developing and developed countries the differential impact of gender and social circumstances on health can be studied. Each setting has a relatively homogeneous older population, however there is a high level of socioeconomic and cultural variability across sites. Tirana, a city of approximately 700,000, is the capital of Albania, a post-communist country and one of the poorest in Europe. Natal in coastal northeastern Brazil, is a provincial capital with a population of approximately 800,000. The research setting in Colombia was Manizales (population 400,000), a mountainous city in a coffee-growing region. Kingston, Canada is a small, primarily English-speaking city of 110,000 with a stable, non-industrial economy while St. Hyacinthe, also in Canada, is a relatively traditional Francophone community of about 50,000 and with an agricultural economic base. Canada is a developed and relatively high-income country while the other three are considered middle-income settings.

Population and Sampling

This analysis uses baseline cross-sectional data from IMIAS, collected in 2012. Community-dwelling men and women, age 65–74 at the time of the initial survey (2012) were recruited via primary care providers’ rosters until 200 women and 200 men were enrolled at each site [Kingston (Ontario, Canada), St. Hyacinthe (Quebec, Canada), Natal (Brazil), Manizales (Colombia) and Tirana (Albania)].

At Canadian sites ethics guidelines precluded approaching perspective participants directly. Instead recruitment letters were sent via a sample of urban family physicians to all patients who met age criteria and lived independently within the city. Letters invited interested recipients to phone a research coordinator. This process yielded a response of approximately 30%. Of those who initiated participation in this manner, approximately 95% were enrolled. At the other sites ethics approval was received to approach randomly selected clinic patients directly, yielding a participation rate of greater than 90% (Colombia) and almost 100% (Brazil and Albania). All respondents were screened for cognitive ability and excluded if they had 4 or more errors on the Leganes Cognitive Test [14].

Participants were interviewed and asked all questions in the appropriate language in their homes, unless they requested a different setting. Questionnaires were available in English, French, Portuguese, Spanish and Albanian. Interviewers across all sites received standardized training by the study coordinator and/or principal investigator. Scales used within the questionnaire were all previously validated. Translations were checked for accuracy via ‘back’ translation and scale translation was validated in pilot studies in Brazil, Colombia, and Canada.[15] All responses were entered immediately by the interviewer into a computer-based program and subsequently reviewed by one of the investigators to check for possible entry errors.

Data Collected

Embedded in a detailed survey of demographics, social circumstances, health status, subjective and objective measures of mobility, physical health, mental function, cognitive status, etc were the following questions of relevance to this paper:

Did your mother, father or both parents die during the first 15 years of your life?

During the first 15 years of your life, would you say that there was a time in which you did not eat enough and that you were hungry?

Did you ever witness physical violence between those close to you (such as between your parents, or your parents and siblings)?

To what extent does your income allow you to meet your needs?

Self-rated health was measured as excellent, very good, good, fair or poor and then dichotomized with the first three categories being collapsed as “good” and the latter two as “poor”. The positioning of the question re SRH within the overall survey was the same for all participants.

Statistical Analyses

Using SPSS (v21) we examined Spearman correlations then built a logistic regression model using dichotomized indicators with SRH as the outcome and first with the 3 adversity measures alone then subsequently controlling for current income sufficiency. The model was then adjusted for other measures of early adversity (divorce of parents, childhood health, physical abuse in childhood). The same regression analyses were then repeated disaggregating women and men. As the number of participants with missing data was less than 1% they were excluded from relevant analyses. P-values < 0.05 were considered statistically significant.

Ethics/consent

The IMIAS project received ethics approval from the Research Centres of the University of Montreal (CR-CHUM), the Albanian Institute of Public Health, and the Research Ethics Boards of the Federal University of Rio Grande do Norte (Brazil), Queen’s University (Kingston) and University of Caldas (Columbia). All study participants provided written informed consent and were told they could withdraw at any time. In Brazil and Colombia, for participants who could not read and write, the consent form was read to them, oral consent was obtained, and the participant placed a sign on the form to acknowledge agreement.

Possible limitations

SRH is a subjective measure but one that has repeatedly been shown to correlate well with actual health [16]. It is a measure that is robust across culture, place and age [17,18]. In general women tend to report poorer health than do men however these gender differences reflect disparities in ability and function [19].

Although recall bias is a potential risk in asking about events that occurred 5 or 6 decades earlier, our main indicator, loss of a parent, is unlikely to be forgotten, whereas early hunger or witnessing violence, and current income sufficiency are more subjective.

With any quantitative research the question of underpowered or overpowered arises. Did we find significant differences only because of a large sample size or miss meaningful differences because the study was underpowered? The lack of previous research on parental loss and SRH in older populations precluded power and sample size planning.

IMIAS is a descriptive study therefore causal inferences cannot be drawn. It would be erroneous to generalize from our findings, which should be viewed as preliminary. Evidence from more settings and across a wider age range is needed before any suggestion of causality would be reasonable. To determine whether IMIAS participants are representative of their communities we compared their socioeconomic and education levels with available data for each city. The Kingston cohort, and particularly women within it, had more education than did non-participants of the same age range in their community. Latin American participants were representative of the SES and educational status of their cities. Never the less, because the populations sampled may not be representative, we have avoided analyzing site-specific data or claiming that findings have external validity.

