Informal caregiving and physical activity among 204,315 adults in 38 low-and middle-income countries: A cross-sectional study

Data on the association between informal caregiving and physical activity (PA) levels are scarce, especially from low- and middle-income countries (LMICs). Furthermore, previous research has yielded con ﬂ icting results. Thus, we investigated this association in adults from 38 LMICs. Data from the World Health Survey (WHS), a cross- sectional, predominantly nationally representative survey conducted in 2002 – 2004, were analyzed. PA was assessed by the International Physical Activity Questionnaire and participants were dichotomized into those who do ( ≥ 150 min of moderate-to-vigorous PA per week) and do not (<150 min = low PA) comply with the World Health Organization PA recommendations. Those who provided help to a relative or friend (adult or child), because this person has a long-term physical or mental illness or disability, or is getting old and weak in the past year were considered to be informal caregivers. Multivariable logistic regression analysis was conducted to assess the associations. There were 204,315 adults aged ≥ 18 years from 38 LMICs included in this study [mean (standard deviation) age 38.6 (16.1) years; 50.7% female]. Overall, the prevalence of caregiving and low PA was 19.5% and 29.9%, respectively. After adjustment for potential confounders, caregivers were at a lower risk for low PA compared to non-caregivers (OR = 0.79; 95% CI = 0.72 – 0.86). Engagement in greater number of caregiving activities was associated with lower odds for low PA dose-dependently. Informal caregiving was associated with higher levels of PA in adults in LMICs. Future studies of longitudinal design are warranted to understand causality and the underlying mechanisms of this association.


Introduction
The prevalence of people aged over 65 years in the world is expected to increase from 9% in 2019 to 16% in 2050, while the number of individuals aged 80 years or over is projected to triple from 143 to 426 million (United Nations, 2019). In addition, disability has steadily expanded in the past two decades, and the number of years lived with disability has increased from 562 million in 1990 to 853 million in 2017 (GBD 2017 Disease and Injury Incidence and Prevalence Collaborators, 2018). The rate of ageing and the prevalence of disability are higher in low-and middle-income countries (LMICs) than in highincome countries (HICs) (Mitra and Sambamoorthi, 2014;Siriwardhana et al., 2018), underlying the importance of the problem associated with population ageing and disability in this part of the world.
In this context, there will be an increase over the upcoming years in the number of individuals requiring formal (e.g., nurse services) and https://doi.org/10.1016/j.ypmed.2020.106007 Received 24 October 2019; Received in revised form 20 January 2020; Accepted 25 January 2020 T informal care (e.g., assistance from a family member). In LMICs, there is limited access to formal care due to several factors such as low capacity of health centers, large geographical distances to the closest health center and insufficient financial means to pay the costs (Peters et al., 2008). Thus, informal caregivers play an essential role in the management of the elderly and people with disability in these countries, and the health of the caregiver is vital to sustain care provision in this setting.
In the past decades, numerous studies have investigated the impact of informal caregiving on several health outcomes. Although the majority of previous reports have focused on adverse health outcomes (e.g., psychological distress, depression) associated with caregiving (Bom et al., 2018), also in LMICs (Thrush and Hyder, 2014), some studies have shown that caregiving may also be associated with positive health outcomes (Buyck et al., 2011;Campen et al., 2013;Freedman et al., 2014;O'Reilly et al., 2017;Roth et al., 2013;Toljamo et al., 2012). Specifically, previous studies have shown that caregiving can increase levels of life-satisfaction, well-being, fulfillment and happiness, and also reduce premature mortality of the caregiver (Roth et al., 2013). Furthermore, it has been reported that caregiving with low levels of burden is associated with better mental and physical health (Buyck et al., 2011). Although the reason why caregiving may be associated with both negative and positive outcomes is largely unknown, it is possible that health behaviors such as physical activity (PA) may be playing a role as the beneficial effects of PA on mental and physical health is well-established (Buchman et al., 2012;Ghaderpanahi et al., 2011;Goodwin, 2003), while there is increasing interest in the promotion of PA as a strategy to improve the well-being of caregivers (Lambert et al., 2016). Caregiving may increase PA via increasing number of daily chores (Lam and Garcia, 2017), increased awareness of the beneficial effects of PA (Lo, 2009), and coping (Litzelman et al., 2018), but it is also possible that caregiving may lead to reduced levels of PA, via burden (del-Pino-Casado et al., 2019), stress (Longacre et al., 2017) and fatigue (Peters et al., 2015). For example, informal caregiving may be physically challenging when informal caregivers help care recipients with household activities (Lam and Garcia, 2017) or when they walk for several hours to reach the nearest health center in the context of LMICs (Blanford et al., 2012). Thus, it is of vital importance to understand PA levels among caregivers for the establishment of effective strategies to improve the well-being of caregivers.
However, to date, there are only a few studies on caregiving and its association with PA with mixed results. These studies have found either higher (Lambert et al., 2017) or lower levels of PA among caregivers compared to non-caregivers (Stacey et al., 2019), or no difference between the two groups (Hoffman et al., 2012;Litzelman et al., 2014;Marquez et al., 2012;Son et al., 2011). Although these studies are of importance, they display several limitations that should be acknowledged. First, some of them were of small sample size, and the lack of significant findings in these studies may have been due to insufficient statistical power. Second, the majority of studies failed to adjust for important potential confounders such as disability or depression, and this may have biased their results. Third, all except one study (Lambert et al., 2017) were conducted in HICs, and therefore, little is known about the informal caregiving-PA relationship in LMICs. Exploring this association in LMICs is of particular interest as for many informal caregivers in this part of the world, their priority is actually to fulfil their own basic human needs (e.g., food, accommodation, income, education) rather than looking after a relative or neighbor (Lempp et al., 2018). Fourth, only one of these studies was nationally representative, and therefore generalizability is limited (Lambert et al., 2017). Finally, many of the studies only focused on one aspect of PA (e.g., leisure-time PA) and data on caregiving and compliance with the World Health Organization (WHO) PA guidelines (≥150 min of moderate-to-vigorous PA per week) are scarce.
Therefore, the goal of this study was to investigate the potential association between informal caregiving and PA (compliance with the WHO PA guidelines) in adults from 38 LMICs using predominantly nationally representative datasets.

