METHODS

Study population

A systematic and representative sample of community based residents aged 19–63 years was drawn from the 1979 census data of a medium size industrial town and two rural municipalities in north east Finland. The result was a sample of 6787 men and women, of whom 5259 (77.5%) answered the baseline survey and formed a study cohort in 1980. According to national census data from the Central Statistical Office of Finland, a total of 340 men and 150 women (9.3% of the study cohort) died between the baseline survey on 1 March 1980 and the follow up survey on 30 September 1996.

The study population consisted of all the men and women who were both alive and between the ages of 55 and 79 years in 1996 (n=1969).

Procedure

A self administered, postal survey provided data on sociodemographic status, perceived health, chronic conditions, functional status, physical activity, smoking, and alcohol consumption. The questionnaire items have proved to be valid in terms of their relations to the existence of chronic diseases,²⁷ prediction of coronary heart disease risk,²⁸ use of physician and hospital services,^29,30 and decreased risk of all cause^29,31,32 and cardiovascular disease^31,32 mortality among middle aged and elderly persons.

The baseline questionnaire was sent to the study cohort at the beginning of 1980. Follow up questionnaires requesting similar information, but with a narrower scope and additional questions on functional status, were sent to all members of the cohort, irrespective of their places of residence in Finland, in 1981, 1990, and to the older members of the cohort in 1996. The response rates for these surveys, after two requests, were 88.0%, 85.4%, and 90.9% of the alive cohort, respectively. A total of 1791 (758 men and 1033 women) responded to the follow up questionnaire in 1996. Of the 178 non-respondents, 133 (6.7% of the alive cohort in 1996) failed to return the questionnaire after two requests, 29 (1.5%) refused to respond, seven (0.3%) did not respond because of a severe health restriction, and nine (0.5%) could not be contacted because of unknown address. The study design of the 16 year prospective follow up study carried out in north east Finland during 1980–1996 is shown in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Study design of the 16 year prospective follow up study carried out in north east Finland during 1980–1996.

The ethics committee of the Urho Kaleva Kekkonen Institute for Health Promotion Research approved the study.

Measurements

Functional status was determined from a self estimate on ability to walk 2 km without rest, climb several flights of stairs without rest, and run a distance of 100 m. The response alternatives were “no difficulties”, “some difficulties”, “severe difficulties”, and “not at all able”. The respondents were classified disabled if they reported difficulties or inability to perform the specified task. The mobility tasks were chosen because of their less gender specific and environment specific nature and their linkage to lower extremity functioning, which has been shown to have a substantial effect on the ability of elderly persons to remain independent.^5,6 The mobility self ratings on walking (π=0.80) and transferring (π=0.58) have been shown to correlate with their objective measures,^33,34 and their reproducibility have been shown to range between fair and good (weighted κ 0.57–0.70) in a representative population sample of middle aged and elderly persons.³⁵ Mobility self ratings, including running, have also been shown to correlate with their objective measures (Pearson’s r=0.69) and with their retest values (Pearson’s r=0.70–0.92) in different clinical populations.³⁶ In both Finland and Sweden, the self rating on ability to run 100 m has been found to discriminate well between different overweight³⁷, age, and gender^37,38 categories in nationally representative populations of middle aged and elderly adults.

Data analysis

All the analyses were conducted separately for the men and women. The respondents were divided into five year birth cohorts with their corresponding age groups for each survey period. Five birth cohorts were identified with birth years 1917–1921, 1922–1926, 1927–1931, 1932–1936, and 1937–1941. The task specific disability rates in the birth cohorts were compared within and between the survey periods (1981, 1990, and 1996) by means of cross tabulation. The task-specific disability rates of similar age groups were also compared between the survey periods. The disability rates were conditional on survival and on inclusion in the 1996 survey.

Generalised estimating equations (GEE) were used to model longitudinal correlated data on the differences in disability rates for the effects of age, period, and cohort. The Oswald software library (version 3.2) in S-PLUS program (version 4.0, release 2) was used for the analysis. The GEE procedure has the advantage of estimating regression coefficients of complete as well as incomplete data without making assumptions of the variance. The models were fitted by means of marginal models, with a logit link function and a compound symmetry covariance structure. It was recognised in fitting the models that the independent effects of age, period, and cohort can not be estimated simultaneously.³⁹ The approach adopted here was to investigate the age period and age cohort models for the data separately. In the age cohort models, tests for trends in disability status were performed to determine whether more recent cohorts are in better or worse health and functioning than previous cohorts. The odds ratios (OR) with 95% confidence intervals were calculated to determine the effect of the factors on disability status.

