Grandparenting, health, and well-being: a systematic literature review

Whether grandparenting is associated with improved health or well-being among older adults is a salient question in present-day aging societies. This systematic review compiles studies that consider the health or well-being outcomes of grandparenting, concerning (1) custodial grandparent families, where grandparents are raising grandchildren without parental presence; (2) three-generation households, where grandparents are living with adult children and grandchildren; and (3) non-coresiding grandparents, who are involved in the lives of their grandchildren. Review was based on literature searches conducted in September 2019 via Web of Science, PubMed, PsycINFO, and Ebsco. We screened 3868 abstracts across four databases, and by following the PRISMA guidelines, we identified 92 relevant articles (117 studies) that were published between 1978 and 2019. In 68% of cases, custodial grandparenting was associated with decreased health or well-being of grandparents. The few studies considering the health or well-being of grandparents living in three-generation households provided mixed findings (39% positive; 39% negative). Finally, in 69% of cases, involvement of non-coresiding grandparents was associated with improved grandparental outcomes; however, there was only limited support for the prediction that involved grandparenting being causally associated with grandparental health or well-being. Despite this, after different robustness checks (counting all nonsignificant results, taking into account the representativeness of the data and causal methodology), the main finding remains the same: the most negative results are found among custodial grandparents and three-generation households and most positive results among non-coresiding grandparents.


Background and objectives
Due to increased life expectancy, the proportion of older adults, including grandparents, has increased on a global scale, and in fact, it has been estimated that currently approximately 13% (one billion) of world population are grandparents (Moore and Rosenthal 2016). Most grandparents play an active role in the lives of their grandchildren. In Europe, for instance, 58% of grandmothers and 49% of grandfathers provide regular grandchild care (Hank and Buber 2009), while approximately 2% of children are raised by their grandparents in the USA meaning that one million grandparents in the USA are the primary caregivers for their grandchildren (Dunifon et al. 2014). Thus, grandparents are often highly involved in their grandchildren's lives and whether grandparenting provides benefits or disadvantages for grandparents in terms of their health or overall wellbeing is a salient question.
Many studies that attempt to detect whether grandparenting is associated with improved health or well-being predict that grandparents benefit from being involved in their grandchildren's lives (e.g., Mahne and Huxhold 2015;Tsai et al. 2013). A counterhypothesis, however, takes the stance that caring for young children is challenging, particularly for older adults with limited reserves of strength (e.g., Baker and Responsible Editor: Marja J. Aartsen.

3
Silverstein 2008a, b; Hughes et al. 2007). According to this perspective, active grandparenting could overburden older adults and lead to grandparents' decreased health and wellbeing. For example, Hertwig (2010, 2011) argue that the association between grandparenting and grandparents' health may result in an inverted U-shaped curve. Based on the Coall-Hertwig hypothesis, moderate grandparental involvement is the most beneficial for grandparents, while negative effects may arise when no grandparental involvement occurs or when it reaches the highest level of involvement (e.g., when grandparents are the primary caregivers of their grandchildren).
Thus, whether grandparenting improves the health or well-being of grandparents may depend on living arrangements that are related to the degree of grandparental involvement, which is why grandparents are commonly separated into three groups: (1) custodial grandparents, (2) grandparents living with their children and grandchildren in threegeneration households, and (3) non-coresiding grandparents (i.e., those involved in their grandchildren's lives without living with them). In households where the grandparents are the primary caregivers of their grandchildren or they live with their descendants in a three-generation household, the high level of their involvement is assumed based on the living arrangements. Among non-coresiding grandparents, however, grandparental involvement is most often measured via childcare support but also the frequency of contact, emotional closeness, and other informal assistance.
