Disability Pension with Special Reference to Sick Leave Track Record, Health Effects, Health Care Utilisation and Survival

Transcription

1 Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine 383 Disability Pension with Special Reference to Sick Leave Track Record, Health Effects, Health Care Utilisation and Survival A Population-based Study THORNE WALLMAN ACTA UNIVERSITATIS UPSALIENSIS UPPSALA 2008 ISSN ISBN urn:nbn:se:uu:diva-9308

2

3 To my family

4 Cover photo: My shadow by the author

5 Original papers This thesis is based upon the following publications, which will be referred to in the text by their Roman numerals I-IV: PAPER I Wallman T, Wedel H, Palmer E, Rosengren A, Johansson S, Eriksson H, Svärdsudd K. Sick-leave track record and other potential determinants of a disability grant: a population based study of 8,218 men and women followed for 16 years. Submitted. PAPER II Wallman T, Burell G, Kullman S, Svärdsudd K. Health care utilisation before and after retirement due to illness: a 13-year population-based follow-up study of prematurely retired men and referents from the general population. Scand J Prim Health Care 2004;22: PAPER III Wallman T, Wedel H, Palmer E, Johansson S, Rosengren A, Eriksson H, Welin L, Svärdsudd K. Quality of life measures among disability pensioners and referents in a longitudinal population-based study. Submitted. PAPER IV Wallman T, Wedel H, Johansson S, Rosengren A, Eriksson H, Welin L, Svärdsudd K. The prognosis for individuals on disability retirement: an 18-year mortality follow-up study of 6,887 men and women sampled from the general population BMC Public Health 2006;6:103. Reproduced with permission from the publishers.

6

7 Contents Prologue...9 Introduction...11 Background...11 A historical review of the development of disability pension in Sweden 11 Background and development...11 Insurance principles...13 Early rehabilitation in the sick leave process...14 International comparisons...14 Assessing disability and rehabilitation potential...16 Aims of the study...18 Scientific questions...18 Methods...19 Study populations...19 Data Collection...19 Disability pension register data...21 Sick leave register data...22 Hospital admission register data...22 Mortality register data...22 Ethical considerations...22 Statistical considerations...23 Results...28 Paper I...28 Paper II...32 Paper III...36 Paper IV...39 Discussion...44 Study design and methodological considerations...44 Discussion by findings...46 Conclusions...52 General conclusions...52 Implications for practice...52

8 Swedish summary...54 Acknowledgements...55 References...57 Terminology Activity compensation: Disability pension for individuals aged years. Term used from Disability pension and Disability benefit: Temporary or permanent disability pension granted to a person who is insured, and has a permanent or prolonged reduction of work capacity. Early retirement: Used in Paper II as a synonym for disability pension. Sickness absence and Sick leave: Used as synonyms for temporary absence from work due to reduced work capacity caused by illness, disease or injury. Sickness benefit: Benefit paid by the Swedish Social Insurance Agency when sick pay from the employer is no longer paid. Sickness certificate: Document issued by a physician to confirm reduced functional work capacity due to illness, disease or injury. Sickness compensation: Disability pension for individuals aged years. Term used from 2003 onwards. Sick pay: Reimbursement for the first 14 days of an illness paid by the employer to individuals employed for at least one month or who have worked for at least fourteen consecutive days. Sickness period and Sick leave period: A continuous period of sick leave days for which a sickness certificate is issued and compensation is paid. Sickness spell and Sick leave spell: Synonyms that include one or more consecutive sick leave periods.

9 Prologue Thirty years ago I was employed at the Swedish Social Insurance Agency, Eskilstuna branch office, analysing long term sickness spells, 90 days or more, and rehabilitation needs for people on long term sick leave. Many of these individuals were young, but hade no, or very little work capacity. They were often not useable on the labour market, and were granted a disability pension. No one knew if these decisions were good or bad for the individuals. In the Swedish decision making process for granting sickness benefit and disability pensions, medical certificates are very important, making clear that people on sick leave are ill and have no work capacity. At that time in my life, I was wondering in what way physicians made these decisions. So I began my medical studies to find the answer. But my question got no answer. Now, more than 25 years later, I meet these people almost daily in my work as general practitioner, and I am still wondering. We still do not know whether or not disability pension is beneficial for the individual. 9

