Research Proposal – Predictors of Return-to-Work After Workplace Injury

Introduction

Workplace injury imposes a substantial and uneven burden on Australian workers, employers and the statutory schemes that fund recovery. Safe Work Australia (2023) records well over 100,000 accepted workers’ compensation claims each year involving a week or more of lost work, at rising cost. Most injured workers resume employment within weeks (Safe Work Australia, 2022), but a persistent minority drift into extended incapacity and consume a disproportionate share of scheme cost and lost labour. Because the chance of a durable return falls sharply the longer a worker stays away, the early weeks of a claim are decisive.

This proposal sets out a prospective cohort study linking de-identified scheme claims data to a nested survey to identify the biological, psychological and workplace predictors of timely versus delayed return to work (RTW), producing Australian evidence that schemes and employers can use to target early intervention.

Background and Problem Statement

Workers’ compensation and delayed recovery

Australian workers’ compensation is federated. Comcare covers Commonwealth employees and licensed national employers, while each state and territory runs its own scheme, including icare in New South Wales, WorkSafe Victoria, WorkCover Queensland and ReturnToWorkSA, all underpinned by the harmonised model Work Health and Safety (WHS) laws and reported nationally through Safe Work Australia. Even for comparable injuries, time lost varies considerably across jurisdictions, implying that system and workplace factors, not pathology alone, shape recovery (Collie et al., 2016); work-related injury also remains a significant contributor to preventable long-term disability (Australian Institute of Health and Welfare, 2023). The problem is duration, not volume: a small proportion of long-term claims consumes most scheme cost.

Psychosocial flags and the limits of current practice

Recovery from common musculoskeletal and psychological injuries is poorly predicted by severity alone. Psychological and social obstacles, the yellow and blue flags, predict prolonged disability more strongly than biomedical variables (Kendall et al., 2009). Low recovery expectations, distress, fear of re-injury and an unsupportive workplace are consistently associated with delayed RTW, and Australian evidence links the stressfulness of claiming itself to poorer recovery (Grant et al., 2014). Yet schemes still triage mainly on injury type and certified incapacity, both weak predictors, and rarely screen early for modifiable obstacles, so problem claims are recognised only once protracted.

Problem statement

Two problems follow. Schemes can identify claims that have already become long-term, but early, prospective identification of who is at risk remains imprecise. And the relative contribution of biological, psychological and workplace factors, measured together in one Australian cohort, is not well quantified, so it is unclear whether early psychosocial screening improves prediction beyond the data schemes already hold. Intervention consequently remains reactive. This study addresses both problems in one design.

Aim and Research Questions

The aim is to identify and quantify the biological, psychological and workplace predictors of time to sustained return to work among workers with accepted claims, and to test whether early psychosocial screening improves prediction beyond routinely collected claims data. Three research questions guide the study:

1. Which biological, psychological and workplace factors, measured within the first weeks of a claim, predict time to sustained return to work?
2. Does the addition of early psychosocial screening data improve predictive performance beyond routinely collected claims variables such as injury type, severity, demographics and jurisdiction?
3. To what extent do workplace factors, in particular the offer of modified duties and perceived supervisor support, moderate the relationship between psychological distress and delayed return to work?

Literature Review

The biopsychosocial model of disability

The biopsychosocial model explains why recovery diverges from tissue healing. Engel (1977) argued that disability cannot be understood through pathology alone, because psychological state and social context shape both symptoms and the capacity to resume work. Applied to work injury, prolonged incapacity becomes the product of interacting biological, psychological and social influences, underpinning the position that work is generally beneficial for health while prolonged worklessness is itself harmful (Waddell & Burton, 2006). Best-evidence syntheses confirm that recovery expectations, workplace factors and psychological status are among the most consistent prognostic factors across injury types (Cancelliere et al., 2016).

Return-to-work coordination

If disability is partly social, intervention in the organisational environment should shorten it. Franche et al. (2005) found that workplace-based interventions, namely early contact with the injured worker, modified or graded duties, and a designated return-to-work coordinator, produced moderate and consistent reductions in time lost. Coordination is an active ingredient because it aligns the clinical, occupational and administrative actors whose fragmentation otherwise delays recovery.

