J Occup Rehabil DOI 10.1007/s10926-014-9505-x
Work Ability Index as Tool to Identify Workers at Risk of Premature Work Exit Corne´ A. M. Roelen • Martijn W. Heymans Jos W. R. Twisk • Jac J. L. van der Klink • Johan W. Groothoff • Willem van Rhenen
•
Ó Springer Science+Business Media New York 2014
Abstract Purpose To investigate the Work Ability Index (WAI) as tool for identifying workers at risk of premature work exit in terms of disability pension, unemployment, or early retirement. Methods Prospective cohort study of 11,537 male construction workers (mean age 45.5 years), who completed the WAI at baseline and reported their work status (employed, unemployed, disability pension, or retired) after mean 2.3 years of follow-up. Associations between WAI scores and work status were investigated by multinomial logistic regression analysis. The ability of the WAI to discriminate between workers at high and low risk of premature work exit was analyzed by the area (AUC) under the receiver operating characteristic curve. Results 9,530 (83 %) construction workers had complete data for analysis. At follow-up, 336 (4 %) workers reported disability pension, 125 (1 %) unemployment, and 255 (3 %)
C. A. M. Roelen (&) W. van Rhenen ArboNed Occupational Health Service, PO Box 85091, 3508 AB Utrecht, The Netherlands e-mail: [email protected] C. A. M. Roelen M. W. Heymans J. W. R. Twisk Division Methodology and Applied Biostatistics, Department of Health Sciences, VU University, Amsterdam, The Netherlands C. A. M. Roelen J. J. L. van der Klink J. W. Groothoff Division Community and Occupational Medicine, Department of Health Sciences, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands M. W. Heymans J. W. R. Twisk EMGO? Institute for Health and Care Research, VU University Medical Center, Amsterdam, The Netherlands W. van Rhenen Center for Leadership and Management Development, Business University Nyenrode, Breukelen, The Netherlands
retirement. WAI scores were prospectively associated with the risk of disability pension at follow-up, but not with the risk of unemployment and early retirement. The WAI showed fair discrimination to identify workers at risk of disability pension [AUC = 0.74; 95 % confidence interval (CI) 0.70–0.77]. The discriminative ability decreased with age from AUC = 0.78 in workers aged 30–39 years to AUC = 0.69 in workers C50 years of age. Discrimination failed for unemployment (AUC = 0.51; 95 % CI 0.47–0.55) and early retirement (AUC = 0.58; 95 % CI 0.53–0.61). Conclusions The WAI can be used to identify construction workers \50 years of age at increased risk of disability pension and invite them for preventive interventions. Keywords Discriminative ability Prognostic research ROC curve Sensitivity and specificity WAI
Introduction The ageing workforce is an increasingly important occupational health issue in developed countries [1–3]. Lower birth rates lead to a declining inflow of young workers and, consequently, older workers have to stay on the job. In addition, many European Union (EU) governments have recently raised the retirement age in reaction to increased population longevity. As a result, people will have to work until older age. Although the EU employment rate for people aged 55–64 years rose from 37 % in 2000 to 47 % in 2011, it is still far below the 69 % employment rate of people aged 20–55 years [2]. The lower employment rate among older workers has been attributed to premature work exits. People above age 50 are often the first to be fired and the last to be hired [4]. Their skills can be out-
123
J Occup Rehabil
dated because employers do not consider them for training and/or because older workers see little benefit to investing their time in training. Poor working conditions and inflexible work schedules may push older workers into early retirement. Such reasons for leaving employment are based on a concept known as ‘organizational age’ i.e., the ageing of individuals in jobs and organizations [5]. Alternative reasons to leave employment are based on the social perceptions of age or on changes in life stages. For example, older workers may decide to leave employment to take extra years of leisure time with family and friends. The concept of ‘functional age’ refers to changes in physical and cognitive capacities with advancing age [5]. Generally, physical capacities decline with age, particularly due to changes in the cardiovascular (e.g., aerobic capacity) and musculoskeletal (e.g., grip strength, lifting strength, limb motility, balance) systems [6–8]. Some cognitive capacities, such as precision and speed of perception also decline with age [7], while others, such as the abilities to reason, deliberate, and comprehend the whole strengthen with age [6]. Functional ageing more or less affects a worker’s performance. Older workers who experience difficulties in performing their work are still encouraged to leave the workforce [4]. In addition, ill health resulting from physical or cognitive deficits contributes to the likelihood of premature work exit of older workers in EU countries [9]. Hence, raising the retirement age may well lead to an increased incidence of premature work exits, particularly in physically demanding jobs, such as construction [8, 10]. The possibility to sustain a work life with advancing age depends on an individual’s work ability [11, 12]. In this paper, we define work ability as having the health, competences, and virtues to manage work demands [13]. The Work Ability Index (WAI) was developed in the 1980s [14] and is nowadays the most widely used tool for measuring work ability [15]. WAI scores have been shown to decline with age [10–12, 16–20]. Women’s WAI scores tend to decrease more with age than do men’s WAI scores [19]. Furthermore, WAI scores have been shown to decrease faster in workers performing physical work, than in those doing mental work [11, 17, 20]. Monteiro et al. [19] reported that the progressive decline in work ability with advancing age increased the risk of work disability and early retirement. Several studies have demonstrated that low or declining WAI scores are associated with an increased risk of premature work exit in terms of work disability [19, 21] or early retirement [19, 22, 23]. Liira et al. [24] investigated WAI scores in relation to both unemployment and disability pension in middle-aged Finnish construction workers. To our knowledge, there are no studies that investigated disability pension, unemployment, and early
123
retirement as different types of premature work exit. Furthermore, the prospective associations found in previous studies do not tell us whether or not the WAI can be used to identify workers at risk of premature work exit. For that purpose, we need to know to what extent the WAI discriminates high-risk from low-risk workers. The objective of the present study was to investigate the discriminative ability of the WAI to assess its usefulness as tool for identifying workers in different age groups at risk of different types of premature work exit.
Methods Prospective cohort study including 11,537 men (age 16–62 years) working in one of 3,069 construction companies who participated in health checks in the period 2005–2007. As part of their health check, the construction workers completed the WAI. After mean 2.3 [standard deviation (SD) 0.1] years, they reported their work status: employed, unemployed, disability pension, or retired. This study was approved by the Medical Ethics Committee of the University Medical Center Groningen (reference METc2012/105). Social Context In The Netherlands, sickness absence is employer-compensated for a maximum period of 2 years. After 2 years of sickness absence, sick-listed employees are examined by a civil service insurance physician [25]. If the insurance physician concludes that medical conditions restrict an individual’s work capacity to such an extent that the person loses more than 35 % of his/her former income, then a (partial) disability pension is granted irrespective of whether disability is work-related. If the insurance physician ascertains that the individual can earn C65 % of his/her income in another job that better suits the person’s remaining functional capacities, then unemployment benefits are granted after 2 sickness absence years. Besides, unemployment benefits are granted when employees are discharged. Until the 1990s, generous early retirement benefits were more popular among ageing workers than disability or unemployment benefits. With the gradual recognition of the problem of population ageing, the policies with regard to premature work exit intensified [26, 27]. As a result, the Dutch government and the Foundation of Labour agreed on a covenant in 1997, replacing the popular early exits paid by the state with capital funded collective pre-pension schemes, which are less generous and, therefore, less attractive for employees. Euwals et al. [27] reported that these policy reforms have reduced the incidence of early retirements.
