Details of the recruitment survey have been presented previously 3233. In summary, the population registers of three general practices from the North Staffordshire Primary Care Research Consortium were used as a population sampling frame to mail 11230 adults aged 50 years and over with a postal questionnaire about health and lifestyle matters. 7259 adults returned questionnaires in which all items in one section (Keele Assessment of Participation (KAP)34) had been completed (overall response = 71.3%) (April 2002). This group as a whole were very similar in general health, age, gender and marital status to UK norms, suggesting that there was no substantial non-response bias 22. 5032 gave written permission for further contact and were eligible for the follow-up study (Figure) 1.

There were 3709 fully completed questionnaires returned at 3 years, giving a follow-up response of 73.7% among persons who were eligible and had consented to the follow-up and representing 51% of the original responding population at baseline. Those lost to follow-up include 294 deaths, 12 departures and 24 exclusions because of terminal illness or severe psychiatric illness; non-responders among the remaining persons for follow-up included 72 people who declined to participate, 29 people who stated ill health as the reason for not completing the questionnaire, 599 people from whom no response was received and 293 returned questionnaires with incomplete KAP items. The study population for the analysis of observed data thus consisted of the 3709 persons who returned fully completed questionnaires on both occasions. Estimated data was calculated for adults who completed the KAP at both time points and those who completed the KAP at baseline and were lost to follow-up (this is defined as those who did not consent to follow-up, non-responders at follow-up, exclusions and those who did not complete the KAP (n = 6965)).

The baseline and follow-up questionnaires collected information on demographic, socioeconomic and general health characteristics, including the Short Form-12 instrument (SF-12) 35. Participation restriction was measured by the KAP 34. This short self-complete instrument is designed to measure participation from the perspective of the individual in 11 aspects of life mapped to domains and sub-domains of the ICF: mobility within the home, mobility outside the home, self-care, looking after belongings, looking after the home, looking after dependents, interpersonal interaction, managing money, work, education, and social activities. Items are phrased to capture performance ("I have"), individual judgement, and the nature and timeliness of participation ("as and when I have wanted"). Responses are on a five point ordinal scale (All/Most/Some/A little/None of the time) and responders were considered restricted in an aspect of life if they did not participate in it "as and when they wanted" for "all" or "most of the time". The number of aspects of life, where responders indicated participation restriction, was calculated for both time points and categorised to (i) any restriction (participation restriction in at least one aspect of life) and (ii) minimal (1–2 restricted aspects of life) or moderate/substantial (3–11 restrictions (moderate and substantial were combined due to small numbers in the substantial category for some items (e.g. work)). The reliability and validity of the KAP have been established as adequate for providing estimates of perceived participation restriction in population studies 34.

Data recorded in the baseline survey was used to assess the effect of (i) loss to follow-up and (ii) death prior to follow-up, by examining differences in baseline demographics (age, gender), socio-economic status (occupation, educational attainment), general health (physical and mental SF-12 scores) and participation restriction. Percentages and means were used to compare groups with different follow-up outcomes.

The following analysis was carried out on the observed data, i.e. on those completing the 3-year follow-up, and to assess the effects of loss to follow-up, we repeated the onset and persistence analyses assuming that those lost at follow-up had the same likelihood of participation restriction (examining onset and persistence separately), within strata defined by age, gender, educational attainment, occupational class and general health (using SF-12 physical and mental health scores), as those who did complete the three year follow-up questionnaire. These six strata-defining variables were selected a priori as they were used to describe the differences between responders and those lost to follow-up. These estimates were calculated using attrition re-weighted logistic regression 36 performed in Stata 9.2.

The frequencies of three-year onset of participation restriction (defined as moving from no restriction at baseline to any restriction at three year follow-up) and three-year persistence (any restriction at both baseline and follow-up) were calculated overall, and by age and gender, and summarized using percentages.

For those with three-year persistence, the proportion that reported (i) the same number and same restrictions at baseline and follow-up, (ii) the same number but different restrictions at follow-up, (iii) a lower number at follow-up and (iv) a greater number at follow-up, were calculated.

Onset of restriction for each of the 11 aspects of life was calculated as the proportion of those not restricted in that aspect of life at baseline who reported restriction in that aspect at 3 years, regardless of whether the individual reported restriction in other items at baseline or three year follow-up. This analysis was also conducted stratified by age group and gender.

