Article Text

Extended report
Predictors of work disability during the first 3 years after diagnosis in a national rheumatoid arthritis inception cohort
  1. Tor Olofsson1,2,
  2. Ingemar F Petersson1,2,
  3. Jonas K Eriksson3,
  4. Martin Englund2,4,
  5. Julia F Simard3,
  6. Jan-Åke Nilsson1,
  7. Pierre Geborek1,
  8. Lennart T H Jacobsson5,6,
  9. Johan Askling3,
  10. Martin Neovius3
  1. 1Section of Rheumatology, Department of Clinical Sciences, Lund University, Lund, Sweden
  2. 2Department of Orthopedics, Musculoskeletal Sciences, Clinical Sciences Lund, Lund University, Lund, Sweden
  3. 3Clinical Epidemiology Unit, Department of Medicine, Karolinska Institutet, Stockholm, Sweden
  4. 4Clinical Epidemiology Research and Training Unit, Boston University School of Medicine, Boston, Massachusetts, USA
  5. 5Department of Rheumatology and Inflammation Research, Institute of Medicine, The Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden
  6. 6Section of Rheumatology, Department of Clinical Sciences, Lund University, Malmö, Sweden
  1. Correspondence to Dr Tor Olofsson, Section of Rheumatology, Department of Clinical Sciences, Lund University, Lund SE-22185, Sweden; torolofsson{at}hotmail.com

Abstract

Objective To identify predictors of sick leave and disability pension in patients with early rheumatoid arthritis (RA).

Methods Individuals aged 19–59 years diagnosed with early RA (≤12 months symptom duration) were identified in the Swedish Rheumatology Quality Register (1999–2007; n=3029). We retrieved days of sick leave and disability pension from the Swedish Social Insurance Agency and baseline predictors of total work days lost during 3 years after RA diagnosis were investigated using linear regression. Due to effect modification by baseline work ability (defined as work days lost the month before diagnosis), analyses were stratified into three categories: full=0 work days lost the month before diagnosis; partial=1–29 work days lost; and none=30 work days lost.

Results 71% of patients with full baseline work ability still had full work ability after 3 years compared with 36% (p<0.001) and 18% (p<0.001) of those with partial and no work ability at baseline, respectively. Elevated baseline levels of HAQ and DAS28, higher age, lower education level and unemployment were associated with more work days lost during 3 years in all strata of baseline work ability (all p<0.05). In a separate analysis, more objective variables (ESR, CRP and swollen joints) were not. Generally, the largest regression coefficients were seen for patients with partial baseline work ability.

Conclusions Work ability at RA diagnosis was the most important predictor of 3-year sick leave and disability pension. Taking this into account, HAQ, DAS28, age and education level were also significant predictors, whereas ESR and CRP were not.

  • Early Rheumatoid Arthritis
  • Epidemiology
  • Disease Activity
  • Outcomes research

Statistics from Altmetric.com

Introduction

Rheumatoid arthritis (RA) is the most common of the rheumatic inflammatory diseases with an estimated prevalence of 0.6–0.8%.12 Without effective treatment, it results in a progressive deterioration of physical function leading to work disability.3

Evidence from the prebiologic era indicates a continual increase in work disability during the disease course with the largest rise the first year after diagnosis.4–6 Since then, the management of RA has undergone marked changes focusing on early, aggressive treatment, including TNF inhibitors.7 ,8 Recent Swedish findings from the biologic era have demonstrated a peak of work disability the year following diagnosis with a decline during the years thereafter.9

Being able to predict which patients are most prone to develop or maintain work disability after diagnosis would have important implications. Results from earlier studies have generally shown that high health assessment questionnaire (HAQ) scores,10–22 old age,12 ,14 ,15 ,17–19 21–25 low education level12 ,13 ,15 ,17 ,18 ,20 ,21 ,24 and high physical job demands14 ,15 ,17–19 ,21 ,22 ,24 ,26 ,27 are strong predictors of work disability. So far, the vast majority of prediction studies are from the prebiologic era and have commonly been carried out on rather small patient cohorts and most often been based upon self-reported measures of work disability associated with non-response and recall bias.

The objective of this study was to investigate predictors of objectively assessed sick leave and disability pension during the first 3 years after diagnosis in a large national inception cohort of Swedish early RA patients diagnosed 1999–2007.