J Occup Rehabil
Predictor Variable: Work Ability Index (WAI) The WAI is a self-administered tool that addresses the demands of work and both the worker’s health status and resources to cope with work demands [28]. In the present study, we use the short version of the WAI, including only 15 diseases instead of 51 in the original WAI [29]. The WAI items were weighted according to the manual. The total WAI score ranges from 7 (poor work ability) to 49 (excellent work ability) and higher scores reflect better work ability [30]. De Zwart et al. [31] investigated WAI test–retest reliability and reported that the mean WAI score and the classification into WAI categories were stable over a 4-week interval. Radkiewich and Widerszal-Bazyl [32] further investigated the psychometric characteristics of the WAI and concluded that it can be treated as internally coherent measure of work ability and can be considered a very predictive, cross-nationally stable instrument. In a systematic review of the literature, Van den Berg et al. [15] found that WAI scores were associated with individual characteristics (older age and obesity), lifestyle (lack of leisure-time physical activity) and work, particularly high physical workload, high mental work demands, and lack of autonomy in work. Information on obesity was not available, but we could adjust the analyses for age, leisure-time physical exercise, physical and mental job demands, as well as job autonomy. Leisure-time physical activity was assessed by the question: ‘How often did you become all sweaty because of vigorous physical exercise or sports?’ with response categories: ‘less than once a month’ (=1), ‘once a month’ (=2), ‘2–3 times a month’ (=3), ‘1–2 times a week’ (=4), and ‘three times a week or more’ (=5). Physical and mental job demands were assessed by the questions ‘Is your work physically demanding?’ and ‘Is your work mentally demanding?’, respectively, with response categories yes (=1) and no (=0). Job autonomy was assessed by the question ‘Can you decide how you do your work?’, with responses yes/no. Outcome Variables Construction workers self-reported their work status: employed, disability pension,1 unemployed,2 or retired after a mean follow-up period of 2.3 (SD = 0.1) years. Workers who were on sick leave at baseline were excluded from the analyses. During the study period, the retirement
1
Disability pension refers to individuals who after 2 years of sickness absence were deemed to be limited in work capacity. 2 Unemployed refers to individuals who were discharged or were deemed after 2 years of sickness absence to not be limited in work capacity.
age in The Netherlands was 65 years; all workers who reported retirement at follow-up were on early retirement. Statistical Analysis Statistical analyses were performed in IBM SPSS Statistics for Windows, Version 20.0 (IBM Corp., Armonk, NY, released August 2011). To assess clustering of observations at the company level, we calculated intraclass coefficients (ICCs) and found ICC = 0.04 for disability pension, ICC = 0.37 for unemployment, and ICC = 0.19 for early retirement, indicating that a considerable amount of the variance in unemployment and early retirement was explained by the clustering of workers into companies. However, 75 % of construction companies staffed\5 workers and 16 % staffed 5–10 workers. Such small clusters lack statistical power and, therefore, we abandoned the idea of multilevel analysis. Instead, we performed multinomial logistic regression, which is a direct extension of binary logistic regression to situations with more than two unordered outcome categories. Associations of WAI scores with outcome categories unemployment and early retirement were non-linear. Therefore, we included baseline WAI scores as categorical independent variable in a multinomial logistic regression model, using previously described WAI categories [14]: poor work ability (score 7–27), moderate work ability (score 28–36), good work ability (score 37–43), and excellent work ability (score 44–49). Multinomial logistic regression estimates odds ratios (ORs) and 95 % CIs relative to outcome category ‘employed’ and adjusted for age, leisure time physical exercise, physical and mental job demands, and job autonomy. Discrimination refers to the ability to distinguish between different outcomes. The area (AUC) under the receiver operating characteristic (ROC) curve is most commonly used to investigate discrimination [33]. If outcome categories are unordered, as in multinomial logistic regression, discrimination cannot be assessed by a single AUC. Biesheuvel et al. [34] compared AUCs of multinomial ROC analysis with AUCs of consecutive binary ROC analyses and found that they were very similar. Discrimination is usually over-optimistic when the data used for ROC analysis are the same as those used to estimate regression coefficients. Bootstrapping is the preferred method to correct for this over-optimism [33]. However, most statistical software packages lack the option to bootstrap multinomial logistic regression models. Therefore, we chose to perform consecutive ROC-analyses based on binary logistic regression for each type of premature work exit and subsequently adjust the resulting AUCs for over-optimism. For ROC analyses, WAI categories were scored as follows: 1 = excellent, 2 = good, 3 = moderate and 4 = poor work ability. An AUC C0.90 indicated excellent,
123
J Occup Rehabil Table 1 Baseline population characteristics
Age (in years)
Included in complete cases analyses (n = 9,530)
Excluded because of missing data (n = 2,007)
Mean
Mean
45.5
SD
N
%
9.5
45.7
SDa
N
Analysis
% P = 0.23b
9.4
Age categories (years) \30 30–39
786
8
169
8
1,270
13
256
13
40–49
3,563
37
728
36
C50
3,911
41
854
43
Years employed in construction Work hours per week
24.3 39.7
12.1 8.7
24.9 38.5
P = 0.26b P \ 0.01b
12.3 8.8
P \ 0.01c
Leisure time physical exercise \19 per month
2,899
37
701
60
19 per month
451
6
46
4
2–39 per month
847
11
79
7
1–29 per week
2,368
30
235
20
C39 per week
1,252
16
114
10
No
3,555
38
616
32
Yes
5,869
62
1,320
68
No
6,042
64
1,271
66
Yes
3,387
36
657
34
7,558
80
1,509
78
1,840
20
419
22
Work Physically demanding P \ 0.01c
Mentally demanding
a
Standard deviation
b
t test for independent samples
c
Chi square test
Autonomy No Yes
0.80–0.89 good, 0.70–079 fair, and 0.60–0.69 poor discrimination; discrimination failed if AUC \0.60 [33]. Over-optimism of the WAI’s discriminative ability was estimated by comparing discrimination in 250 bootstrap samples with discrimination in the original sample, using Harrell’s regression modeling strategies (RMS) package in R [35]. Throughout the study, we present over-optimism adjusted AUCs, which generalize to other working populations within the construction sector.
Results Of the 11,537 construction workers, 2,007 (17 %) had missing WAI data, particularly on the items sickness absence within the last year (1,106 missings) and estimation of work ability in 2 years’ time (357 missings). Construction workers with complete WAI data worked more hours per week and reported more frequent leisure time physical exercise than did construction workers with missing WAI data (Table 1). Construction workers with complete WAI data less often mentioned physically demanding work and less often experienced autonomy in
123
P = 0.12c
P = 0.03c
work than workers with missing WAI data. The mean WAI score was 40.1 (SD 4.9) and decreased with age from 42.5 (SD = 3.8) in construction workers \30 years to 41.5 (SD = 4.1) in those aged 30–39 years, 40.3 (SD = 4.6) in the age group 40–49 years, and 38.9 (SD = 5.2) in workers C50 years of age (ANOVA P \ 0.01). At follow-up, 336 (4 %) construction workers reported disability pension, 125 (1 %) unemployment, and 255 (3 %) early retirement. The odds of disability pension increased with decreasing WAI categories to OR 16.08 for construction workers reporting poor work ability compared to those reporting excellent work ability (Table 2). The odds of unemployment also increased with decreasing work ability categories, but differences relative to excellent work ability were not significant. The WAI categories were not significantly associated with the odds of early retirement among male construction workers. Bootstrapping techniques showed 1.2, 14.2, and 4.4 % over-optimism for the WAI’s ability to discriminate between construction workers at high and low risk of disability pension, unemployment, and early retirement, respectively. After adjustment for over-optimism, the WAI fairly discriminated between construction workers with and
J Occup Rehabil Table 2 Multinomial analysis of work ability and premature work exit relative to being employed at follow-up Work Ability Index
N
%
Disability pension OR
a
Unemployment b
95 % CI
OR
a
Early retirement b
ORa
95 % CI
95 % CIb
Excellent (44–49)
2,391
25
1.00
–
1.00
–
1.00
–
Good (37–43)
5,255
55
1.68
1.04–2.07
0.53
0.32–0.88
1.22
0.77–1.93
Moderate (28–36)
1,689
18
8.29
5.20–13.20
1.06
0.58–1.92
1.50
0.90–2.49
195
2
16.08
8.98–28.79
2.03
0.75–5.51
0.75
0.26–2.20
Poor (7–27) a
Odds ratio adjusted for age, physical exercise, physical and mental job demands, and job autonomy
b
Confidence interval
a
Percentage per age group
b
Area under receiver operating characteristic (ROC) curve
c
Confidence interval
Age
Disability pension a
N (% )
Unemployment
b
c
AUC (95 % CI )
a
N (% )
Early retirement b
c
AUC (95 % CI )
N (%a)
AUCb (95 % CIc)
\30 years
1 (0)
Not analyzed
7 (1)
0.58 (0.35–0.81)
0
Not analyzed
30–39 years
5 (0)
0.78 (0.30–1.00)
12 (1)
0.59 (0.42–0.76)
0
Not analyzed
40–49 years
89 (2)
0.74 (0.67–0.80)
59 (2)
0.52 (0.44–0.60)
14 (0)
0.60 (0.51–1.00)
C50 years
241 (6)
0.69 (0.66–0.74)
47 (1)
0.52 (0.43–0.62)
241 (6)
0.55 (0.51–0.59)
Total
336 (4)
0.74 (0.70–0.77)
125 (1)
0.51 (0.47–0.55)
255 (3)
0.58 (0.53–0.61)
without disability pension (AUC = 0.74; 95 % CI 0.70–0.77). The discriminative ability was highest in male construction workers aged 30–39 years and decreased with age (Table 3). Discrimination failed for both unemployment (AUC = 0.51; 95 % CI 0.47–0.55) and early retirement (AUC = 0.58; 95 % CI 0.53–0.61) as is shown in Fig. 1. The optimal WAI cut off for identifying construction worker at risk of disability pension was between good and moderate work ability, with sensitivity 0.63 and specificity 0.83. At a cut off between poor and moderate work ability, sensitivity was 0.12 indicating that 88 % of workers with disability pension at follow-up would be missed when only those reporting poor work ability at baseline would be considered at increased risk of disability pension.
