This study establishes older adult profiles according to the four AA pillars in Spain, and studies relationships between AA profiles and personal and contextual factors, well-being and quality of life. By doing so, it fills a gap in previous research: in scientific literature, the lack of consensus on formulating the AA model has been conditioned by discrepancies in materials and methods from a multidimensional perspective (59, 60).
As regards the first objective, consideration was given to AA's multidimensionality based on the construction of four pillar-related indicators: Health, Lifelong Learning, Participation and Security. The aim was to overcome reductionist approaches that do not address the theoretical model or others that do not distinguish between construct criteria and determinants (64), in order to consider other analytical or methodological approaches, such as the generation of the active ageing index (82) the empirical validation of the AA model (18) or the study of AA and its impact on survival (24).
Each pillar was built with multiple indicators from the SHARE dataset (63, 65, 66). Other authors, working with the same survey (24, 74, 83) or with other data (18), arrived at a similar selection of health domain-related indicators (diseases, dependency and physical or cognitive functioning) (84). This paper has expanded the health-related indicators by adding others linked to the use of medical services (85), nutrition (86) and alcohol consumption (87); all of them have been related to functionality, morbidity and/or mortality (85-87). As regards the Participation pillar, SHARE provides information on the type of activities carried out and their frequency, usually considered by most authors (88-92). For the Lifelong Learning pillar, consideration was given to basic skills (reading, writing or computer use) and involvement in training activities (93), while in the Security pillar, measures related mainly to financial security were considered, due to the limitation of the source. Other studies have approached Security as a pillar of manageable living conditions, such as physical security in the face of dependency (94), or the intuitive and lay understanding of older adults themselves (41).
The analytical procedure for handling all the information was planned and executed in successive phases, starting with the identification of patterns of relationships between variables based on principal component analysis. The cluster analysis conducted for each AA pillar has reflected how different older adults are as they grow older, revealing a wide variety of old age states, as the result of a process in which opportunities are used unequally within each pillar, as stated in the very definition of AA (17, 42). Furthermore, this very powerful analytical technique is influenced by the data set used and the research strategy (95). The resultant classification is similar to that obtained in other available reports (96). For instance, more than half of the subjects were classified in the "moderate health" cluster of the Health pillar. According to the National Health Survey (97), 45.5% of older adults in Spain regard their health to be good or very good. Slightly more than half of the older adults were grouped in the "low competence" cluster of the Lifelong Learning pillar; according to the same report, there is a predominance of older adults with primary education and no education (96). As for the Security pillar classification, the fact that almost 60% of the subjects are grouped in the "Self-assessed low economic status" cluster could be explained by the volume of inactive population in the sample studied (around 24% were still employed) and the effect of retirement on income (98). This is in line, firstly, with the individuals' own self-perception of their financial resources (99) and the reported difficulties in making ends meet, which are particularly noticeable in Eastern and Southern European countries (100), and, secondly, with the fact that they are close to the poverty line (96) which could be a limiting factor in promoting AA and enhancing the quality of life of older adults.
An optimal combination of the Health, Lifelong Learning, Participation and Security pillars will be key to achieving AA. Most authors hold this assumption and there have been attempts to build it into empirical models (101-103), yet few have succeeded in showing the interdependence between some pillars and others to show profiles of older adults ageing along diverse trajectories. This paper has demonstrated the interdependence of the pillars, giving five main profiles, which in turn were related to personal and contextual factors as well as to measures of well-being and quality of life.
Worth noting is that engaging in activities is present in all profiles except the one defined only by the Health pillar (profile II). Health, through the multiple measures used, either as a factor or as an outcome, is another area that is closely linked with the activity profiles. Indeed, a low perception of health, a limited level of functioning and unhealthy habits lead to less activity among older adults (28, 104), while more favourable health conditions encourage leisure and participation activities, such as volunteer work (33, 105). This is probably due to the fact that Participation is a cornerstone of the AA framework (92, 106) and is a defining element, as opposed to other related terms such as successful, productive, or positive (47, 84). Moderate or high participation is related to moderate health and economic conditions. These variables are associated in a multitude of studies and their interdependence is clearly evident (52, 107, 108). Health and Security seem to be the necessary elements (22, 41) underpinning Participation (106). In addition, better conditions in the Lifelong Learning pillar (15) were related to higher activity profiles.
By analysing profiles of older adults in Spain, this research has identified a wide range of factors that give them interpretative consistency, but which do not always match those offered in AA literature, either because they do not follow the same theoretical basis or because they do not always use the same analytical methodology. That is why the AA profiles are constructed with quantitative methods that combine independent variables, generally at the individual level, with others that express the results of the AA process or other related facts, such as quality of life or subjective well-being as outcome variables. As a consequence, it is often hard to clearly distinguish between dimensions, determinants and outcomes, because the analyses are too closely linked to the available data. However, so of the many different factors that influence older adults' active behaviours are far more prevalent than others.
