Changes in work organisations affect the job characteristics, health, and wellbeing of workers.1 Recently, emphasis on the role of self employment and entrepreneurship in economic growth has increased. The corporate trends behind these changes are outsourcing, subcontracting, franchising, and concentrating on core businesses. Differences in entrepreneurial activity may explain one third to one half of international differences in economic growth.2 Entrepreneurs created most of the wealth in the United States in the past 20 years, while the share of the workforce employed by the largest corporations decreased significantly.3 Small enterprises are also one of the dominant features of the European economy, with one third of all workers in Europe being employed by “micro” enterprises with fewer than 10 employees.4 Farming, the most traditional way of self employment, has gone through dramatic changes in western countries and farmers are increasingly seen as entrepreneurs.

Self employment poses health risks because of difficulties in regulation and lack of resources and motivation for health and safety promotion.4 The entrepreneur is personally responsible for almost everything, especially in small scale enterprises. In many cases he himself does the same practical work as his employees and performs the most difficult jobs—in addition to administrative tasks and management. Psychological work stress may carry over into domestic hours, and the self employed person can easily ignore regulatory systems for workplace safety and health promotion if he/she so wishes.4 However, the work characteristics of salary earners have also changed during recent years and become increasingly entrepreneur-like. Organisations demand more personal productivity and responsibility, and have increased the flexibility, diversity, and worker’s control over his working hours, consequently mixing paid and domestic working hours.5 There are very few studies investigating how these changes in work characteristics affect quality of life among different working populations.

Quality of life

Quality of life (QoL) is an increasingly important measure of wellbeing and is also increasingly used as an outcome measure in health care.6 This is emphasised at working ages when prevalence of fatal conditions is low in comparison with milder impairments. Today’s interest in QoL reflects the increased significance of subjective wellbeing and subjective evaluations of health over objective, purely physical or economic indicators.

There are two main approaches to QoL: subjective quality of life and health-related quality of life. The former usually means a more subjective, global rating of quality of life, whereas the latter is more functional capacity orientated and tries to encompass the areas of QoL that health and health care can influence.7 Health-related quality of life (HRQoL) measurement is increasingly important in health economics, as it can be used to estimate quality adjusted life years (QALYs), which combine morbidity with mortality. Generic HRQoL measurement is important for evidence based health policy, as it enables comparison across conditions and treatments.8 9 For example, the UK’s National Institute for Health and Clinical Excellence (NICE) recommends QALYs as their preferred outcome measure in cost effectiveness analysis.10

From the macroeconomic point of view; however, the preferred outcome of health care is often the preservation of the ability to work. Work ability is a complex factor to measure: it is based on both objective findings and on a person’s subjective estimations of his resources in relation to work demands.11 The monetary implications of work ability can be directly estimated. Using HRQoL as an outcome measure requires a separate valuation exercise to determine the values of different health states, and determining an acceptable price for QALY. Work ability and utility based HRQoL measurements are essential components of health economics, but combining them is not easy. For example, using work ability as outcome is likely to emphasise the cost benefit of interventions aimed at people of working ages, whereas using HRQoL may not show any benefits in populations who are initially in good health. How work ability and HRQoL interact and whether they measure the same construct is largely unknown.12

The study aims to compare the work ability, HRQoL, and QoL between salary earners and the self employed in a representative population sample of Finns. As a group, the self employed are heterogeneous, and so we analysed entrepreneurs, with or without personnel, and farmers separately.

METHODS

The survey

The study is based on the Health 2000 survey data, which represents the Finnish population aged 30 and over. The methods and basic results of the survey have been published13 14 and are available at www.ktl.fi/health2000. Briefly, the survey was based on a two stage, stratified cluster sampling design and conducted between August 2000 and July 2001. Our study is limited to a sample of 5834 people aged between 30 and 64. The study was approved by the ethics committee of the National Public Health Institute. Written informed consent was obtained.

Socioeconomic factors and chronic conditions

Information on socioeconomic factors, work status, and somatic conditions were collected using structured home interviews, with a participation rate of 89% (5184 people). People were asked whether they had ever been diagnosed by a physician with any of 43 specified diseases and conditions. Education was classified as basic, secondary, or higher education. Family income was adjusted for family size using the OECD equivalence scale, where the first adult of a family is weighted as 1.0, other adults as 0.7, and children less than 18 years old as 0.5.15 Marital status was classified as married or cohabiting, and others. We asked whether the subjects had full time employment and what type work of work they did (entrepreneur, farming, salaried). Entrepreneurs were asked whether they employ other people (yes/no). We also classified those working on a family farm without salary as farmers and those working in the family enterprise without salary as entrepreneurs. Our study concerns only those who worked full time.