Discussion The WAI identified male construction workers at risk of disability pension, although the discriminative ability decreased with age. The WAI did not discriminate between workers at high and low risk of unemployment or early retirement. Work Ability and Disability Pension The significant association between WAI scores and disability pension confirms the findings of previous studies [21, 22, 36]. Liira et al. [24] showed that the WAI was predictive of disability pension among 961 Finnish construction workers aged C40 years (55 % 40–49 years and
1.00 0.90 0.80 0.70
sensitivity
Table 3 WAI as risk marker of premature work exit stratified by age group
0.60 0.50 0.40 0.30 0.20 0.10 0.00 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
1–specificity
Fig. 1 Discriminative ability of Work Ability Index (WAI). The figure shows the receiver operating characteristic curve representing the ability of WAI categories (1 = excellent, 2 = good, 3 = moderate, and 4 = poor work ability) to discriminate construction workers at high and low risk of disability pension (black line), unemployment (dark grey line), and early retirement (light grey line); the diagonal indicates no discrimination above chance
45 % C50 years). The authors found that workers with poor work ability had a 10.7 times higher risk of disability pension as compared to those with excellent work ability; 36 % of the workers with poor WAI scores and 16 % of those with moderate WAI scores received disability pension at 4-year follow-up. Alavinia et al. [21] found that poor to moderate WAI scores were highly predictive of disability pension after on average 23 months of follow-up
123
J Occup Rehabil
of 850 Dutch construction workers aged C40 years (mean age 48.4 years). Our study of 9,530 Dutch construction workers (aged 16–62 years; mean age 45.5 years, with 37 % aged 40–49 years and 41 % C50 years) added that the WAI correctly discriminated between workers at high and low disability pension risk in 74 % of the cases. Furthermore, we found that the discriminative ability decreased with age. This was probably due to the fact that WAI scores also decrease with age. It is conceivable that low WAI scores are less discriminative in age groups where lower work ability is more common. Work Ability, Unemployment and Early Retirement The risk of unemployment increased with decreasing WAI categories, but associations were not statistically significant. Previously, Liira et al. [24] showed that baseline WAI scores did not predict unemployment at 4-year follow-up. We found that the WAI poorly identified male construction workers at increased risk of unemployment. It should be reminded here that 37 % of the variance in unemployment in our study was explained by clustering at company level. Possibly, employees were discharged because of organizational downsizing or bankruptcy rather than transferred to unemployment because of their incapacity to do construction work. This could explain why the WAI did not identify workers at risk of unemployment. Furthermore, the shared variance at the company level implicates that the discriminative ability of the WAI may differ across companies. It is conceivable that the WAI can identify workers at risk of unemployment in some companies, for example economically stable and profitable companies. The WAI was not associated with early retirement, which contrasts the results of Salonen et al. [23], who reported that low WAI scores, poor physical work ability, and heavy physical workload were associated with early work exit of employees in the Finnish food industry. Our results showed that the WAI failed to discriminate between construction workers at high and low risk of early retirement. Possibly, workers retired because they knew that prepension benefits would soon become capital-funded and less generous. Dutch retirement policies were reformed in 1997, but pre-pension benefits remained attractive for workers in arduous jobs (e.g., construction work) for up to 10 years afterwards. Unfortunately, information on why construction workers had retired and what type of prepension benefits (state- or capital-funded) they were receiving was not available. Study Strengths and Limitations The prospective design and the large sample size are strengths of the study. The study population consisted of
123
construction workers employed at different companies all over The Netherlands. Thus, it is unlikely that results were biased by organizational or regional policies and practices. Older construction workers were over-represented in the study population, which is probably due to the fact that trade unions particularly stimulate older construction workers to participate in health checks. Associations were adjusted for age, but the overall discriminative ability of the WAI may have been attenuated by an over-representation of older workers. We dealt with this problem by presenting the discriminative ability of the WAI stratified by age groups. Previous research has shown a ‘healthy volunteer effect’ implicating that healthy individuals are more likely to participate in health surveys than are workers experiencing poor health [37]. The fact that construction workers with complete WAI data reported more frequent leisure time physical exercise than those with missing data might be indicative of such a healthy volunteer effect, which could have weakened associations between WAI scores and premature work exit. A further limitation of the study is the use of selfreported measures. Outcomes such as disability pension and unemployment are negatively valued in the Dutch economy and may have been under-reported. This type of ‘social desirability bias’ [38] could have attenuated associations of WAI scores with disability pension and unemployment. In a large-scale Swedish study, the overall agreement between self-reported and registered disability pension was 96 %; specificity was 99 %, while the sensitivity was 70 % indicating that 30 % of employees with registered disability pension did not report disability pension [39]. Early retirement is less stigmatized and the association between the WAI and early retirement is less likely to be affected by social desirability bias. Practical Implications The present results show that the WAI fairly discriminates between construction workers at high and low disability pension risk. This implicates that the WAI could be used to screen construction workers for risk of disability pension. At a cut off between moderate and good work ability, the specificity was 0.83 and sensitivity 0.63, meaning that 37 % of construction workers who reported disability pension at follow-up had good or excellent WAI scores at baseline. At a cutoff between poor and moderate work ability, specificity increased to 0.98 restricting false-positive rates to 2 %, but sensitivity was low and most construction workers at risk of disability pension would be missed when only those reporting poor work ability were identified as being at risk.
J Occup Rehabil
The discriminative ability of the WAI decreased with age. It should be acknowledged that the incidence of disability pension was low in the age groups \30 and 30–39 years. Furthermore, the present results apply only to male construction workers. Thus, studies in larger as well as other working populations are required to further investigate the WAI’s ability for identifying workers at risk of disability pension. If larger-scale studies confirm that the WAI’s discriminative ability declines with age to poor discrimination in workers aged C50 years, then employers and occupational health providers should be recommended to measure work ability with the WAI in workers aged \50 years. Subsequently, workers with poor to moderate WAI scores could be motivated to participate in workplace health promotion (WHP) programs to improve or restore work ability [40]. It has been shown that WHP programs increase work ability and decrease sickness absence and disability leaves [41, 42]. A recent meta-analysis of WHP effectiveness, however, showed that the overall effect of WHP programs is small [43]. This might indicate that WHP programs are not adequately targeted at the employees who need them. Considering the aforementioned healthy volunteer effect, employees with health complaints or unhealthy lifestyle behaviors may be less inclined to participate in WHP programs. WHP programs could be more effective when better targeted to high-risk groups. The WAI could be used to identify high-risk workers and specifically invite them for WHP programs. Another practical implication may be to use the WAI to identify workers who could benefit from occupational rehabilitation services. Bethge et al. [44] showed that reduced work ability does not always lead to the utilization of rehabilitation services. The authors discussed insufficient information by social insurance agencies and lack of appropriate counseling by general practitioners as major barriers for utilizing rehabilitation services. The authors also found indications that employees refrain from requesting rehabilitation when working conditions are precarious and insecure. The use of the WAI to measure work ability at the workplace could overcome these barriers.