The survey
The World Health Survey (WHS) was a cross-sectional survey conducted in 70 countries in [2002][2003][2004]. Survey details are available from the WHO (http://www.who.int/healthinfo/survey/en/). Briefly, singlestage random sampling was carried out in 10 countries, while the remaining 60 countries used stratified multi-stage random cluster sampling. All adults aged ≥18 years with a valid home address were assigned a non-zero chance of inclusion. Standard translation procedures for the survey questionnaire were followed to ensure comparability across countries. Face-to-face interviews and telephone interviews were conducted by trained interviewers. Individual level response rates were over 82%. Post-stratification corrections were made to sampling weights to adjust for non-response and the population distribution reported by the United Nations Statistical Division. Ethical boards at each study site provided ethical approval for the survey with all participants providing informed consent.
In 2003, LMICs were countries with gross national income (GNI) per capita ≤$9385 (low-income countries ≤$765; lower-middle income countries: $766-$3035; upper-middle income countries $3036-$9385) (The World Bank, 2020). Data from 69 countries were publically available in the WHS. In the current analysis, a total of 38 LMICs (n = 204,315) were included (reasons for the exclusion of countries are provided in Fig. S1). These countries corresponded to 16 low-income countries (LICs) (n = 83,905) and 22 middle-income countries (MICs) (n = 120,410). The data were nationally representative for all countries with the exception of China, Comoros, India, Ivory Coast, and Russia.

Physical activity (outcome variable)
Items from the International Physical Activity Questionnaire were used to categorize PA (Stubbs et al., , 2016. Specifically, participants were asked how many days over the past week on average they engaged in moderate and vigorous PA. Secondly, participants were asked for how many minutes on average, they engage in PA at a moderate and vigorous level. The total amount of moderate-to-vigorous PA over the last week was calculated and those scoring ≥150 min were classified as meeting the recommended guidelines (coded 0), and those scoring < 150 min (low PA) were classified as not meeting the recommended guidelines (coded 1) (World Health Organization, 2010).