Cultural variation in living arrangements, filial norms, and grandparental involvement are substantial (Shwalb and Hossain 2017). Living in three-generation families or being a custodial grandparent is much more common in many Asian countries than in Western ones. This variation has been partly explained by the influence of Confucianism, which promotes a tradition of filial responsibility (Burr and Mutchler 1999;Speare and Avery 1993). Also reciprocity may be an important factor in Asian countries with strong filial obligations (e.g., Sheng and Settles 2006). Cultural traditions could also influence on whether custodial grandparenting or living in three-generation household is associated with positive or negative outcomes among grandparents because in Asian countries grandparents living with grandchildren are not as selected group as they are in Western countries. In addition, due to the lack of publicly provided old age support grandparents need to rely on their children and thus living with them or with grandchildren could provide benefits to grandparents themselves. The expectation is, that the negative effects of highly involved grandparenting observed in Western countries are not present or could be even positive in Asian countries.
This review makes a novel contribution to the literature by compiling research on all three contexts of grandparenting and revealing how in each context the involved grandparenting is associated with grandparental health or well-being. Cultural context of grandparenting is taken into account as the review observes also the distribution of studies and results by various countries. In addition, the review investigates whether previous studies have provided convincing causal evidence for the possible association.

Aim of the review: to reveal a grandparenting effect
Our main aim is to investigate whether grandparenting (or grandparental involvement) is associated with the health or well-being of grandparents and whether this association is positive or negative. Grandparent outcomes have been measured with several variables which we can summarize into two rough categories: health and well-being. The health category includes, for instance, longevity, cognitive skills, mental health, depressive symptoms, stress levels, physical health, frailty index, self-rated health, preventive health behavior, and limitations in activities of daily living. Well-being category in turn includes variables such as happiness, life satisfaction, subjective well-being (SWB), and perceived quality of life. Of course, these broad categories are not mutually exclusive but rather interrelated. Being in good health is probably associated with increased well-being and vice versa. Studies detecting the effect of grandparental involvement on grandparent outcomes have commonly used one or several of these outcomes and to be as comprehensive as possible, we try to take all such studies into account.
Three contexts of grandparenting (i.e., custodial grandparents, grandparents in three-generation households, and non-coresiding grandparents) have been considered, respectively. In addition, the geographic and cultural context of grandparenting has been observed. As we conducted a database of studies included in the review, we marked each study's result as being either positive, negative, or nonsignificant, depending on the association and its statistical significance between grandparents' involvement and their health or well-being. Some studies that included more than one context of grandparenting may have been marked as providing nonsignificant results in one context (e.g., custodial grandparents) but positive results in another (e.g., noncoresiding grandparents; Choi and Zhang 2018). Likewise, some studies may provide negative results in one context and positive results in another (e.g., Hughes et al. 2007).
In several cases, more than one health or well-being outcome was investigated in a single study. We marked the result of a study as being either positive or negative, even if there was one positive or negative association revealed, and we marked a study as being both positive and negative if it contained both results. The latter was often the case if the results were separated according to gender (e.g., Hughes et al. 2007) or ethnicity (e.g., Goodman andSilverstein 2002, 2006). Also, a study was marked as nonsignificant if all the results in specific grandparent groups showed nonsignificant associations (e.g., Ates 2017;Hsu and Chang 2015). Thus, the total number of positive, negative, or negligible results exceeded the number of studies included in the review (see Table 2). However, in the review, we also counted the total number of all the results in the studies, which was substantially higher than counting only the positive, negative, and nonsignificant results overall, as one study may have several positive, negative, or nonsignificant results due to multiple outcome measures and separations (see Table 4).

Search strategy
On September 27, 2019, we conducted a systematic literature search in three databases: Web of Science, PubMed, and PsycINFO. Then, on September 30, 2019, we included one more database, Ebsco, in the review process. We limited the search to peer-reviewed articles in English that employed a quantitative method and were published between 1970 and 2019. In practice, the first study in our sample is from 1978 (Wood and Robertson 1978) because before this there were none eligible studies.