Conclusion We conclude that the WAI identifies male construction workers at risk of future disability pension, but not those at risk of unemployment or early retirement. Workers with less than good WAI scores could be invited for tailor-made multidimensional interventions to restore or improve work ability.
References 1. Organisation for Economic Co-operation and Development. Maintaining prosperity in an ageing society. http://www.oecd. org/els/public-pensions/2429430.pdf. Accessed 05 Feb 2014. 2. Eurostat. http://epp.eurostat.ec.europa.eu/statistics_explained/index. php/Labour_markets_at_regional_level. Accessed 02 Mar 2014. 3. Bureau of Labor Statistics 2008. Older workers: are there more older people in the workplace? http://www.bls.gov/spotlight/ 2008/older_workers/. Accessed 20 Dec 2013. 4. Organisation for Economic Co-operation and Development. Older workers: living longer, working longer. http://www.oecd. org/social/family/35961390.pdf. Accessed 05 Feb 2014. 5. Kooij D, De Lange A, Jansen P, Dikkers J. Older workers’ motivation to continue to work: five meanings of age. A conceptual review. J Manag Psychol. 2008;23:364–94. 6. Ilmarinen J. Aging workers. Occup Environ Med. 2001;58: 546–52. 7. Charness N. Aging and human performance. Hum Factors. 2008;50:548–55. 8. Kenny GP, Yardley JE, Martineau L, Jay O. Physical capacity in older adults: implications for the aging worker. Am J Ind Med. 2008;51:610–25. 9. Schuring M, Burdorf L, Kunst A, Mackenbach J. The effects if ill health on entering and maintaining paid employment: evidence in European countries. J Epidemiol Community Health. 2007;61:597–604. 10. Sluiter JK. High-demand jobs: age-related diversity in work ability? Appl Ergon. 2006;37:429–40. 11. Ilmarinen J, Tuomi K. Work ability of ageing workers. Scand J Work Environ Health. 1992;18(Suppl 2):8–10. 12. Ilmarinen J. Toward a longer worklife: ageing and the quality of worklife in the European Union. Helsinki: Finnish Institute of Occupational Health; 2005. 13. Tengland PA. The concept of work ability. J Occup Rehabil. 2011;21:275–85. 14. Tuomi K, Ilmarinen J, Jahkola A, Katajarinne L, Tulkki A. Work ability index, second revised version. Helsinki: Finnish Institute of Occupational Health; 1998. 15. Van den Berg TI, Elders LA, de Zwart BC, Burdorf A. The effects of work-related and individual factors on the Work Ability Index: a systematic review. Occup Environ Med. 2009;66:211–20. 16. Tuomi K, Ilmarinen J, Martikainen R, Aalto L, Klockars M. Aging, work, life style and work ability among Finnish municipal workers. Scand J Work Environ Health. 1997;23(Suppl 1):58–65. 17. Ilmarinen J, Tuomi K, Klockars M. Changes in the work ability of active employees over an 11-year period. Scand J Work Environ Health. 1997;23(Suppl 1):49–57. 18. Pohjonen T. Perceived work ability of home care workers in relation to individual and work-related factors in different age groups. Occup Med. 2001;51:209–17. 19. Monteiro MS, Ilmarinen J, Correˆa Filho HR. Work ability of workers in different age groups in a public health institution in Brazil. Int J Occup Saf Ergon. 2006;12:417–27. 20. Safari S, Akbari J, Kazemi M, Mououdi MA, Mahaki B. Personnel’s health surveillance at work: effect of age, body mass index, and shift work on mental work load and work ability index. J Environ Public Health. 2013;28:94–8. 21. Alavinia SM, de Boer AGEM, van Duivenbooden JC, FringsDresen MH, Burdorf A. Determinants of work ability and its predictive value for disability. Occup Med. 2009;59:32–7. 22. Sell L, Bu¨ltmann U, Rugulies R, Villadsen E, Faber A, Søgaard K. Predicting long-term sickness absence and early retirement pension from self-reported work ability. Int Arch Occup Environ Health. 2009;82:1133–8.
123
J Occup Rehabil 23. Salonen P, Arola H, Nyga˚rd CH, Huhtala H, Koivisto AM. Factors associated with premature departure from working life among ageing food industry employees. Occup Med. 2003;53:65–8. 24. Liira J, Matikainen E, Leino-Arjas P, Malmivaara A, Mutanen P, Rytko¨nen H, Juntunen J. Work ability of middle-aged Finnish construction workers—a follow-up study in 1991–1995. Int J Ind Ergon. 2000;25:477–81. 25. Organisation for Economic Co-operation and Development. Sickness and disability schemes in the Netherlands. http://www. oecd.org/social/soc/41429917.pdf. Accessed 20 Dec 2013. 26. Van Oorschot W. Narrowing pathways to early retirement in the Netherlands. Policy Press. 2007;15:247–55. 27. Euwals R, van Vuuren D, Wolthoff R. Early retirement behaviour in the Netherlands: evidence from a policy reform. Economist. 2010;158:209–36. 28. Ilmarinen J. The work ability index (WAI). Occup Med. 2007;57:160. 29. Nu¨bling M, Hasselhorn HM, Seitsamo J, et al. Comparing the use of the short and the long disease list in the Work Ability Index Questionnaire. In: Proceedings of the second international symposium on work ability. Verona: ICOH; 2004. p. 74. 30. Ilmarinen J, Tuomi K. Past, present and future of work ability. In: Ilmarinen J, Lehtinen S, editors. People and work—research reports 65. Helsinki: Finnish Institute of Occupational Health; 2004. 31. De Zwart BCH, Frings-Dresen MCH, van Duivenbooden JC. Test-retest reliability of the Work Ability Index questionnaire. Occup Med. 2002;52:177–81. 32. Radkiewich P, Widerszal-Bazyl M. Psychometric properties of work ability index in the light of comparative survey study. In: International congress series 1280. Amsterdam: Elsevier; 2005. p. 304–9. 33. Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models. Epidemiology. 2010;21:128–38. 34. Biesheuvel CJ, Vergouwe Y, Steyerberg EW, Grobbee DE, Moons KGM. Polytomous logistic regression analysis could be
123
35. 36.
37. 38. 39.
40. 41.
42.
43.
44.
applied more often in diagnostic research. J Clin Epidemiol. 2008;61:125–34. Harrell FE. Regression modeling strategies. http://cran.r-project. org/web/packages/rms. Accessed 20 Dec 2013. Tuomi K, Toikkanen J, Eskelinen L, et al. Mortality, disability and changes in occupation among aging municipal employees. Scand J Work Environ Health. 1991;17(Suppl 1):58–66. Etter JF, Perneger TV. Analysis of non-response bias in a mailed health survey. J Clin Epidemiol. 1997;50:1123–8. Fisher RJ, Katz JE. Social-desirability bias and the validity of self-reported values. Psychol Mark. 2000;17:105–20. Svedberg P, Ropponen A, Lichtenstein P, Alexanderson K. Are self-report of disability pension and long-term sickness absence accurate? Comparisons of self-reported interview data with national register data in a Swedish twin cohort. BMC Public Health. 2010;10:763. Gould R, Ilmarinen J, Ja¨rvisalo J, et al. Dimensions of work ability. Helsinki: Finnish Centre for Pensions; 2008. Proper KI, Staal BJ, Hildebrandt VH, van der Beek AJ, van Mechelen W. Effectiveness of physical activity programs at worksites with respect to work-related outcomes. Scand J Work Environ Health. 2002;28:75–84. Kuoppala J, Lamminpaa A, Husman P. Work health promotion, job well-being, and sickness absences—a systematic review and meta-analysis. J Occup Environ Med. 2008;50:1216–27. Rongen A, Robroek SJW, van Lenthe FJ, Burdorf A. Workplace health promotion: a meta-analysis of effectiveness. Am J Prev Med. 2013;44:406–15. Bethge M, Radoschewski FM, Gutenbrunner C. The work ability index as a screening tool to identify the need for rehabilitation: longitudinal findings from the second German sociomedical panel of employees. J Rehabil Med. 2012;44:980–7.