Interviews in homes meant that in settings outside Canada there were often family members present. As a result, respondents may have been guarded about describing, for example, current family violence. However, the responses about childhood experiences analyzed in this paper are unlikely to have been answered differently had there been privacy.

The multi-country nature of the study ensures socioeconomic and cultural diversity among participants. However, this variability can only be accurately captured in analyses within measures of specific characteristics. Site specific reporting has been avoided for several reasons. We cannot assume that samples are representative of each community’s population nor can we make inferences about possible community characteristics that give rise to or mediate self-rated health. Furthermore, this study lacks the power to assess associations between parental loss and SRH by site.

Strengths and Interpretations

In bivariate analyses each of the three ACEs predicted poorer self-rated adult health, as did insufficient current income. Death of a parent remained a significant predictor after controlling for other ACEs and for current income sufficiency. Early death of a parent may rob a child of nurturing, stability, a sense of normalcy, and the income that parent would have provided. Although each of these deprivations, alone, can have a significant and long-lasting impact on mental and physical health, to date, parental loss has not specifically been studied as a predictor of adult self-rated health. Our study indicates that death of a parent in childhood may create a real and ongoing wound. Its negative health impact is only minimally ameliorated by income sufficiency in later life. This would suggest that parental loss may be an independent predictor of SRH decades later and that its effect is not mediated via current socioeconomic status.

The impacts of parental loss and early hunger on later health are both statistically significant but different. Those with adequate intake in childhood are more likely to rate their current health as good (63% vs 57% of the whole cohort) while early hunger predicts poorer SRH than that described by the cohort, overall (69% vs 43%). On the other hand, growing up with 2 parents alive does not particularly improve adult SRH (59% of good SRH among those with two parents alive vs 57% overall) while loss of a parent is clearly associated with poorer health (54% poor SRH among those with parental loss vs 43% of total cohort). Data, alone cannot explain why having two parents in childhood does not confer measureable benefits while parental loss translates into poorer lifetime self-rated health. It may be that in the face of this adversity some children bounce back and thrive, that is, they develop a resilience that their two parent counterparts lack, and that this resilience mitigates the health harms of parental loss. The health detriment of early hunger may be more direct than that of a parent’s death and not corrected by developing resilience. This could explain why the absence of hunger bestows a health benefit not seen for the absence of parental death.

The impact of ACEs on adult health was markedly different for women and men. A disproportion of women report poor SRH overall, in keeping with findings across other studies and settings [19,20]. However the relative impact of early parental loss on later SRH is much greater for men than women, whereas both suffer significant long-term consequences of early hunger regardless of current socioeconomic status. The gender difference in long-term health is in keeping with findings of increased cortisol responses among men relative to women aged < 55 following early parental loss [21]. We hypothesize that the harmful impact of this particular adversity is modified by unidentified individual and/or social circumstances that, themselves, vary by gender. It is not early parental death, alone, that writes the story of harm on the body foecades to come but the interplay of this loss and individual and socially mediated strengths, constraints, expectations and opportunities of those who suffered the loss. Some unidentified circumstance(s) operating in conjunction with but not limited to death of a parent, appear to decrease the harmful impact on women’s health and/or magnify it among men. Our data cannot further define this gender difference. It may be that young men are more susceptible to the loss of nurturing that parental death brings and/or that women are more able to translate loss into inner strength and resilience [21]. As Krieger states in describing eco-social and embodiment theory: “how people resist injustice and its health-harming effects, individually and collectively, and the resilience that enables them to do so also must be examined” [5].

Existing research on gender differences in ACEs is limited but suggests that women tend to be more prone to later harm from early adversity [22]. The somewhat greater long-term effect of early hunger on health among the women in our cohort is in keeping with this. However, loss of a parent in childhood is not among the ACEs previously studied. Our results provide preliminary evidence that the harm to later health arising from death of a parent may be mediated by a different set of gendered characteristics than those described for other ACEs.

This study is the first to report that parental loss should be considered as a childhood adversity with lifelong consequences for self-rated health and, further, that these consequences are and can be ameliorated or modified [23]. By identifying what decreases the impact of loss on subsequent health for women relative to men it may be possible to develop individual or community-based interventions that augment those beneficial characteristics in general. The observed gender differences on health long after parental loss suggest that childhood adversities do not inevitably harm but that other aspects of one’s life, environment and circumstances modify and mitigate this association. What those aspects are remains unidentified. We refer to them collectively as resilience and in subsequent IMIAS data collection will include measures of current resilience. However such measures give only an immediate snapshot of characteristics when resilience measured at a time more proximal to the loss itself might be of greater explanatory value. The traits that build resilience vary over time and most meaningful would be measures of these done much closer to the time of parental loss. Such measures cannot be collected retrospectively. This makes the argument for including measures of resilience in prospective, longitudinal population-based health surveys to actually identify factors linked to resilience and whether their presence moderates the ills associated with adversity.

In summary, we have documented the significant and long-lasting harm to self-rated health associated with loss of a parent in childhood. The health impact of this loss is greater among men than women. This gender difference suggests that the detrimental effect of death of a parent in childhood may be modified by individual and/or social traits.