Cross-sectional and longitudinal factors influencing physical activity of 65 to 75-year-olds: a pan European cohort study based on the survey of health, ageing and retirement in Europe (SHARE)

Study design and population

This study is conducted using data of the interdisciplinary panel Survey of Health, Ageing and Retirement in Europe (SHARE), which is performed in 19 countries of the European Union and Israel. The aim of the SHARE-project is to provide an overall picture of ageing in Europe and it gathered data on health, socio-economic status and social as well as family networks [43]. In SHARE, non-institutionalised people aged 50+ underwent a short physical examination and were interviewed with computer assisted personal interviews (CAPI). Wave one took place in 2004/2005, wave two in 2006/2007, wave three in 2008/2009 and wave four in 2010/2011. SHARE has data of more than 60,000 individuals, the response rate in wave one was around 62%, in waves two, three and four 73%, 77% und 56% [43]. The ten countries participating in all first four waves of SHARE were Belgium, Denmark, Germany, the Netherlands, France, Austria, Italy, Switzerland, Sweden and Spain. For further methodological details of SHARE see Börsch-Supan et al. (2013) [43]. This study is following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement [44]. An additional file shows this in more detail [see Additional file 1]. Inclusion criteria for this study were participation in all four waves of SHARE and belonging to the age group of 65-to-75–year-olds at the time of wave four. This resulted in a sample size of n = 3846 (male n = 1761, female n = 2085).

Outcome and country of residence

Physical activity was assessed based on the following two questions: ‘How often do you engage in vigorous physical activity, such as sports, heavy housework, or a job that involves physical labour?’ and ‘How often do you engage in activities that require a low or moderate level of energy such as gardening, cleaning the car or doing a walk?’. Given possible answers were ‘More than once a week’, ‘once a week’, ‘one to three times a month’ and ‘hardly ever’ or ‘never’. Participants, who stated to engage in vigorous physical activity once a week or less were defined as insufficient physical activity (IPA) (wave 4: n = 2582; 67.2%). Participants engaging in vigorous physical activity more than once a week are categorized as sufficient PA and serve as control group. The WHO used IPA as measurement and defined it as less than 150 min of moderate physical activity or less than 75 min of vigorous physical activity per week [45]. The variable country is defined as the country of residence of the survey participants.

Covariables – Intrapersonal factors

Age was calculated from birth month and year of the participants and time of the interview in wave four. The exact age in months was divided by 12 to generate age in years. Level of education was classified by the International Standard Classification of Education (ISCED) [46]. For parental education, the number of books present in the household that the participants lived in at the age of ten was used [47].The variable was dichotomised. Zero to eleven books were classified as ‘low parental education’ and more than eleven books as ‘higher parental education’. The financial situation was assessed with the question whether a household has trouble or not to make ends meet with the available monthly income. Four answer categories were used: ‘with great difficulty’, ‘with some difficulty’, ‘fairly easily’ and ‘easily’. The variable was dichotomised. All participants answering ‘with great difficulty’ and ‘with some difficulty’ were graded as ‘without difficulty’ and ‘fairly easily’ and ‘easily’ as without difficulties. The number of chronic diseases was based on a multiple answer question. The numeric variable was truncated at a number of three chronic diseases. Depression was examined with the EURO-D symptom scale, an index consisting of twelve items: depressed mood, pessimism, suicidality, guilt, sleep, interest, irritability, appetite, fatigue, concentration, enjoyment and tearfulness. The scale ranges from zero ‘not depressed’ to twelve ‘very depressed’. The variable was dichotomised according to Dewey and Prince into ‘no depression’ (0–3) and ‘depression’ (4–12) [48]. Hospitalization in the last twelve months shows whether participants were in a medical, surgical, psychiatric or any other specialized hospital overnight during the last twelve months before the interview. Two grip strength measurements on each hand were recorded using a dynamometer (Smedley, S Dynamometer, TTM, Tokyo, 100 kg). The variable shows the maximum grip strength of all four measurements. To describe limitations in activities of daily living (ADL) the activities of daily living index (ADLI) was used [49]. The ADLI is the sum of the five tasks dressing, bathing or showering, eating, cutting up food, walking across a room and getting into or out of bed. The higher the index the more difficulties exist with these activities. The variable was dichotomised. All participants with the index value zero were graded as ‘without limitations’ and with the index value one to five as with limitations. Body Mass Index (BMI) is based on self-reported values and is calculated as: BMI = weight in kilogram (kg) / (height in meter (m)) ² .