Work ability index

The work ability index (WAI) developed by the Finnish Institute of Occupational Health16 is a widely used self administered questionnaire to assess an individual’s work ability. WAI has been shown to have good test-retest reliability17 and poor work ability has been shown to predict sick leaves among younger workers and early retirement among workers over 50.11 18 WAI is calculated as the sum of the scores of seven items (range of scores in parenthesis): subjective estimation of current work ability compared with lifetime best (0–10), subjective work ability in relation to both physical and mental demands of the work (2–10), number of current diseases diagnosed by a physician (1–7), subjective estimation of work impairment as a result of diseases (1–6), sickness absence during past year (1–5), own prognosis of work ability after two years (1–7), and mental resources (1–4). WAI can be divided into four categories: poor (7–27 points), moderate (28–36 points), good (37–43 points), and excellent work ability (44–49 points).16

Quality of life measurement

We used self administered questionnaires after the home interview19 to collect information on QoL. Subjective QoL was measured by asking the respondents to rate their quality of life as a whole in the last 30 days, on a scale from 0 to 10.

We used two established, generic, preference based HRQoL measures: the EQ-5D and the 15D. As there is no gold standard of HRQoL measurement and different measures can yield different results,20 we have included two HRQoL instruments. Preference based HRQoL measures allow numeric valuation of HRQoL as a single index score, commonly referred to as health utility, which is the quality component of QALYs.

The 15D (available at www.15d-instrument.net), includes 15 dimensions with five categories of severity: mobility, vision, hearing, breathing, sleeping, eating, speech, elimination, usual activities, mental function, discomfort and symptoms, depression, distress, vitality, and sexual activity.21^–23 We used the Finnish multi-attribute utility tariff to calculate the 15D index.22 The 15D index gives values between 1 (full health) and 0 (dead). The minimally important difference (MID) on the 15D index is 0.02–0.03.21 We included subjects with 12 or more completed 15D dimensions, and imputed missing values as recommended by the author of the measure.21

The EQ-5D,24^–26 available at www.euroqol.org, has five dimensions with three categories of severity: mobility, self care, usual activities, pain or discomfort, and anxiety or depression. We used the most widely used tariff, the UK time-trade-off (TTO) index,27 to estimate the health utility scores. The EQ-5D TTO index gives values between 1 (full health) and −0.59 (0 = dead). The EQ-5D is among the most extensively evaluated of HRQoL measures.6 For EQ-5D, a MID of 0.07 has been proposed.28 Only respondents fully completing the EQ-5D questionnaire were included in the analysis.

The 15D compares favourably with similar HRQoL instruments in most of the important properties.21^–23 29 30 The EQ-5D is problematic in general population samples, owing to low sensitivity in detecting deviations from full health,19 so we concentrated on the 15D results.

Statistical methods

To estimate the impact of self employment status on the quality of life and work ability, we created separate multiple regression models for 15D, EQ-5D, WAI, and QoL as dependent variables, and self employment status, age, sex, marital status, income, education, and number of chronic conditions as independent variables. To estimate the modifying effects of the socioeconomic variables, we also created models adjusting for only age and sex. Linear regression for survey data was used to analyse WAI and QoL. As the HRQoL measures have a ceiling effect (62% of respondents scored full health on the EQ-5D and 20% on the 15D) we used a Tobit model to account for this censoring.31 32 We report the marginal effects of the unconditional expected value of the HRQoL score, valued at the means of the explanatory variables.33 These marginal effects are interpreted comparably to beta coefficients of linear regression—that is, as the change in HRQoL score (health utility) associated with the self employment group in question, adjusted for socioeconomic factors and chronic conditions and compared to the average full time working Finn. We also tested for possible interactions between sex and occupational status and age and occupational status.

To estimate which components of WAI, 15D, and EQ-5D differed between the employed groups, we compared both the unadjusted and age and sex adjusted means of the WAI domains. For 15D, we compared the age and sex adjusted domain scores between the different employment groups, and with all respondents aged 30–65, irrespective of employment status. The 15D preference based scoring system scales all dimensions to values between 0 and 1, making the different domains directly comparable. As reporting serious problems on any of the EQ-5D domains was rare, we dichotomised the EQ-5D components and used logistic regression to compare the age and sex adjusted odds ratios of reporting any problems on the EQ-5D domains.

The regression analyses included only those working full time. All analyses accounted for the complex sampling design and non-participation using the survey procedures of Stata 8.2 statistical software.34

RESULTS

Of the responding 5184 people aged between 30 and 64, 3536 worked full time. Of these, 88% (3113) completed 15D, 87% (3059) EQ-5D, 93% (3304) QoL, and 91% (3221) WAI.