10

11 Introduction Background In Sweden, with 9,200,000 inhabitants, and a labour force of 5,500,000 individuals, some 50,000 women (58%) and men (42%), years of age, are granted disability pensions annually [1]. In 2008, 550,000 persons (10% of the labour force) had a disability pension. Moreover, 200,000 (4%) people daily are on sick leave, which means that about 14% of the labour force is unable to work because of illness, resulting in high costs for society [2] in the form of pension benefit payments and loss of production [1]. Most people who are granted disability pensions, especially the younger ones, have participated in rehabilitational interventions. A disability pension is thus to be regarded as a final solution for a difficult-to-solve ill health problem [3-8]. The WHO Ottawa Charter for Health Promotion states: Work and recreation should be a source of health for people. The way in which a society organizes work should aid in creating a healthy society [9]. Work satisfies many human needs. It is hence one of the most important resources for human development and quality of life. From this perspective unemployment and under-employment are, to an ever-increasing extent, a public health problem. Certain groups are affected more than other. These include immigrants, women with minimal education, and adults with disabilities [10-13]. A historical review of the development of disability pension in Sweden Background and development In the 1850s, proposals were put forward in the Swedish parliament, the Riksdag, on the introduction of a pension system [14]. In the 1884 session of the Riksdag, Adolf Hedin ( ), liberal politician and journalist, proposed an investigation of accident and old age insurance for workers. The intention was to provide the elderly and invalids who could no longer support themselves due to reduction or loss of their working capacities, with financial security in a more dignified manner than that afforded by the poor relief (social services) or private charities of the day. The proposal was ap- 11

12 proved with the addition of the phrase workers and comparable individuals. This amendment was meant to pave the way for future insurance for small company owners, primarily farmers. The proposal resulted in the 1913 Act on a Government Funded Pension, upon which the current Swedish pension system is based. According to the act, in principle, the entire population of the country, both women and men, were insured for an old age and disability pension. Major changes were made in the government insurance legislation in 1938, but these were mostly with regard to current economic conditions [14]. After World War II, work began on reforming the health insurance and the national old age pension schemes. All Swedish residents were to be guaranteed reasonable security against the financial consequences of illness, disability and old age. The champions of reform eyed the British Beveridge Plan, which was aimed at giving the UK an effective social security system. The Swedish Social Security Committee of 1958 produced, in 1961, a report on issues including disability pensions and health insurance [15]. It contained coordination between health insurance and the pension system. The result of this work was the National Social Insurance Act, effective as of 1 January Since 1963, part time disability pension has been possible. Three groups of disability were defined. Group I were those at least 5/6 incapacitated, Group II were those who had partial disability of at least 2/3 but less than 5/6, and Group III were those who had a disability of at least 1/2 but not as much as 2/3. A 1/2, 2/3 or full disability pension could be granted. Since 1993, 25%, 50%, 75%, or 100% disability pensions may be granted [3]. Until 1970, a disability pension could only be given for medical reasons. In reality other factors such as unemployment, business conditions and geographical living circumstances could affect the decision. For people 63 years or older, medical and labour market conditions could be weighed into disability pension decisions, and in 1976 the age limit was reduced to 60 years. Further liberalisation occurred in 1977, when a disability pension could be granted regardless of the reason for incapacity to work. Even people with long-term alcohol or drug problems could thus receive disability pensions [16]. The number of disability pension applications increased during the 1970s and 1980s. In 1991, the right to receive a disability pension for causes related to the labour market was rescinded, and since 1997 the possibility of combining labour market related and medical reasons has been removed [16]. In January 2003, the disability pension was replaced for individuals between with activity compensation, while for individuals between years old it is called sickness compensation. The previous levels were retained [3]. 12

13 Insurance principles The Swedish social insurance act states that an individual whose work capacity is reduced due to illness is to be granted an allowance for her/his support in the form of sick leave compensation, pension or a life annuity [15]. Work is defined as: A lasting and well-adapted human activity that has another purpose than the pleasure afforded by the actual occupation. It can be of two kinds an unremunerated job or a salaried job (an activity for which a remuneration is paid) [17]. An individual s capacity for work is the result of the interaction between the individual s capacities and the job s demands on the individual. The concept of incapacity for work is insufficiently defined, and common instruments for assessing capacity for work are lacking. This is true for work capacity in general, but especially for mental and social work capacity [18]. The preparatory work for the Social Insurance Act in 1962 established that long-term incapacity to work caused by illness should be the basic principle for a disability pension [3]. The individual s medical condition is central. The purpose of a disability pension is to ensure citizens financial security when there is a premature loss or reduction of the work or incomeearning capacity. The principle is not to compensate for illness or physical injury, nor for conditions resulting in these, without consideration of how the capacity for support is affected [15]. Rehabilitation efforts are to precede the assessment for disability pension, and the results of these attempts shall serve as guidance for decisions on disability pension. There is no definition available for the term completely fit for work. Decisions are based on the individual and the situation concerning whether the person may have some remaining work capacity that can be utilized. The work tasks should be adapted to the ill individual s strength and skills [15]. Pension assessment is to be based on a doctor s certificate. The certificate is to be accompanied by the collected data on occupational and social factors which, together with the certificate, make up the basis for determining whether there is cause for a disability pension [15]. Until June 2008 a temporary disability pension (a time-limited disability pension for one, two or three years) could be granted if the person s work capacity was assessed not to be permanent, but to last for a period of at least one year [15]. From 1 July 2008 this time limitation is no longer a possibility for sickness compensation decisions, which are now to be permanent and last until the person is eligible for an old age pension [19]. From 1 January 2003, sickness compensation and activity compensation were maximized to three years, in accordance with the preparatory work for the legislation, after which it was to be reconsidered impartially again [3]. Activity compensation is to include a continuing rehabilitation plan for the individual. The insurance levels are the same as for previous disability pen- 13