Covariables – Interpersonal factors

To determine marital status the existing response categories ‘married and living together with spouse’, ‘married, living separated from spouse’ and ‘registered partnership’ were summarised in the category ‘married and registered partnership’. The dichotomous variable partner in household indicates whether a participant was living together with his partner in a household. To measure the size of family network, a score was calculated from the number of children alive, the number of parents alive and the number of siblings alive. The numeric variable number of children was truncated at a number of four children. For four or more children, four children were included into the score. A high score value indicates a large family network. From wave four of SHARE on, information on up to seven individuals with whom respondents discussed important things most often during the 12 months before the interview is available. The size of social network may include family members, friends, neighbours and others and was truncated at three network partners. The dichotomous variable perceived social support shows whether the respondent had received instrumental social support within the last 12 month. The perceived quality of social network was assessed on a scale ranging from ‘0’ (completely dissatisfied) to ‘10’ (completely satisfied), including respondents who reported to have no social network.

Statistical analyses

To identify cross-sectional associations of intra- and interpersonal factors with IPA in wave four, prevalence odds ratios (POR) and 95%-confidence intervals (95%-CI) were calculated using binary logistic regression models.

To analyse the longitudinal association over time, we calculated hazard rations (HR) with 95% confidence intervals (95%-CI) using Cox regression models in the group of sufficient PA in wave 1. IPA at wave 4 served as the outcome. To account for heterogeneity between the observed countries, we included country as random effect in our models. We omitted stratification by sex for intrapersonal factors, because in the cross-sectional analyses the effects were similar for men and women. For interpersonal factors this was not the case, therefore we stratified the analyses by sex. Model building was based on the Wald statistic. But previously we calculated all other inclusion methods (backward, forward and expected inclusion) with the result that the method did not affect the result of regression after adjustment.

All analyses were performed using the statistical analysis software IBM SPSS statistics for Windows, version 20.0 (IBM Corp. Released 2011. Armonk, NY).

Cross-sectional associations

Regarding to the intrapersonal level, women were less physically active than men. These results are in line with previous research: A cross-sectional study using objective measurements of PA showed that 31% of older men and 20% of older women meet the WHO recommendations of 2.5 h moderate PA per week [50]; in questionnaire-based studies 16–21% in men and 11–21% in women were categorized as sufficiently physically active [15, 51]. In the SHARE data, we found an increasing proportion of IPA with age in both sexes. The influence of age on PA was confirmed in several previous studies [20–23]. Other intrapersonal factors of IPA that reached statistical significance were low educational level and financial difficulties. There is ample evidence for an association of socio-economic status with PA in the general population and this is as well true for older adults [20, 22]. Interestingly, we were also able to establish an association between parental education and IPA in older adults. Even in adulthood, the influence of parental education on several lifestyle factors was confirmed in studies [27, 28]. A previous study reported an association between childhood socioeconomic position and physical capability levels in adulthood [52]. Likewise, a number of health-related and fitness-related factors were associated to IPA. Many previous studies showed that good health in general is positively associated to PA in older adults [20, 22, 24, 25]. There are fewer studies on fitness-related factors in this age group. However, our findings of a positive association between higher grip strength and PA in older adults are consistent with a previous study [53].