In individuals with participation restriction at baseline, estimates of persistence within each of the 11 aspects of life were calculated overall, and by age and gender. Persistence was calculated as those with restriction in an aspect of life at both baseline and follow-up.

To examine the link between onset of restriction in each aspect of life and amount of restriction at baseline (0, 1–2, 3–11), we used logistic regression, unadjusted, adjusted for gender and, separately, adjusted for age (10 year bands). Associations are presented as odds ratios, with 95% confidence intervals.

Persons who completed all KAP items (n = 3709) at 3-year follow-up, compared with the 3256 persons lost to follow-up, had higher baseline levels of physical health (SF-12 physical health score: 42.3 v 39.9) (Table 1) and mental health (SF-12 mental health score: 49.8 v 48.3), and were more likely to be younger (aged 50 to 69: 72.7% v 55.4%), have gone on to further education (41.1% v 25.6%) and be classified as having a non-manual occupation (48.7% v 34.7%). They were also less likely to report participation restriction (any restriction: 44.7% v 57.6%). In turn, compared with 294 persons who died between baseline and follow-up, persons lost to follow-up had higher levels of physical health (39.9 v 32.1), and mental health (48.3 v 44.3), and were more likely to be younger (aged 50 to 69: 55.4% v 27.9%). They were less likely to report participation restriction (any restriction: 57.8% v 76.5%).

There was little difference between the attrition re-weighted estimates in any of the onset and persistence analyses. Therefore, all estimates presented here are for the attrition re-weighted analyses. The observed data are available in additional file 1 Observed onset and persistence of restriction in any and each aspect of life at three years.

Among the whole study population of this analysis (n = 6965), 69% did not change their overall participation status, whilst 31% either developed or recovered from participation restriction. The estimated proportion of 3431 participants with no restriction at baseline, who would report restriction at three-year follow-up was 29.8% (95% confidence interval: 27.6%, 32.0%) (Table 2). This estimated proportion was similar in women and men (29.0% (26.1%, 32.1%) v 30.6% (27.6%, 33.8%), and increased with age in both genders, although this trend was more pronounced in women. Of those with no restriction at baseline, 26.5% (24.4%, 28.6%) would indicate restriction in one to three items and 3.3% (2.5%, 4.3%) in four to eleven items, at three-year follow-up.

The estimated proportion of 3534 participants with any restriction at baseline, who would report restriction at three-year follow-up was 68.8% (66.2%, 71.3%). The proportion with persistence of restriction was higher in women than men (72.4% (69.1%, 75.4%) v 63.9% (59.8%, 67.8%), and increased with age in both genders, although more so in women than men.

However the overall picture in the persistent group hides variation in the change in status of individual participation items. Of those who would indicate three-year persistence, 12.3% (10.3%, 14.6%) would indicate the same restrictions at both time points, 20.1% (17.5%, 22.9%) would indicate the same number of restrictions but different areas would be restricted, 35.1% (32.0%, 38.4%) would indicate a lower number of restrictions and 33.2% (29.5%, 35.8%) would indicate a greater number of restrictions at follow-up.

Onset of restriction at 3 years in individual aspects of life ranged from 3.0% (2.5%, 3.7%) (work) to 12.2% (10.9%, 13.6%) (mobility outside the home) (Table 3). Onset increased with age for 8 aspects of life and decreased with age for the remaining 3 aspects of life (looking after dependents, work, education). Onset was more frequent in women than in men for two aspects of life (mobility outside the home, interpersonal interaction) (data not shown).

Onset of restriction in individual aspects of life was more frequent in those who had restriction in other aspects of life at baseline (Table 4). These relationships were not affected by adjustment for age or for gender. The association with onset was stronger as baseline restriction increased for all aspects of life, other than for looking after dependents and for work and education.

Persistence of restriction in individual aspects of life, among persons with baseline participation restriction, was highest for mobility outside the home (63.7% (59.6%, 67.6%)) and was lowest for work (10.9% (6.7%, 17.3%)) (Table 3). Persistence increased with age for one aspect of life (mobility outside the home) and decreased with age in one aspect (looking after dependants). There was no relationship between persistence and age for the remaining nine aspects of life. Persistence was greater in women than men for four aspects of life (looking after the home, looking after belongings, interpersonal interaction, social activities) (data not shown).