Functional outcome for older adults with movement disabilities : A cross-sectional study

Functional

outcome

for

older adults with

movement

disabilities

MAIN FIELD: Gerontology AUTHOR: Nynne Olsen SUPERVISOR: Deborah Finkel JÖNKÖPING 2021 May

Summary

Background: Previous research have found that different factors have associations with the level of

function. Only few studies investigate the population of older adults with movement disabilities.

Objective: The objective was to investigate how childhood socioeconomic status, education, gender,

rural/urban living, and cohort affects functioning in late adulthood for people with movement disabilities. Six hypotheses were tested.

Method: The sample was selected from the Swedish Adoption/Twin Study of Aging, and consisted of

n=69 older adults with self-reported movement disability, mean age 78 years. They have all participated in motor function testing, which is an objective measure of function. Mann-Whitney U test, Kruskal-Wallis H test and binary logistic regressions were performed.

Result: No significant difference were found between low/high childhood socioeconomic status,

low/high education, men/women, rural/urban, and early/late cohort. Associations were found between age, urban living, later cohort and poorer functional level.

Conclusion: The results indicate that the older adults from Sweden aging with a movement disability

might have equal opportunity to develop and maintain functional ability. The participants all have a movement disability and it is possible that the disability is the main factor determining the functional level.

Keywords: Cohort, childhood socioeconomic status, gender, rural/urban, The Swedish Adoption/Twin

Sammanfattning

Funktionellt utfall för äldre med rörelsenedsättning

en tvärsnittsstudie

Bakgrund: Tidigare studier har funnit olika faktorer som har associationer med funktionell nivå. Få

studier har undersökt en population av äldre med rörelsenedsättning.

Syfte: Syftet var att undersöka hur socioekonomisk status i barndomen, utbildning, kön,

landsbygd/stad boende, och kohort påverka funktion hos äldre med rörelsenedsättning. Sex hypoteser blev testade.

Metod: Urvalet bestod av en mindre del av deltagarna från the Swedish Adoption/Twin Study of Aging,

bestående av n=69 äldre med självrapporterade rörelsenedsättning, medelålder 78 år. De har alla deltagit i test av motorfunktionen, vilket är ett objektivt mått på funktion. Mann-Whitney U test, Kruskal-Wallis H test och binary logistic regressions test genomfördes.

Resultat: Ingen signifikant skillnad hittades mellan låg/hög socioekonomisk status i barndomen,

låg/hög utbildning, män/kvinnor, boende i landsbygd/stad och tidig/sen kohort. Associationer hittades mellan ålder, boende i stad, sen kohort och sämre funktionsnivå.

Slutsats: Resultaten pekar på att äldre vuxna från Sverige som åldras med rörelsenedsättning kan ha

samma möjligheter att utveckla och bevara funktionell förmåga. Alla deltagare har rörelsenedsättning, och det är möjligt att detta är den faktor som i högst utsträckning påverkar funktionsnivån.

Nyckelord: Kohort, socioekonomisk status i barndomen, kön, landsbygd / stad, The Swedish

Content

2.2 Health and wellbeing living with a movement disability ... 2

2.3 Factors to investigate ... 3

2.3.1 Socioeconomic status and education ... 3

2.3.2 Gender ... 3 2.3.3 Rural vs. Urban ... 4 2.3.4 Cohort ... 5 2.4 Measures of function ... 5 2.5 Theoretical framework ... 6 3. Objective ...7 4. Method ... 8 4.1 Research design ... 8 4.2 Data collection ... 8 4.3 Sample ... 8 4.5 Measures ... 9 4.6 Statistical tests ... 10 4.7 Ethical considerations ... 11 5. Result ... 12 5.1 Descriptive statistics ... 12 5.2 Results H1-H5 ... 13 5.3 Results H6 ... 18 6. Discussion ... 22 6.1 Method discussion ... 22 6.1.1 Research design... 22 6.1.2 The sample ... 22 6.1.3 Measures ... 23

6.1.4 The statistical tests ... 23

6.2 Result discussion ... 24

6.3 Future research ... 26

7. Conclusion ... 27

8. Acknowledgements ... 27

9. References ... 28

Abbreviations

ADL Activity of Daily Living

IADL Instrumental Activities of Daily Living

IPT In-person testing

IQR Interquartile range

OR Odds ratio

Q Questionnaire

SATSA the Swedish Adoption/Twin Study of Aging

SES Socioeconomic status

WHO World Health Organization

1

1. Introduction

The populations around the world are getting older, people live longer and a greater percentage of the population is 65+ years. In 2000 12.4 % of the American population was 65+ years and it is estimated by 2050 that 20.2 % will be 65+ (Halaweish & Alam, 2015). The European population is old compared to world average and is still getting older. For instance in Germany 20.7 % of the population was 65+ in 2010 and is projected to be 32.3 % by 2050, in Sweden it was 18.1 % and projected to 24.5%, and France was 16.6 % and projected to be 26 % by 2050 (Eurostat, 2011). Aging, or increasing age, is associated with physical deterioration, reduced function, more dependencies and disability (Aijänseppä et al., 2005). However, the aging population is a heterogenetic group and has large individual differences when it comes to physical function and health. In other words, aging has an association with decreased function, but chronological age has only a loose association (WHO, 2015). Concepts such as healthy aging and successful aging try to capture and describe an old age with independence and function. Healthy aging as defined by WHO, is the process of developing and maintaining the functional ability

that enables well-being in older age (WHO, 2015, p. 28 ). Successful aging has no standard definition,

however the focus is on prolonging physical, functional, social, and psychological health (Urtamo et al., 2019).

Aging with a movement disability might create more challenges to keep a sufficient functional level and reach healthy and successful aging. The concept of double jeopardy describes two disadvantages occurring simultaneously. In healthcare double jeopardy can mean that a minority group will experience an increase of health inequality with aging (Ferraro & Farmer, 1996). The physical deterioration associated with aging (Dehlin & Rundgren, 2014) and aging with a movement disability could create a double jeopardy. In other words, older adults with movement disability may face extra burdens and challenges as age-related changes in physical functioning progress.

The populations around the world are getting older and there is an interest in investigating the older populations about physical function, health and wellbeing. This study aims to focus on the group of older adults with a movement disability: to understand how movement disability and other factors might interact and affect functional ability in old age. This could potentially illuminate aspects of Healthy aging for people aging with a movement disability.

2. Background

2.1 Movement disabilities

Movement disabilities can arise from a number of different causes. People might be born with birth defects, deformities or diseases causing movement difficulties, or get/develop a disease later in life such as multiple sclerosis, polio, spinal cord injury, cardiovascular diseases, stroke, arthritis, other diseases, or amputation which can cause movement disabilities. It can also be caused by restrictions in movement

2

pattern due to pain, tendon contractures, muscle weakness, or obesity. Measures of movement disability can be self-reported, having a diagnosis causing the disability or having limitations in certain activities (Rosenberg et al., 2011). People aging with movement disabilities have lower physical activity compared to people aging without movement disabilities (Rosenberg et al., 2011). Developing and retaining function when aging with a movement disability might be more challenging compared to maintaining function when aging without a movement disability. Therefore, it is of interest to identify risk factors that can affect the functional outcome for an aging population with movement disabilities. Different factors can be of interest, and will be introduced below.