Screening eligibility and inclusion criteria
The review's search yielded 19,246 records in total, but we excluded the following articles: duplicates (n = 8189), those that covered other topics based on the article's name (n = 7030), those that were not scientific or peer-reviewed (n = 99), and those that were in other languages besides English (n = 60). Thus, we included 3868 abstracts for screening, and afterward, we excluded articles that did not concern grandparenting (n = 2735) or include indicators about grandparental health or well-being (n = 465). We also excluded those that were not peer-reviewed (n = 217), only employed qualitative methods (n = 68), were in other languages besides English (n = 60), were reviews (n = 15), and those specifically concerned with being/becoming a grandparent (n = 8) (Fig. 1).
We assessed the full text for 300 articles, excluding those that were not based on quantitative research (n = 39), did not have grandparental investment as an independent variable (n = 38), included no grandparental health or well-being outcomes (n = 15), and those that only generally covered caregivers but did not distinguish between grandparents and other types of caregivers (n = 7). Finally, we excluded studies that only concerned custodial grandparents (n = 99) or only three-generational households (n = 9) if they did not include a comparison group of either non-custodial or noncoresiding grandparent populations. The exceptions included studies that compared grandparents' health or well-being before and after they were custodial grandparents or before and after they began living in three-generational households. Finally, one study could not be accessed, and based on the abstract, we were unable to evaluate whether it should have been included in the review (Minkler and Fuller-Thomson 2001).
Based on these selections, a total of 92 articles (i.e., peer-reviewed, published research reports) were included in the final sample. We classified all the studies (i.e., part of an article that covers one of the abovementioned types of familial circumstances) based on whether they concerned custodial grandparents (n = 34), three-generational households (n = 18), or non-coresiding grandparents (n = 65). Some articles covered more than one type of family circumstance; thus, the total number of studies included was higher (n = 117) than the number of articles in the final database. In this paper, the term, result, refers to a single finding that concerns grandparental involvement and an outcome measure (whether article had at least one positive and/or negative or negligible result n = 141; all results n = 452). In this sense, an article can contain a maximum of three studies, while a single study can contain several results.

Descriptive findings
The populations that were studied in the review are shown in Table 1. Most research was conducted with data from the USA (n = 48), especially in the custodial grandparent group. Also, many studies, especially in the non-coresiding grandparent group, were conducted with data from European countries (n = 21), and of these, a significant number utilized data from multiple European countries by using the Survey of Health, Ageing and Retirement in Europe (SHARE) data (n = 12). However, studies using data from European countries are missing from the custodial grandparent group, which is likely because the number of custodial grandparent 1 3 households is much lower in Europe (particularly Western Europe) than in the USA or Asian countries (Shwalb and Hossain 2017). In addition, there were 29 studies utilizing data from Asian countries, 7 studies from Australia, 2 from South America, and 10 from other countries (4 from Turkey, 4 from Kenia and 2 from Israel).

Custodial grandparents
Custodial grandparents are the primary caregivers for their grandchildren, and recently, the number of these "skipped-generation households," or "grandfamilies," has increased in many Western countries. Currently, this population group is highest in the USA, where approximately 2% of children are raised by their grandparents (Dunifon et al. 2014). While grandparents may be responsible for raising their grandchildren for many reasons, among the most common in Western countries are parental teenage pregnancy, drug addiction, mental/physical health problems, incarceration, distance employment, relationship breakdown, and death (Hayslip et al. 2017). In Asian countries, however, grandparents mostly raise their grandchildren due to distance 1 3 employment, especially in China, where parents often leave their children with their parents due to temporary migration, which refers to working in other locations (Chen and Liu 2012;Cong and Silverstein 2008).