The socioeconomic characteristics of the sample are presented in table 1. Of the Finnish working aged population, 6.7% were full time entrepreneurs, 2.5% full time farmers, and 55.6% full time salary earners. There were more male entrepreneurs (71%) and farmers (64%) than salary earners (52%). Farmers and entrepreneurs were older than the salary earners. Farmers and entrepreneurs without personnel had lower education and lower income, adjusted for family size, than the salary earners. Entrepreneurs with personnel had the highest income.

The unadjusted HRQoL, QoL, and WAI results are presented in table 2. Entrepreneurs with personnel scored the highest on all measures other than 15D, on which the mean scores were fairly even between entrepreneurs and salary earners. Farmers had the lowest scores on all measures. When the WAI was dichotomised to poor/moderate and good/excellent work ability (cut-off 43/44 points) 36% of farmers, 16% of salary earners, 16% of entrepreneurs without personnel, and 12% of entrepreneurs with personnel had poor or moderate WAI scores.

Table 3 presents the results of the regression models comparing participants with different occupational status, adjusting for age and sex or age, sex, marital status, income, education, and number of chronic conditions.

After controlling for all the background variables, being a farmer was associated with statistically significantly worse 15D compared to salary earners and the entrepreneurs without personnel (analyses with other comparison groups not shown). For EQ-5D, farmers had worse scores than salary earners. In the case of subjective QoL, farmers had worse scores than salary earners and entrepreneurs with personnel. Farmers had worse work ability scores than all the other groups, whereas entrepreneurs with personnel had higher scores than salary earners. No differences between the two groups of entrepreneurs were statistically significant.

In fully adjusted models, age was related to worse HRQoL and WAI, but not to subjective QoL. Being married or cohabiting was associated with small improvements in 15D and WAI and relatively large improvements in QoL. Income was associated with improvement on all scales. Education improved the WAI score, but the associations with QoL and 15D were small. Chronic conditions decreased HRQoL and QoL. As chronic conditions decrease the WAI score a priori because they are a part of the measure, they were not included in the regression model. The only background variable with statistically significant but conflicting results between the measures was male sex, which was associated with improved 15D and WAI, but with decreased QoL. No statistically significant interactions between age or sex and occupational status were found in any of the measures.

The results of the models controlling for only age and sex were similar to the fully controlled models—that is, the differences in education, income, chronic conditions, or marital status did not explain the findings. The model fit (R²) clearly improved with the inclusion of all the covariates (data not shown), but even in the fully adjusted models the R² was relatively low.

Supplementary table 1 (see OEM website, http://.oem.bmj.com/supplemental) presents the scores of each component of WAI. Although farmers scored statistically significantly lower than all other groups on five of the seven components, most of the difference in the total WAI score was caused by poor scores on the two dimensions of current work ability compared with lifetime best, and work ability in relation to demands of the work. This finding did not change when age and sex were adjusted for (data not shown).

Figure 1 presents the age and sex adjusted 15D profiles of the self employed groups. The profiles are presented as age and sex adjusted deviations from salary earners. Farmers scored statistically significantly lower (only p<0.01 considered statistically significant due to multiple testing) than salary earners on domains measuring psychosocial functioning: usual activities and mental function, and the domain of general physical discomfort and symptoms. No other differences were statistically significant.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Age and sex adjusted 15D profiles of salary earners, entrepreneurs with and without personnel, and farmers.

Supplementary table 2 (see OEM website, http://.oem.bmj.com/supplemental) presents the unadjusted proportion of respondents reporting no problems on any of the EQ-5D domains and table 4 presents the age and sex adjusted odds ratios for reporting problems on any of the EQ-5D domains. Farmers report problems statistically significantly more often on three of the five domains of the EQ-5D: 17% of farmers report problems of mobility, 16% of usual activities, and 51% of pain and discomfort, compared to 8–9%, 5–6%, and 29–39% for other groups, respectively. No differences were found for domains of self care and anxiety and depression. The findings were similar when age, sex, socioeconomic factors, and number of chronic conditions were controlled for. Farmers were twice as likely to report problems of mobility or pain and discomfort and almost three times more likely to report problems of usual activities than salary earners. Adding socioeconomic status and chronic conditions to regression models controlling for age and sex changed the results only marginally (data not shown).

DISCUSSION

We conducted a general population survey to compare the work ability, HRQoL, and QoL of salaried and self employed Finns.