2.2 Health and wellbeing living with a movement disability

The definitions of health and wellbeing affect how to view health and wellbeing for people living with movement disabilities. If health is defined with a holistic perspective as healthy aging “developing and

maintaining the functional ability that enables well-being” (WHO, 2015), then health can be achieved

by people with movement disabilities. Wellbeing is a concept often used but hard to define. Dodge et al. (2012) describes how various examples are presented in the literature to describe wellbeing, examples such as happiness, positive affect, life satisfaction, positive psychological functioning, human development and quality of life. As these terms merely describe wellbeing, they propose to define wellbeing as “the balance point between an individual’s resource pool and the challenges faced” (Dodge et al., 2012, p. 230). Resources and challenges refer to psychological, social and physical factors. Lack of challenges will lead to stagnation and thereby impact the balance point, also meaning the wellbeing (Dodge et al., 2012). People with movement disability can achieve wellbeing if they have the resources to deal with their challenges. For example using walking aids, living in an environment where they can manage moving around, adapting or coping with their movement disability.

The physical aspect of health and wellbeing is affected by the physical disability, but so are the aspects of psychological and social wellbeing. Rosso et al. (2013) found that low mobility both with and without disability was associated with lower social engagement in a large sample of older adults (n=676). Mobility was measured using the Life-Space Assessment, and disability was assessed by ADL and IADL. Social engagement was measured as usage of senior center and social engagement in organizations, frequency of phone conversations with friends and family, and usage of internet (Rosso et al., 2013). Social engagement can be important for wellbeing, but the quantity might not express the quality of the social engagement. A previous study investigated groups of older adults, one group only with diseases (n=186) and the other with diseases and disability (n=168). They found that the group with both diseases and disability was less likely to participate in social activities and that they were less satisfied with life compared to the group of older adults only having diseases. However, for the group of older adults with disability and diseases the association between life satisfaction and social engagement was strong and significant (Jang et al., 2004).

3

2.3 Factors to investigate

2.3.1 Socioeconomic status and education

Socioeconomic status (SES) and education can affect health outcomes in aging. SES can be measured in different ways and education is often used as a marker of SES. A previous study based on older adults from China investigated functional decline measured as ADL and IADL decline over a 3-year period. They found an association between lower SES factor and higher incidence proportions of decline in IADL function, when measuring SES factor as a composite of education and household income (Beydoun & Popkin, 2005). Data collected among older adults from Italian showed lower SES, as < 5 years of education, was associated with lower gait speed, the association effect was decreased after controlling for six physiological measures of impairment but remained significant. However, function measured with the Short Physical Performance Battery did not remain significant after controlling for the physiological measures (Coppin et al., 2006). A study testing both level of education and household income as measures of SES in 11 European countries, found that socioeconomic inequalities in health persist into old age. Health included three measures, self-reported health, decrease in ADL and long-term disabilities. These inequalities differ between countries, gender, age, educational level and net household income (Huisman et al., 2003). A study using data from the English Longitudinal Study of Aging (ELSA) found that lower SES is associated with a more rapid decline of functions even when testing for covariates such as childhood SES and education among others. They measured SES as wealth (financial, housing and physical) and function as physical, cognitive and social measures in six different domains (Steptoe & Zaninotto, 2020). A Swedish study investigated income, education, social class, occupational complexity, and a composite measure of SES to explore their relative ability and importance to determine health in old age. Longitudinal data was used from the Swedish Level of Living Surveys (LNU) and the Swedish Longitudinal Study of Living Conditions of the Oldest Old (SWEOLD), and health was determined as self-reported measures of mobility limitations, limitations in ADL, and psychological distresses. They found that all measures of SES were significantly associated with late life health, but income had a strong association and the only one to remain significant after adjusting for the other SES variables (Darin-Mattsson et al., 2017). Childhood SES can be a covariate when measuring SES (Steptoe & Zaninotto, 2020), but also a measure to explore when investigating health in old age. A previous study using data from the Swedish Adoption/Twin Study of Aging (SATSA) investigated associations between cognitive aging and childhood SES. The level of cognitive performance was associated with childhood SES, and education predicted some of the association (Ericsson et al., 2017). There are many different ways to measure SES and they are connected, but might also have independent associations with health and functional outcomes in aging (Avlund et al., 2003; Darin-Mattsson et al., 2017).

2.3.2 Gender

Gender can also affect health and functional level in aging. Women compared to men have poorer functional outcomes measured in performance-based motor function factors; however, in old age men

4

develop the same amount of difficulties as the women (Bravell et al., 2017). Education and income affect health in aging differently between genders. The odds ratio (OR) of developing long-term disabilities among women with low income and lower education decreased with age (Huisman et al., 2003), which could indicate the age-as-a-leveller theory that inequality between low and high income/education decreased in aging. However, for the men the OR of developing long-term disabilities does not follow the same pattern and shows increases in old age (Huisman et al., 2003), which could indicate double jeopardy meaning that men with low education or low income experience more health inequality in aging. Among men and women 80+ years, men with the lowest income have significantly higher risk for developing long-term disabilities compared to women in the same income group; this gender differences is also shown for the group with the lowest education (Huisman et al., 2003). A study measuring ADL, IADL and mobility in 8 categories found that women were more likely to report functional limitations and had higher levels of disability. However, no significant differences were found in function between the group of men and women who had limitations (Murtagh & Hubert, 2004). Based on these findings, the hypothesis would be that there are no gender differences in the level of function when investigating a population of older adult with movement disabilities.

2.3.3 Rural vs. Urban

Urban and rural differences in function have been studied. A previous study found poorer outcomes in function for older adults from China living rural compared to living urban (Yiengprugsawan et al., 2019), function was measured with the WHO Disability Assessment Schedule 2.0 (WHO DAS 2.0), which covers six domains of functioning (WHO, 2010). Another Chinese study investigated the healthcare access effect on function, measured as IADL and ADL, in an urban and rural setting. They found that older adults with inadequate access to healthcare had higher levels of functional disability, and this association was stronger in a rural setting, in addition to that, a higher percentage of older adults living rural had inadequate access to healthcare (Zhang et al., 2017). The lack of access to specialised healthcare also causes lower levels of physically-related Quality of Life for people living with multiple sclerosis rural in America (Buchanan et al., 2008). A Finnish study found less health-related behaviour, in terms of eating habits and exercise, in older adults living rural compared to semi-urban and urban (Fogelholm et al., 2006). However, a Australian study found an association between being adequately physical active and living rural (Lim & Taylor, 2005). Transportation options have significance for mobility in an older population. The ability to drive, the amount of public transportation, or services available and affordable affect movement outside the home. An American study found that older adults living rural are less likely to believe they have adequate transport options, and problems with the public transportation identified by the urban population are concerns about crime, getting to the stops and inadequate shelter while waiting (Mattson, 2011). The urban/rural differences in function, as found in previous studies, can be a result of different factors including access to healthcare, health-related behaviour, ability to drive and transportation options, but also that older adults living rural tend to have lower SES and lower education compared to the urban population (Fogelholm et al., 2006). When investigating physical function and health these urban/rural differences have been measured around

5

the world, nevertheless the differences can vary between provinces and countries, as a result of different social and health policies.

