Disability Management and Return to Work

Definition

Disability management is the set of employer policies and behaviors designed to restore an injured worker's productive capacity and maintain their employment after an occupational injury. In the workers' compensation (WC) context, the dominant outcome measure has historically been the first return-to-work rate — whether an injured worker ever returns after an absence — but Butler et al. (1995) demonstrate that this metric conflates initial return with stable post-injury employment, which are distinct outcomes with distinct determinants.

Key Ideas

The first return-to-work rate ( $\approx 85\%$ in Ontario workers with permanent impairments) dramatically overstates success; roughly half of injured workers achieve stable, continuous post-injury employment.
Four mutually exclusive post-injury employment patterns exist among workers who return: single absence/successful, single absence/unsuccessful, multiple absences/successful, multiple absences/unsuccessful.
Employer accommodations — reduced hours, modified equipment, light-duty work — are the dominant predictors of stable employment, dwarfing the effects of individual worker characteristics.
The WC replacement rate discourages initial return to work but has no significant effect on whether workers who do return achieve stable employment.
Back injuries disproportionately produce multiple-absence patterns and are most severe for older, lower-skilled workers.

How It Works — The Nagi Framework

Butler et al. (1995) adopt the Nagi (1969) disability model, which distinguishes three levels:

Impairment — a physiological or anatomical loss or abnormality (e.g., herniated disc, partial limb loss)
Functional limitation — a restriction in basic physical or mental action (e.g., inability to lift $> 20$ lbs)
Work disability — inability to perform job duties as a result of the functional limitation

Work disability D is not fixed by the injury — it depends on the gap between the worker's functional capacity and the demands of the available jobs. Employer accommodations reduce D directly by lowering job demands (modified equipment, reduced hours) or by matching the worker to lighter tasks (light-duty work). This is why accommodations are the strongest predictor of stable return: they operate at the disability-determination margin, not the impairment or limitation margin.

Four Employment Patterns (Butler et al. 1995)

Data: Survey of Ontario Workers with Permanent Impairments (1990), $n = 1{,}850$ with complete post-injury history (from $\approx 11{,}000$ surveyed), injured 1974–1987. Among 2,870 who returned to work:

Pattern	Description	Share
1	Single absence, successful return	39%
2	Single absence, unsuccessful return	29%
3	Multiple absences, successful return	21%
4	Multiple absences, unsuccessful return	11%

Pattern 4 (omitted category in the multinomial logit) is the worst outcome — chronic cycling through spells of return and re-absence without stable employment. 40% of those who initially returned were not employed due to injury effects at the 1990 interview date.

Employer Accommodations as the Dominant Predictor

Semi-elasticities from the multinomial logit (effects on Pattern 1 and Pattern 4 probability):

Accommodation	Effect on Pattern 1 (best)	Effect on Pattern 4 (worst)
Reduced hours	+37%	−61%
Modified equipment	+35%	−38%
Light-duty work	+18%	−55%

For comparison, individual worker effects on P1 vs. P4: higher education +28%, public sector +44%, male sex positive. Accommodations dominate individual characteristics in magnitude.

Replacement Rate: Entry vs. Stability Effects

The WC replacement rate discourages initial return to work — higher benefits lower the opportunity cost of remaining absent. But it has no significant effect on whether workers who do return achieve stable employment (no significant effect on Pattern 1 vs. 2, or Pattern 3 vs. 4 probabilities). The standard moral hazard concern about WC benefit generosity is real at the entry margin but irrelevant to the stability margin. Policy focused exclusively on reducing benefits to encourage return ignores the dominant driver of post-return failure: lack of accommodation.

Why It Matters

Measurement: First return-to-work rates are used in program evaluation, employer benchmarking, and litigation. Butler et al. show they are systematically misleading — programs with identical first-return rates can have very different post-return stability outcomes.
Policy design: Employer accommodations are the dominant lever. WC policy should consider accommodation incentives (experience-rating adjustments, subsidies, mandates) rather than focusing primarily on benefit structure.
Back injuries: The excess multiple-absence risk from back conditions disproportionately affects older, lower-education workers — the same population most vulnerable to Social Security Disability Insurance (DI) entry via the occupational injury pathway documented by O'Leary et al. (2012).
Interaction with DI: Workers cycling through multiple absences without stable employment are the most likely to eventually apply for DI. Accommodations that intercept this trajectory could reduce the WC-to-DI pipeline.