Work Ability Index as Tool to Identify Workers at Risk of Premature Work Exit Corne´ A. M. Roelen • Martijn W. Heymans Jos W. R. Twisk • Jac J. L. van der Klink • Johan W. Groothoff • Willem van Rhenen
•
Ó Springer Science+Business Media New York 2014
Abstract Purpose To investigate the Work Ability Index (WAI) as tool for identifying workers at risk of premature work exit in terms of disability pension, unemployment, or early retirement. Methods Prospective cohort study of 11,537 male construction workers (mean age 45.5 years), who completed the WAI at baseline and reported their work status (employed, unemployed, disability pension, or retired) after mean 2.3 years of follow-up. Associations between WAI scores and work status were investigated by multinomial logistic regression analysis. The ability of the WAI to discriminate between workers at high and low risk of premature work exit was analyzed by the area (AUC) under the receiver operating characteristic curve. Results 9,530 (83 %) construction workers had complete data for analysis. At follow-up, 336 (4 %) workers reported disability pension, 125 (1 %) unemployment, and 255 (3 %)
C. A. M. Roelen (&) W. van Rhenen ArboNed Occupational Health Service, PO Box 85091, 3508 AB Utrecht, The Netherlands e-mail: [email protected] C. A. M. Roelen M. W. Heymans J. W. R. Twisk Division Methodology and Applied Biostatistics, Department of Health Sciences, VU University, Amsterdam, The Netherlands C. A. M. Roelen J. J. L. van der Klink J. W. Groothoff Division Community and Occupational Medicine, Department of Health Sciences, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands M. W. Heymans J. W. R. Twisk EMGO? Institute for Health and Care Research, VU University Medical Center, Amsterdam, The Netherlands W. van Rhenen Center for Leadership and Management Development, Business University Nyenrode, Breukelen, The Netherlands
retirement. WAI scores were prospectively associated with the risk of disability pension at follow-up, but not with the risk of unemployment and early retirement. The WAI showed fair discrimination to identify workers at risk of disability pension [AUC = 0.74; 95 % confidence interval (CI) 0.70–0.77]. The discriminative ability decreased with age from AUC = 0.78 in workers aged 30–39 years to AUC = 0.69 in workers C50 years of age. Discrimination failed for unemployment (AUC = 0.51; 95 % CI 0.47–0.55) and early retirement (AUC = 0.58; 95 % CI 0.53–0.61). Conclusions The WAI can be used to identify construction workers \50 years of age at increased risk of disability pension and invite them for preventive interventions. Keywords Discriminative ability Prognostic research ROC curve Sensitivity and specificity WAI
Introduction The ageing workforce is an increasingly important occupational health issue in developed countries [1–3]. Lower birth rates lead to a declining inflow of young workers and, consequently, older workers have to stay on the job. In addition, many European Union (EU) governments have recently raised the retirement age in reaction to increased population longevity. As a result, people will have to work until older age. Although the EU employment rate for people aged 55–64 years rose from 37 % in 2000 to 47 % in 2011, it is still far below the 69 % employment rate of people aged 20–55 years [2]. The lower employment rate among older workers has been attributed to premature work exits. People above age 50 are often the first to be fired and the last to be hired [4]. Their skills can be out-
123
J Occup Rehabil
dated because employers do not consider them for training and/or because older workers see little benefit to investing their time in training. Poor working conditions and inflexible work schedules may push older workers into early retirement. Such reasons for leaving employment are based on a concept known as ‘organizational age’ i.e., the ageing of individuals in jobs and organizations [5]. Alternative reasons to leave employment are based on the social perceptions of age or on changes in life stages. For example, older workers may decide to leave employment to take extra years of leisure time with family and friends. The concept of ‘functional age’ refers to changes in physical and cognitive capacities with advancing age [5]. Generally, physical capacities decline with age, particularly due to changes in the cardiovascular (e.g., aerobic capacity) and musculoskeletal (e.g., grip strength, lifting strength, limb motility, balance) systems [6–8]. Some cognitive capacities, such as precision and speed of perception also decline with age [7], while others, such as the abilities to reason, deliberate, and comprehend the whole strengthen with age [6]. Functional ageing more or less affects a worker’s performance. Older workers who experience difficulties in performing their work are still encouraged to leave the workforce [4]. In addition, ill health resulting from physical or cognitive deficits contributes to the likelihood of premature work exit of older workers in EU countries [9]. Hence, raising the retirement age may well lead to an increased incidence of premature work exits, particularly in physically demanding jobs, such as construction [8, 10]. The possibility to sustain a work life with advancing age depends on an individual’s work ability [11, 12]. In this paper, we define work ability as having the health, competences, and virtues to manage work demands [13]. The Work Ability Index (WAI) was developed in the 1980s [14] and is nowadays the most widely used tool for measuring work ability [15]. WAI scores have been shown to decline with age [10–12, 16–20]. Women’s WAI scores tend to decrease more with age than do men’s WAI scores [19]. Furthermore, WAI scores have been shown to decrease faster in workers performing physical work, than in those doing mental work [11, 17, 20]. Monteiro et al. [19] reported that the progressive decline in work ability with advancing age increased the risk of work disability and early retirement. Several studies have demonstrated that low or declining WAI scores are associated with an increased risk of premature work exit in terms of work disability [19, 21] or early retirement [19, 22, 23]. Liira et al. [24] investigated WAI scores in relation to both unemployment and disability pension in middle-aged Finnish construction workers. To our knowledge, there are no studies that investigated disability pension, unemployment, and early
123
retirement as different types of premature work exit. Furthermore, the prospective associations found in previous studies do not tell us whether or not the WAI can be used to identify workers at risk of premature work exit. For that purpose, we need to know to what extent the WAI discriminates high-risk from low-risk workers. The objective of the present study was to investigate the discriminative ability of the WAI to assess its usefulness as tool for identifying workers in different age groups at risk of different types of premature work exit.
Methods Prospective cohort study including 11,537 men (age 16–62 years) working in one of 3,069 construction companies who participated in health checks in the period 2005–2007. As part of their health check, the construction workers completed the WAI. After mean 2.3 [standard deviation (SD) 0.1] years, they reported their work status: employed, unemployed, disability pension, or retired. This study was approved by the Medical Ethics Committee of the University Medical Center Groningen (reference METc2012/105). Social Context In The Netherlands, sickness absence is employer-compensated for a maximum period of 2 years. After 2 years of sickness absence, sick-listed employees are examined by a civil service insurance physician [25]. If the insurance physician concludes that medical conditions restrict an individual’s work capacity to such an extent that the person loses more than 35 % of his/her former income, then a (partial) disability pension is granted irrespective of whether disability is work-related. If the insurance physician ascertains that the individual can earn C65 % of his/her income in another job that better suits the person’s remaining functional capacities, then unemployment benefits are granted after 2 sickness absence years. Besides, unemployment benefits are granted when employees are discharged. Until the 1990s, generous early retirement benefits were more popular among ageing workers than disability or unemployment benefits. With the gradual recognition of the problem of population ageing, the policies with regard to premature work exit intensified [26, 27]. As a result, the Dutch government and the Foundation of Labour agreed on a covenant in 1997, replacing the popular early exits paid by the state with capital funded collective pre-pension schemes, which are less generous and, therefore, less attractive for employees. Euwals et al. [27] reported that these policy reforms have reduced the incidence of early retirements.