2.3.4 Cohort

Cohort can affect health and functional outcome in late adulthood, for the reasons that different events such as war, pandemic, and policies on economics, healthcare and education affect people differently depending on when in the individual’s life these events occur. Previous studies have found improvements in function over succeeding cohorts with function measured as: five aspects of ADL (Spiers et al., 1996), self-reported disabilities in performing climbing several flights of stairs and running 100 meters (Malmberg et al., 2002), and with self-administered questionnaires on capacity to perform ADL with focus on self-care and mobility (Aijänseppä et al., 2005). Functional impairment increases with aging but decreases over succeeding cohorts (Aijänseppä et al., 2005). However, previous studies have also found opposite results. In a Finnish study, the non-institutionalized older population was found to significantly decrease mobility and increase dependencies in ADL within a 10-year interval, in which time the number of non-institutionalized older adults increased 1.5 –fold (Anttila, 1991). A UK study found higher levels of self-reported disabilities, especially conditions such as arthritis and chronic airway obstruction, with succeeding cohorts (Jagger et al., 2007). Factors that can affect an increase or decrease in function among the aging population include health-related behaviours such as alcohol consumption, smoking, eating and exercise habits, which can change among cohorts (Mawditt et al., 2016). The physical activity/sedentary behaviours people have in mid-life are predictors for physical activity later in life (Hamer et al., 2012), which potentially change between cohorts. A study using longitudinal data from SATSA found that low and high education affected functional ability differently across early and later cohorts. The low and high education groups in the early cohort had medium trajectories for functional decline, and the low education in the later cohort had the worst trajectory for functional decline and the high education group in the late cohort had the best functional trajectory (Finkel & Ernsth Bravell, 2020). This means different factors can affect functional or health outcomes in different cohorts.

2.4 Measures of function

The previous studies have generally found that poorer function seems to be more prevalent among people with lower SES, lower education and living rural however, the results are equivocal. The function or functional level are measured in many different ways. Commonly used are self-reported measures of ADL and IADL (Aijänseppä et al., 2005; Beydoun & Popkin, 2005; Huisman et al., 2003; Murtagh & Hubert, 2004; Zhang et al., 2017). There are also self-reported measures of climbing stairs, running 100 meters (Malmberg et al., 2002), physical activity (Steptoe & Zaninotto, 2020) and WHO DAS 2.0 (Yiengprugsawan et al., 2019). However, there are also a number of objective functional measures such as gait speed, grip strength and chair stand (Coppin et al., 2006; Steptoe & Zaninotto, 2020), and performances based motor function. A series of studies used measures of motor function comprised of 20 functional tests collected in 3 factors; the balance factor, the fine motor factor and the flexibility factor

6

(Bravell et al., 2017; Finkel & Ernsth Bravell, 2020). The objective measures provide higher quality information, and this study used the 3 factors in the performance-based motor function. Only a few of the previous studies focus on people with disabilities (e.g. Coppin et al., 2006; Rosenberg et al., 2011), which was the focus in this study.

2.5 Theoretical framework

Healthy aging and the public-health framework for healthy aging as compiled by WHO have a holistic

view. The aging population is a heterogenic group and aging is an individual experience, therefore the concept healthy aging needs to be inclusive of all people. Older people as a group are more likely to have one or more diseases. However, how the disease affects and interacts with their functioning is far more interesting than the disease itself. Functioning in healthy aging refers to physical and mental capacities, and interaction with the environment. Personal characteristics such as gender, education, occupation and wealth, can impact an individual’s ability to develop and maintain functioning (WHO, 2015). This study wanted to illuminate associations between personal characteristics and physical function for people with movement disabilities.

The concept of double jeopardy and the contrary concept age-as-a-leveller are both interesting when investigating health and aging. Double jeopardy refers to double disadvantage when aging in a minority group, and age-as-a-leveller refers to health disadvantages levelling out in aging (Ferraro & Farmer, 1996). These concepts have been used to examine the interaction of race, socioeconomic position, gender, sexuality and health when aging (Barrett & Barbee, 2017; Chatters et al., 2020; Ferraro & Farmer, 1996; Green & Benzeval, 2011).

As this study used cross-sectional data from a group of older adults with movement disability, this study cannot investigate if the minority group of older adults with a movement disability experience the double jeopardy or age-as-a-leveller. This research can use the concepts to investigate if, in the group of older adults with movement disability, there are personal characteristics such as childhood SES, education, gender, rural/urban living, and cohort that affect functional ability and because of the cross-sectional data only indicate possible inequalities in the sample. This research may be the first attempt to examine how aging and disability interact to affect outcomes.

7

3. Objective

The objective of this study was to investigate how childhood SES, education, gender, urban/rural, and cohort affect functioning in late adulthood for people with movement disabilities. To investigate this objective six hypotheses were tested.

H1: Older adults with movement disabilities who had higher childhood SES have higher functional level H2: Older adults with movement disabilities who have higher education have higher functional level H3: There are no gender differences in functional level among older adults with movement disabilities H4: For older adults who have movement disabilities, the functional level is higher for people living in

urban rather than rural settings

H5: For older adults who have a movement disability, the functional level is higher for people born in a

later cohort than those born in an earlier cohort

8

4. Method

4.1 Research design

A quantitative research approach was used, meaning that the research question was investigated through objective measures and statistical testing (Leung & Shek, 2018). More specifically using the hypothetico-deductive design, where the hypotheses are set up according to existing theories and previous research, and the hypothesis were either verified or falsified through statistical testing (Haig, 2012; Park et al., 2020). H1-H5 have a non-experimental comparative design (Leung & Shek, 2018) and aim to compare different subgroups of people with movement disabilities and therefore cross-sectional data was used to answer this objective. Cross-sectional data are gathered from participants´ at one point in time, which provides an image of the participants´ life at that particular moment of data collections (Liu, 2011). H6 has a non-experimental correlational design investigating the relationship between the different moderators and the outcome variable (Leung & Shek, 2018), still using the same cross-sectional data sample.