In the review, a total of 68% (23/34) of the custodial grandparent studies were conducted with data from the USA (Table 1 and Appendix Table 6). According to the research describing custodial grandparents' characteristics in the USA, they are more often concerned with women than men, and more often, they focus on the maternal side (Dunifon et al. 2014;Hayslip et al. 2017). In most cases, custodial grandparents in the USA are members of lower socioeconomic classes, single women Heywood 1999;Minkler and Fuller-Thomson 2000), African-Americans, and between ages 50 and 59, whereas only very few are under 40 or over 80 (Ellis and Simmons 2014).
Thus, in the USA, custodial grandparent families are predominantly a selected group, which inevitably affects any comparison between custodial grandparents and noncustodial grandparents. According to the studies in this review, grandparents raising grandchildren have a higher risk of various health problems. Among 68% (30/44) of the results, the health or well-being of custodial grandparents was, at least in some grandparent subpopulations, poorer than non-custodial grandparents, their counterparts (Table 2 and Appendix Table 6). For instance, custodial grandparents have a higher risk of being limited in their daily activities as well as having depressive symptoms, elevated stress levels, and poorer self-rated health than their counterparts (e.g., Baker and Silverstein 2008a;Blustein et al. 2004;Fuller-Thomson 1999, 2005;Musil et al. 2011). Most studies in this group have used measurements related to grandparental health but also some used measurements related to well-being (e.g., life satisfaction, quality of life) and also these revealed mostly negative associations (Bowles and Myers 1999;Wilmoth et al. 2018;Yalcin et al. 2018). Many of the detrimental effects on grandparents' health or well-being in skipped-generation households are likely due to their characteristics and history rather than their caring responsibilities exclusively.
Although many studies on custodial grandparents have been conducted with cross-sectional data and could reveal selection effects, numerous investigations also contain longitudinal data (n = 16 studies, n = 21 results; Table 3). In these studies, a negative association is also apparent, as a grandparent who begins raising a grandchild often suffers from a decline in health (e.g., Baker and Silverstein 2008a, b;Musil et al. 2011). Thus, it might not only be selection that explains the negative association between custodial grandparenting and grandparental health.
Among the results, only 27% (12/44) showed positive outcomes for custodial grandparents (Table 2 and Appendix  Table 6). Twelve studies report at least one positive result between being a custodial grandparent and grandparental health, including those from the USA (7), South Korea (1), Taiwan (1), Kenya (2), and Thailand (1). Most of these  studies, however, also report some negative results regarding an outcome or grandparent group (Appendix Table 6), and importantly, only two studies report solely positive results (Chung and Park 2018;Ku et al. 2013).
In summary, there are two options for interpreting the results, which showed that, in most cases, custodial grandparents reported poorer health when compared to their noncustodial counterparts. The first option is that a decline in health occurs when one enters the role of a custodial grandparent since it increases the burden of caregiving. However, one study shows that grandmothers, who had been raising and continued to raise their grandchild, were more likely to have preventive health behaviors (Baker and Silverstein 2008b), meaning that the health decline may not be considered long-lasting. The second option is that these findings are based on selection effects, and custodial grandparentsespecially in Western societies-are typically considered part of a disadvantaged group. In Asian countries, however, custodial grandparents do not constitute a disadvantaged group, so the results may differ. In this review, 5/34 studies concerning custodial grandparents were conducted with data from Asian countries, and the results were either nonsignificant (Chen and Liu 2012;Choi and Zhang 2018), positive (Chung and Park 2018;Ku et al. 2013), or both positive and negative (Komonpaisarn and Loichinger 2019).

Grandparents in three-generation households
The terms, "three-generation" and "multigenerational" families, refer to a living arrangement whereby children, parents, and grandparents live together in the same household. The number of three-generation households varies remarkably between countries. For instance, approximately 25% (or more) of adolescents in Southern European countries live in three-generation households, whereas the number of children living in multigenerational households is less than 5% in Scandinavian countries (Kreidl and Hubatkova 2014). Meanwhile, by age five, almost a fourth of children in the USA live in three-generation families, while 8% and 11% do so in the UK and Australia, respectively (Pilkauskas and Martinson 2014).
