Factors associated with duration of disability benefi ts : a cohort study

OBJECTIVE: To analyze factors associated with the duration of disability benefi ts due to work-related upper-limb musculoskeletal disorders. METHODS: Ambispective cohort study conducted with 563 insured workers from the General Social Security System who received temporary disability benefi ts due to work-related upper-limb musculoskeletal disorders in the city of Salvador, Northeastern Brazil, in 2008. The data came from an inquiry performed by the Regional Audit of the National Social Security Institute and from administrative records. Sociodemographic and work-related variables were analyzed, as well as characteristics of the health problem and aspects related to social security. Factors associated with time until the cessation of the benefi t were identifi ed through survival analysis techniques. RESULTS: Low socioeconomic position (RR=1.29; 95%CI 1.02; 1.64), age below 39 years (RR=1.23;95%CI 1.03; 1.47), income replacement by the National Social Security Institute ≤ 100% (RR=1.24; 95%CI 1.04; 1.47) and high expectation of returning to work (RR=1.20; 95%CI 1.00; 1.44) are the categories related to higher rate of cessation of the benefi t and with its shorter duration. CONCLUSIONS: Factors that are not strictly medical, like socioeconomic position, age, expectation of returning to work and level of income replacement by the National Social Security Institute, seem to infl uence the benefi t’s duration. These hypotheses need to be tested with further confi rmatory studies in order to improve the understanding of the process of determining incapacity for work. DESCRIPTORS: Sick Leave. Musculoskeletal Diseases. Insurance, Disability, utilization. Insurance Benefi ts. Insurance, Health. Occupational Health. Cohort Studies.


INTRODUCTION
Incapacity for work caused by health-related injuries is a problem all over the world, with considerable social and economic cost.Approximately 3% to 10% of the economically active population in developed countries received benefi ts due to disability caused by health problems in 2003. 14The grant of these benefi ts has been vigorously increasing in Brazil, remarkably for temporary disability, which grew 138% (from 909,476 to 2,163,063) between 2000 and 2008. a Studies use measures of the duration of benefi ts receipt as indicators of the burden of disability in the population.The results show low correlation between physical limitation and duration of benefi ts receipt, mainly concerning chronic diseases.Factors beyond the specifi c to trigger or sustain health problems may explain differences in return to work or in the benefi t's duration. 5,10,17terature reviews have shown distinct predictors of the duration of incapacity for work, like individual characteristics of the worker, of the disease, of the health system, of social security, of work's demands, besides other factors such as discrimination, legal framework and unionization, which may infl uence the duration of the disability and the odds of returning to work. 7,8,17The work's high physical and psychological demand and low social support have been associated with 20% lower rates of return to work in workers from California who suffered from low-back pain. 16Even after adjustment by severity of the event, Cheadle et al 4 (2004) found association between age above 30 years and female sex and duration of the disability benefi t.A study with workers on leave due to musculoskeletal diseases found that a positive expectation of return to work was a predictor of the benefi t's duration. 12n inverse relation between income and duration of work leave was observed in workers who received benefi ts due to low-back pain in the United States. 24onversely, a study carried out in Brazil with all the health problems found a positive association between the monthly value of the benefi t (which indirectly measures the worker's salary) and its duration.b Disability studies investigate mainly musculoskeletal disorders (MSD).These diseases represent the largest cause of receipt of social security compensations in diverse countries, remarkably when related to work. 2,18,19In Brazil, 48.2% of the benefi ts granted due to work-related diseases were granted due to MSD in 2006.c Thus, this research focuses on neck and/or upper-limb musculoskeletal disorders (ULMSD) of occupational origin, as a way to reduce differences in characteristics and clinical evolution and to highlight aspects related to prevention capacity.
The present study aimed to analyze factors associated with the duration of disability benefi ts due to workrelated ULMSD.