Informal caregiving (exposure variable)
Those who answered affirmatively to the question "During the past year, did you provide help to a relative or friend (adult or child), because this person has a long-term physical or mental illness or disability, or is getting old and weak?" were considered to be informal caregivers (Hosseinpoor et al., 2013). This question is comparable to those used in previous surveys to identify informal caregivers (Smith et al., 2014). Furthermore, questions on five types of caregiving activities (i.e., personal care, medical care, household activities, supervision, transport/mobility) with "Yes" and "No" options were asked to caregivers (see Table S1 for actual questions). The number of caregiving activities was summed. Non-caregivers were assigned a score of 0.

Control variables
The selection of the control variables used in this analysis was based on past literature (Koyanagi et al., 2018a(Koyanagi et al., , 2018b and included age, sex, marital status (married/cohabiting, never married, separated/divorced/widowed), wealth, highest education attained (no formal education, primary education, secondary or high school completed, and tertiary education completed), household size (1, 2, 3-5, ≥6), employment status (not working for pay or currently in paid employment), disability, and depression. Country-wise wealth quintiles were created using principal component analysis based on 15-20 assets. Individuals who had severe/extreme difficulty in either moving around, performing self-care, concentrating/remembering things, or seeing and recognizing a person across the road in the past 30 days were considered to have disability (Mitra and Sambamoorthi, 2014). Depression was assessed with the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV algorithm based on duration and persistence of depressive symptoms in the past 12 months (Cifuentes et al., 2008;Loerbroks et al., 2012). The algorithms used are provided in Table S2.

Statistical analysis
Statistical analyses were performed with Stata 14.1 (Stata Corp LP, College station, Texas). Multivariable logistic regression analysis was conducted to assess the association between informal caregiving (exposure variable) and low PA (outcome variable). Analyses using the overall sample and those stratified by age groups (18-44, 45-64, ≥65 years) and sex were conducted. These analyses adjusted for age, sex (apart from the sex-stratified analysis), marital status, wealth, education, household size, employment status, disability, depression, and country. Adjustment for country was conducted by including dummy variables for each country as in previous WHS publications . Using the overall sample, we repeated a similar analysis with the number of caregiving activities as the exposure variable.
Finally, to assess the generalizability of the findings based on the pooled sample across all countries, we conducted country-wise regression analyses for the association between caregiving and low PA adjusting for age and sex. A pooled estimate was obtained by meta-analysis with random effects. To assess the level of between-country heterogeneity, the Higgins's I 2 statistic was calculated. This represents the degree of heterogeneity that is not explained by sampling error with values of 25%, 50%, and 75% often being considered to represent low, moderate, and high levels of heterogeneity, respectively (Higgins et al., 2003).
All variables were included in the models as categorical variables with the exception of age (continuous variable). Taylor linearization methods were used in all analyses to account for the sample weighting and complex study design. Results from the logistic regression analyses are presented as odds ratios (ORs) with 95% confidence intervals (CIs).

Association between informal caregiving and low physical activity
Overall, after adjustment for a variety of potential confounders (i.e., age, sex, marital status, wealth, education, household size, employment status, disability, depression, country), caregivers were at a lower risk for low PA compared to non-caregivers (OR = 0.79; 95% CI = 0.72-0.86) ( Table 3). Among those aged ≥65 years, this effect was particularly pronounced with the OR being 0.70 (95% CI = 0.55-0.90). The association between the number of caregiving activities and low PA estimated by multivariable logistic regression is shown in Fig. 1. Compared to non-caregivers, there was a gradual decrease in the OR for low PA with increasing number of caregiving activities although the estimates for 2 and 3, or 4 and 5 caregiving activities were similar. Specifically, engaging in one caregiving activity was associated with 13% lower odds for low PA while this figure was 32%-34% for 4-5 caregiving activities. The country-wise associations between informal caregiving and low PA are shown in Fig. 2. Informal caregiving was inversely associated with low PA in the vast majority of the countries. The pooled estimate based on a meta-analysis was 0.70 (OR = 0.63-0.77) with a high-level of between-country heterogeneity (I 2 = 78.7%). 4. Discussion

Main findings
Among a sample of > 204,300 adults from 38 LMICs, the overall prevalence of informal caregiving and low PA was around 20% and 30%, respectively. After adjusting for several potential confounders, there was a negative association between informal caregiving and low Data are weighted %. PA (OR = 0.79; 95% CI = 0.72-0.86). Furthermore, a dose-response relationship was observed with increasing number of caregiving activities. These findings were corroborated in the country-wise analysis although a high level of country-wise heterogeneity was observed.