Few investigations examine whether living in three-generation households is associated with improved or impaired outcomes among grandparents, (Dunifon et al. 2014) and, similar to custodial grandparenting, this population group is primarily studied in the USA (Dunifon et al. 2016). In our review, 56% (10/18) of the studies were conducted with US data, while 39% (7/18) was data from Asian countries and one study included data from Turkey.
Based on the results, living in a three-generation household is as likely to be beneficial as detrimental for grandparents. Of the results, 39% (9/23) showed a positive association, but 39% (9/23) also showed a negative association. Meanwhile, 22% (5/23) of the results showed a nonsignificant association (Table 2 and Appendix Table 7). Many of the articles examining grandparents' health or well-being while living in three-generational households also consider custodial grandparents, so in these cases, comparisons are often made between these two groups (e.g., Blustein et al. 2004;Goodman and Silverstein 2002;. The outcome measures that were utilized include, for instance, the following that can be counted as health measurements: depressive symptoms, self-rated health, functional/mobility limitations, and different stress factors and the following that can be counted as well-being measurements: happiness, quality of life, and life satisfaction (e.g., Tsai et al. 2013;Musil and Ahmad 2002;Ku et al. 2013;Hsu and Chang 2015;Goodman 2003;Yalcin et al. 2018).
Like the results concerning custodial grandparents, those involving grandparents in three-generation households may reflect the selection effect, meaning that grandparents living in three-generation households may already have poorer health than those in different living arrangements (Hughes et al. 2007). However, studies with longitudinal data (n = 11 studies, n = 13 results; Table 3) indicate that negative (e.g., Chen and Liu 2012;Hughes et al. 2007;Musil 2000) and positive (Tsai et al. 2013;Musil et al. 2011;Hughes et al. 2007) associations exist, even when the health or well-being of a grandparent is measured over time.
Positive and negative results were found in studies that were conducted with data from both the USA and Asia (e.g., 1 3 Chen et al. 2015;Chen and Liu 2012;Hughes et al. 2007;Tsai et al. 2013). Thus, the positive/negative results were not solely related to the country of residence; however, the results that were solely positive were more often found in studies with data from Asian countries (Guo et al. 2008;Ku et al. 2013;Tsai et al. 2013). Among three-generation household studies, 44% of them were conducted with representative data (8/18) and 56% with non-representative data (10/18). Among positive results 33% (3/9), among negative results 33% (3/9), and among nonsignificant results 60% (3/5), the data used were representative.

Grandparents living separately from their grandchildren
The largest group of caregiving grandparents, particularly in Western countries, includes those who do not live with their grandchildren but provide them with different kinds of support relatively frequently. In recent decades, an increasing number of studies have investigated the associations between active grandparenting and the health and well-being of noncoresiding grandparents. Most studies that focus on noncoresiding grandparents are conducted with European data (32%; 21/65), and over half of these (12 studies) contain data from multiple European countries. Meanwhile, 26% of the studies (17/65) were conducted with data from Asian countries, 23% (15/65) utilized data from the USA, 9% (6/65) utilized data from Australia, and the remaining 9% (6/65) involved data from other countries (e.g., Israel or Turkey) ( Table 1 and Appendix Table 8).
Non-coresiding grandparents are involved in the life of their grandchildren in many ways, and the most common measure of their involvement is grandparental childcare assistance (e.g., Ates 2017; Grundy et al. 2012;Xu et al. 2012). Grandparental involvement measures also include the frequency of contact between grandparent and grandchild (e.g., Taylor 2012, 2016;Danielsbacka and Tanskanen 2016;García-Campos et al. 2010), financial help, and other informal forms of assistance or emotional support (e.g., Fujiwara and Lee 2008). Also, grandparental health has been measured with various variables, such as self-rated health (e.g., Choi Danielsbacka and Tanskanen 2016;Nimrod 2008). In many cases, several measurements from both groups (health and well-being) are utilized in the same study.