METHODS
Ambispective cohort study conducted with 563 insured workers from Regime Geral da Previdência Social (RGPS -General Social Security System) who received temporary disability benefi ts due to workrelated ULMSD in the city of Salvador, Northeastern Brazil, in 2008.Eligible individuals were benefi ciaries with fi rst entrance in the system in 2008, start of the benefi t between 1/1/2008 and 12/31/2008 and diagnoses contained in the International Classifi cation of Diseases (ICD-10) compatible with ULMSD.d Injury resulting from acute trauma and previous history of benefi t receipt due to work-related ULMSD in any period were exclusion criteria.The follow-up lasted until 12/31/2009 and the benefi t's cessation due to medical decision was considered an event.
The secondary data came from two sources: Sistema Único de Benefícios (SUB, National Benefi ts System, which is the record of administrative data of Instituto Nacional do Seguro Social [INSS -National Social Security Institute], in which each benefit grant is processed) and a project about temporary disability benefi ts due to ULMSD developed by the Regional Audit of INSS in Salvador.This project included the RGPS workers whose ULMSD benefi ts were granted in the Agências da Previdência Social (APS -Social Security Agencies) of Salvador in the period.After the grant of the benefi t, the insured workers were summoned by the Audit to answer a questionnaire containing sociodemographic information, occupational history, work demands, characteristics of the current health problem and others.The use of these data was authorized by the management of the Regional Audit of Salvador.The data and the reason for the benefi t's cessation were provided by the Regional Audit in January 2010, due to the fact that they were requested after the date that had been set for the termination of the follow-up of the insured workers.The access to the other data occurred in April 2009.
A total of 1,738 benefi ts were granted due to workrelated ULMSD (32% of the occupational benefi ts, including occupational diseases and accidents).These benefi ts corresponded to 1,692 workers, among whom 46 received two temporary benefi ts due to ULMSD in the year and 1,016 were considered ineligible (972 with previous receipt and diagnosis compatible with work-related ULMSD and 44 with injury deriving from acute trauma).
Of the 676 eligible insured workers, 83.3% answered the questionnaire administered by the Regional Audit of the INSS.Of the 16.7% that did not participate, 61 did not go to the Regional Audit after communication b Santana V, Bouzón A, Bouzas Filho J, Nobre L, Campos M, Silva M. Estimativa da carga e custos indiretos com os acidentes de trabalho: relatório de pesquisa by letter or by telephone and 52 were not located.There was a higher proportion of women and older workers among participants in relation to non-participants, but no statistically signifi cant difference was observed in the monthly income paid by the INSS.The median of the benefi t's duration was higher to participants compared to non-participants.
The response variable was duration of the disability benefi t (interval in days between the beginning and the cessation of the benefi t due to the decision of the medical expert of the INSS).
The predictor variables were selected based on literature reviews about return to work and disability benefi ts.
The sociodemographic variables were age, sex, level of schooling, skin color, marital status, employment situation, unionization and socioeconomic position -measured by the ownership of assets and level of schooling of the head of household e : low (classes C, D and E -scores from 0 to 22) and high (classes A and B -scores from 23 to 46).Psychosocial work demands were measured through the Job Content Questionnaire (JCQ), f translated and adapted by Araújo g (2000).Scores were estimated for its three components: psychological demand, control and social support.Each component was analyzed in a dichotomous way (high; low), based on its respective medians.This variable was created through the combination of these three measures: 6 low (high support/high control, high control/low demand or high support/low demand); and high (low control/ low support, low control/high demand or low support/high demand).Job dissatisfaction, whose evaluation was based on questions of the JCQ adapted by Fernandes h (2004), was measured by the scores and dichotomously analyzed (high; low) based on the medians.Physical work demands were defined based on the answers to the questions proposed in the adapted instrument h referring to repetitive gestures, muscular strength with arms or hands, lifting loads and elevation of the arms above the height of the shoulders.Scores were built by adding the points corresponding to each answer and were analyzed as low exposition and high exposition, with cut-off point in the median.
Other work-related factors were analyzed: employees' turnover in the company and time of employment in the last company (in days).
Factors related to the health problem and to health in general approached pain intensity, measured by the individual him/herself based on his/her perception (scale of zero -absence of pain -to ten -unbearable pain), 3 classified into low (< median) and high (≥ median); duration of the problem (in days) before the receipt of the benefi t; having undergone surgery due to the current disease; depression symptoms assessed by the administration of the Patient Health Questionnaire; 22 undergoing psychological therapy; comorbidities; and utilization of the Sistema Único de Saúde (SUS -National Health System) to treat the health problem.
The insured worker's expectation regarding return to work was assessed through the question adapted from the Work Ability Index: i "Considering your health, do you think that in two years' time you'll be able to perform your current work?"The expectation was classifi ed into low ("it's improbable" or "I'm not sure") and high ("probable" or "quite probable").
Social security factors involved monthly income paid by the INSS and level of income replacement by the social security (percentage of the net value of the insured worker's last salary replaced by the INSS, classifi ed into up to 100% and higher than 100%).
The factors associated with time up to benefi t cessation were identifi ed with survival analysis techniques.The Kaplan-Meier method was used to estimate the survival functions.The identifi cation of factors associated with the duration of the benefi t was based on the hazard ratio, which is equivalent to the risk ratio (RR) and respective 95% confi dence intervals, and the Cox model was applied.RR > 1 indicated higher rate of benefi t cessation and shorter duration of the benefi t compared to the reference group.RR were estimated to each descriptor variable, taking a category as referent in bivariate analysis to select the variables to the multiple analysis.The multiple analysis was based on a backward approach, starting from a model with all the variables that presented statistically signifi cant crude risk ratios for α = 0.20.The statistical signifi cance level α = 0.05 was adopted as the criterion for the variables to remain in the model.Level of schooling and monthly income paid by the INSS were excluded from the initial model because they are constructs that directly or indirectly constitute the socioeconomic position.The likelihood ratio test was employed to compare the multiple models.The presupposition of risk proportionality, necessary to the Cox model, was evaluated with the graphical approach and with a statistical test based on the analysis of Schoenfeld residuals. 15The fi nal model was stratifi ed by job dissatisfaction due to the violation of this presupposition.
The management of the databases was performed with EpiInfo, version 6.0 and Stata 7.0 for the conduction of the analysis.