4.2 Data collection

The sample in this research have been selected from a database containing data from the Swedish Adoption/Twin Study of Aging (SATSA). SATSA was a longitudinal study that started collecting data in 1984 where the first questionnaire (Q) was sent to 2854 individuals (Pedersen et al., 1991). The participants were twins from the Swedish Twin Registry, where a noticeable amount of the twins were separated in early childhood. This made it possible to study genetic and environmental impact on aging (Finkel & Pedersen, 2004). The sample contains a base of twins reared apart and a control sample of twins reared together (Pedersen et al., 1991). The study collected data through a comprehensive questionnaire, Q1 in 1984, Q2 in 1987, Q3 in 1990, Q4 in 1993, Q5 in 2004, Q6 in 2007 and Q7 in 2010 (Finkel & Pedersen, 2004; Karolinska Institutet, n.d.). The Q contain questions concerning health and health-related behaviours, personality, work environment, rearing environment and adult family environment (Finkel & Pedersen, 2004; Pedersen et al., 1991). Data were also collected through in-person testing (IPT) the first one in 1986-1988 (IPT1), IPT2 in 1989-1991, IPT3 1992-1994, IPT4 1995-1997, IPT5 1999-2001, IPT6 2002-2004, IPT7 2005-2007, IPT8 2008-2010, IPT9 2010-2012 and IPT10 2012-2014 (Finkel & Pedersen, 2004; Karolinska Institutet, n.d.). The IPT had a biomedical component containing general health, performing and functional capacity, and a cognitive component testing crystallized and fluid intelligence, and memory. The IPT was performed by nurses and took on average 4 hours (Finkel & Pedersen, 2004; Pedersen et al., 1991). The database has collected information from n=2209 (pairs=758). However the IPT have collected data from n=859 (pairs=403).

4.3 Sample

The inclusion criteria for this research are older adults with movement disabilities, who have participated in at least one IPT. The movement disability is self-reported by answering ´Do you have a

9

movement disability?´ yes or no. If yes, there was a follow-up question ´How much does the movement

disability complicate your daily life? Answered, 1 = not at all, 2 = a little, or 3 = a lot. Only 69

participants met these criteria and therefore the sample consisted of n=69. The data in this sample consisted of information collected from the first Q and the last IPT of participation, meaning that the data have been collected at different Q and IPT for the participants. Using data from the last IPT of participation and not from one specific IPT was chosen to maximize the sample size and to maximize participants’ age.

4.5 Measures

Different moderators, as childhood SES, education, gender, living rural/urban and cohort, were tested to investigate the effect on functional level for older adults with movement disabilities. In the literature there are different examples on how to measure childhood social class, for example parental occupation and a measure of household social class (Ericsson et al., 2017), or a calculated variable composed of 3 items as residents/room ratio in childhood home, how many sharing their sanitation facility and parents occupation (Staff et al., 2018). In this study childhood SES was calculated from 6 items: education and occupation of parents, housing density in childhood (rooms/person), parents had a boat or summer cottage, the family´s economic situation compared to others, and if the family´s income met their needs. These 6 items were standardized, so when summed each item added equally to the childhood SES variable. The childhood SES was divided at the median into low SES and high SES. It was chosen to calculate the childhood SES from these 6 items because they are believed to give a better understanding of the childhood social class.

The participants self-reported their level of education including primary education (n=43), lower secondary education (n=19), upper secondary education (n=1) and tertiary education (n=3). This sample provided very small numbers of participants with upper secondary- and tertiary education, and therefore, the sample was divided in low education including primary education, and high education including lower secondary-, upper secondary - and tertiary education.

Information was collected about the participants living location, as countryside, small town, city or large city (e.g. Stockholm). No definitions of countryside, small town, or city were provided and therefore the participants’ responses are based on their own interpretation of these possible answers. Based on the information available it was logical to divide the data into rural including countryside and small town, and urban including city and large city.

The participants were born from 1900-1943, and were divided in an early cohort (<1925) and a later cohort (≥ 1925). Dividing the cohorts this way it based on a fairly even birth-year interval and that a previous study investigating cohort differences in a SATSA sample used this cohort division (Finkel & Ernsth Bravell, 2020), which makes it easier to compare with previous findings.

10

The functional level was measured in 3 factors of motor functioning: balance factor, fine motor factor and flexibility factor. These 3 factors will not be added together as one functional level, or as one measure, but be used as 3 independent measures of function. The balance factor was calculated from 10 function variables including for instance 1) standing up from a chair with the arms crossed in front, 2) lifting a drinking glass, and 3) touch the right toes with the left hand. The fine motor factor was calculated from 8 function variables including task such as 1) pouring water from one glass to another with the dominant hand, 2) inserting a key into a lock and turn, and 3) screw in a light bulb. The flexibility factor consist of 2 variables of function including 1) reach right hand behind head to touch left earlobe, and 2) reach left hand behind head to touch right earlobe (Bravell et al., 2017). See all the 20 motor function variables in appendix 1. The functional variables were performed by the participants while observed by a nurse ranking the performances (1) without difficulty, (2) with some difficulty, or (3) impossible. This means that the best performance in the balance factors gives a value of 10 and the worst performances a value of 30. The best performance in the fine motor factor have a value of 8 and the worst performance a value of 24. In the flexibility factor 2 is the best performance and 6 the worst. The factors values goes from low (indicating ability to perform) to high (indicating difficulties to perform) (Bravell et al., 2017). A low balance factor score indicates a better balance factor function and a high balance factor score indicates more limitations when performing the function variables.

4.6 Statistical tests

The hypotheses H1-H5 were tested by comparing independent groups with the nonparametric Mann-Whitney U test (Lærd statistics, n.d.-b). The nonparametric test was chosen based on the data samples not being normally distributed and the small sample size, see table 2. The test on H1-H5 was performed by comparing one functional factor at the time, using 3 tests per hypotheses. First, a comparison of motor function between the two groups identified by the subject factors (e.g., education). Second, additional tests were performed to investigate if the level of movement disability might have affected the results by performing a Mann-Whitney U test to compare the independent group’s ´a little´ and ´a lot´ complication in daily life. Third, the ´level of movement disability´ was combined with the independent variables and Kruskal-Wallis H Test was conducted in the same way as described below for the test of H6.

Two different tests were performed to test H6. First, each factor was considered in combination with cohort. The nonparametric Kruskal-Wallis H test was performed to investigate associations between the outcome variable and the independent variables. Four new variables were created. Childhood SES and Cohort grouped as, 1 - low childhood SES and early cohort, 2- low childhood SES and late cohort, 3- high childhood SES and early cohort, 4- high childhood SES and late cohort. Education and Cohort grouped as 1-low education and early cohort, 2-low education and late cohort, 3-high education and early cohort, 4-high education and late cohort. Gender and Cohort grouped as 1-man and early cohort, 2-man and late cohort, 3-woman and early cohort, 4-woman and late cohort. Rural/urban and Cohort grouped as 1-rural and early cohort, 2-rural and late cohort, 3-urban and early cohort, 4-urban and late cohort. These variables were created so that the test could be performed, because the Kruskal-Wallis H test

11

determine if there are any statistical significant difference between two or more groups of an independent variable (Lærd statistics, n.d.-a). A significant Kruskal-Wallis H test would in this case mean that the cohort interacts with the other independent variable, which is the aim of H6. Second, all subject factors including cohort were considered in a single analytical approach. The binary logistic regression was performed to look for associations between childhood SES, education, gender, rural/urban living, age and functional level measured in balance, fine motor and flexibility factor, and to investigate if these associations differed between cohorts. The logistic regression was chosen based on the sample size and the skewed data for the balance, fine motor and flexibility factor (Barmark & Djurfeldt, 2009). The functional factors were altered into dichotomous variables of high and low ability to perform, this was done by dividing the groups at the median: high function being ≤ median and low function > median. The regression was performed in three steps, first with the variables childhood SES, education, gender, rural/urban living and age, and secondly adding cohort to observe changes in the associations. Third, interaction terms, for variables having significant or marginally significant associations with the dependent variable, were added to a new regression model. The p-value to determine a statistical significant result was set at p ≤ 0.05 and a marginally significant result have a p ≤ 0.10. All the statistical tests were performed in SPSS Statistics 27.