The basic demographic variables, age and gender, are part of the most common interpretative construct because they feature in all studies, whatever their type. Age plays a preferential role, yet it tends to act in two directions: firstly, by appearing in the least active groups (109) and, secondly, by influencing the reduction in the number and type of activities as the population ages (28, 52), although it is not always documented to work this way (110). Similarly, the age variable shows a different association by type of active elderly, depending on whether the activities are carried out at home (at older ages) such as family help or home maintenance, or outside the home (at younger ages) such as volunteer work (95). The fact is that, by including other age-related variables, this activity trajectory is also related to living without a partner, with lower economic income (104) and a decrease in personal well-being (50).
Women's involvement is greater in some specific profiles, such as those involving caring for people or activities in the home, or less when it comes to profiles of people still linked to the world of work or volunteer work (95, 109-111). In the case of Spain, the life trajectory of these post-Civil War (post-1939) generations, marks an appreciable difference in gender roles, although recently women seem to be more interested in carrying out 'novel' and motivating activities, which are more rewarding and which allow them to recover a role hitherto not usually assigned to them (29). Men of these generations behave more conservatively and are more attached to the closer and less active social community space.
The level of education, measured by the number of years spent in the system and the level attained, is another factor that conditions the activity profile, through general rules: a lower educational level tends to be associated with less activity (28, 110) and less rewarding or motivating activities, but of a compulsory nature in the family sphere (110). The profiles obtained also show intergenerational educational level-related gains. For instance, profiles IV and V are more defined by the Lifelong Learning profile, with younger ages and a higher level of education. Different studies point to the country's older adults having higher levels of education, making it possible to reduce the gender gap in old age (112).
From a life course perspective, the population studied includes people who are old enough to be retired from work or who are carrying out household tasks, as the main activity-related groups. Both can guide their transition into retirement through a variety of possibilities (113), from those requiring remuneration to those undertaken on a voluntary basis (114) or to maintain intergenerational care relationships (115). However, activity-relatedness is not a factor in many AA studies, probably because of the limited ability to discriminate if the vast majority of the population is already retired or because it is mediated by other variables such as age (116). However, this factor becomes relevant when analysed together with many others to relate AA to quality of life (48, 56).
Another way of influencing activity is through concomitant variables, such as level of income, so that education and economy are associated in determining activity profiles (105), or marital status to indicate that people who live alone and have a low level of education behave in a similar way (95). Precisely, beyond marital status, the form of cohabitation, the size of the household and having children and grandchildren are relevant variables in the differential characterisation of activity profiles. The key could be found in whether there are children (or even grandchildren) in the household, or within the family network but living outside the household, in more or less close environments and with more or less frequent contacts in an ascending familialism or supportive-at-distance typology (117). In the first case, a larger household size and reporting having few children and grandchildren is consistent with a profile of younger people, and, in general, men, people living in a couple and with others, possibly children yet to be emancipated, who maintain a diverse and balanced activity (profile IV). Something similar happens with profile V, but in this case they would be women. At the same time, having more children and grandchildren corresponds to low activity profiles (profiles I and II): people living alone, in smaller households, older and, above all, women. Yet having more children and grandchildren could also tend to lead to activities in the home or family care environment that compete with other leisure and participation activities for the person's available time, in order to reconcile tasks of different types and nature (110). The latter could be the case of the profile of limiting conditions for AA (profile III), which is observed among not very old women who say that they have more children and grandchildren both inside and outside the home, and which would also correspond to a descending familialism typology and activity based on intergenerational family solidarity provided by older women (118). In any case, profiles I, II and III show higher reported loneliness, compared to lower scores for profiles IV and V, which would be related not so much to the size of the family network but rather to other factors such as increasing age and changes and lost in marital status, income, self-rate health, cognitive functioning and depression (119, 120), aspects that are also related to limiting conditions for AA and maintaining a good quality of life (121).
Personal motivation (or a lack thereof) as well as personal rewards (life satisfaction) and social rewards (social networks, avoidance of loneliness) also contribute to understanding the active behaviour of older adults (106). It has been found that having a higher number of people in one's social network is associated with higher levels of activity, while a less dense network is associated with lower activity, although perceived support may act in the opposite direction (28, 104). On the other hand, the importance of the social and community environment in which the activities are carried out must be assessed as a mechanism for reinforcing them (106).
As regards other contextual conditions, older adults tend to reside in cities, especially medium-sized ones(122), which mirrors the process of urbanisation and demographic ageing (123-125). In this study, no homogeneous pattern has been observed according to the two large profile groups, such that both profile I and V subjects reside in large urban and metropolitan areas, while the remaining ones do so in smaller cities. In any case, the trend towards urbanisation has led to the development of a specific city friendliness programme in order to optimise the living conditions and quality of life of older adults (30, 126).