RESULTS
The mean age of the population was 39 years (Table 1).The majority of the insured workers was of the female sex (71%), non-white (83%), lived with a partner (65%), came from Class C (64.3%), had studied up to the middle or the end of Secondary School (56.7%) and was composed of employed workers (95%).The mean monthly income paid by the INSS was R$ 944.23.To 46% of the workers, the income paid by the INSS was higher than their last net salary.The benefi ciaries had high control of the work, high psychological demand, high social support, high physical demand and low job dissatisfaction.Workers with negative expectation of return to work predominated (62%).
The mean pain intensity that was reported was 9.28, the mean duration of the health problem before the beginning of the benefi t was more than two years, the majority did not report having undergone surgery (85%), 42% presented depression symptoms, and 37% used the SUS for medical treatment (Table 2).
The insured workers' probability of continuing to receive the benefi t was higher than 70% in the fi rst 90 days and 14% after one year (Figure 1).The duration of the benefi t varied from six to 730 days, with median of 139 days and mean of 188 days (SD = 151).Considering that the censored cases were included in the calculation of the mean, there was underestimation.At the end of the follow-up, 95% of the insured workers had their benefi t terminated, 24 continued on work leave and two had their sickness allowance transformed into disability retirement.
The rate of benefit cessation was 35% higher for younger insured workers than for older ones (RR =  4).These were measures of residual effect, after adjustment by all the potential predictors.