Among non-coresiding grandparents, most of the results were positive (69%; 51/74). Meanwhile, only 19% (14/74) reported a negative association between grandparental involvement and well-being, while 14% (9/74) showed a negligible association (Table 2 and Appendix Table 8). The positive results were found from the data of European countries (e.g., Arpino and Bordone 2014;Mahne and Huxhold 2015), Asian countries (e.g., Luo et al. 2019;Park 2018), the USA (e.g., Hughes et al. 2007;Xu et al. 2017), and other countries (e.g., Grundy et al. 2012;Thiele and Whelan 2008). Thus, the positive results were not solely restricted to certain geographic regions. Furthermore, they were found among studies that contained cross-sectional data (e.g., Conde-Sala et al. 2017), longitudinal data (e.g., Di Gessa et al. 2016a), and methods for detecting causal relations [e.g., the IV approach (Arpino and Bordone 2014) or panel fixed-effect models ]. Since most studies utilize grandparental childcare support as an independent variable, this is the most common explanatory variable among the studies with positive results. Grandparental health or well-being were measured with several outcome variables, and thus, the positive associations were not restricted to certain health or well-being outcomes.
The negative results were most commonly accompanied with positive results (n = 9), and in these cases, the negative associations only applied to a certain grandparent group or outcome. Results that were solely negative were only found in five studies, which included associations between grandfathers' frequency of contact with a grandchild and decreased life satisfaction (Sener et al. 2008), a grandparent's centrality role and decreased psychological well-being (Muller and Litwin 2011), and grandparental childcare and increased depressive symptoms (Brunello and Rocco 2019).
Only nonsignificant results were found most likely among the studies that were not specifically focused on associations between grandparental involvement and well-being but considered a wider range of social connections or caregiving roles (i.e., caring for grandchildren was one measurement among others) (Hsu and Chang 2015;Nimrod 2008;O'Loughlin et al. 2017;Ward et al. 2019;Young and Denson 2014).

Discussion and implications
The present review includes articles that consider the associations between grandparenting and grandparents' health or well-being. In 68% of cases, custodial grandparenting was associated with decreased health or well-being of grandparents. Studies considering grandparents' health or wellbeing who live in three-generation households provided mixed results (39% positive; 39% negative). The involvement of non-coresiding grandparents was associated with improved grandparental outcomes in 69% of the results. Thus, the most negative results were present in the case of custodial grandparents, the most mixed results were among those that involved grandparents living in three-generation households, and most positive results concerned the case of non-coresiding grandparents ( Table 2).
We also considered whether the results were based on representative rather than non-representative data. Nonrepresentative data were most commonly used in studies focusing on three-generation households (56%) whereas representative data were utilized mostly in studies of noncoresiding grandparents (65%). Among custodial grandparent studies, positive results were most commonly achieved with representative data (50%, 6/12), in three-generation households representative data constituted 60% (3/5) of nonsignificant results and in non-coresiding grandparent group also nonsignificant results were most likely conducted with representative data (67%, 6/9). Two last mentioned proportions are from the category that had overall lowest number of results.
However, when we consider all the results that were included in the studies in this review, the overall proportion of positive, negative and nonsignificant results appear different (Table 4). As in many studies, several results were investigated that were either due to a differentiation in the grandparent subgroups (e.g., the grandparents according to gender) or multiple outcomes, so the same study may include several positive, negative, or negligible results. When all the results were considered (n = 452), the most common in every grandparent group was nonsignificant. In the case of custodial grandparents, 44% (67/151) of the results were nonsignificant, 62% (53/85) in the case of three-generation households, and 51% (111/216) for non-coresiding grandparents (Table 4). Although the number and proportion of nonsignificant results increased after all the results were counted, the proportion of negative results remains the highest among custodial grandparents (43%), while that of the positive results were those of the non-coresiding grandparents (36%). However, it is evident that after counting all the results that address the association between grandparental involvement and grandparental health or well-being, the overall evidence for significant results (either positive or negative) becomes weaker.