4.7 Ethical considerations

There are different ethical considerations in this project. When it comes to the data collection SATSA has an ethical approval from the Ethical committee at Karolinska Institutet and the Regional Ethics Review Board in Stockholm (Bravell et al., 2017). For this project, no new ethical approval was needed, but there are some ethical concerns. The participants in the SATSA sample gave consent to participate and for their data to be registered and used in research, but they have no knowledge about their data being used in this research project. Data from national quality registers have similar ethical concerns even though different rules apply for consent and use of data (Johansson, 2014). Using data from databases or registers also have some advantage by not wasting resources on collecting new data, and that databases often have larger samples, which improves the quality of the research. To get access to the sample of interest, an abstract of intent was sent to the directors of SATSA, and approved by the directors. The sample sent to the author of this project was de-identified and only containing variables and information relevant for these specific research questions. The data is stored on a computer with a personal password and is not accessible to unauthorized persons. The data will be removed from the authors’ computer after the research is completed.

12

5. Result

5.1 Descriptive statistics

The 69 participants are older adults with self-reported movement disabilities; their age at the last participation was 42-100 years old and the samples mean age was 78.36 years. All of the participants have self-reported movement disabilities but 10 of the participants reported known diseases, which likely caused the disability. Table 1 shows the samples distribution, consisting of 17 men and 52 women, 24 participants having low childhood SES, 41 having high childhood SES, and data missing from 4 participants. None of the participants are born in 1925 and therefore the division before and after 1925 does not include the year 1925, 49 participants are born before 1925 and 20 are born after 1925.

Table 1. Sample distribution

Variables Total n=69 (100 %) Age, M (SD) 78.36 (10.062) Movement disability, n (%) Self-reported 69 (100) Diseases, n (%) Epilepsy 1 (1.4) Parkinson´s disease 3 (4.3) Multiple sclerosis 4 (5.7) Polio 2 (2.8)

Complicates daily life, n (%)

Not at all 1 (1.4) A little 39 (56.5) A lot 21 (30.4) Missing values 8 (11.5) SES, n (%) Low 24 (34.8) High 41 (59.4) Missing values 4 (5.8) Education, n (%) Low 43 (62.3) High 23 (33.3) Missing values 3 (4.3) Gender, n (%) Men 17 (24.6) Women 52 (75.4) Missing values 0 (0) Living, n (%) Rural 23 (33.3) Urban 43 (62.3) Missing values 3 (4.3) Cohort, n (%) Before 1925 49 (71) After 1925 20 (29) Missing values 0 (0)

13

Table 2 illustrates the three functional factors and the group division of the participants and shows the mean functional outcome and the number of data points available (n) in the group. In the balance factor the low childhood SES group (n=22) have a mean of 17.08 and the high childhood SES group (n=40) have a mean on 15.7. This shows that the high childhood SES group on an average have better functional ability in the balance factor, because a low value shows better function and having better functional ability. However, this is only by looking at the descriptive statistic and is therefore not showing if this difference is statistically significant. The flexibility factor for the men (n=16) has a mean 2.81 and the woman (n=51) have 2.9, this indicates a very similar function among men and women in the flexibility factor.

Table 2. Descriptive statistics

Balance Fine Motor Flexibility

N Mean (SD) N Mean (SD) N Mean (SD)

Low childhood SES 22 17.08 (1.474) 24 10.83 (.954) 23 3 (.267) High childhood SES 40 15.7 (.897) 41 9.83 (.366) 40 2.75 (.178) Low education 41 17.25 (.981) 43 10.34 (.564) 42 3 (.196) High education 22 14.71 (1.158) 23 10.06 (.597) 22 2.68 (.222) Men 16 15.14 (1.588) 17 9.16 (.270) 16 2.81 (.228) Women 50 16.61 (.840) 52 10.58 (.519) 51 2.9 (.175) Rural 22 14.95 (1.084) 23 10.52 (.816) 22 2.55 (.194) Urban 41 17.12 (1.012) 43 10.09 (.480) 42 3.07 (.2) Early Cohort 46 16.32 (.944) 49 10.33 (.544) 47 2.94 (.191) Late Cohort 20 16.11 (1.166) 20 10 (.404) 20 2.75 (.176)

5.2 Results H1-H5

H1-H5 was tested by comparing the median between groups, low & high childhood SES (H1), low & high education (H2), men & women (H3), rural & urban (H4), and early & late cohort (H5). The results from the balance factor are presented in figure 1 and table 3. As shown in figure 1, the high childhood SES group had better function in the balance factor compared to the low childhood SES group; however, this difference is not significant, see table 3. The high education group had better function in the balance factor compared to the low education group, this difference was marginally significant meaning a p-value ≤ 0.10, but not to a sufficient p-p-value ≤ 0.05. The men show better ability to perform the functional variables in the balance factor compared to the women. Surprisingly the rural vs. urban groups show opposite results of the expected. The rural group have better function in the balance factor compared to the urban group. The early cohort also shows better function compared to the later cohort. However, none of these differences are significant, see table 3.

14

Figure 1. The balance median

† = p-value ≤ 0.10

Table 3. Balance factor median and statistical testing (H1, H2, H3, H4 & H5)

Median (IQR) P-value

Low Childhood SES 15.64 (10)

0.533

High Childhood SES 13.5 (9)

Low education 17 (9) .087† High education 12.93 (8) Men 12.14 (8) .198 Woman 16 (8) Rural 13.5 (7) .255 Urban 17.14 (10) Early cohort 14.14 (8) .888 Late cohort 17 (10)

Note: the p-values are from the comparison test of low/high childhood SES, low/high education, men/women, rural/urban, and Early/late cohort. A p-value ≤ 0.05 represents a statistical significant difference between the two groups. † = p-value ≤ 0.10.

The results from the fine motor factor are presented in figure 2 and table 4. The median of the fine motor factor is 9 in all groups but the late cohort which has a median of 10. The best performance in the fine motor factor gives a value of 8, and therefore a median on 9 or 10 indicates a good performance in the fine motor factor. In table 4 the interquartile range (IQR) is shown and have a value of 2, 3 or 4, indicating that 50 % in that group is placed within a range of 2, 3 or 4 point from the median. For the low and high childhood SES, the Q1 is 8 and Q3 is 11, and therefore the IQR is 3. The IQR of 2, 3 or 4, and the median close to the best performance possible indicates that the groups’ overall have good ability to perform the fine motor factor. The difference between the early and late cohort is not significant, see table 4.