J Occup Rehabil
Predictor Variable: Work Ability Index (WAI) The WAI is a self-administered tool that addresses the demands of work and both the worker’s health status and resources to cope with work demands [28]. In the present study, we use the short version of the WAI, including only 15 diseases instead of 51 in the original WAI [29]. The WAI items were weighted according to the manual. The total WAI score ranges from 7 (poor work ability) to 49 (excellent work ability) and higher scores reflect better work ability [30]. De Zwart et al. [31] investigated WAI test–retest reliability and reported that the mean WAI score and the classification into WAI categories were stable over a 4-week interval. Radkiewich and Widerszal-Bazyl [32] further investigated the psychometric characteristics of the WAI and concluded that it can be treated as internally coherent measure of work ability and can be considered a very predictive, cross-nationally stable instrument. In a systematic review of the literature, Van den Berg et al. [15] found that WAI scores were associated with individual characteristics (older age and obesity), lifestyle (lack of leisure-time physical activity) and work, particularly high physical workload, high mental work demands, and lack of autonomy in work. Information on obesity was not available, but we could adjust the analyses for age, leisure-time physical exercise, physical and mental job demands, as well as job autonomy. Leisure-time physical activity was assessed by the question: ‘How often did you become all sweaty because of vigorous physical exercise or sports?’ with response categories: ‘less than once a month’ (=1), ‘once a month’ (=2), ‘2–3 times a month’ (=3), ‘1–2 times a week’ (=4), and ‘three times a week or more’ (=5). Physical and mental job demands were assessed by the questions ‘Is your work physically demanding?’ and ‘Is your work mentally demanding?’, respectively, with response categories yes (=1) and no (=0). Job autonomy was assessed by the question ‘Can you decide how you do your work?’, with responses yes/no. Outcome Variables Construction workers self-reported their work status: employed, disability pension,1 unemployed,2 or retired after a mean follow-up period of 2.3 (SD = 0.1) years. Workers who were on sick leave at baseline were excluded from the analyses. During the study period, the retirement
1
Disability pension refers to individuals who after 2 years of sickness absence were deemed to be limited in work capacity. 2 Unemployed refers to individuals who were discharged or were deemed after 2 years of sickness absence to not be limited in work capacity.
age in The Netherlands was 65 years; all workers who reported retirement at follow-up were on early retirement. Statistical Analysis Statistical analyses were performed in IBM SPSS Statistics for Windows, Version 20.0 (IBM Corp., Armonk, NY, released August 2011). To assess clustering of observations at the company level, we calculated intraclass coefficients (ICCs) and found ICC = 0.04 for disability pension, ICC = 0.37 for unemployment, and ICC = 0.19 for early retirement, indicating that a considerable amount of the variance in unemployment and early retirement was explained by the clustering of workers into companies. However, 75 % of construction companies staffed\5 workers and 16 % staffed 5–10 workers. Such small clusters lack statistical power and, therefore, we abandoned the idea of multilevel analysis. Instead, we performed multinomial logistic regression, which is a direct extension of binary logistic regression to situations with more than two unordered outcome categories. Associations of WAI scores with outcome categories unemployment and early retirement were non-linear. Therefore, we included baseline WAI scores as categorical independent variable in a multinomial logistic regression model, using previously described WAI categories [14]: poor work ability (score 7–27), moderate work ability (score 28–36), good work ability (score 37–43), and excellent work ability (score 44–49). Multinomial logistic regression estimates odds ratios (ORs) and 95 % CIs relative to outcome category ‘employed’ and adjusted for age, leisure time physical exercise, physical and mental job demands, and job autonomy. Discrimination refers to the ability to distinguish between different outcomes. The area (AUC) under the receiver operating characteristic (ROC) curve is most commonly used to investigate discrimination [33]. If outcome categories are unordered, as in multinomial logistic regression, discrimination cannot be assessed by a single AUC. Biesheuvel et al. [34] compared AUCs of multinomial ROC analysis with AUCs of consecutive binary ROC analyses and found that they were very similar. Discrimination is usually over-optimistic when the data used for ROC analysis are the same as those used to estimate regression coefficients. Bootstrapping is the preferred method to correct for this over-optimism [33]. However, most statistical software packages lack the option to bootstrap multinomial logistic regression models. Therefore, we chose to perform consecutive ROC-analyses based on binary logistic regression for each type of premature work exit and subsequently adjust the resulting AUCs for over-optimism. For ROC analyses, WAI categories were scored as follows: 1 = excellent, 2 = good, 3 = moderate and 4 = poor work ability. An AUC C0.90 indicated excellent,
123
J Occup Rehabil Table 1 Baseline population characteristics
Age (in years)
Included in complete cases analyses (n = 9,530)
Excluded because of missing data (n = 2,007)
Mean
Mean
45.5
SD
N
%
9.5
45.7
SDa
N
Analysis
% P = 0.23b
9.4
Age categories (years) \30 30–39
786
8
169
8
1,270
13
256
13
40–49
3,563
37
728
36
C50
3,911
41
854
43
Years employed in construction Work hours per week
24.3 39.7
12.1 8.7
24.9 38.5
P = 0.26b P \ 0.01b
12.3 8.8
P \ 0.01c
Leisure time physical exercise \19 per month
2,899
37
701
60
19 per month
451
6
46
4
2–39 per month
847
11
79
7
1–29 per week
2,368
30
235
20
C39 per week
1,252
16
114
10
No
3,555
38
616
32
Yes
5,869
62
1,320
68
No
6,042
64
1,271
66
Yes
3,387
36
657
34
7,558
80
1,509
78
1,840
20
419
22
Work Physically demanding P \ 0.01c
Mentally demanding
a
Standard deviation
b
t test for independent samples
c
Chi square test
Autonomy No Yes
0.80–0.89 good, 0.70–079 fair, and 0.60–0.69 poor discrimination; discrimination failed if AUC \0.60 [33]. Over-optimism of the WAI’s discriminative ability was estimated by comparing discrimination in 250 bootstrap samples with discrimination in the original sample, using Harrell’s regression modeling strategies (RMS) package in R [35]. Throughout the study, we present over-optimism adjusted AUCs, which generalize to other working populations within the construction sector.
Results Of the 11,537 construction workers, 2,007 (17 %) had missing WAI data, particularly on the items sickness absence within the last year (1,106 missings) and estimation of work ability in 2 years’ time (357 missings). Construction workers with complete WAI data worked more hours per week and reported more frequent leisure time physical exercise than did construction workers with missing WAI data (Table 1). Construction workers with complete WAI data less often mentioned physically demanding work and less often experienced autonomy in
123
P = 0.12c
P = 0.03c
work than workers with missing WAI data. The mean WAI score was 40.1 (SD 4.9) and decreased with age from 42.5 (SD = 3.8) in construction workers \30 years to 41.5 (SD = 4.1) in those aged 30–39 years, 40.3 (SD = 4.6) in the age group 40–49 years, and 38.9 (SD = 5.2) in workers C50 years of age (ANOVA P \ 0.01). At follow-up, 336 (4 %) construction workers reported disability pension, 125 (1 %) unemployment, and 255 (3 %) early retirement. The odds of disability pension increased with decreasing WAI categories to OR 16.08 for construction workers reporting poor work ability compared to those reporting excellent work ability (Table 2). The odds of unemployment also increased with decreasing work ability categories, but differences relative to excellent work ability were not significant. The WAI categories were not significantly associated with the odds of early retirement among male construction workers. Bootstrapping techniques showed 1.2, 14.2, and 4.4 % over-optimism for the WAI’s ability to discriminate between construction workers at high and low risk of disability pension, unemployment, and early retirement, respectively. After adjustment for over-optimism, the WAI fairly discriminated between construction workers with and
J Occup Rehabil Table 2 Multinomial analysis of work ability and premature work exit relative to being employed at follow-up Work Ability Index
N
%
Disability pension OR
a
Unemployment b
95 % CI
OR
a
Early retirement b
ORa
95 % CI
95 % CIb
Excellent (44–49)
2,391
25
1.00
–
1.00
–
1.00
–
Good (37–43)
5,255
55
1.68
1.04–2.07
0.53
0.32–0.88
1.22
0.77–1.93
Moderate (28–36)
1,689
18
8.29
5.20–13.20
1.06
0.58–1.92
1.50
0.90–2.49
195
2
16.08
8.98–28.79
2.03
0.75–5.51
0.75
0.26–2.20
Poor (7–27) a
Odds ratio adjusted for age, physical exercise, physical and mental job demands, and job autonomy
b
Confidence interval
a
Percentage per age group
b
Area under receiver operating characteristic (ROC) curve
c
Confidence interval
Age
Disability pension a
N (% )
Unemployment
b
c
AUC (95 % CI )
a
N (% )
Early retirement b
c
AUC (95 % CI )
N (%a)
AUCb (95 % CIc)
\30 years
1 (0)
Not analyzed
7 (1)
0.58 (0.35–0.81)
0
Not analyzed
30–39 years
5 (0)
0.78 (0.30–1.00)
12 (1)
0.59 (0.42–0.76)
0
Not analyzed
40–49 years
89 (2)
0.74 (0.67–0.80)
59 (2)
0.52 (0.44–0.60)
14 (0)
0.60 (0.51–1.00)
C50 years
241 (6)
0.69 (0.66–0.74)
47 (1)
0.52 (0.43–0.62)
241 (6)
0.55 (0.51–0.59)
Total
336 (4)
0.74 (0.70–0.77)
125 (1)
0.51 (0.47–0.55)
255 (3)
0.58 (0.53–0.61)
without disability pension (AUC = 0.74; 95 % CI 0.70–0.77). The discriminative ability was highest in male construction workers aged 30–39 years and decreased with age (Table 3). Discrimination failed for both unemployment (AUC = 0.51; 95 % CI 0.47–0.55) and early retirement (AUC = 0.58; 95 % CI 0.53–0.61) as is shown in Fig. 1. The optimal WAI cut off for identifying construction worker at risk of disability pension was between good and moderate work ability, with sensitivity 0.63 and specificity 0.83. At a cut off between poor and moderate work ability, sensitivity was 0.12 indicating that 88 % of workers with disability pension at follow-up would be missed when only those reporting poor work ability at baseline would be considered at increased risk of disability pension.