Interpretation of the findings
Previous studies focusing on the association between informal caregiving and PA have yielded contradicting findings. One large multicountry study including three LMICs found that caregivers engage in significantly greater amount of PA in India and Russia but not in Ghana (Lambert et al., 2017). Furthermore, a Korean study observed that physical inactivity was less frequent in caregivers than in controls, but this result did not reach statistical significance (Son et al., 2011). Next, a large study including 546,156 caregivers and 246,927 non-caregivers from the US found that the metabolic equivalent of task-minutes per week did not significantly differ between the caregiving and control groups (2031 versus 1762) (Litzelman et al., 2014). Finally, a population-based longitudinal biomedical cohort study conducted in Southern Australia showed that the odds for sufficient PA were significantly lower in caregivers than in non-caregivers (Stacey et al., 2019). The discrepancy between our findings and that of some of the previous studies may be explained by methodological differences such as the definition of PA (e.g., overall PA, leisure-time PA), the tools used to measure PA, the representativeness of the sample (e.g., nationally representative or subgroups) or the setting in which the study was conducted.
We propose three hypotheses that may explain the negative association between caregiving and low PA in LMICs found in our study. First, several types of informal caregiving (e.g., personal care, household activities, transport/mobility) may be considered as physical activities per se. For example, household activities were a frequent type of Fig. 2. Country-wise associations between informal caregiving (exposure) and low physical activity (outcome). Abbreviation: OR odds ratio; CI confidence interval. Overall estimate was obtained by meta-analysis with random effects. The model is adjusted for age and sex.
informal caregiving in this study, and this is in line with previous research. A US study including > 19,500 participants revealed that caregivers spent more time on daily housework (24 min more) than their counterparts not involved in caregiving (Lam and Garcia, 2017). Next, a substantial proportion of informal caregivers may be involved in long-distance informal caregiving. A study including child adult-older parent dyads from China found that between 26.0% and 32.4% of caregivers lived in or beyond the same province as their parents (Zuo et al., 2014). Help with transport and mobility may be particularly challenging for many people in LMICs. For example, walking for several hours to reach the nearest hospital is not rare in this setting. A study conducted in Niger found that 61% and 76% of the population needed to walk for at least 1 h to access health facilities during the dry and wet season, respectively (Blanford et al., 2012). Second, informal caregivers may be more aware of the benefits of practicing health-promoting behaviors than their counterparts not involved in caregiving. Interestingly, it was observed in a Taiwanese cross-sectional study of 137 participants that caregivers had healthier lifestyles than non-caregivers in several domains (i.e., health responsibility, PA, stress management) (Lo, 2009). Another study with 657 participants also showed that PA increased in around 24% of the family and friends of cancer patients after a cancer diagnosis, suggesting that a cancer diagnosis may favor changes in health behaviors not only in the patients but also relatives and friends (Humpel et al., 2007). This may be explained by increased recognition of the importance of healthy lifestyles following the illness of a close relative/friend in need of care, or more frequent contact with health professionals. Third, engaging in PA may be a coping strategy in informal caregivers. A study including almost 1500 caregivers of cancer survivors showed that there was a positive association between problem-focused coping and PA (Litzelman et al., 2018). Similar results were found in a US study of 162 family caregivers, where leisure-time PA participation was a significant moderator in the relationship between type of caregiving and daily psychological distress (Wong et al., 2016).
Another interesting finding of this multi-country study was the high between-country heterogeneity in the association between informal caregiving and low PA. This between-country heterogeneity may be explained by differences in factors such as quality of social service, intensity or content of caregiving, family size, and underlying disease of the care recipient. For example, personal care such as moving around was less frequent in Ghana (23.7%) and India (45.4%) than in Russia (67.9%) (Lambert et al., 2017), and thus participants from the first two countries may have been involved in caregiving activities of less intensity when compared with those from Russia. Furthermore, in countries with low fertility rates, it is possible that people have fewer sibling or other family members to share the caregiving task and this might also influence levels of PA associated with caregiving. Finally, the burden on the informal caregivers may also depend on the underlying illness of the care-recipient (Ringer et al., 2016), and the betweencountry heterogeneity may also be explained by different disease profiles of the care-recipient across countries.