Employer and workplace factors

Two workplace factors recur. Suitable modified duties allow a graded resumption that preserves occupational role and identity, while perceived supervisor and co-worker support shapes whether a worker feels able to return. In an Australian review, Iles et al. (2008) found that low recovery expectations and fear predicted failure to return to work in non-chronic low back pain more strongly than clinical severity. Because these levers lie within the control of employers and schemes, they are of direct policy interest under the model WHS framework, which places positive duties on employers to support recovery.

Synthesis and gap

The international literature identifies a fairly stable set of prognostic factors, but three gaps persist locally: much evidence is clinical or overseas; many Australian studies use administrative claims data alone and cannot measure psychosocial constructs; and few prospective cohorts model return to work as a time-to-event outcome while combining routine scheme data with early screening. No recent Australian study has quantified, in one linked cohort, whether early screening adds predictive value, which is this study’s contribution.

Conceptual Framework

The framework in Figure 1 applies the biopsychosocial model to the return-to-work pathway. Biological, psychological and workplace factors present at claim onset are theorised to influence time to sustained return to work both directly and through the quality of return-to-work coordination, which mediates by buffering or amplifying risk, all within a scheme and system context of jurisdiction and claim process. It generates the study’s central hypothesis: that psychological and workplace factors carry predictive weight beyond biological severity, and that coordination moderates their effect.

Methodology

Design and setting

A prospective inception cohort will follow workers from claim acceptance, drawn from de-identified administrative claims data supplied by one or more participating Australian schemes. Eligible claims are those accepted for musculoskeletal or psychological injury with at least one week of certified incapacity, giving an anticipated base cohort of about 1,200 workers followed for 12 months. A nested survey administered to consenting members three to six weeks after acceptance will capture the psychological and workplace constructs that claims data cannot measure.

Cohort and nested survey

Time to sustained return to work is a time-to-event outcome, so power depends on the number of events observed rather than the sample size alone. With 1,200 accepted claims and an expected 85 per cent returning to sustained work within the 12-month window, the expected number of events is:

Expected events = 1,200 x 0.85 = 1,020

The events required to detect a hazard ratio (HR) of 1.30 for a balanced binary predictor, at a two-sided alpha of .05 and 80 per cent power, follow the Schoenfeld formula:

d = (z_1-alpha/2 + z_1-beta)² / [p(1 – p)(ln HR)²]

d = (1.96 + 0.84)² / [0.5 x 0.5 x (ln 1.30)²] = 7.84 / (0.25 x 0.0688) = 456

The 1,020 expected events far exceed the required 456, so the study is well powered to detect associations of modest size. At a minimum of ten events per estimated parameter, 1,020 events support around 100, far beyond the roughly 15 candidate predictors in Table 1. A response rate near 60 per cent would still yield some 700 workers with complete psychosocial data.

Measures and variables

Table 1 lists the outcome and candidate predictors by biopsychosocial domain, with measurement source and variable type. Claims variables cover the full cohort; psychological and most workplace variables come from the nested survey, using instruments with established Australian use.

Table 1: Outcome and candidate predictors by biopsychosocial domain, measurement source and variable type

Note. Sustained return to work is defined as cessation of wage-replacement payments maintained for at least four consecutive weeks, distinguishing durable return from brief unsuccessful attempts. Workers who exit the scheme without returning are treated as a competing outcome in sensitivity analyses.

Statistical analysis

The primary analysis will model time from claim acceptance to sustained return to work using Cox proportional hazards regression, with the proportional-hazards assumption tested through scaled Schoenfeld residuals. Claims that close without a return will be treated as a competing risk. For the second research question, discrimination will be compared between a claims-only model and an extended model adding the psychosocial variables, using Harrell’s concordance statistic; a meaningful gain would show that early screening adds value. The third question will be tested through interaction terms (distress by modified duties, and distress by supervisor support). The secondary six-month outcome will be analysed by multivariable logistic regression, with multiple imputation for missing survey data.

Ethics and data governance

Approval will be sought from the administering university’s Human Research Ethics Committee, with the study conducted under the National Statement on Ethical Conduct in Human Research (National Health and Medical Research Council, 2023). Claims data will be accessed through a data-sharing agreement with each scheme, identifiers separated from analytical variables so the team works only with a de-identified linked dataset in a secure environment. Because the survey collects identifiable self-report, participants will give informed consent, and no individual response will be returned to employers or insurers. A distress protocol will provide support-service contacts, and governance of the data linkage will be specified in the ethics application before any extraction.