Discussion The WAI identified male construction workers at risk of disability pension, although the discriminative ability decreased with age. The WAI did not discriminate between workers at high and low risk of unemployment or early retirement. Work Ability and Disability Pension The significant association between WAI scores and disability pension confirms the findings of previous studies [21, 22, 36]. Liira et al. [24] showed that the WAI was predictive of disability pension among 961 Finnish construction workers aged C40 years (55 % 40–49 years and
1.00 0.90 0.80 0.70
sensitivity
Table 3 WAI as risk marker of premature work exit stratified by age group
0.60 0.50 0.40 0.30 0.20 0.10 0.00 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
1–specificity
Fig. 1 Discriminative ability of Work Ability Index (WAI). The figure shows the receiver operating characteristic curve representing the ability of WAI categories (1 = excellent, 2 = good, 3 = moderate, and 4 = poor work ability) to discriminate construction workers at high and low risk of disability pension (black line), unemployment (dark grey line), and early retirement (light grey line); the diagonal indicates no discrimination above chance
45 % C50 years). The authors found that workers with poor work ability had a 10.7 times higher risk of disability pension as compared to those with excellent work ability; 36 % of the workers with poor WAI scores and 16 % of those with moderate WAI scores received disability pension at 4-year follow-up. Alavinia et al. [21] found that poor to moderate WAI scores were highly predictive of disability pension after on average 23 months of follow-up
123
J Occup Rehabil
of 850 Dutch construction workers aged C40 years (mean age 48.4 years). Our study of 9,530 Dutch construction workers (aged 16–62 years; mean age 45.5 years, with 37 % aged 40–49 years and 41 % C50 years) added that the WAI correctly discriminated between workers at high and low disability pension risk in 74 % of the cases. Furthermore, we found that the discriminative ability decreased with age. This was probably due to the fact that WAI scores also decrease with age. It is conceivable that low WAI scores are less discriminative in age groups where lower work ability is more common. Work Ability, Unemployment and Early Retirement The risk of unemployment increased with decreasing WAI categories, but associations were not statistically significant. Previously, Liira et al. [24] showed that baseline WAI scores did not predict unemployment at 4-year follow-up. We found that the WAI poorly identified male construction workers at increased risk of unemployment. It should be reminded here that 37 % of the variance in unemployment in our study was explained by clustering at company level. Possibly, employees were discharged because of organizational downsizing or bankruptcy rather than transferred to unemployment because of their incapacity to do construction work. This could explain why the WAI did not identify workers at risk of unemployment. Furthermore, the shared variance at the company level implicates that the discriminative ability of the WAI may differ across companies. It is conceivable that the WAI can identify workers at risk of unemployment in some companies, for example economically stable and profitable companies. The WAI was not associated with early retirement, which contrasts the results of Salonen et al. [23], who reported that low WAI scores, poor physical work ability, and heavy physical workload were associated with early work exit of employees in the Finnish food industry. Our results showed that the WAI failed to discriminate between construction workers at high and low risk of early retirement. Possibly, workers retired because they knew that prepension benefits would soon become capital-funded and less generous. Dutch retirement policies were reformed in 1997, but pre-pension benefits remained attractive for workers in arduous jobs (e.g., construction work) for up to 10 years afterwards. Unfortunately, information on why construction workers had retired and what type of prepension benefits (state- or capital-funded) they were receiving was not available. Study Strengths and Limitations The prospective design and the large sample size are strengths of the study. The study population consisted of
123
construction workers employed at different companies all over The Netherlands. Thus, it is unlikely that results were biased by organizational or regional policies and practices. Older construction workers were over-represented in the study population, which is probably due to the fact that trade unions particularly stimulate older construction workers to participate in health checks. Associations were adjusted for age, but the overall discriminative ability of the WAI may have been attenuated by an over-representation of older workers. We dealt with this problem by presenting the discriminative ability of the WAI stratified by age groups. Previous research has shown a ‘healthy volunteer effect’ implicating that healthy individuals are more likely to participate in health surveys than are workers experiencing poor health [37]. The fact that construction workers with complete WAI data reported more frequent leisure time physical exercise than those with missing data might be indicative of such a healthy volunteer effect, which could have weakened associations between WAI scores and premature work exit. A further limitation of the study is the use of selfreported measures. Outcomes such as disability pension and unemployment are negatively valued in the Dutch economy and may have been under-reported. This type of ‘social desirability bias’ [38] could have attenuated associations of WAI scores with disability pension and unemployment. In a large-scale Swedish study, the overall agreement between self-reported and registered disability pension was 96 %; specificity was 99 %, while the sensitivity was 70 % indicating that 30 % of employees with registered disability pension did not report disability pension [39]. Early retirement is less stigmatized and the association between the WAI and early retirement is less likely to be affected by social desirability bias. Practical Implications The present results show that the WAI fairly discriminates between construction workers at high and low disability pension risk. This implicates that the WAI could be used to screen construction workers for risk of disability pension. At a cut off between moderate and good work ability, the specificity was 0.83 and sensitivity 0.63, meaning that 37 % of construction workers who reported disability pension at follow-up had good or excellent WAI scores at baseline. At a cutoff between poor and moderate work ability, specificity increased to 0.98 restricting false-positive rates to 2 %, but sensitivity was low and most construction workers at risk of disability pension would be missed when only those reporting poor work ability were identified as being at risk.
J Occup Rehabil
The discriminative ability of the WAI decreased with age. It should be acknowledged that the incidence of disability pension was low in the age groups \30 and 30–39 years. Furthermore, the present results apply only to male construction workers. Thus, studies in larger as well as other working populations are required to further investigate the WAI’s ability for identifying workers at risk of disability pension. If larger-scale studies confirm that the WAI’s discriminative ability declines with age to poor discrimination in workers aged C50 years, then employers and occupational health providers should be recommended to measure work ability with the WAI in workers aged \50 years. Subsequently, workers with poor to moderate WAI scores could be motivated to participate in workplace health promotion (WHP) programs to improve or restore work ability [40]. It has been shown that WHP programs increase work ability and decrease sickness absence and disability leaves [41, 42]. A recent meta-analysis of WHP effectiveness, however, showed that the overall effect of WHP programs is small [43]. This might indicate that WHP programs are not adequately targeted at the employees who need them. Considering the aforementioned healthy volunteer effect, employees with health complaints or unhealthy lifestyle behaviors may be less inclined to participate in WHP programs. WHP programs could be more effective when better targeted to high-risk groups. The WAI could be used to identify high-risk workers and specifically invite them for WHP programs. Another practical implication may be to use the WAI to identify workers who could benefit from occupational rehabilitation services. Bethge et al. [44] showed that reduced work ability does not always lead to the utilization of rehabilitation services. The authors discussed insufficient information by social insurance agencies and lack of appropriate counseling by general practitioners as major barriers for utilizing rehabilitation services. The authors also found indications that employees refrain from requesting rehabilitation when working conditions are precarious and insecure. The use of the WAI to measure work ability at the workplace could overcome these barriers.