14 sion and temporary disability pensions, i.e. 100%, 75%, 50%, and 25%. The 2/3 pension no longer exists [3]. Early rehabilitation in the sick leave process The basic idea underpinning the Social Insurance Act is to give priority to rehabilitation efforts and not simply grant a disability pension, so as to enable the individual to return to work [15]. Rehabilitation is defined as covering all the interventions that can be called upon for the purpose of restoring the physical or mental functional capacities of ill or injured person, or for helping her/him to overcome remaining disabilities and, when feasible, become independent and self-supporting members of society [20]. Quite early it became apparent that the pension scheme would involve high costs for the state. Pension costs might be reduced by facilitating return to work for individuals on sick leave by means of rehabilitation interventions. As early as in 1915 it was decided that rehabilitation interventions should be available for those who had the right to receive disability pension. The National Pensions Board was commissioned to handle disabilitypreventing health care and occupational training [14]. In 1954 the National Board of Health and Welfare proposed the establishment of special rehabilitation clinics at the three state rehabilitation hospitals, for the purpose of medical rehabilitation [21]. They were to offer physiotherapy, work physiology testing, and the services of a welfare officer. Direct occupational rehabilitation was taken care of by the Occupation Rehabilitation Unit (Arbetsvården), a section of the Swedish Public Employment Service, using methods like work tests and occupational rehabilitation. The right to receive rehabilitation was established, as was an obligation for the person on sick leave to accept and undergo rehabilitation [21]. For employed persons, the employers had the primary rehabilitation responsibility until This included investigation and planning of how the individual could be able to return to work, preferably with the present employer. The Riksdag adopted a decision in 2007 that the primary responsibility for rehabilitation was with the Swedish Social Insurance Agency and no longer with the employer [1]. For the unemployed, the Swedish Social Insurance Agency continued to have the primary responsibility in cooperation with the Swedish Public Employment Service. The Swedish Social Insurance Agency has the ultimate responsibility for seeing to it that the needs of the individual are met [16]. International comparisons How basic economic security is organized can be described using either Beveridge s or Bismarck s criteria. According to Beveridge, the state guar- 14

15 Figure 1. Disability pension recipiency rate in 2002, per 1000 persons aged Dean D. "Learning from others": about partial disability pension programs. New Brunswick, NJ: Rutgers University, With the author s, Dr David Dean, permission. antees a minimum level, leaving room for additional trade union insurance coverage. Bismarck s criteria are based on income-related insurance financed by direct charges [16]. It is difficult to compare disability pension systems in various countries [22, 23]. The pension systems have varying degrees of compensation and requirements for qualifications. In some countries disability refers to reduced working capacity, in other countries to reduced earning capacity and in still other countries to both [22]. Moreover, there are varying criteria for a full disability pension. In Australia it requires 20 points on a point rating scale of body systems and inability to work 30 hour/week. In Germany the requirement is inability to engage in three or more hours work per day, and in Japan it is meeting one of 11 disability conditions classifying the person as First Class (totally incapacitated). In the Netherlands, earnings incapacity rating of greater than 80% has to be assessed by a medical advisor and a labour market expert, and in Norway and Sweden social insurance board assessment of 100% earnings incapacity is required. There is similar diversity regarding partial disability pension criteria. In Germany the requirement is that individuals can work only between three and six hours per day. In Japan Second Class (17 conditions) entitles the person to 80% of full national pension, and Third Class (14 conditions) only to a pension paid by the employer. The Netherlands uses six levels of 15