Furthermore, as previously discussed, the associations that were found may not be causal in nature, but rather, they may reflect the selection of different caregiving groups. To observe how well the abovementioned studies capture the causal nature of the associations, we have compiled a table of the results that are only based on longitudinal data and/ or methods that can detect causality. The ones that are most commonly utilized include panel fixed-effect models and instrumental variable approaches (Table 3). Overall, approximately 50% (59/117) of studies utilized longitudinal data and/or causal methods, and based on these, the most negative results were still found among custodial grandparents (57%; 12/21) and the positive among non-coresiding grandparents (72%; 26/36). However, a more detailed investigation reveals that only approximately 20% (22/117) of studies' methods can actually address the question of causality (Table 5). Still, based on these studies, the negative effects were most commonly found among custodial grandparents (50%; 3/6) and three-generation households (60%; 3/5), while positive effects were found among non-coresiding grandparents (50%; 6/12).
As was assumed, based on the hypothesis of Hertwig (2010, 2011), moderate grandparental involvement (e.g., that of non-coresiding grandparents) seems the most beneficial for grandparents, while negative effects were more common when grandparental involvement reached the highest level, like when grandparents became the primary caregivers for their grandchildren. However, as more detailed investigation has revealed, the most common result in all grandparenting contexts was nonsignificant. Also, although approximately half of the studies utilized longitudinal data, Negative results found among custodial grandparents and grandparents living in three-generation households may reflect selection effects (concerning poorer health), as previously discussed. However, we also found negative results in the longitudinal data that was conducted with methods able to detect causal associations. At least in some circumstances, therefore, the interpretation that becoming a custodial grandparent or living in a three-generation household is detrimental for grandparental health or well-being seems to be robust.
In the case of three-generation households, it is important to consider the reason behind these living arrangements, as grandparental co-residence could either be a result of their poor health (i.e., they need daily support) or stem from a need to take care of their grandchildren. For the former, grandparents may receive significant support from their adult children, which can improve their health and wellbeing, but they are in poor health to begin with. Regarding the latter, the grandparents are likely in reasonably satisfactory health to begin with but may become constant "nannies" for their grandchildren, causing extra strain that may have a negative health effect.
One of our aims was to investigate whether the results in three groups of grandparents would differ according to study population, i.e., whether they were dependent on cultural context. Our prediction was that being a custodial grandparent or living in three-generation household could be less detrimental or even positive for grandparents in Asian countries. However, the positive/negative results regarding custodial grandparents or three-generation households were not solely related to the country of residence, although the results that were solely positive were more often found in studies with data from Asian countries (Chung and Park 2018;Guo et al. 2008;Ku et al. 2013;Tsai et al. 2013). Regarding non-coresiding grandparents, the positive results in this group were also found from the data of European countries, Asian countries, the USA, and other countries meaning that the positive results were not solely restricted to certain geographic regions. To conclude, we did find some support for the prediction that living with grandchildren would have less detrimental effects for grandparents in Asian countries, but also that the distribution of positive and negative results did not follow strictly the geographic or cultural distinctions.
Although an extensive and increasing number of studies have investigated whether grandparenting is associated with the health or well-being of grandparents, some gaps still exist in the research. While studies with longitudinal data are well-represented, more studies are needed that analyze the causal nature of the associations. Furthermore, studies using longitudinal data (especially with several follow-ups), and even those with causal methods, cannot disregard that a health decline is inevitable among older adults. Thus, studies should concentrate on the relative health decline rather than health improvement (e.g., Chen and Liu 2012). Thus, a hypothesis may be that moderately involved grandparents would suffer from a slower health decline than their counterparts. However, one problem with longitudinal designs is that they may suffer from selective attrition over time meaning that people who experience health decline drop out from the survey. Another relevant direction would include studies that use cross-sectional data and causal methods (e.g., instrumental variable approach), as they may capture the causal effect more accurately without involving the aging effect or selective attrition over time.