15

Figure 2. The fine motor median

Table 4. Fine motor factor median and statistical testing (H1, H2, H3, H4 & H5)

Median (IQR) P-value

Low Childhood SES 9 (3)

.895

High Childhood SES 9 (3)

Low education 9 (3) .967 High education 9 (3) Men 9 (2) _.301 Woman 9 (4) Rural 9 (4) .847 Urban 9 (3) Early cohort 9 (2) .456 Late cohort 10 (3)

Note: the p-values are from the comparison test of low/high childhood SES, low/high education, men/women, rural/urban, and Early/late cohort. A p-value ≤ 0.05 represents a statistical significant difference between the two groups.

The results from the flexibility factor are presented in figure 3 and table 5. The high childhood SES group have a better ability to perform the functional tests in the flexibility factor compared to the low childhood SES as shown in figure 3. There are no difference between the low and high education group. The women perform better than the men, and the early cohort better than the later cohort does. None of the differences are significant, but the difference between rural and urban is marginally significant. This implies that there are some effect, meaning that the rural group might have better performances in the flexibility factor compared to the urban group. The value of 2 is the best performances possible in the flexibility factor, and the worst performances would give a value of 6 (Bravell et al., 2017). The groups in this sample have a median on 2 or 3, indicating a good performance in this factor.

16

Figure 3. The flexibility median

† = p-value ≤ 0.10

Table 5. Flexibility factor median and statistical testing (H1, H2, H3, H4 & H5)

Median (IQR) P-value

Low Childhood SES 3 (2)

.389

High Childhood SES 2 (1)

Low education 2 (2) .383 High education 2 (1) Men 3 (1) .784 Woman 2 (2) Rural 2 (1) _{.102 †} Urban 3 (2) Early cohort 2 (2) .887 Late cohort 3 (1)

Note: the p-values are from the comparison test of low/high childhood SES, low/high education, men/women, rural/urban, and Early/late cohort. A p-value ≤ 0.05 represents a statistical significant difference between the two groups. † = p-value ≤ 0.10. Through the statistical testing using the Mann-Whitney U test the hypotheses are accepted or rejected.

H1: Older adults with movement disabilities who had higher childhood SES have higher functional

level is rejected, because there was no significant differences between the groups in the functional

factors, and therefore the conclusion is that there are no difference in functional outcome for the groups of low and high childhood SES among older adults with movement disabilities.

H2: Older adults with movement disabilities who have higher education have higher functional level is rejected and there are no significant differences in functional outcome for the groups of low and high education among the sample of older adults with movement disabilities.

17

H3: There are no gender differences in functional level among older adults with movement disabilities is accepted. Because there was no significant difference between the men and women in any of the functional measures.

H4: For older adults, who have movement disabilities, the functional level is higher for people living

in urban rather than rural settings is rejected. There was no significant differences between the rural

and urban group. However, the difference in the flexibility factor was marginal significant, and there was a trend which turned out to be opposite of the predicted, meaning a trend that the group living rural having better function than those living urban.

H5: For older adults who have a movement disability, the functional level is higher for people born in

a later cohort than those born in an earlier cohort is rejected based on none of the functional factors

being significantly different. When looking at figure 1, 2 and 3, there seems to be a trend that the earlier cohort actually seems to have a better function compared to the later cohort, but when looking at table 2, which shows the mean of the sample, no such trend seems to exist.

These results were so different from what was expected and therefore additional tests were performed to investigate if the level of movement disability might have had an influence on the results. First investigating if there was a significant difference in function between the groups replying that their movement disability complicates their daily life a little compared to those responding that it complicates their daily life a lot. There was no significant difference between these groups. Then new variables were made combining the independent variables (childhood SES, education, gender, rural vs. urban and cohort) and the level of movement disability, complicating daily life a little or a lot. Even when adding the level of movement disability no significant differences existed between the groups, see table 6.

Table 6. Level of disability interacting with other variables, comparing independent groups

Variable Balance factor

p-value

Fine motor factor p-value

Flexibility factor p-value

Childhood SES and level of disability .915 .881 .724

Education and level of disability .611 .772 .802

Gender and level of disability .665 .102 † .793

Rural/urban and level of disability .100 † .744 .296

Cohort and level of disability .693 .653 .506

18

5.3 Results H6

H6 was tested through two different statistical tests. The results of the Kruskal-Wallis H tests are shown in table 7. The new variables combining the independent variables; childhood SES, education, gender and rural/urban living, and the independent variable cohort into 4 groups show no significant differences among any of the groups in the combined variables for all 3 of the functional factors. This suggest that cohort does not interact with the other independent variables.

Table 7. Cohort interacting with other variables, comparing independent groups

Variable Balance factor

p-value

Fine motor factor p-value

Flexibility factor p-value

Childhood SES and Cohort .459 .542 .710

Education and Cohort .341 .625 .846

Gender and Cohort .564 .555 .871

Rural/urban and Cohort .556 .873 .336

Three binary logistic regression were performed, two regression models in all three functional factors and adding one model to the functional factors there had significant associations in model 1, model 2 or both. The first model investigating associations between the functional factor and the independent variables childhood SES, education, gender, rural/urban living and age. In the second model cohort was added to the regression. The third model added interaction terms.

The results for the regression on the flexibility factor are shown in table 8 and table 9. In table 8, the B-value is negative for childhood SES this should be interpreted as when the independent variable childhood SES decreases (lower childhood SES) there is an association with the increase of the dependent variable (difficulties in functional ability increase). However this association is not significant, and none of the associations are significant in model 1, but age is marginally significant with a p-value of 0.069, implying an association between aging and decreased function in the flexibility factor with an OR of 1.064, so a very small increase, see table 8. The R2_{=0.155 meaning that the independent}

variables in model 1 explained 15.5 % of the variance in the dependent variable. This indicates that model 1 might not be a good model to explain the flexibility factor. When adding cohort the model becomes a much better fit and the independent variables explain 30.9 % of the variance in the dependent variable. Model 2 is a much better model and 3 of the independent variables have a significant association with the flexibility factor. The rural/urban variable has a positive association meaning that when living urban the individual is more likely to have a higher flexibility factor, which indicates worse ability to perform the two functional variables. The OR is 4.633 and this indicate 4.6 higher odds of having worse function in the flexibility factor when living urban compared to rural. This says nothing about the causal relationship, if a person living urban is more likely to have worse function or if a person with more limited function are more likely to live urban, it only shows an association between the two variables. Age has a significant association with the flexibility factor in model 2. Cohort has a significant association, which indicates an association between later cohort and worse function in the flexibility factor, with a large OR 11.623.