The investigated interpersonal factors display several statistically significant associations with IPA, especially in men. The size of the social network as well as the size of the family network are both negatively associated with IPA, i.e. the larger these networks are the less likely it is to perform IPA. This is in line with previous research, which stated social networks and social support from friends and peers as influential factors for PA in older adults [20, 29]. However, we found that the perceived quality of the social network has no association with IPA. The perceived social support is even positively associated with IPA. One reason for this result might be that the SHARE data only includes received instrumental social support. Additional analyses showed that this result is rather a statistical artefact that can be fully explained by partner in household in men and by frailty status in women (data not shown). A recent Australian study showed that the association between social network and PA might be more complex and dependent on the type of activity [54], e.g. household activities and leisure-time activities. Distinctive gender patterns were present regarding family indicators, marital status, partner in household and number of grandchildren. In men, being married, presence of a partner in household and number of grandchildren showed all a statistically significant positive association with IPA. In women, the only interpersonal factor being statistically significant associated with IPA was being divorced. A clear gender difference of the health effects of divorce or being single was reported before [55].

In this study, the prevalence of IPA in 65–75-year-olds varied widely between countries, ranging from 55.4% to 83.3% in women and from 46.6% to 73.7% in men. A comparison of the IPA prevalence of the sub set used in this study and all participants of SHARE in this age group showed that most prevalence rates are in line with each other. Only the IPA prevalence for Germany, Austria and Denmark only in women differed (see Additional file 1). A German health survey showed an IPA prevalence in Germany of 60% in men and 63% in women in the age group of 60 to 69 years, which are higher values compared to those presented in this study [56]. For Austria and Denmark no prevalence rates of IPA where found for the age group 65 to 75 years. Therefore, the prevalence rates of the mentioned countries must be considered with caution. Women and men were more likely to meet the WHO recommendations of PA in northern and central European countries, as Sweden or Germany, and compared to southern European countries, as Italy and Spain. The reported increase in IPA from Northern and Central Europe to Southern Europe are in line with findings of a study of the European Commission in 2014 [14]. Climate, environmental, infrastructural and policy conditions for PA can differ highly within countries such as weather, topography, culture, street layout and crime rate. In our study, older adults from the Netherland, Germany, Denmark and Switzerland showed lower rates of overall IPA. A possible explanation is that these countries are all good practice examples for high levels of cycling and street safety [57], setting the ground for physical activity in older adults.

Longitudinal analyses

Longitudinally, educational level, number of chronic diseases, presence of depression, grip strength and BMI were statistically significant predictors of IPA at the age of 65 to 75 in previously active persons. Of the investigated interpersonal factors, only living together with a partner in one household showed a statistically significant influence in men, but not in women.

In the multivariable model, only grip strength and BMI remained statistically significant predictors for IPA in older adults. Previous cross-sectional studies that were not able to distinguish between cause and effect showed that BMI is associated with PA in older adults [20, 23, 50, 58].

Limitations and strengths

There are several limitations to the study. One limitation is the measurement of PA by questionnaire. Objective methods using accelerometers are more valid for PA measurements [59], but many large population studies did not use accelerometers due to high costs. In SHARE, only information on the frequency but not on the duration of moderate and vigorous PA is available. Thus, it can only be determined with some certainty that the WHO recommendations are not met, if persons engage in PA at most once a week. Moreover, there is a certain risk of social desirability bias due to the interview method (computer assisted personal interview) and the interviewee knowing that sufficient PA is desirable [60]. Previous studies have found that questionnaire-based surveys are likely to overestimate duration and intensity of PA compared to objective measurements [59]. A further limitation is the restricted set of variables that leaves out further putative interesting factors from investigation as e.g. social support as noted above or extrapersonal variables, which could help explaining cross-country differences, as built environment, weather conditions, cultural or policy factors [31, 36–39]. Finally, we restricted our analyses to a subset of the SHARE participants (participation in all 4 waves) to be able to analyse longitudinal effects. A questionnaire survey of non-responder in Germany showed only small indication for a non-response bias with regard to household composition and health status [43, 61]. Additionally, subgroup analyses indicated only small differences in sex and age in survey participation and retention [43, 53, 62].

The study also has several strengths. SHARE is conducted in several European countries with a highly standardized study protocol involving physical measurements as e.g. grip strength and BMI [43]. SHARE provides panel data of high quality and representativeness due to the high degree of standardization in data collection. It allows comparing several European countries and the sample size is sufficient to detect smaller effects and has high participation rates as well as only small evidence for attrition bias [53]. The longitudinal nature of the data enables identification of temporality of associations as e.g. BMI and physical activity [63].