We found that, measured with unadjusted HRQoL, QoL, and WAI scores, the entrepreneurs with personnel seem to be the best off, and farmers the worst. When socioeconomic factors were adjusted for, most of the differences between entrepreneurs and salary earners disappeared, but farmers still scored statistically significantly lower on all scales. The results were similar whether all socioeconomic variables or only age and sex were controlled for, suggesting the differences between the employment groups (or lack of them) were not the result of differences in education, income, marital status, or chronic conditions. The problems of farmers seem to stem mostly from the psychosocial functioning and physical discomfort domains of HRQoL, and manifest themselves as poor self estimated work ability in relation to lifetime best or current work demands.

Comparison with previous studies

Farmers from industrialised countries have been shown to have poorer physical and mental health than the general population,35 36 although there are also reports to the contrary.37 38 Several reasons are suggested for this: ageing, economic difficulties, risk of injuries, work related illnesses, and exposure to pesticides.36 Our findings emphasise the severity of farmers’ problems, as they report poor HRQoL, poor QoL, and poor work ability even after chronic conditions, age, and socioeconomic factors were controlled for.

We are aware of no general population studies comparing the HRQoL of entrepreneurs, farmers, and salary earners using utility based instruments (comparable instruments being the HUI, EQ-5D, 15D, AQoL, and SF-6D). As conceptualisations of HRQoL vary, so do the results,20 making direct comparison of studies using measures other than utility based ones difficult. A small study of British farmers39 found primary farmers rated significantly lower on the EQ-5D index than the general population (mean score 0.785 for farmers vs 0.850 for non-farmers), but there was no difference in self rated general health (EQ VAS (range 0–100) valuations of 79.4 for farmers and 80.1 for non-farmers). Unfortunately, no adjustments for sex or socioeconomic factors were conducted. Respondents in our study were younger, at least partly explaining the higher raw EQ-5D scores, but generally the results were similar.

Strengths and weaknesses of the study

The strengths of our study lie in the large, representative general population sample, high response rate, and the use of validated, commonly used instruments. However, the subgroups of the self employed were relatively small, which most probably lowered the statistical significance of the findings. A cross sectional study can only show associations, so no conclusions about causality can be drawn.

Our study only deals with those working full time. Thus it is possible that the relation between work ability and QoL/HRQoL is different in the population as a whole. Some people cannot work for health reasons, so the age adjusted HRQoL of those who work is likely to be better than those who do not work. It is possible that farmers find it more difficult to leave their job and retire, as they may be reluctant to sell or pass on what is still commonly a family farm, leaving more farmers with ill health working full time. However, this could equally be the case for other entrepreneurs. Similarly, comparing sickness absence during the past year between employment groups is problematic, as the practical and economic consequences of sickness absence vary between groups. The WAI cannot be used for non-working populations and, even theoretically, defining the work ability of the heterogeneous group of non-working people in a meaningful way would be difficult.

Possible selection bias must also always be borne in mind: it is possible that, as farms are still commonly inherited in Finland, people choosing full time farming are different from the general population.

Conclusion

The dominant theories explaining the impact of work on health and wellbeing are variants of job demand control (support) JDC(S) models.40 41 JDC theories state that having control over one’s job (decision latitude) decreases stress, especially if the job is highly demanding. Low control-high demand jobs are the worst for one’s health and wellbeing, but high demand-high control jobs may actually be the most motivating. JDCS models claim that worksite social integration can buffer the negative impact of high demands. The worst job for health and wellbeing would be one with low control, low support, and high demands. The self employed and entrepreneurs could generally be considered as having more control over their jobs than salary earners, but also more work related stress. A large review of empirical studies on the JDCS theories and wellbeing42 did not find consistent support for the buffering effect of control on psychological wellbeing or job satisfaction. This is in line with our findings of comparable WAI, HRQoL, and QoL between salary earners and entrepreneurs. Our study shows how farmers have poor work ability, health-related and subjective quality of life when compared to other working populations. This finding remained even after controlling for age, sex, socioeconomic factors, and chronic conditions. The differences observed between farmers and entrepreneurs demonstrates how being self employed does not, in itself, guarantee good work ability, HRQoL, or even good subjective QoL. It is likely that the demands of even modern farming are still very different from other forms of enterprise. The finding that the problems of farmers centred on the psychosocial domains on HRQoL and not on physical health could support the “low control, low support, and high demand” nature of current Finnish farming. Entrepreneurs as a group did not differ clearly from salary earners, although the entrepreneurs with personnel were better off than those working alone. In conclusion, it seems that health-related and subjective quality of life are best optimised, if one is a healthy, married female—who is not a farmer.