In several studies, grandparental outcomes have been separated by gender or ethnicity, so some additional segregation or interactions could be relevant. For instance, it is wellknown that socioeconomic status is associated with health (e.g., Kim and Durden 2007), but few studies examine the interaction between socioeconomic status and grandparental involvement and its association with grandparental health or well-being (e.g., Chung and Park 2018;Mahne and Huxhold 2015). It is also well-known that lineage (i.e., whether a grandparent is from the maternal or paternal side) is strongly associated with grandparental childcare and being a custodial or coresiding grandparent . Still, surprisingly few studies consider this while studying the association between involved grandparenting and grandparental health or well-being (e.g., Danielsbacka and Tanskanen 2016). Number of grandchildren varies a lot across studies and also depends on the context of grandparenting. Custodial grandparents and grandparents living in three-generation household are commonly involved with grandchildren of one of their child whereas non-coresiding grandparents can be involved with grandchildren via several adult children. Not only the intensity of grandchild care but also the number of grandchildren to be cared for may affect grandparental outcomes. Thus, the number of grandchildren, especially the number of grandchildren via different children, is relevant factor to be considered in future studies.
Moreover, in the case of non-coresiding grandparents, the most common measure for grandparental involvement is childcare that is provided by grandparents. Looking after grandchildren without a parental presence may not capture all the positive aspects of being an involved grandparent. Thus, contact frequency or emotional closeness with grandchildren could be a more relevant measurement to understand the association between involved grandparenting and grandparental health or well-being. In addition, we have concentrated on studies that use the intensity of grandparental involvement as an explanatory variable but there are also other ways to compare grandparent types in respect to their health or well-being. For instance, the different styles of grandparenting (e.g., Neugarten and Weinstein 1964;Cherlin and Furstenberg 1985), different levels of reserves and strengths, or different role identities (e.g., Drew and Silverstein 2004) could lead different outcomes measured as grandparent health or well-being indicators.
The impact of grandparental involvement on grandparental health or well-being has been measured with varying measurements across the studies included in this review. This could be regarded as a limitation because the effects might differ regarding different outcomes. However, we did not find any clear biases on whether the associations would have been positive, negative, or negligible according to health or well-being measure used. This indicates that grandparental involvement may be similarly associated with various measures of health and well-being.
The question of how to increase the healthy years of one's life is crucial in contemporary aging societies, so whether time spent with grandchildren could promote health or wellbeing remains relevant. Policy implications concerning this review's findings are threefold. First, grandparents in custodial circumstances and three-generational households are the most vulnerable grandparent group, which policymakers should recognize. For them, caring responsibilities are not beneficial (although they are not solely detrimental either). Moreover, based on scant causal evidence, negative associations are not merely due to selection, which means that becoming or continuing a custodial/coresiding role as a grandparent could deteriorate health and well-being. However, especially in the case of these grandparent groups, cultural differences do exist and thus it is important to take into account the study population when the results are considered.
Moreover, among non-coresiding grandparents, their involvement is associated with improved health and wellbeing, although this association is not unequivocal. Despite the paucity of strong causal evidence, moderate grandparental involvement of non-coresiding grandparents should still be encouraged and enabled in terms of social policy decisions. Finally, we need more studies that can detect the causal nature of this association, as the lack of causal evidence concerns all three contexts of grandparenting.

Appendix
See Tables 6, 7 and 8.   Musil (2000) USA 74 grandmothers living in the same home as grandchild(ren), 49 primary caregiver grandmothers, and 25 with partial/supplemental responsibility; 39-72 years old; non-representative Self-assessed health, depression, parenting stress, anxiety, coping and social support neg. (parenting stress)       Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.