Project Timeline

The study will be completed over 18 months, as set out in Table 2. Deliberate overlap between cohort assembly and survey fieldwork keeps the design feasible within the funded period without compromising the prospective follow-up.

Table 2: Eighteen-month project timeline

Significance of the Study

For compensation schemes, the study offers an empirically validated basis for identifying, within the first weeks of a claim, the workers most likely to experience delayed recovery. If early screening improves prediction beyond routine data, schemes such as Comcare and the state-based insurers could justify embedding brief screening at lodgement and shifting case-management effort from entrenched claims towards early support. For employers, the analysis will indicate which controllable levers, particularly modified duties and supervisor support, most influence recovery, informing how positive obligations under the model WHS framework are met. Theoretically, it contributes rare Australian prospective evidence on the biopsychosocial predictors of return to work in a linked cohort.

Limitations and Delimitations

Four constraints qualify the design. First, return to work is defined administratively through cessation of wage-replacement payments, which may misclassify workers who return to reduced hours or exit the scheme for other reasons; the four-week threshold and competing-risks analysis reduce, but do not remove, this risk. Second, the survey depends on voluntary participation, so selection and non-response may bias the psychosocial estimates, addressed through multiple imputation and comparison of respondents with the full cohort on claims variables. Third, recruitment from one or a few schemes limits generalisability across Australia’s jurisdictions, given the known variation in scheme design (Collie et al., 2016). Fourth, observational data cannot establish causation, so associations are prognostic rather than proof that intervention will change outcomes. The study is further delimited to accepted claims, excluding unreported injuries and rejected claims, whose pathways may differ.

Conclusion

Delayed recovery after workplace injury is costly, damaging to workers and only weakly predicted by the injury alone. Decades of biopsychosocial research point to psychological and workplace factors as decisive, yet Australian schemes still triage on weak predictors and rarely screen for the obstacles that matter most. This proposal sets out a feasible 18-month prospective cohort study that links de-identified claims data to an early nested survey and models time to sustained return to work directly, testing whether psychosocial screening adds predictive value and whether workplace support moderates the effect of distress. By quantifying these predictors in a single linked Australian cohort, it aims to give schemes and employers an evidence-based means of acting early, when action is most likely to keep an injured worker connected to work.

References

Australian Institute of Health and Welfare. (2023). Injury in Australia. AIHW.

Cancelliere, C., Donovan, J., Stochkendahl, M. J., Biscardi, M., Ammendolia, C., Myburgh, C., & Cassidy, J. D. (2016). Factors affecting return to work after injury or illness: Best evidence synthesis of systematic reviews. Chiropractic & Manual Therapies, 24(1), 32.

Collie, A., Lane, T. J., Hassani-Mahmooei, B., Thompson, J., & McLeod, C. (2016). Does time off work after injury vary by jurisdiction? A comparison of workers’ compensation systems in Australia and New Zealand. BMJ Open, 6(9), e010910.

Engel, G. L. (1977). The need for a new medical model: A challenge for biomedicine. Science, 196(4286), 129-136.

Franche, R. L., Cullen, K., Clarke, J., Irvin, E., Sinclair, S., & Frank, J. (2005). Workplace-based return-to-work interventions: A systematic review of the quantitative literature. Journal of Occupational Rehabilitation, 15(4), 607-631.

Grant, G. M., O’Donnell, M. L., Spittal, M. J., Creamer, M., & Studdert, D. M. (2014). Relationship between stressfulness of claiming for injury compensation and long-term recovery. JAMA Psychiatry, 71(4), 446-453.

Iles, R. A., Davidson, M., & Taylor, N. F. (2008). Psychosocial predictors of failure to return to work in non-chronic non-specific low back pain: A systematic review. Occupational and Environmental Medicine, 65(8), 507-517.

Kendall, N. A. S., Burton, A. K., Main, C. J., & Watson, P. J. (2009). Tackling musculoskeletal problems: A guide for clinic and workplace. TSO.

National Health and Medical Research Council. (2023). National statement on ethical conduct in human research. NHMRC.

Safe Work Australia. (2022). National return to work survey: Summary report. Safe Work Australia.

Safe Work Australia. (2023). Australian workers’ compensation statistics 2021-22. Safe Work Australia.