Public health implications and areas for future research
In our study, although informal caregivers were more likely to engage in PA than non-caregivers, 24.6% of caregivers did not reach the PA targets recommended by the WHO, while caregiving was not associated with increased PA in some countries, and thus, further efforts to improve PA levels in caregivers may lead to improvements in various aspects of health. Assessment of PA (including type) among informal caregivers and provision of adequate advice on PA by health care workers may be important given increasing evidence that promotion of PA among informal caregivers may be an effective strategy to improve health outcomes (Farran et al., 2008). In particular, although caregivers may already be engaging in more PA, it is important to assess the type of PA that they are engaging in as some types of PA may be deleterious to health (e.g., strenuous activities which may lead to injury and musculoskeletal symptoms), and provide advice on how to increase other types of PA (e.g., leisure-time PA) which are more likely to be beneficial for health. Future studies, ideally of longitudinal design, should seek to understand the underlying mechanisms of our findings. In addition, the role of PA in the association between caregiving and health outcomes should be investigated particularly in LMICs. This is particularly relevant given the recent increasing interest in promoting PA among caregivers as a strategy to improve well-being, quality of life, and self-efficacy (Lambert et al., 2016).

Strengths and limitations
The strengths of the study include the large number of participants and countries, and the use of predominantly nationally representative datasets. However, the findings should be interpreted in light of several limitations. First, there was no information on caregiving intensity, although there may be important differences in PA between low-and high-intensity caregivers. Second, more detailed variables on PA (e.g., type, frequency, context) may have led to a better understanding of the informal caregiving-PA relationship. Third, although the IPAQ is a tool to assess PA that has been validated in a variety of settings (Deng et al., 2008;Hallal et al., 2010;Helou et al., 2017), it still relies on self-report, which entails some limitations. Thus, future research should employ more sophisticated tools (e.g., accelerometry) to obtain more accurate measures on PA levels. Finally, this was a cross-sectional study, and it was thus not possible to determine causality and temporality in the association between informal caregiving and PA. For example, it is possible that those who are healthier with greater levels of PA are more likely to engage in caregiving activities.

Conclusions
There was an inverse relationship between informal caregiving and low PA in adults in LMICs. Given the increasing interest in PA promotion as a strategy to improve well-being among caregivers, future studies should investigate the underlying mechanisms that link informal caregiving and PA, while the role of PA in the association between caregiving and various health outcomes should also be investigated. Ethical boards at each study site provided ethical approval for the survey with all participants providing informed consent.

Availability of data and materials
The datasets supporting the conclusions of this article are available in the World Health Survey repository, available at http://www.who. int/healthinfo/survey/en/.
Funding Ai Koyanagi's work is supported by the PI15/00862 project, integrated into the National R + D + I and funded by the ISCIII -General Branch Evaluation and Promotion of Health Research -and the European Regional Development Fund (ERDF-FEDER). Brendon Stubbs is supported by a Clinical Lectureship (ICA-CL-2017-03-001) jointly funded by Health Education England (HEE) and the National Institute for Health Research (NIHR). Brendon Stubbs is part funded by the NIHR Biomedical Research Centre at South London and Maudsley NHS Foundation Trust. Brendon Stubbs is also supported by the Maudsley Charity, King's College London and the NIHR South London Collaboration for Leadership in Applied Health Research and Care -Greater Manchester (CLAHRC) funding. This paper presents independent research. The views expressed in this publication are those of the authors and not necessarily those of the acknowledged institutions.
CRediT authorship contribution statement

Declaration of competing interest
The authors have no conflicts of interest to disclose.