16 loss in earnings capacity, from 15-25% through 65-80%. In Norway, 50-95% of lost earnings capacity in five percent increments is used [24]. As shown in Figure 1 the disability pension prevalence rates vary among countries, from only a few pensions per 1000 individuals in Japan, to 100 per 1000 individuals in Norway. Assessing disability and rehabilitation potential The health care services have medical responsibility for patient rehabilitation [25], with the goal of optimizing the patient s health. If this is not feasible, the secondary goal is to treat the symptoms of the illness or disability so that the patient attains as good a health status as possible. The Swedish Social Insurance Agency assesses the individual s work capacity and has administered the rehabilitation via a seven-step model for the last decade, with the goal that the individual should be able to return to work [26]. The assessment criteria are: Can the individual carry out the tasks of her/his present job after necessary treatment or convalescence? Can the individual carry out the tasks of her/his present job after some rehabilitation or adaptation of the tasks? Can the individual find or carry out other tasks with the same employer without extra effort? Can the individual find other tasks with the same employer after some training, adaptation of tasks, or similar rehabilitation interventions? Can the individual manage to do any other job that is normally available on the labour market, without extra effort? Can the individual manage to do any other job that is normally available on the labour market after some rehabilitating interventions, for example education or re-training? Is the individual permanently incapacitated for work or incapacitated for an appreciable period of time? Since 1 July 2008 the seven-step model has been changed to a rehabilitation chain [27]. If the individual is not capable of doing her/his ordinary work within the first three month of the sickness spell, the individual is granted sickness benefit. If the individual is unable to do other jobs at the ordinary place of work, the Social Insurance Agency pays sickness benefit for up to six month. After six month of sickness benefit the individual must be unable to work on the labour market all to get further sickness benefit, with a maximum time limit of one year. After that, the Swedish Social Insurance Agency can change the person from sickness benefit to disability pension, or stop the compensation for the sick spell. 16

17 One of the problems involved in cases of long-term sick leave is assessing the potential for return to work. Potential prognostic factors, such as age, sex, socio-economic factors, time on sick leave, and others have been investigated in several studies [28-51]. From a scientific point of view it is a complication that the utilisation of the social insurance benefit is affected by political decisions, such as the compensation level, and by economic factors, such as the business cycle, with more liberal utilisation during good times than during recessions. The Swedish Council on Technology Assessment in Health Care performed an extensive literature review in 2003, showing that there were relatively few well carried out published studies on potential determinants of disability grants [52]. Most published studies are small or focussed on particular occupational groups, special medical problems, sick listing practices, or specific risk factors. We therefore decided to perform a large-scale prospective study on determinants and consequences of disability pension. 17

18 Aims of the study The aims of the thesis were to gain knowledge about disability pensions and to study the medical consequences of disability pension. Scientific questions 1. Can the probability of a disability pension be predicted? What are the determinants and at what point in the course of events can the risk of a disability pension be high enough to motivate preventive actions being taken? 2. Does having a disability pension affect health care utilisation and how long do these effects continue? 3. Does having a disability pension affect a person s quality of life? If so, is quality of life affected already before the disability pension is granted or not until afterwards? 4. Does having a disability pension affect survival? If so, is survival affected by the underlying disease or are there other factors at work? 18

19 Methods Study populations Data from ten ongoing population cohort studies in Gothenburg, Eskilstuna, and Uppsala, Sweden, with baseline investigations performed between 1980 and 2003, were used for this study. The names of the studies, investigation year, sex distribution, age range, sample sizes, response rates, and investigation procedures are presented in Table 1. Random samples of the specified age and sex segment of the local general populations were drawn from the national population register for a total of 14,538 subjects. Of these, 10,678 subjects (5,238 women and 5,440 men) participated in the baseline examinations in the various surveys. The overall response rate was 73.5%. In Papers I and IV the study population included the cohorts BEDA I, Men Born in 1933, ESKIL, Men Born in 1943 examined in 1993, and Public Health Cohort Uppsala. In Paper II, ESKIL and all 80 men years at age, who were granted disability pension during the period of 1 November 1980 to 31 October 1981, and all 135 men granted disability pension during the period of 1 November 1985 to 31 October 1986 for the first time, and who were living in the city of Eskilstuna, Sweden, were sampled in November The age range used was chosen to cover men assumed to be fully trained for their occupations and established on the labour market, but not very near old age pension age. The response rate was 70.7% among the pensioners and 73.5% among the referents. In Paper III the cohorts BEDA II, Men Born in 1933, ESKIL, Men Born in 1943 and examined in 1993 and 2003, Women Born in 1953, Men Born in 1953 and Public Health Cohort Uppsala were used. Data Collection Postal questionnaires were sent to some of these subpopulations, while others answered questionnaires on location in connection with a medical examination. Questionnaire data used in these studies were identical in all subpopulations where they were measured. The information obtained included sex and age at baseline examination, marital status, and number of people in household unit, educational level, smoking habits, and quality of life meas- 19