Productivity Loss Decomposition (Butler, Baldwin & Johnson 2006)

The 2006 paper extends the 1995 findings by quantifying how large post-injury productivity losses are even after workers return. The method decomposes the gap between expected and actual earnings into two components:

Absentee effect $= \log(\text{preinjury weeks worked} / \text{postinjury weeks worked})$ — losses from spells absent
Earnings effect $= \log(\text{preinjury weekly wage} / \text{postinjury weekly wage})$ — losses from reduced wage or hours while present

Key results from Ontario workers with permanent partial disabilities (PPDs) ( $n = 800$ ; three-year complete history):

Group	Year 1 loss	Year 2 loss	Year 3 loss	Dominant component
Stable ( $N = 322$ )	43%	6%	16%	Y1: absentee; Y2–3: earnings
Unstable ( $N = 478$ )	58%	29%	52%	Absentee throughout

The earnings-effect finding for stable workers is critical: workers who return and stay still earn 16% less than their preinjury benchmark three years out, primarily through lower wages or reduced hours while present. This is invisible to standard return-to-work metrics.

Different determinants by stability group:

Stable workers: specific human capital (5+ years tenure) and high school education reduce losses; flexible-schedule accommodation significantly reduces losses years 1–2
Unstable workers: vocational training (portable skills) reduces losses $\approx 15\%$ in all three years; light work reduces losses year 3; but flexible schedules can increase absentee losses in year 2 (trial-work failure hypothesis: accommodated workers re-absent after failing the accommodation test)

Age and Job Performance in Disability Management (Ng & Feldman 2008)

Ng & Feldman's meta-analysis of 380 studies provides an important complement to the WC return-to-work literature by documenting how age relates to absenteeism across its subtypes:

General absence (objective): age is strongly negatively related ( $-0.26$ ) — older workers have fewer total missed days in archival records.
Sickness absence (objective): near-zero ( $0.02$ to $0.04$ , slightly positive) — the disability-relevant absence category shows essentially no age gradient.
Voluntary/nonsickness absence (objective): negative ( $-0.10$ ) — older workers are reliably more punctual on the discretionary margin.
Self-rated absence: near-zero ( $0.01$ ) regardless of type — self-report systematically underestimates the age–absence gap relative to archival records.

The sickness/voluntary distinction is critical for interpreting DI research: studies that aggregate all absence conflate discretionary withdrawal (strongly age-negative) with health-driven absence (age-neutral). The dominant age effect in general absence statistics reflects the voluntary component, not the sickness component that is relevant to disability. This means age-related absenteeism improvements cannot be used as evidence that older workers' health-driven work capacity is increasing.

Additionally, Ng & Feldman document a positive age–organizational citizenship behavior (OCB) relationship ( $0.06$ – $0.21$ depending on target and rater) and a negative age–supervisor-rated task performance relationship for workers 40+ that is likely inflated by supervisor age bias rather than actual cognitive decline. These patterns are directly relevant to understanding why employers may underaccommodate injured older workers: biased performance evaluations create a disincentive to accommodate workers who would, by objective standards, be good candidates for return.

Open Questions

Do accommodation effects generalize across U.S. WC systems, which vary enormously by state and differ from Ontario's universal healthcare context?
How has the Americans with Disabilities Act (ADA, 1990) accommodation mandate interacted with WC disability management? The paper predates the ADA.
Does employer experience-rating under WC create adequate incentives for accommodation, or do employers strategically underaccommodate to shift long-term costs to DI?
Are the four-pattern findings stable across different injury types, industries, and macroeconomic conditions?
Does the weakening of the age–injury relationship over time (Ng & Feldman moderator finding) reflect genuine occupational safety improvements for older workers, or survivor bias as injured older workers exit to DI?