Conclusion We conclude that the WAI identifies male construction workers at risk of future disability pension, but not those at risk of unemployment or early retirement. Workers with less than good WAI scores could be invited for tailor-made multidimensional interventions to restore or improve work ability.
References 1. Organisation for Economic Co-operation and Development. Maintaining prosperity in an ageing society. http://www.oecd. org/els/public-pensions/2429430.pdf. Accessed 05 Feb 2014. 2. Eurostat. http://epp.eurostat.ec.europa.eu/statistics_explained/index. php/Labour_markets_at_regional_level. Accessed 02 Mar 2014. 3. Bureau of Labor Statistics 2008. Older workers: are there more older people in the workplace? http://www.bls.gov/spotlight/ 2008/older_workers/. Accessed 20 Dec 2013. 4. Organisation for Economic Co-operation and Development. Older workers: living longer, working longer. http://www.oecd. org/social/family/35961390.pdf. Accessed 05 Feb 2014. 5. Kooij D, De Lange A, Jansen P, Dikkers J. Older workers’ motivation to continue to work: five meanings of age. A conceptual review. J Manag Psychol. 2008;23:364–94. 6. Ilmarinen J. Aging workers. Occup Environ Med. 2001;58: 546–52. 7. Charness N. Aging and human performance. Hum Factors. 2008;50:548–55. 8. Kenny GP, Yardley JE, Martineau L, Jay O. Physical capacity in older adults: implications for the aging worker. Am J Ind Med. 2008;51:610–25. 9. Schuring M, Burdorf L, Kunst A, Mackenbach J. The effects if ill health on entering and maintaining paid employment: evidence in European countries. J Epidemiol Community Health. 2007;61:597–604. 10. Sluiter JK. High-demand jobs: age-related diversity in work ability? Appl Ergon. 2006;37:429–40. 11. Ilmarinen J, Tuomi K. Work ability of ageing workers. Scand J Work Environ Health. 1992;18(Suppl 2):8–10. 12. Ilmarinen J. Toward a longer worklife: ageing and the quality of worklife in the European Union. Helsinki: Finnish Institute of Occupational Health; 2005. 13. Tengland PA. The concept of work ability. J Occup Rehabil. 2011;21:275–85. 14. Tuomi K, Ilmarinen J, Jahkola A, Katajarinne L, Tulkki A. Work ability index, second revised version. Helsinki: Finnish Institute of Occupational Health; 1998. 15. Van den Berg TI, Elders LA, de Zwart BC, Burdorf A. The effects of work-related and individual factors on the Work Ability Index: a systematic review. Occup Environ Med. 2009;66:211–20. 16. Tuomi K, Ilmarinen J, Martikainen R, Aalto L, Klockars M. Aging, work, life style and work ability among Finnish municipal workers. Scand J Work Environ Health. 1997;23(Suppl 1):58–65. 17. Ilmarinen J, Tuomi K, Klockars M. Changes in the work ability of active employees over an 11-year period. Scand J Work Environ Health. 1997;23(Suppl 1):49–57. 18. Pohjonen T. Perceived work ability of home care workers in relation to individual and work-related factors in different age groups. Occup Med. 2001;51:209–17. 19. Monteiro MS, Ilmarinen J, Correˆa Filho HR. Work ability of workers in different age groups in a public health institution in Brazil. Int J Occup Saf Ergon. 2006;12:417–27. 20. Safari S, Akbari J, Kazemi M, Mououdi MA, Mahaki B. Personnel’s health surveillance at work: effect of age, body mass index, and shift work on mental work load and work ability index. J Environ Public Health. 2013;28:94–8. 21. Alavinia SM, de Boer AGEM, van Duivenbooden JC, FringsDresen MH, Burdorf A. Determinants of work ability and its predictive value for disability. Occup Med. 2009;59:32–7. 22. Sell L, Bu¨ltmann U, Rugulies R, Villadsen E, Faber A, Søgaard K. Predicting long-term sickness absence and early retirement pension from self-reported work ability. Int Arch Occup Environ Health. 2009;82:1133–8.
123
J Occup Rehabil 23. Salonen P, Arola H, Nyga˚rd CH, Huhtala H, Koivisto AM. Factors associated with premature departure from working life among ageing food industry employees. Occup Med. 2003;53:65–8. 24. Liira J, Matikainen E, Leino-Arjas P, Malmivaara A, Mutanen P, Rytko¨nen H, Juntunen J. Work ability of middle-aged Finnish construction workers—a follow-up study in 1991–1995. Int J Ind Ergon. 2000;25:477–81. 25. Organisation for Economic Co-operation and Development. Sickness and disability schemes in the Netherlands. http://www. oecd.org/social/soc/41429917.pdf. Accessed 20 Dec 2013. 26. Van Oorschot W. Narrowing pathways to early retirement in the Netherlands. Policy Press. 2007;15:247–55. 27. Euwals R, van Vuuren D, Wolthoff R. Early retirement behaviour in the Netherlands: evidence from a policy reform. Economist. 2010;158:209–36. 28. Ilmarinen J. The work ability index (WAI). Occup Med. 2007;57:160. 29. Nu¨bling M, Hasselhorn HM, Seitsamo J, et al. Comparing the use of the short and the long disease list in the Work Ability Index Questionnaire. In: Proceedings of the second international symposium on work ability. Verona: ICOH; 2004. p. 74. 30. Ilmarinen J, Tuomi K. Past, present and future of work ability. In: Ilmarinen J, Lehtinen S, editors. People and work—research reports 65. Helsinki: Finnish Institute of Occupational Health; 2004. 31. De Zwart BCH, Frings-Dresen MCH, van Duivenbooden JC. Test-retest reliability of the Work Ability Index questionnaire. Occup Med. 2002;52:177–81. 32. Radkiewich P, Widerszal-Bazyl M. Psychometric properties of work ability index in the light of comparative survey study. In: International congress series 1280. Amsterdam: Elsevier; 2005. p. 304–9. 33. Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models. Epidemiology. 2010;21:128–38. 34. Biesheuvel CJ, Vergouwe Y, Steyerberg EW, Grobbee DE, Moons KGM. Polytomous logistic regression analysis could be
123
35. 36.
37. 38. 39.
40. 41.
42.
43.
44.
applied more often in diagnostic research. J Clin Epidemiol. 2008;61:125–34. Harrell FE. Regression modeling strategies. http://cran.r-project. org/web/packages/rms. Accessed 20 Dec 2013. Tuomi K, Toikkanen J, Eskelinen L, et al. Mortality, disability and changes in occupation among aging municipal employees. Scand J Work Environ Health. 1991;17(Suppl 1):58–66. Etter JF, Perneger TV. Analysis of non-response bias in a mailed health survey. J Clin Epidemiol. 1997;50:1123–8. Fisher RJ, Katz JE. Social-desirability bias and the validity of self-reported values. Psychol Mark. 2000;17:105–20. Svedberg P, Ropponen A, Lichtenstein P, Alexanderson K. Are self-report of disability pension and long-term sickness absence accurate? Comparisons of self-reported interview data with national register data in a Swedish twin cohort. BMC Public Health. 2010;10:763. Gould R, Ilmarinen J, Ja¨rvisalo J, et al. Dimensions of work ability. Helsinki: Finnish Centre for Pensions; 2008. Proper KI, Staal BJ, Hildebrandt VH, van der Beek AJ, van Mechelen W. Effectiveness of physical activity programs at worksites with respect to work-related outcomes. Scand J Work Environ Health. 2002;28:75–84. Kuoppala J, Lamminpaa A, Husman P. Work health promotion, job well-being, and sickness absences—a systematic review and meta-analysis. J Occup Environ Med. 2008;50:1216–27. Rongen A, Robroek SJW, van Lenthe FJ, Burdorf A. Workplace health promotion: a meta-analysis of effectiveness. Am J Prev Med. 2013;44:406–15. Bethge M, Radoschewski FM, Gutenbrunner C. The work ability index as a screening tool to identify the need for rehabilitation: longitudinal findings from the second German sociomedical panel of employees. J Rehabil Med. 2012;44:980–7.