20 Table 1. Data source, examination year, sex, age range at baseline, sample size, response rate and investigation procedure for the subpopulations Subpopulations Investigation Sex Age range Sample size Responders Response rate <65 years at baseline year place n n % n Investigation procedure BEDA I 1980 Gothenburg women q 1) + me 2) BEDA II 1997 Gothenburg women q 1) + me 2) Men born in Gothenburg men q 1) + me 2) ESKIL 1986 Eskilstuna men pq 3) Men born in Gothenburg men q 1) + me 2) Men born in Gothenburg men q 1) + me 2) Women born in Gothenburg women q 1) + me 2) Men born in Gothenburg men q 1) + me 2) Public Health Cohort 1993 Uppsala women men pq 3) 1) questionnaires 2) medical examination 3) postal questionnaire

21 ures. For this thesis, marital status was classified as married/cohabiting or not. Educational level was classified on a 4-point scale ranging from mandatory school only (= 1) to college or university education (= 4). Smoking habits were classified as currently smoking or not smoking. In addition, for eight of the cohorts a 5-degree scale was available (never smoked, ex-smoker, currently smoking less than 15 cigarettes per day, cigarettes per day or 25 cigarettes per day or more). The Gothenburg Quality of Life Instrument with its three sub-scales Complaint Score, Well-being and Activity Score, was used to measure quality of life aspects [53]. Complaint Score lists 30 general symptoms. The respondents were asked to indicate which of these they had experienced during the last three months. Possible responses were yes or no. The subinstrument is not intended to measure presence of disease but rather the tendency to report complaints. In the Well-being sub-scale, nine items are listed and the responses are given on Likert scales ranging from "very bad" (=1) to "excellent, could not be better" (=7), with no verbal description of the intervening steps. The items used in Paper III were health, self-esteem, sleep, energy, mood and patience. The respondents were asked to indicate their present situation for each of these items. The Activity Score sub-scale lists 32 specified leisure time activities covering six areas and two open alternatives. The subjects were asked to indicate which of these activities they had performed during the last two months with response alternatives never (0), occasionally (1) or often or regularly (2). The scores were summed across the area and to an overall activity score, high scores indicating active lifestyle. In Paper II, the postal questionnaire included, beyond the variables mentioned above, number of visits to hospital emergency rooms, hospital outpatient clinics, general practitioners (GPs ) or district nurses offices, home calls by GPs or district nurses, occupational health service clinics, and private physicians offices during the previous 12 months. Moreover, information was obtained on employment status (employed or unemployed), among the referents in 1986, and among the retired men at the time of their disability pensions. Disability pension register data The Swedish Social Insurance Agency administers all disability benefits. Information on whether the subjects in the study populations (except Men and Women Born in 1953) had been granted a disability pension at any time from 1971 until 2001 was obtained from the Agency. The data obtained included decision date, diagnoses, extent (100%, 75%, 67%, 50% or 25%) and type (temporary or permanent). More than one decision could be taken, for instance first for a temporary and then for a permanent disability pension, or 21

22 first for a part time one and then for a full time one. The number of decisions ranged from 1 to 6. In Paper IV data from all decisions were used, in all other papers only the first decision was used. The four largest disability pension diagnosis groups were musculoskeletal disorders, psychiatric disorders, cardiovascular disorders, and neurological disorders. Among the disability pensioners the time interval between the date of the disability pension decision and the date of the baseline examination was computed as the number of days between decision and baseline, negative for those who were disability pensioners at baseline, and positive for those who became so after baseline. The time interval was used only in Paper III. In the cohort Men and Women Born in 1953, information on disability pension was obtained from questionnaire data. However, no interval could be calculated since no information on decision date was available. Sick leave register data Information on all compensated days of sick leave for each individual in the study populations (except Men and Women Born in 1953) from 1 January 1986 until 31 December 2002 was obtained from the Swedish Social Insurance Agency and was complete. The data included the first and last day of each sickness spell, the type of sick leave benefit (compensation for sickness, work injury, or rehabilitation), and extent of sick leave (25%, 50%, 75% or 100%). Hospital admission register data Data on all hospital admissions from 1971 until December 31, 2002 was obtained from the National Hospital Inpatient Registry and were complete. The data obtained included day of admission, day of discharge and diagnoses. Mortality register data Data on cause-specific mortality from 1971 until December 31, 2002 was obtained from the National Causes of Death Register. The data obtained were: date of death, underlying cause of death and place of death. The reasons for disability pension, discharge diagnoses, and causes of death were classified according to the ICD versions 8-10 [54]. Ethical considerations Informed consent for participation in the studies was obtained from all participants, oral in the early studies and written later on, as required first by the 22