Methods

Setting

In 2009 the Swedish population was 9.3 million, 60% being 19–64 years of age. The Swedish healthcare system offers universal access, and RA patients are generally diagnosed and treated by rheumatologists. The retirement age is 65 years.

The Swedish social insurance system

All residents in Sweden aged 16–64 years can be granted economic security from the Social Insurance Agency in case of sickness, disability or injury. The compensation can take the form of sick leave or disability pension and it is possible to have both compensation forms simultaneously, though never exceeding 100% together.

RA patients and general population comparators

Those eligible for this study were patients in the Swedish Rheumatology Quality Register (SRQ) diagnosed with RA during 1999–2007, being 19–59 years and having symptom duration of ≤1 year at diagnosis. The SRQ is a national register comprising both the register for follow-up of incident RA (according to 1987 American College of Rheumatology (ACR) criteria28), started in the mid-1990s, and the Swedish biologics register (Anti-Rheumatic Treatment In Sweden; ARTIS) set up in 1999.29

Through each patient's unique personal identification number data were linked to the Social Insurance Agency register including data on sick leave and disability pension (1996–2010) and to the Longitudinal Integration Database for Health Insurance and Labour Market Studies including data on education level and unemployment benefits (1996–2007). Data on place of residence, civil status and vital status were retrieved from the Population and Housing Census database at Statistics Sweden. From the same register, five general population comparators were sampled for each RA patient, using age (±1 year), sex, county and education level (≤9, 10–12 and >12 years of schooling) as matching factors.

Outcome

Net days of sick leave and disability pension were used as primary outcome because both compensation forms may be complete (100%) or partial (eg, 25%, 50% or 75%). We calculated net days by multiplying the degree of compensation with the number of days. For example, 1 net day may equal 1 day with 100% sick leave, 2 days with 50% sick leave or 1 day with 50% sick leave plus 1 day with 50% disability pension. Since patients often move from sick leave to disability pension, the two compensation forms were analysed jointly in this study. The maximum number of days was set as 30 per month. The mean number of net days was chosen as main metric since the mean multiplied by the number of patients represents the overall burden on society. Because of the non-normal distribution of days, median values and proportions were also calculated.

Sick leave

During the study period (1996–2010), the first day of sickness (‘the qualifying day’) was unpaid. From day 2–14, the compensation was obtained directly from the employer (‘sick pay period’; day 2–28 between 1 January 1997 and 31 March 1998). Only disease periods >14 days were registered by the Social Insurance Agency. However, when an episode exceeds 14 days all days (from day 1 and on) were registered. Furthermore, an episode occurring within 4 days of a previous one did not require a new sick pay period before registration.

Disability pension

Disability pension requires at least 25% reduction in work ability expected to remain for at least 1 year and may be time-limited or permanent.

Statistics

Statistical analyses were conducted using SPSS (V.20.0). Differences in baseline characteristics between categories of work ability at diagnosis were assessed by one-way analysis of variance (ANOVA) for continuous variables and χ2 tests for ordinal ones.

Linear regression models were used to assess baseline predictors, with cumulative days of sick leave and disability pension during the 3-year period following RA diagnosis as outcome. Due to effect modification by baseline level of work ability, the analyses were stratified into three categories: ‘full work ability’=0 days with sick leave and disability pension the month before diagnosis; ‘partial work ability’=1–29 days; and ‘no work ability’=30 days (maximum possible). Because of the skewed distribution in the outcome variable, complementary categorical analyses were also undertaken (see online supplementary appendix 1).

Independent variables included in the prediction models were chosen based on subject matter knowledge. Continuous variables were checked for multi-collinearity using Spearman correlation accepting correlation coefficients ≤0.3 and ≥−0.3 otherwise omitting the variable least correlated to the outcome measure. HAQ and disease activity score using 28 joint counts (DAS28) were both considered essential but due to multi-collinearity analysed in separate regression models. Results for the other variables in the main analysis were retrieved from the models including HAQ. Patients who died or emigrated during the 3-year period after RA diagnosis or had missing data for any of the independent variables were excluded from the regression analyses. For detailed information on the regression models see online supplementary appendix 2.

Data on sick leave and disability pension for the last 3 months of 2010 were not available at the time of linkage, resulting in 61 patients and 303 general population comparators having between 33 and 35 months of follow-up instead of 36. This was managed through annualisation, deriving values for October to December 2010 by applying the monthly mean for the rest of the last calendar year for these individuals.