With respect to the residential environment, home ownership is the most significant regime in Spain compared to other neighbouring countries (127), and among the older population it reaches higher proportions in line with their age and the time they have had to acquire it (128). The results show that people with the worst AA conditions (profiles I, II and III, located in the low scores of dimension 1 of the perceptual map) showed slightly lower percentages of ownership compared to the profiles of better positioned subjects, in accordance with their greater purchasing power. In relation to the type of residence dwelling, two situations were observed; on the one hand, older adults with a moderate active profile, living to a greater extent on a farm or in family housing, in line with their location in smaller residential areas, and, on the other hand, the profiles of younger people with better AA conditions, living in housing in block buildings in line with their settlement in large cities and metropolitan areas. In Spain, part of the older population faces the problems of an ageing housing stock characterised by a lower level of amenities (lifts, heating, air conditioning) and the need for renovations, which worsen their isolation, hinder the desire to grow old at home with autonomy and independence, and jeopardise the promotion of AA (128, 129).
Other factors may also influence the level of activity, but their effects are not differentiated because they are incorporated into the more general variables. Something very similar happens when we try to measure the impact of carrying out more or less activities of one type or another on personal well-being, quality of life or satisfaction with it. These are very general social and multidimensional constructs, in which it is not the influence of all their conditioning factors is not easily identifiable, and their effects may be contradictory depending on the research design and the data used (50, 130). The relationship between AA and personal well-being (including life satisfaction, quality of life, satisfaction with social networks, absence of perceived loneliness) has been highlighted in the profiles of older adults who are more competent and with better personal and contextual conditions to have a high level of activity, in line with the high association of these constructs (19, 92).
Constructing an ageing model based on a broad set of variables, in order to identify profiles of older adults with different degrees of activity, is a significantly increasing trend in the literature, and one that uses a methodology based on individual data with multidimensional variables: some that measure different activities, the "process" variables (130) while others measure the person's situation and which the AA model accepts as determining factors. Yet the multidimensional approach is also entails far more complex, as this paper has shown with regard to the construction of AA profiles. The use of quantitative data, from SHARE or other European and North American databases, has highlighted the potential of this classification strategy, both in terms of the activities analysed and the determinants that serve to explain the types of activity and/or profiles of older adults, measured from different perspectives (individual, countries) and supported by different theories (28, 38, 50, 52, 92, 109, 110, 131-133). This paper has also revealed a far from negligible diversity of results influenced by the population samples and the variables selected and available for analysis (104, 133, 134). Furthermore, one must not lose sight of the interpretative capacity of using qualitative information in the study of AA profiles (29). The tendency, however, is that the WHO AA model is not usually considered as the reference to be followed in studies on activity profiles and older adults, and when it is, not all dimensions and determinants are covered (133). It is much more common to use various unidimensional, multidimensional or behavioural models, according to Boudiny (135), based on successful, healthy or productive ageing theories, using specific sources that do not make it easy to standardise results. The sample of studies cited above are good evidence of this.
Limitations and future lines
It must be noted that this research was subject to certain limitations. The first stems from the difficulty of finding data on active ageing (136). This study used a database, the SHARE project, which is characterised by its rich multidimensional design, and the fact that it studies a large number of countries, thus permitting cross-sectional and longitudinal comparative studies. However, this survey is not designed to specifically survey AA. So, from a thematic approach, this dataset does not offer all the information defined in the AA paradigm (17, 42). In this regard, an unequal number of variables have been used per pillar, which also conditions the different number of variables involved in its construction, on the one hand, and a possible bias in the results, on the other.
The larger number of indicators available in the SHARE survey matches the areas of greatest scientific development within AA, namely Health and Participation, with a lower presence of questions related to the Lifelong Learning and Security pillars, despite their proven relevance in positive ageing trajectories (14, 103, 137).
As regards the variables selected in the Participation pillar, almost 30% of the participants had missing values. Therefore, during the PCA of this pillar, these values were replaced by the mean of the variable. This may have influenced the results obtained for this pillar in the first CLA run, as almost 95% of the cases were grouped around a single cluster. This was the reason why, for this set of variables, the initial centroids were user-defined.
Despite these limitations, the research also has certain strengths, including the methodological design to address the study of a large dataset of different types of data. Consequently, the successive analytical procedure phases have been expressly planned and executed for the proposed objectives.
In the multiple and diverse AA studies, there is room for future and novel developments stemming from their conceptualisation and progress and from the aforementioned limitations to achieve more precise diagnoses. Some of the possible improvements in these studies should come from the need to establish comparative frameworks between countries, differentiated by their social, cultural and political model, thus overcoming the reductionism imposed by research anchored in, for example, developed countries. Although this is an increasingly widespread trend, there are two other areas that would require more attention, such as cross-referencing and further triangulation studies. Both would stem from longitudinal type analyses, albeit constrained by the availability of adequate data, and the use of combined quantitative-qualitative methodologies, which would make it easier to compare the two visions and provide deeper insight into the views and experiences of older adults.
Finally, there is another area for improvement in AA research, derived from the use of pre-post methodology, which enables social interventions to first assess and then improve the behaviour of older adults. This knowledge would underpin the application of public policies aimed at promoting active ageing as a mechanism for consolidating quality of life in the ageing process.