23 Research Ethics Committees at Uppsala and Gothenburg Universities and later by the National Research Ethics Board. The Committees and the Board approved the study on several occasions during the data collection process. Statistical considerations Data were analysed with the SAS [55] and JMP [56] statistical programme packages. Data on age, sex, pension status, sick leave data, disability pension data, hospital admission data and survival data were 100% complete. Data on socio-economic factors, such as marital status, educational level, household size and smoking habits were 98.1% complete among respondents. Quality of life data completeness varied since all variables were not measured in all subpopulations. Summary statistics, such as means and measures of dispersion, were computed with traditional parametric methods. Simple differences between the groups regarding continuous data were tested with Student's t-test or analysis of variance, and nominal or ordinal data with the chi-square test. Since the distributions of all sick leave data were skewed towards high values, all analyses were performed on original as well as on log transformed data. Since the results were almost identical, only results based on original data are presented. In Paper I a guiding pilot study was performed based on a nested casereferent design. Cases were defined as individuals who were granted a disability pension during follow up (n=1498). For individuals who did not have a disability pension at baseline and had not received one during follow up, and who survived until the cases became disability pensioners, two sets of referents were matched to the cases, one by age, sex, and geographical area (the cities of Gothenburg, Eskilstuna, or Uppsala), and the other by age and sex only (total referent n=2996). The time at which the cases received their disability pension was defined as time 0 for both cases and their referents. Mean and median sick leave duration for each month from time 0 and backwards until baseline was then calculated in the two groups. The longest follow-up periods covered 192 months. To avoid problems owing to right truncation of unfinished sick leave at time 0, and bias due to the effects of cases waiting for their disability pension decision and on unlimited sick leave during this time, the follow-up time was ended two years before time 0. As shown in Figure 2, cases and referents had fairly similar mean sickness spell durations during the first few months. The differences then increased progressively during follow up, (similarly for women and men), indicating that measures of sick leave track record increased over time in both groups, particularly among those who became disability pensioners, in turn indicating that logistic regression or proportional hazards regression might be appropriate analytical methods. 23

24 Individual sickness spell duration, days Disability pensioners, monthly means Disability pensioners, trend line Referents, monthly means Referents, trend line Follow-up time, months Figure 2. Mean sick leave duration among female and male disability pensioners and referents per month during follow up until two years before disability pension (follow-up time 0) of the cases. Results from pilot testing In the main study design of Paper I, the study population was treated as a cohort followed from 1 January 1986 until 31 December First, the monthly proportion of individuals on sick leave in relation to individuals open to the possibility of sick leave certification (still alive, age less than 65, not on disability pension) was calculated and successively adjusted for nonexposure. As shown in Figure 3, there was evident variation over time, similar in the three geographical areas. The levels increased slowly during , then fell until 1998 after which they increased again. During the period there was strong seasonal variation, with lower levels during the summer than during the rest of the year, less pronounced from 1993 on, owing to change of regulations. The variation of the proportion of individuals on sick leave across calendar time indicates the need to adjust for calendar time period as a proxy for sick leave position in the business cycle and for the effects of political decisions. In the analyses, a time variable based on calendar time period was used, coded as , , and The analyses of the influence of length of sick leave and other potential determinants on the probability of receiving a disability pension during follow up were performed using logistic regression. Four analytical models (I- IV) were used: being granted a disability pension was the outcome variable in all models, and one of the measure, sickness spell duration (I), sickness spell interval (II), cumulative annual sick leave days (III), or cumulative two- 24

25 Proportion on sick leave, % Gothenburg Eskilstuna Uppsala Calendar month Figure 3. Proportion (%) of individuals on sick leave in the cities of Gothenburg, Eskilstuna, and Uppsala per month during the 16-year study follow-up time year sick leave days (IV), was used as exposure variable in each model. In addition, the variables age at baseline, marital status, educational level, household size, smoking habits, calendar time period, and geographical area were entered in all models. The analyses were stratified for sex. Follow-up time was defined as number of days from baseline until disability pension, death, or end of follow up, whichever came first. In Paper II, adjustments for the influence of possible confounding gave minor effects. Inpatient health care utilisation was based on data for the complete samples and measured as the proportion admitted to hospital, analysed with Cox s proportional hazards regression, and number of admissions for those admitted, analysed with Student s t-test of log-transformed data, adjusted for mortality-caused differences in length of follow up. The effect of adjustment for potential confounding due to patient characteristics (based on questionnaire data) was small. For this reason unadjusted health care utilisation data are presented. For the analyses reported in Table 7, the study population was subdivided into four subgroups, those who were less than 65 years of age at baseline and who had no disability pension during the study period, those who were less than 65 years of age and had a disability pension at baseline, those who were less than 65 years of age at baseline and were granted a disability pension after baseline, and finally those who were 65 years or older at baseline and among whom the vast majority were old age pensioners. 25