Two-tailed p values <0.05 were considered statistically significant.

Results

We identified 3029 early RA patients. Overall, the patient cohort had a high disease activity at baseline with mean DAS28 5.1 and mean HAQ score 1.01. The mean/median symptom duration at diagnosis was 6/6 months. After the initial consultation visit recorded in the SRQ, when the vast majority of patients were also diagnosed, 88% had at least one non-biologic disease-modifying antirheumatic drug (DMARD) (table 1).

Table 1

Characteristics of the early RA cohort at diagnosis shown for different categories of baseline work ability

Differences between baseline work ability groups

Patients were followed with respect to work disability from 3 years before until 3 years after diagnosis. The mean number of days with sick leave and disability pension the month before diagnosis was 13 (SD 13; median 7). The days were unevenly distributed: 1349 patients (44%) had 0 days (=full baseline work ability), 779 patients (26%) had 1–29 days (=partial baseline work ability) and 901 patients (30%) had 30 days (=no baseline work ability). Among general population comparators the corresponding percentages for full, partial and no baseline work ability was 81%, 7% and 12%, respectively.

Compared with patients with partial or no baseline work ability, patients with full baseline work ability had significantly lower HAQ scores, lower values of global visual analogue scale (VAS) and pain VAS, less tender and swollen joints, and lower levels of erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP), but did not have shorter symptom duration at diagnosis than the other groups (table 1).

The three groups also showed large differences in the probability of having full work ability after 3 years. In all, 71% of the patients with full work ability at diagnosis still had full work ability after 3 years compared with 36% and 18% among patients with partial or no work ability at diagnosis, respectively (p<0.001). Conversely, only 3% of patients with full work ability at diagnosis were totally work-disabled after 3 years compared with 10% and 46% among patients with partial and no work ability at diagnosis, respectively (p<0.001; figure 1).

Figure 1

Annual distribution of days with sick leave and disability pension from 2 years before until 3 years after day of rheumatoid arthritis diagnosis for patients diagnosed during 1999–2007, stratified on level of work ability at diagnosis: ‘full work ability’=0 days with sick leave and disability pension the month before diagnosis; ‘partial work ability’=1–29 days; and ‘no work ability’=30 days. The maximum number of days was set as 30 for a month and 360 for a year.

Descriptive analysis of work days lost across the predictors

The development of mean monthly days of sick leave and disability pension from 3 years before until 3 years after RA diagnosis showed distinct variations across segments of the investigated predictor variables (figure 2). The trajectory for the general population comparators was similar to that of the RA patients until 6 months before diagnosis, when the curves started to diverge. After diagnosis a certain convergence was seen.

Figure 2

Mean days of sick leave and disability pension per month from 3 years before until 3 years after day of rheumatoid arthritis (RA) diagnosis for patients diagnosed during 1999–2007 and matched comparators from the general population (GenPop, 5 per RA patient), stratified on selected predictors. In the bottom right panel, ‘full’=0 days of sick leave and disability pension the month before diagnosis; ‘partial’=1–29 days; and ‘none’=30 days (maximum possible).

There were large and statistically significant differences between segments of the investigated predictor variables when assessing 3-year sick leave and disability pension after diagnosis (figure 3). The highest variation was seen for baseline work ability, ranging from mean 98 days (95% CI 87–109; median 0) for full work ability to mean 763 days (95% CI 701–825; median 935) for no work ability.

Figure 3

Unadjusted cumulative days of sick leave and disability pension during the first 3 years after day of diagnosis for rheumatoid arthritis (RA) patients diagnosed during 1999–2007 and matched comparators from the general population (GenPop, 5 per RA patient), given as mean with 95% CIs and as median. Regarding work ability at diagnosis, ‘full’=0 days with sick leave and disability pension the month before diagnosis; ‘partial’=1–29 days; and ‘none’=30 days (maximum possible). Overall, 44 RA patients lacked GenPop comparators and were therefore omitted in this analysis. Among the remaining patients, 20 died and seven emigrated during the 3-year period after diagnosis and were also omitted.

Multivariable predictor analysis

Disease related variables

In multivariable analysis, higher values of HAQ and DAS28 were both associated with more cumulative days of sick leave and disability pension during the 3-year period after RA diagnosis in all three strata of baseline work ability. Longer symptom duration at baseline (range 0–12 months) was a significant predictor of more work days lost only for patients having no baseline work ability (table 2).