DISCUSSION
Low socioeconomic position, age below 39 years, high expectation of returning to work, and income replacement by the INSS lower than or equal to 100% are related to higher benefi t cessation rate and to shorter benefi t duration.The shorter duration of the benefi t among workers of low socioeconomic position compared to those of high socioeconomic position, which was shown in this study, differs from the fi ndings of investigations carried out in other countries, in which the association between socioeconomic position and benefi t duration is observed in the opposite direction.Volinn et al 24 (1991) examined benefi ts granted due to low-back pain in the United States and found an inverse relation between income and duration of the work leave.Katz et al 13 analyzed factors related to return to work among workers who underwent surgery for carpal tunnel syndrome and observed that individuals with lower income had odds almost four times higher of remaining on leave compared to those with higher income.A literature review showed a strong and consistent inverse association between social class and duration of work leave due to low-back pain, mainly among men. 25However, as in the present study, preliminary analyses of the database of the INSS in Bahia in 2000 showed a positive association between the monthly value of the disability benefi t, which indirectly measures the worker's salary, and benefi t duration.j These results can refl ect inequalities in the local social security, in which the economically privileged individuals have more lasting benefi ts.Individuals with privileged socioeconomic position would have greater dialog capacity with their employers and with State institutions, and greater knowledge of their rights and of the functioning of social security, which might determine a difference in the utilization of the social security benefi ts.
The association between expectation of returning to work and benefi t duration agrees with other studies about MSD.This association has been found in Canada by Hogg-Johnson & Cole, 12 who studied workers with work-related MSD in Ontario, and by Gross & Battié, 9 who analyzed patients with chronic low-back pain in Alberta.Expectation of returning to work among patients submitted to surgery due to carpal tunnel syndrome was the most important predictor of absence from work. 13The reason that infl uences workers' expectation is not exactly known, nor the mechanisms of its association with duration of work leave.A qualitative study with workers with occupational low-back pain showed that the expectation of returning to work refl ects four domains: fi nancial and job security, possibility of recurrence of the health problem, social support at the workplace and self-image. 21e level of income replacement by the INSS was a predictor of benefi t duration.In the Brazilian social security, the value of the disability benefi t received by the insured worker corresponds to 91% of the benefi t salary, which corresponds to the mean of the 80% highest salaries of the entire contribution period, duly corrected.Thus, the benefi t's value can be lower, equal or higher compared to the insured worker's last salary before the leave.Individuals who received benefi ts equal to or lower than the value of the last salary before the leave had shorter benefi ts in relation to those who received benefi ts with higher value.The higher value of the benefi t compared to the salary may be a direct incentive to the worker, thus becoming an attractive condition and infl uencing the longer duration of the benefi t. there is an effect of the level of income replacement by the insurance on the incidence and duration of disability benefi ts.Another review concluded that the results concerning the association between level of income replacement by the insurance and benefi t duration are controversial, possibly due to differences in the contexts in which the studies were conducted. 11nefi ts of younger insured workers lasted less, which indicates lower recovery diffi culty in this group.Age is one of the most consistent predictors of disability due to occupational health problems. 23ychological demand, work control, comorbidities and the period in which the health problem was present before the receipt of the benefi t were associated with benefi t duration, but lost signifi cance in the fi nal model.One of the possible explanations may be the insuffi cient size of the studied population to detect differences in benefi t duration caused by these factors.Analytical approaches with multiple predictors frequently demand an increase in the size of the studied sample. 1 The results that were found should be analyzed in light of the limitations of the database; after all, this study used secondary data from the INSS, which may negatively affect the quality of the information.In the present study, the losses represented 16.7% of the selected insured workers.The median of benefi t duration was higher to the participants in comparison to non-participants and the difference between the survival curves was statistically signifi cant.The utilization of information restricted to the fi rst benefi t of the insured worker was another limitation.Such procedure might underestimate the duration of the disability, because the ULMSD may cause recurrent leaves from work because they are chronic diseases.
Non-strictly medical factors, like socioeconomic position, age, expectation related to returning to work and level of income replacement by the INSS infl uence the duration of the benefi t.These hypotheses should be tested in confi rmatory studies.Enhancing the understanding of the process that determines the utilization of the social security/ incapacity for work is fundamental to the development of interventions that reduce the economic, social and personal costs of disability, as well as to face the possibility of social inequality in the utilization of the insurance.
This research was approved by the Research Ethics Committee of the Institute of Public Health of Universidade Federal da Bahia (Process no.014/2009).Access to the nominal data was restricted to the fi rst author.

Table 2 .
Clinical characteristics of the cases and health care.

Table 3 .
Means of benefit duration and results of the bivariate analysis for the association of the variables: sociodemographic, work, social security, and cases' clinical characteristics with the duration of the disability benefi t.
SUS: Sistema Único de Saúde (Unifi ed National Health System); INSS: Instituto Nacional do Seguro Social (National Social Security Institute).a Risk Ratio higher than 1 -increased odds of benefi t cessation/ return to work and, consequently, of shorter duration of the benefi t.
20,20On the other hand, inadequate income replacement by the social security system, with value lower than the salary, may make the worker return prematurely to work, with important physical limitations.This does not allow him/ her to recover adequately, which may lead to relapses and other work leaves.Loeser et al20concluded that Survival curve (likelihood of remaining in benefi t, in days) of insured workers with work-related neck and/or upper limb musculoskeletal disorders. Salvaor, Northeastern Brazil, 2008.(N = 563) j Santana V, Bouzón A, Bouzas Filho J, Nobre L, Campos M, Silva M. Estimativa da carga e custos indiretos com os acidentes de trabalho: relatório de pesquisa.Salvador: Instituto de Saúde Coletiva da Universidade Federal da Bahia; 2004 [cited 2012Jan 16].Available from: http://www.2pontos.net/preview/pisat/hp/upload/custos_relCOSAT2007.rar

Table 4 .
Factors associated with the duration of the disability benefi t.Salvador, Northeastern Brazil, 2008.(N = 563) Risk Ratio higher than 1-increased odds of benefi t cessation/ return to work and, consequently, of shorter duration of the benefi t.Note: Final model stratifi ed by the variable "job dissatisfaction" a