RESULTS

The number and proportion of respondents in the surveys of the study is presented by birth cohort and gender in table 1. The female to male ratio was somewhat higher in the oldest birth cohort (I) compared with the younger birth cohorts (II-V): 1.9 versus 1.3, 1.3, 1.4, and 1.2. In addition, the proportion of female respondents who completed the follow up was somewhat lower in the two oldest birth cohorts (I and II) compared with the younger birth cohorts (III-V): 86% versus 92%–94%. In 1981, the age adjusted prevalence of disability in stair climbing among the respondents to the 1990 and the 1996 surveys, respectively, was 32%, whereas the corresponding rate among the non-respondents was 42%.

The age related prevalence of disability in walking, stair climbing, and running is presented by gender, birth cohort, and survey period in figures 2, 3, and 4. There was, with few exceptions, a linear increase in the prevalence of disability in walking, stair climbing, and running across the cohorts and the periods among the men and the women. The disability rates were lowest in walking and highest in running, and they were consistently lower among the men than among the women. As described in tables 2, 3, and 4, the prevalence of disability in similar age groups decreased across the survey periods, especially in terms of stair climbing and running disability. For example (bold figures in table 3), 47% of the men and 53% of the women in cohort II (aged 54–58 years in 1981) were disabled in stair climbing compared with 30% of the men and 47% of the women in cohort IV (aged 53–57 years in 1990) and 24% of the men and 41% of the women in cohort V (aged 54–58 years in 1996). The corresponding figures were 45% of the men and 65% of the women in cohort I (aged 59–63 years in 1981) compared with 37% of the men and 50% of the women in cohort III (aged 58–62 years in 1990) and 35% of the men and 53% of the women in cohort IV (aged 59–63 years in 1996).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Prevalence of stair climbing disability under the age, period, and cohort model among men (n=758) and women (n=1033).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Prevalence of running disability under the age, period, and cohort model among men (n=758) and women (n=1033)

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Prevalence of walking disability under the age, period, and cohort model among men (n=758) and women (n=1033).

An age-period model was applied to quantify the differences in age related prevalence of disability between the survey periods (table 5). Higher age predicted statistically significantly disability in walking, stair climbing, and running among the men (ORs 1.09 to 1.12) and the women (ORs 1.09 to 1.11). The higher odds of stair climbing and running (not walking) disability with age were countered by statistically significantly lower odds of disability between the survey periods, suggesting that the age related decline in the proportion of men and women with disability during follow up was strongly attributed to the time of the survey periods.

An age-cohort model was applied to quantify the differences in age related prevalence of disability between the birth cohorts (table 6). As in the age-period model, higher age predicted disability in walking, stair climbing, and running among the men (ORs 1.05 to 1.10) and the women (ORs 1.05 to 1.09). The higher odds of stair climbing and running (not walking) disability with age were countered by lower odds of disability between the birth cohorts. There was a consistent, declining trend in the odds of disability in stair climbing (OR 0.79 and 95% CI 0.70 to 0.88 for men and OR 0.85 and 95% CI 0.77 to 0.93 for women) and in running (OR 0.88 and 95% CI 0.78 to 0.98 for men and OR 0.85 and 95% CI 0.76 to 0.94 for women) with succeeding birth cohorts. For the women, the decline in the odds of stair climbing disability was notable in cohort II, III, and IV as compared with cohort I, but even more so in cohort V as compared with all other cohorts in terms of both the decline in stair climbing disability and running disability. For the men, the corresponding disability decline was notable in cohort III and IV as compared with cohorts I and II, and in cohort V as compared with all other cohorts.

DISCUSSION

This 16 year longitudinal investigation among middle aged and elderly men and women indicates that the age related prevalence of disability as assessed by difficulties in stair climbing and running declined in the 1980s and the early 1990s. The extent of the decline varied with time of birth, with smallest relative declines in the earliest birth cohort and largest declines in the latest birth cohort. Consistent findings in the disability rates between similar age groups across the survey periods and the age-period analysis pointed at a period effect, whereas consistent findings in disability rates between similar age groups across the birth cohorts and the age-cohort analysis pointed at a birth cohort effect. Although the age effect on the prevalence of walking disability in the age-period and the age-cohort analysis was of similar or greater magnitude than the age effect on stair climbing and running disability, there were no statistically significant differences in the age related prevalence of walking disability between the birth cohorts over the 16 year follow up.