19

Table 8. Logistic regression results in the flexibility factor

B S.E. p-value OR Childhood SES -.192 .595 .747 .825 Education -.186 .620 .763 .830 Gender -.886 .707 .210 .412 Rural/Urban .981 .638 .124 2.666 Age .062 .034 .069† 1.064 R2 Model 1 .155 B S.E. p-value OR Childhood SES -.330 .641 .606 .719 Education -.547 .663 .409 .579 Gender -.591 .775 .446 .554 Rural/urban 1.533 .721 .033* 4.633 Age .129 .047 .007** 1.138 Cohort 2.453 .948 .010* 11.623 R2 Model 2 .309

Model 1: without cohort, Model 2: with cohort. R2_{, Nagelkerke R}2_{. †P ≤ 0.10, *P < 0.05,**P < 0.01}

Interaction terms were added to the regression to investigate interactions between cohort and the independent variables, which had a significant or marginally significant association, and education based on a previous study that found an interaction between education and cohort (Finkel & Ernsth Bravell, 2020). In table 9 the regression with interaction terms are presented for the flexibility factor. None of the interaction terms are significant and only the Cohort x rural/urban is marginal significant.

Table 9. Logistic regression results in the flexibility factor with interaction

B S.E. p-value OR Childhood SES -.425 .678 .531 .654 Education -1.207 2.227 .588 .299 Gender -.345 .850 .585 .708 Rural/Urban -2.985 2.453 .224 .051 Age .158 .159 .322 1.171 Cohort -3.641 10.217 .722 .026 Cohort x education .613 1.624 .706 1.845 Cohort x rural/urban 3.480 1.925 .071† 32.470 Cohort x age .003 .111 .978 1.003 R2 Model 3 .399 R2 _{= Nagelkerke R}2_{. †P ≤ 0.10}

20

In the balance factor, age has a marginally significant association with the function in the model 1, see table 10. Model 1 explains 14.5 % of the variance in the dependent variable and model 2 explains 24.6 %. Again, the model becomes much better after adding cohort. In model 2 rural/urban is marginally significant, and age and cohort have a significant association with the balance factor. Age with an OR 1.109, so a quite small increase for poorer function and cohort OR 6.128 which is a high OR indicating association between later cohort and poorer function, see table 10.

Table 10. Logistic regression results in the balance factor

B S.E. p-value OR Childhood SES .567 .605 .348 1.763 Education -.664 .613 .279 .515 Gender .013 .684 .985 1.013 Rural/Urban .818 .613 .182 2.266 Age .059 .034 .080† 1.061 R2 Model 1 .145 B S.E. p-value OR Childhood SES .571 .637 .370 1.769 Education -1.029 .666 .122 .357 Gender .338 .751 .652 1.403 Rural/urban 1.181 .666 .076† 3.256 Age .104 .042 .014* 1.109 Cohort 1.813 .852 .033* 6.128 R2 Model 2 .246

Model 1: without cohort, Model 2: with cohort. R2_{, Nagelkerke R}2_{. †P ≤ 0.10,*P < 0.05,**P <0 .01.}

The regression with the interaction terms for the balance factor shows no significant associations and age remains marginally significant, see table 11.

Table 11. Logistic regression results in the balance factor with interaction

B S.E. p-value OR Childhood SES .608 .670 .364 1.838 Education .295 2.099 .888 1.343 Gender .627 .821 .446 1.871 Rural/Urban -1.576 2.214 .476 .207 Age .227 .136 .095† 1.255 Cohort 5.838 7.332 .426 343.034 Cohort x education -.916 1.448 .527 .400 Cohort x rural/urban 1.923 1.606 .231 6.842 Cohort x age -.073 .084 .389 .930 R2 Model 3 0.303 R2_{, Nagelkerke R}2_{. †P ≤ 0.10}

21

The regression performed on the fine motor factor shows no significant or marginally significant associations. Model 1 explained 3.2 % and model 2 explained 7.9 % of the variance in the dependent variable, which mean these models explain the fine motor factor to an insufficient level, see table 12. No regression was performed with interaction terms for the fine motor factor.

Table 12. Logistic regression results in the fine motor factor

B S.E. p-value OR Childhood SES .563 .583 .334 1.756 Education .005 .566 .993 1.005 Gender -.267 .637 .675 .766 Rural/Urban .335 .584 .566 1.398 Age .015 .028 .585 1.016 R2 Model 1 .032 B S.E. p-value OR Childhood SES .568 .594 .339 1.765 Education -.207 .595 .727 .813 Gender -.118 .662 .859 .889 Rural/urban .516 .605 .394 1.676 Age .041 .034 .221 1.042 Cohort 1.096 .737 .137 2.993 R2 Model 2 .079

Model 1: without cohort, Model 2: with cohort. R2_{, Nagelkerke R}2_.

Cohort did add significantly to the associations with the balance - and flexibility factor, indicating a main effect of cohort on functioning. However, because there was no strong evidence that cohort interacted with the other subject variables, either in the median comparison or in the regression analysis, the H6

the effects of childhood SES, education, gender and urban vs. rural will differ between cohorts is

22

6. Discussion

6.1 Method discussion

6.1.1 Research design

A quantitative deductive research design has been applied to investigate how childhood SES, education, gender, rural/urban living, and cohort affects functioning in late adulthood for people with movement disabilities. The quantitative design provide objective measures and statistical analysis compared to a qualitative design investigating personal experiences (Henricson & Billhult, 2017; Park et al., 2020). Functioning in aging is a well-researched area and by applying a deductive approach, previous research and theories were used when designing this project and setting up hypotheses to test (Priebe & Landström, 2017). This project investigated older adults with movement disabilities, and thereby differed from the previous studies focusing mainly on the general population of older adults. Using a cross-sectional sample is adequate to investigate the hypotheses; however, inadequate to determine whether the older adults with movement disability experience double jeopardy or age-as-a-leveller in aging (Ferraro & Farmer, 1996), as well as to determine if the older adults with movement disability experience healthy aging as developing and maintaining functional ability (WHO, 2015).

6.1.2 The sample

The SATSA project has a large sample of older adults and has collected objective measures of physical performance, which made SATSA an excellent database to select the sample to investigate this projects objective. Even though SATSA has collected objective physical performance measures on n=859, only n=69 of these adults answered yes to having a movement disability. This constitutes a quite small accessible sample and for the actual statistical testing, the sample becomes smaller due to missing data. The missing data is both missing information about characteristics such as childhood SES n=4, education n=3, urban/rural n=3, or missing data on the physical performance. The missing data makes the groups to compare for the statistical testing small, but remaining adequate to perform the tests. The sample size is a limiting factor in this project.

The sample consisted of data from the participants first Q and last IPT. In other words, the sample consist of data collected in different Qs’ and IPTs’. The different times for data collection is not believed to affect the validity of this project, as SATSA has been following the same procedure throughout. If the participants last IPT was IPT4, which was collected over the phone (Finkel & Pedersen, 2004), then data from IPT3 was used.

The participants were between 42-100 years old at their last IPT. Only 3 participants were under 65 years being 42, 49, and 58 at their last IPT. These 3 participants was kept in the sample to not further reduce the sample. However, the choice to keep them raises the question at which age are the participants’ older adults.

