Association between number of children and incident heart disease and stroke in parents – results from the Survey of Health, Ageing and Retirement in Europe (SHARE)

Sample

We used data from the Survey of Health, Ageing and Retirement in Europe (SHARE) [19]. This longitudinal, harmonized panel contains data on health, socio-economic situation, and social networks of elderly people in 28 European countries and Israel. Our analyses were based on data from waves 1 to 7 of this survey, which were collected between 2004 and 2017. We used participants’ answers from the interview on children in each wave to define the exposure. The outcome was defined using information from the regular SHARE interviews on physical health and—in case of death of a participant – the standard SHARE end-of-life interview with the participant’s relative in waves 5 to 7.

The working sample consisted of 133 198 participants with at least one valid interview in waves 1 to 7 (except for wave 3). Participants without information on the number of children were excluded from further analyses (n = 18 677). After that, all participants were excluded who reported at study entry having had HDS (n = 17 543), taking heart medication (drugs for high blood cholesterol, high blood pressure, coronary or cerebrovascular diseases, and other heart diseases) (n = 38 563), or who had given an invalid answer regarding this (n = 441) when first answering the question. After further exclusion of subjects with no follow-up data about HDS (n = 15 899), the study population consisted of 42 075 subjects (18 080 men; 23 995 women) aged 24 to 102 years (Supplementary Fig. 1). Of these, only individuals without missing values on marital status and with an educational level according to the 1997 International Standard Classification of Education (ISCED 97), codes 0 to 6, were considered in the main analysis (n = 41 699).

Heart disease and stroke

The outcome ‘heart disease and/or stroke (HDS)’ was self-reported and assessed via questionnaires. All participants were shown a list of 16 health outcomes and asked at wave 1: ‘Has a doctor ever told you that you had any of the conditions on this card?’. In waves 2 to 7 (except wave 3), the question was changed as follows:’Has a doctor ever told you that you had/Do you currently have any of the conditions on this card?’ The outcome HDS was defined as ‘heart attack including myocardial infarction or coronary thrombosis or any other heart problem including congestive heart failure’ and/or ‘stroke or cerebral vascular disease’. In addition, in the case of the death of a participant, in waves 5, 6, and 7, end-of-life interviews were conducted with family or household members who were asked: What was the main cause of [his/her] death? Variables were coded dichotomously, meaning with or without disease.

Number of children

The exposure ‘number of children’ was self-reported by one family member. The family respondent was asked the following question: ‘How many children do you have that are still alive? Please count all natural children, fostered, adopted and stepchildren.’ The first report of the number of children was taken as the exposure variable and adopted for the partner. Partners of the family respondent were identified using a created CoupleID. Depending on the analysis strategy, the self-reported number of children was used continuously, categorized as none, 1, 2, 3, 4, 5, 6, and 7 + children, or categorized in groups of 1–2, 3–4, and 5 + children. The 2-child and 1–2-child groups were used as references because the incidence of HDS was lowest in this group.

Covariates

The selection of covariates for statistical adjustment was based on previous literature [20,21,22,23,24]. All covariates were self-reported by the respondents in baseline and follow-up interviews. Covariates included baseline age, sex, education, region, marital status, age at first birth, and health status until age 15. For descriptive analyses, respondents were categorized into one of seven age categories (< 35, 35–44, 45–54, 55–64, 65–74, 75–84, ≥ 85) based on their reported age. Since both exposure and outcome were expected to be linearly related to age, baseline age was continuously included in the model. All analyses were sex-stratified.

The level of education reported by respondents in the baseline interview was categorized according to ISCED 97 and included in the main model (code 0–6). For further analyses, education was grouped into: ‘primary’ (pre-primary and primary education), ‘secondary’ (lower secondary, upper secondary, and post-secondary non-tertiary education), and ‘tertiary’ (first and second stages of tertiary education). Subjects with ‘no degree or other’ were excluded from statistical analyses (n = 190).

The final study population consisted of participants from 20 countries. In the model, these countries were categorized into four regions: ‘Northern Europe’ (Sweden, Denmark), ‘Southern Europe’ (Spain, Italy, Greece, Israel, Portugal), ‘Western Europe’ (Austria, Germany, Netherlands, France, Switzerland, Belgium, Luxembourg), and ‘Eastern Europe’ (Czech Republic, Poland, Hungary, Slovenia, Estonia, Croatia).

The first report of marital status was categorized into three groups for analysis: ‘married’ (married and living together or separated from spouse or registered partnership), ‘never married’, and ‘divorced or widowed’.

The variable ‘age at first birth’ was calculated from the first valid information of the year of birth of the first child, and the year of birth of the participant given in the baseline interview. If the age at first birth was less than 12 years (n = 119) or if a person reported no children but still had a child’s birth year noted (n = 296), it was set as missing. For descriptive analyses, the variable ‘age at first birth’ was categorized into < 20, 20–29, 30–39, 40–49, and ≥ 50 years and continuously included in the model.

Statistical analyses

Baseline characteristics were reported for the whole study population and separately for men and women. Sex-specific risks of incident cases of HDS were reported by number of children, by categories of baseline age and age at first birth, by health status until age 15, by education, marital status, and region. The frequency of reporting different numbers of children during the study period was calculated.

To examine the association between number of children and parental incident HDS logistic regression analyses were performed and odds ratios (OR) with 95%-confidence intervals (95% CI) estimated.

To control for potential confounding, two models were conducted in addition to the unadjusted crude model: Model 1 adjusted for baseline age and sex; and Model 2 (main model) additionally adjusted for education, region, and marital status. There were 11 943 and 8 370 missing data for age at first birth and health status until age 15, respectively, but these variables were potential confounders. However, adding these variables to Model 2 led to a change estimate of < 5%, and therefore were not included in the adjustment set. All models were additionally stratified by sex. Having two children was used as reference. Further sensitivity analyses were carried out. Model 2 was also calculated with people included in the analyses who were only taking medication for high blood pressure or elevated cholesterol to check that no large group of people was excluded from the analyses that would strongly influence the outcome. Furthermore, a sensitivity analysis was performed for model 2 in which the influence of the number of children on the outcomes heart disease and stroke was calculated separately.

Stratified analyses by sex and education or region were conducted to estimate ORs and 95% CIs for the association between number of children (0, 1–2, 3–4 or 5 +) and parental incident HDS. All of these analyses were adjusted for baseline age, marital status, and, depending on the model, for region or education. Persons with one or two children were used as reference.

All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).