The findings of this study differ in part from those of others. However, the comparison of time trends in disability between studies is complicated by differences in sampling strategies, methods of data collection and analyses, measures of disability, age of subjects, and length of intervals defining birth cohorts and follow up times. This study comprised a sample of middle aged and elderly men and women with a wide age range (39–63 years at baseline) and a long follow up time (16 years). In a similar study to this improved mobility status was observed between 1979 and 1989 by comparing two Finnish birth cohorts of men between the ages 65 and 69 years and women between the ages 60 and 64 years.⁴⁰ In the USA, the Framingham Health Study offspring cohort aged 55–70 years in 1994 reported also less need for help in mobility than did the similarly aged members of the original cohort in 1977,²⁰ a finding consistent with the US Bureau of the Census’s Survey of Income and Program Participation among persons aged 50 years and older⁴¹ and the National Long Term Care Survey among persons aged 65 years and older.⁴² However, no differences in mobility status were found between two independent samples each of middle aged⁴³ and elderly Finnish men and women⁴⁴ between 1972 and 1992 and between 1988 and 1996, respectively. The minor differences in mobility status between the cohorts in the former study and between the cohorts in the present study may be explained by a ceiling effect: 70% of the 59–63-year-olds in 1981, 1990, and 1996 in this study (table 2) and 80% of the 66 year olds in 1972 and 1992 in the study by Pohjolainen et al⁴³ reported no difficulty with walking. In line with a Canadian study,²³ two other Finnish studies, using 10 year and 20 year data over the same time period than this study, found also no notable differences in disability status between two independent study populations aged 75 years and over.^22,24 However, all three studies differed methodologically from this study and they concerned older populations. In this study the age effect on functional status was disentangled from the period and the cohort effect. Accordingly, functional status improved in the study population both across the survey periods and birth cohorts. This finding is in line with the Melton Mowbray Study concerning British men and women 75 years and over.²⁶ However, in the Zutphen Elderly Study, only a strong age effect on functional status was found among Dutch men 70 years and over.²⁵ The lack of a period and a cohort effect may in part be explained by the comparatively short follow up time in the study.

The interpretation of the cohort and period differences in age related prevalence of disability should take into account the possibility of selection and measurement bias. The non-response is unlikely to affect the findings because it was very low in the three surveys. The higher age standardised rates in walking disability at baseline (in 1981) among those who died than among those who survived until the end of the 16 year follow up (33% versus 18%) may suggest an effect of selective drop out because of death. Such bias would tend to underestimate the baseline prevalence of disability in the study population, which would decrease the cohort and period differences in disability and limit generalisability of the findings. As the disability proportions were conditional on survival and on participation in the 1996 survey, the interpretation of the findings should bear this effect of a selective survival in mind. Bias because of the survey measures, or bias because of prevailing attitudes to the survey measures independent of functional status, may also explain the cohort and period differences in age related disability. However, mobility self ratings have been shown to correlate moderately with their objective measures^45,46 and to be highly consistent over six months among older persons, independent of age and disability status.⁴⁷ In addition, the mobility self ratings have provided consistently higher disability rates than interviewer administered surveys, but only for surveys that differ in methodology.⁴⁸ In this study the survey methods were comparable and the data on functional status were gathered from mobility self ratings on everyday physical tasks such as walking, climbing, and running, which are not sensitive to cultural expectations about roles and changes in the environment.⁴⁹ The survey respondents also shared similar life circumstances. They were born between the first world war and second world war and they were aging in a technically advanced welfare state. A slightly different introductory text for the mobility self ratings was used in the 1981 survey than in the 1990 and 1996 surveys, which may have caused a higher number of missing values in the 1981 survey compared with the other two surveys. The introductory text in the 1981 survey read: “ . . .Circle the task and response alternative . . .”. The text in the 1990 and the 1996 survey read: “ . . .Circle for each task the response alternative . . .”. Levels of missing data were 22%–32% in 1981 compared with 10%–11% in 1990 and 3%–5% in 1996, depending on the task. The effect of the missing data on the findings is minimal because of a compound symmetry covariance structure that was used to link the disability data between the surveys of the study. In addition, a separate analysis on the impact of the missing values on the odds of disability did not change the results in the age-period model.

Even with the limitations mentioned above, the findings of this study are in line with others showing a gradual improvement in functional status with succeeding birth cohorts. Although the findings provide only limited support for the compression of morbidity hypothesis, they may have implications for future health and social care planning. Given the continuing growth in the older population, an improved time trend in functional status may serve to postpone the need of health and social care services associated with disabilities in the aging population. In terms of public healthcare spending, the findings may provide guidance to health and social service producers how to allocate resources to institutional and community care, to treatment and prevention strategies of disabilities, and to early diagnosis of functional status decline. However, additional population based studies are needed to confirm whether the improved time trend in functional status reported here reflect the national experience. A more direct analysis of risk factors for functional status decline would also enhance the understanding of preventive strategies that drive the time trend in functional status among the middle aged and elderly persons.