26 The mean levels of the various quality of life measures in the four study population groups were computed with multivariate linear regression, adjusting for the influence of age, sex, marital status, education, smoking habits, and household size. The same technique was used for significance tests of complaint score and activity variables between the study groups, whereas significance of the well-being variables was tested with multivariate ordinal logistic regression, using the same adjustment variables as mentioned above. The regression line in Figure 8 was computed with multivariate linear regression technique, modelled as a continuous function of the time dependent variable by connected linear and quadratic pieces in specified intervals. In Paper IV, January 1 of the examination year or the year of the postal questionnaire mailing was used as baseline and December was the last day of follow up. Those who had a disability pension at baseline were regarded as disability pensioners throughout the study. Those who received a disability pension during follow up were regarded as non-retired from baseline until the day before they received their pension, and from then on as disability pensioners. If a new pension decision was taken after the baseline date the observation period of the old decision was terminated the day before the new decision and a new observation period was started on the day of the new decision. For those with more than one decision before baseline, the one in effect at baseline was counted as the first. Subjects who were referents throughout the study thus had only one observation period, from baseline to death or end of follow up. Those who were disability pensioners at baseline or became disability pensioners during follow up had one or more observation periods, the first one starting at baseline and succeeding periods starting when disability pension status changed. The underlying hazards function and cumulative mortality rate were analyzed with a time dependent Poisson regression procedure adapted to the SAS programme package. In these analyses the relation between the hazards function of the various groups was found to be approximately proportional. Hazards ratios (HR) and their 95% confidence intervals (95%CI) were therefore also computed with Cox's proportional hazards regression. Censoring events were termination of a disability pension decision (because a new one was about to take effect) or no event until the end of follow up. The outcome variable was death from any cause during follow up. Follow-up time was right truncated at the end of the 18th year because of small numbers. The effects of potential confounders and effect modifiers on mortality were adjusted for by including these variables as covariates in the model. The results were checked with survival analyses in strata of these variables. No power calculations were made before the studies were initiated, because it was anticipated by experience in the research group that the statistical power would be sufficient. This was verified by a post hoc power analysis that showed a power of more than 96% to detect the differences in hospital admission rate between disability pensioners and referents found in Paper 26

27 II. The corresponding statistical power to detect the mortality rate differences between the disability pensioners and referents found in paper IV was more than 99.9%. The power was of the same order of magnitude in papers I and III. Thus, none of the papers was underpowered. Only two-tailed tests were used. Generally, p-values < 0.05 were regarded as statistically significant. However, in some of the papers a large number of analyses were performed. To account for multiple testing a modified form of Bonferoni adjustment was used, where the p-level sought was obtained by dividing 0.05 by the square root of the number of analyses performed. 27

28 Results Paper I The characteristics of the study population are presented in Table 2. There was a strong relationship between individual sickness spell duration (model I) and cumulative annual days of sick leave (model III) on the one hand and being granted a disability pension on the other, among both women and men. The interval between sickness spells (model II) showed a corresponding inverse relationship. Sickness spell duration (model I) and the cumulative number of sick leave days increased with age. The mean and median sickness spell durations were 16.4 and 3 days among those aged 34 or less, 24.5 and 4 days among those aged years, 34.0 and 4 days among those age 45-54, and 33.3 and 4 days among those aged years. The corresponding means and medians for cumulative annual sick leave days were 15.7 and 0 days, 17.5 and 0 days, 26.0 and 0 days, and 30.3 and 1 day. The interval (II) means showed a corresponding difference between disability pensioners and referents as the duration, but in the reverse direction. The cumulative annual sick leave days (III) showed similar variations between disability pensioners and referents as the sick leave duration data, and so did cumulative two-year sick leave days (IV). All four sick leave measures showed significant differences between disability pensioners and referents, and between the two sexes. For both women and men, sickness spell duration, age and calendar time period independently increased the probability of disability pension, while high educational level decreased the probability (Table 3, upper half). Among women, the most powerful predictors were duration, calendar time period and age in that order, and among men duration, age and calendar time period. The model involving interval between sickness spells (II) showed similar results as the one with sickness spell duration, except that the effect of the interval variable was the reverse of that of duration. The effects of the model involving cumulative annual sick leave days (III) are shown in the lower part of Table 3. The results were similar to those for sickness spell duration. The model, using two-year cumulative sick leave days, showed similar results. The strength of the predictors is reflected by the chi-square values (all with 1 degree of freedom). The exposure variables 28

29 Table 2. Characteristics of study population Mean or % Disability beneficiaries Referents Women Men Women Men SD Mean or % No. of subjects No. of person years No. of sick leave spells Mean follow-up time, years < < Sickness spell duration, days Mean < < Median < < Sickness spell interval Mean < < Median < < Cumulative one-year sick leave days Mean < < Median < < Age at baseline < < Married, % < ns Mandatory education only, % < < Household size ns ns Smokers, % <0.001 ns SD Mean or % SD Mean or % SD Benefit status p Sex