23

There is a risk that the sample over represents the healthy and functioning older adults with movement disability. However, the nurses’ also collected data in peoples home and nursing homes’ to allow for a more representative sample. In other words, older adults with poorer health or function could participate.

6.1.3 Measures

The movement disability is a self-reported measure. An objective measure for example limitation in ADL or having a diagnosis would be preferred. However using an objective measure would only provide a sample of n=10 from the SATSA database. The subjective self-reported measure will possibly provide a larger diversity in types of movement disability and thereby the sample will potentially be more representative of the population with movement disabilities, instead of focusing on only one or a few types of disability. Therefore, the self-reported measure will give the results of this study a better generalizability. Based on the sample this study’s results can be generalized to the population of older adults with movement disability living in Sweden.

Previous studies measured SES in a number of different ways. This study measured SES in two variables; childhood SES and education. The measure of childhood SES could be a good measure, not only for the association with education and adult life SES (Ericsson et al., 2017), but also for early life opportunities for habilitation and rehabilitation if the movement disability emerged early in life. Unfortunately, information about onset of movement disability was not available. Education is an accepted measure of SES, but might also have an independent role in health as education could give the individual the ability to access and apply health information (Mirowsky & Ross, 2003).

This study’s focus on cohort difference and how cohort might interact with other variables is a strength, because previous studies have found that cohort interact with other variables (Finkel & Ernsth Bravell, 2020) and leaving out the cohort analysis could create a misinterpretation of the results.

It is a strength that the analysis were performed on objective measures of function. The performance-based motor functioning factors were developed from the functional tests performed in the SATSA project and have high internal consistency (Bravell et al., 2017).

6.1.4 The statistical tests

The small, not normally distributed and independent groups where compared using the Mann-Whitney U and Kruskal-Wallis test. The Kruskal-Wallis test was used to compare 4 independent groups, and caution should be applied when interpreting a significant result, as the number of groups effect the risk of a type I-error (Dai, 2018). None of the tests performed with the Kruskal-Wallis test resulted in a significant difference between groups. Therefore, no extra caution was applied when interpreting the results. The Kruskal-Wallis test cannot give satisfying information about associations nor investigate multiple independent variables, and therefore a regression was performed (Dai, 2018).

24

In the regression, associations between the dependent variable and the independent variables does not explain the causal relationship (Egerton, 2018), and therefore the results of the regression are interpreted with caution, and only plausible variables were tested to minimize random associations. When performing a logistic regression it is important to have the right number of variables, and the right independent variables (Egerton, 2018). Adding variables will make the model explain more of the variance of the dependent variable, and therefore only meaningful variables should be included. In this research, the independent variables tested in the regression were the variables from the hypotheses, but age was added as age plausibly affected functional ability. The regression might not have included all meaningful variables, for example marital status, living alone or in a shared household could have associations with functional ability (Joutsenniemi et al., 2006). It is a possible limitation that other independent variables could have been meaningful for the regression.

The sample size tested in this research has likely affected the result of the regressions (Egerton, 2018). In other words, associations between the dependent and independent variable might not be significant in these tests because of the sample size, but further research is needed to determine whether this is the case or not.

6.2 Result discussion

In the balance factor, there was a trend that adults with lower childhood SES and lower education had poorer functional ability, but none of the tested variables had a significant difference between groups (low/high, men/woman, rural/urban, early/late cohort) in the balance factor. Living urban was marginally associated with poorer functional ability, and aging and later cohort was significantly associated with poorer functional ability in the balance factor. No differences between groups or associations were found for the fine motor factor. A marginally significant difference existed between living rural and urban in the flexibility factor, but no difference when testing childhood SES, education, gender and cohort. In the regression urban living, aging and later cohort had associations with poorer functional outcome in the flexibility factor. The interaction term cohort x rural/urban was marginally significant indicating an interaction between later cohort, urban living and poorer function in the flexibility factor.

The finding of no gender difference was expected, as a previous study found no gender difference when controlling for health conditions (Murtagh & Hubert, 2004). However, overall these results were unexpected. The sample should be representative of the population of interest, and the sample size should allow for statistical test and acceptable results. Therefore, an attempt was made to investigate if the level of disability (how much the disability complicates daily life) would have affected the results; the measure of level of disability was self-reported and did not explain these unexpected results.

The childhood SES and education can both be measures of SES and previous studies have found that lower SES has an association with poorer physical function (Beydoun & Popkin, 2005; Coppin et al.,

25

2006; Huisman et al., 2003; Steptoe & Zaninotto, 2020). However, Coppin et al. (2006) did find that the association decreased or disappeared when controlling for physical measures of impairment, depending on how physical function was measured. So perhaps the movement disability levels out the SES differences that normally can be observed in a representative sample of the general older population. How the physical function and SES is measured can also affect the results (Darin-Mattsson et al., 2017). Therefore, a measure of income might have been a better SES indicator. However, the sample has disabilities, and the time the disability emerged and the level of disability could have affected the work opportunities and thereby the income level. The household income could have been an option to use as a SES indicator, and might have yielded a different result.

Rural living have been associated with poorer functional ability (Yiengprugsawan et al., 2019). This study found no difference between rural and urban, but there was a contrary trend indicating that people living urban had lower functional ability and an association between living urban and poorer function. This trend and association could possibly be because living rural became too difficult when having more physical limitations and that the person therefore chose to move urban. Living urban might also give better access to healthcare and better transportation options (Buchanan et al., 2008; Mattson, 2011; Zhang et al., 2017).

This research expected to find cohort differences, a better function in the later cohort and that the independent variables of childhood SES, education, gender, and rural/urban living would differ between cohorts. However, a trend emerged that later cohort was associated with poorer functional outcome; this trend is supported as previous studies have found equivocal results indicating both functional increase (Aijänseppä et al., 2005; Malmberg et al., 2002; Spiers et al., 1996) and decrease (Anttila, 1991; Jagger et al., 2007) over succeeding cohort. The Finnish study found decreased function in non-institutionalized older adults, but also that the number of non-non-institutionalized older adults increased a lot (Anttila, 1991). Jagger et al. (2007) found higher levels of self-reported disability over succeeding cohort. However, the results indicating a decrease in function over succeeding cohort might be because people live longer and gets older. The life expectancy is increasing (Dehlin & Rundgren, 2014). In other words, it is possible that the trend that later cohort have poorer functional outcome in this study is a result of more people with movement disabilities surviving and living longer, and that the disease or accident that caused the movement disability did not have a fatal outcome. Alternatively, there might be something else affecting the functional outcome for the later cohort of older adults with movement disability.

Finkel and Ernsth Bravell (2020) found that older adults in the later cohort with high education had the best trajectory of maintaining the functional ability, and that the later cohort with low education had the worst trajectory of maintaining the functional ability in the balance and flexibility factor. The trajectories between the low and high education in the early cohort did not differ much and was a medium trajectory. Finkel and Ernsth Bravell (2020) investigated a much larger sample from SATSA and used longitudinal data. Even though this study used cross-sectional data and a smaller sample from SATSA, it was