Table 2

Baseline predictors of total days of sick leave and disability pension during the first 3 years after RA diagnosis, stratified on level of work ability at diagnosis

Concerning the components of DAS28, it was shown that pain VAS, global VAS and tender joint count were significant predictors in all three baseline work ability strata, whereas ESR, CRP and swollen joint count were not significant in any of the strata (table 3).

Table 3

Baseline DAS28 components as predictors of total days of sick leave and disability pension during the first 3 years after RA diagnosis, stratified on level of work ability at diagnosis

Generally, the largest regression coefficients were seen for the group with partial work ability at baseline and the smallest for the group with full work ability at baseline. A 1-unit higher HAQ score (eg, the difference between 1.0 and 2.0) was associated with an average increase of 45 days (p<0.001) off work during the 3-year period after RA diagnosis if having full baseline work ability, 137 days (p<0.001) if having partial baseline work ability and 106 days (p<0.001) if having no baseline work ability.

Demographic variables

Higher age, lower education level and unemployment were significant predictors of more 3-year sick leave and disability pension days in all three strata of baseline work ability, whereas sex reached statistical significance only in patients with partial or no baseline work ability (table 2).

Discussion

In this study of early RA patients in a real-life setting, the most important predictor of 3-year sick leave and disability pension after RA diagnosis was work ability the month before diagnosis. This variable modified the effect of the other predictors investigated. The results showed that more than 70% of patients with full work ability at RA diagnosis remained fully capable of work after 3 years, whereas less than 5% with full work ability at diagnosis were totally work-disabled after 3 years. By contrast, less than 20% of patients with no work ability at diagnosis were fully capable of work after 3 years, while almost 50% of patients with no work ability at diagnosis were totally work-disabled after 3 years. Patients with no work ability the month before diagnosis who returned to full work ability after 3 years had much fewer average compensation days during the 12 months preceding diagnosis than those not returning to full work ability after 3 years (see online supplementary appendix 3).

Previous research

These results are similar to those of a regional Swedish study from the prebiologic era of 120 patients with recent onset RA (≤1 year symptom duration) identified during 1996–1998, showing sick leave and disability pension the year before diagnosis as the strongest predictor of sick leave and disability pension after 3 years.4 This variable was also shown to be a strong predictor in a recent Finnish study of 7831 patients, though without inclusion of any clinical or socioeconomic data in the analysis.25 Beyond this, reliable data on history of work disability in RA have been scarce and generally not taken into account in predictor studies, although this approach has been used in predictor studies in related conditions30 ,31 and, according to our results, appears to be important.

When using multivariable analysis and stratifying on baseline work ability, the prediction effect of the other variables was generally largest for the patient group with partial work ability at baseline and smallest for the group with no baseline work ability. The results were confirmed in categorical analyses (see online supplementary appendix 1).

Modifiable disease variables such as HAQ, DAS28, global VAS, pain VAS and tender joint count were significant predictors of total days in all three strata of baseline work ability. This was also the case for age, education level and unemployment status. Concerning HAQ, age and education level, our results are congruent with findings from the prebiologic era,13 ,15 ,16 ,19–22 confirming the register-based findings of Puolakka et al12 in FIN-RACo, a Finnish therapy trial of 162 patients with early RA (<2 years) recruited between 1993 and 1995, but in our case on a larger scale and in a contemporary setting during the biologic era. Our results for pain VAS and tender joint count correspond with those of two medium-sized earlier studies,14 ,18 although a few smaller studies have found no such association.19 ,22

Interestingly, the more objective disease variables (CRP, ESR and swollen joint count) proved not to be significant predictors, irrespective of baseline work ability. Some studies from the prebiologic era have, however, shown results differing from ours.14 ,18 In a British multi-centre study of early RA patients (≤2 years symptom duration) recruited 1987–1993, Young et al14 found ESR to be a significant predictor of work disability over 5-year follow-up, though with self-reported job loss as outcome and analyses performed solely on patients employed at baseline (245 patients included in the model). Similar results were shown by Wolfe and Hawley18 in a prospective 18-year study of 456 patients, adjusting only for sex, age and disease duration in the regression models and also differing from the present study by a substantially longer disease duration at inclusion (mean 5 years). Neither of these studies performed separate analyses by strata of baseline work ability.