30 Disability pension probability, % Annual cumulative days, men and women Two-year cumulative days, men and women Duration of individual sickness spell periods, women Duration of individual sickness spell periods, men Sick leave days Figure 4. Cumulative distribution of the proportion of women and men combined receiving disability pension by annual and two-year sick leave days and for women and men separately by sickness spell duration during the 16-year follow-up time with no further adjustment in model I-IV explained 94-97% of the disability pension variance, depending on sex and sick leave measure used. The crude influence of various levels of sickness spell duration, cumulative annual sick leave days and cumulative two-year sick leave days on cumulative disability pension distribution is presented in Figure 4. For a 50% disability pension probability, a sickness spell duration of 390 days among women and 480 days among men was needed. The corresponding number of cumulative annual sick leave days was 245 days, and for cumulative twoyear sick leave days 365 days, for both women and men. The corresponding number of days needed to reach an 80% probability were 830 and 730 sick leave days, 360 annual sick leave days, and 540 cumulative two-year days. Cumulative annual sick leave days thus appeared to be the most powerful predictor of receiving a disability pension. The probability of receiving a disability pension described in terms of annual sick leave days, age and sex, adjusted for the influence of marital status, educational level, household size, smoking habit and calendar time period is presented in Figure 5. The probability increased in terms of annual sick leave days and age, but not in terms of sex. Among the youngest subjects, the disability benefit probability ranged from 0.1% with few annual sick leave days 30

31 to 14.9% for individuals with the maximum number of days. The corresponding probabilities among the oldest subjects were 1.6% and 72.1%. Only the two oldest age groups reached a 50% probability of disability pension. Table 3. Effects of possible determinants for the prediction of being granted a disability pension during the study follow-up time. In the two logistic regression models disability pensioner-to-be (yes/no) was used as dependent variable and the listed variables as independent variables OR Women 99.5%CI OR Men 99.5%CI Chisquare Chisquare Single sickness spell duration Duration by 10 day-periods Age by 10-year age groups Calendar time period versus versus Educational level Married Household size < Smokers <0.1 Gothenburg versus Uppsala Eskilstuna versus Uppsala Cumulative annual sick leave days Sick leave days by 10-day periods Age by 10-year age groups Calendar time period versus versus Educational level Married Household size Smokers Gothenburg versus Uppsala Eskilstuna versus Uppsala

32 Disability pension probability, % 75 Age 60 at baseline Age 50 at baseline Age 40 at baseline Age 30 at baseline Annual sick leave days Figure 5. Probability of being granted a disability pension in relation to number of annual sick leave days and age during the 16-year follow-up time, adjusted for the influence of sex, marital status, education, household size, smoking habits, and business cycle Paper II The men who were granted a disability pension were somewhat older than the referents, and unmarried or divorced to a larger extent, Table 4. They had less education, were more often living in one-person households and were more often smokers and unemployed at the time of retirement. The differences in characteristics between the groups were independent of the age differences. The dominant diagnostic group among disability pensioners was musculoskeletal disorders. The disability pensioners reported significantly more of all sorts of health care utilisation except for appointments with occupational health service centres and doctor and district nurse home calls. The odds of being in touch with outpatient health care services during the past 12 months were 8.2 (95%CI ) times higher among the retired men than among the referents. The men who had been retired for 5 years had insignificantly lower odds then those retired for one year or less, (OR 0.80, 95%CI ). 32

33 Table 4. Characteristics of study population at baseline of follow-up Retired men Referents n mean or % 95%CI n mean or % 95%CI p Age, years < Married, % <0.001 Mandatory education only, % < One-person households, % < Smokers, % < Unemployed, % < Retirement diagnosis, % Cardiovascular disorders Psychiatric disorders Musculoskeletal disorders Miscellaneous

34 Table 5. First hospital admission rate among men on disability pension and referents and number of admissions among those admitted, 1 July 1986 until 31 December 1998 Main discharge diagnosis First admission to hospital Admitted persons Retired men Referents RR 95% CI Mean number of admissions among n rate n rate retired men referents ratio 95% CI Infections Tumours Endocrine disorders Psychiatric disorders Alcohol or drug abuse Other Neurological disorders Cardiovascular disorders Coronary heart disease Stroke Other Respiratory disorders Gastrointestinal disorders Urinary and genital disorders Dermatological disorders Musculoskeletal disorders Trauma Observation Symptom diagnoses Miscellaneous disorders All causes n= number of persons admitted to hospital, rate = number of admitted persons/year of exposure, RR = rate ratio, 95%CI = 95% confidence interval, ratio = ratio of number of admissions per person and year of exposure among retired men in relation to referents for those admitted to hospital