Symptom duration

Mean monthly days of sick leave and disability pension started to rise sharply approximately 6 months before diagnosis, coinciding with the average symptom duration at diagnosis. By changing the baseline from RA diagnosis to symptom start, we could show that the level of mean monthly days was stable until symptom start and then increased markedly (see online supplementary appendix 4). Despite this, no association between symptom duration at diagnosis and baseline work ability group was shown. Patients with no baseline work ability did not have longer symptom duration than the other two groups. Furthermore, longer symptom duration was a significant predictor of more 3-year sick leave and disability pension only for patients having no baseline work ability in adjusted analysis.

Strengths and limitations

A strength of this study was the objectively assessed data on sick leave and disability pension from an independent source, eliminating non-response and recall bias.32 Sick leave and disability pension could also be analysed jointly, which has been proven important in earlier studies.4 ,9 ,33 A further strength was access to data on sick leave and disability pension from the period preceding diagnosis, allowing for inclusion of this important information in the analyses.

This study also has limitations. First, sick leave episodes ≤14 days were generally not contained in the Social Insurance Agency register, unless they appeared within 4 days of a previous sick leave episode. A Finnish study reported only 3% of sick leave episodes to be <10 days in RA patients, constituting 0.2% of total days,33 indicating that underestimation in their data was likely to be small.

Second, the adjusted R-squared values of the regression models indicate that the included demographic and clinical variables only to a limited degree explain work disability during the 3-year period after diagnosis and that there are important additional explanatory variables. Earlier studies have shown that physical job demand is one such variable.10 ,14 ,15 ,18 ,21 ,22 ,24 ,27 Additional important variables to which we did not have access were working hours,24 ,27 work environment,15 ,18 financial situation,27 coping ability15 ,21 and catastrophising (tendency to magnify negative events).34 No information was available on smoking and obesity status, two strong predictors of disability pension in the general population.35

Third, although SRQ is a nationwide register and ARTIS, its subregister for biologics-treated patients, covers an estimated 87% of all target patients36 there are currently no published estimates of the coverage and generalisability of SRQ's subregister for follow-up of incident RA.

Finally, the Swedish welfare system is generous in an international perspective, potentially limiting generalisability to other countries. However, a systematic review of productivity losses in RA found that rates of work disability were largely similar in the USA and in Northern European countries despite variations in social systems and study methodologies.6

Conclusions

The most important predictor of total days of sick leave and disability pension was baseline work ability. As much as 70% of individuals who develop RA against a background of full work ability are likely to remain fully capable of work after 3 years, whereas less than 5% of these patients become totally work-disabled. HAQ, global VAS, pain VAS and tender joint count as well as age, education level and unemployment status were also significant predictors when stratifying on baseline work ability. This was not the case for less subjective variables such as ESR, CRP and swollen joint count, suggesting a focus on the more subjective components of DAS28 when predicting future work ability. The results also indicate a role for baseline work ability as an important prognostic instrument and suggest that interventions might have the greatest impact on future sick leave and disability pension in patients with partial work ability at diagnosis.

Acknowledgments

The authors would like to express their gratitude to all rheumatologists contributing data to the Swedish Rheumatology Quality Registers as well as to members of the ARTIS study group who were not coauthors of this paper: E Baecklund, L Coster, H Forsblad, N Feltelius, L Klareskog, LE Kristensen, S Lindblad, S Rantapaa-Dahlqvist and R van Vollenhoven.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Handling editor Tore K Kvien

  • Funding This study was funded by unrestricted grants from Lund University, Region Skåne and The Swedish Social Insurance Agency.

  • Competing interests IP has received fees for speaking and consultancy from Abbott, Pfizer, UCB Pharma and Wyeth. ME has received lecturing fees from Pfizer. LTHJ has received speaking and advisory board fees from Abbott, BMS, MSD, Pfizer and UCB Pharma. JA has received speaking fees from BMS, MSD and Pfizer and research grants from AstraZeneca and Pfizer for research unrelated to the current work. MN has received fees for advisory board participation by Abbott and Pfizer. The other authors have not declared any potential competing interests.

  • Contributors TO, MN, IP, LTHJ and PG planned the study. JE and MN provided the data. TO and MN performed the analysis and drafted the paper. All the other authors contributed to editing the draft for content and approved of the final version. All authors had full access to all the data (including statistical reports and tables) in the study and take responsibility for the integrity of the data and accuracy of the data analysis.

  • Ethics approval This study was approved by the ethics committee of Karolinska Institutet, Stockholm, Sweden.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.