Work absence tends to perpetuate itself: that is, the longer someone is off work, the less likely they become ever to return.
If the person is off work for:
- 20 days the chance of ever getting back to work is 70%;
- 45 days the chance of ever getting back to work is 50%; and
- 70 days the chance of ever getting back to work is 35%.
The statements are referenced to a study for the Victorian WorkCover Authority by Johnson and Fry published in 2002.2 However, the reference does not contain the statements or results that could support them.
The statements are being repeated by New Zealand3 and Australian4,5 government agencies, in the explanatory memorandum for a bill to amend the Safety, Rehabilitation and Compensation Act in Australia,6 non-government organisations7 and the commercial sector, including insurance.8 They were presented to the New Zealand Government’s Welfare Working Group Forum in the context of influencing government policy.3 The statements are frequently referenced to Johnson and Fry.2 They have appeared in international literature,9 also referenced to Johnson and Fry.2
Misinterpretation of survival curves
The conclusions appear to be based on the misinterpretation of survival curves. A presentation by Dr Robin Chase (President AFOEM) and Dr Mary Wyatt (Chair Policy and Advocacy Committee AFOEM, Co-chair and Australian lead of the working group which contributing substantially to the position statement) shows the statements associated with a figure that looks like figure 5.3 (Survivor Functions: Males, Timeliness) from the position statement’s referenced study.10 A webpage by Dr Mary Wyatt shows what appears to be same figure and states, “This graph shows that the likelihood of return to work goes down the longer the person is off work”, together with the same statements about the chance of ever getting back to work, as the statements quoted above from the position statement.11
All of the other occurrences of these statements appear to postdate the AFOEM Position Statement, which seems likely to be their source.
Figure 1 is the graph from Chase and Wyatt presentation10 and appears the same as the graph on Wyatt’s web page.11 What the graph shows is that if the person is off work for 20, 45 or 70 days, the chance of ever getting back to work is close to 100%. The graph is the fitted survival distribution from a Weibull model which, as time becomes infinite, has this property. The graph does not show “the longer someone is off work, the less likely they become ever to return”.
In figure 5.3 of the Johnson and Fry publication, the vertical axis of the graph was labelled Pr(T>=t).2 This is the probability that the time to the event is greater than or equal to the time on the horizontal axis. The event is the cessation of weekly payments, treated synonymously as return to work. Time is the number of days off work, after an initial 10 days. The data comes from the Victorian WorkCover Authority’s administrative database of injured workers. Only those off work for more than 10 days are included, 100% of those have a time to return to work greater than 10 days, the Pr(T>=t) = 1.0 (the vertical axis) at 0 days. About 70% have a time to return to work greater than 20 days, about 50% greater than 45 days, and about 35% greater than 70 days. Only about one or two percent have, from the model fitted, a time to return to work greater than one year.
The curves in the figure are for males with sets of characteristics, rather than an overall curve for all males. Males with different characteristics—and females—will have different curves. Those seriously injured have, not surprisingly, much lower probabilities of having returned to work within a year (see figure 5.5 in Johnson and Fry2). The figures were presented in the publication to illustrate the differences in probabilities under a range of circumstances and are not suitable for estimating the overall chance of ever getting back to work.
The graph does show that the longer someone is off work the less likely they are to return in the next time period; however, this aspect of the shape of the curve is hard to see.
When I read the statements I thought them unlikely to be true, so I looked for their source, the position statement,1 and checked the referenced study,2 which shows them to be without foundation. Presumably the authors and repeaters have found them believable, for example Dr David Bratt, Principal Health Advisor, Ministry of Social Development, New Zealand, said at the Welfare Working Group Forum, that the figures come from Australia, but that he knows that figures would show exactly the same thing happens here.12
The statements have been presented in non-injury contexts, for example being out of work12 or mental illness.7
The statements use
The statements are being used to support statements like: “Urgent action is required if a person is not back at work within a matter of weeks. If a person is not back at work within three weeks urgent attention is needed”11 even though the data is for time after an initial 10 days off work.
The incorrect statements about the chance of ever getting back to work are being presented to general practitioners (GPs) continuing medical education conferences in the context certifying people as unfit for work, together with statements like the ‘benefit’ is “an addictive debilitating drug with significant adverse effects to both the patient and their family (whānau)”.13 They are being presented to GPs in the context of assisting patients to safely stay at work or return to work early.4 These appear to be encouraging GPs to assess injured and unwell patients as having capacity for work and not issuing medical certificates for work incapacity. This could result in the cessation of welfare benefits or injury compensation. When these patients lack the capacity to work, they could experience increased financial hardship. For example, people might move from injury compensation to an unemployment benefit, and those without benefit entitlements to no income. There are also consequential beneficiaries of these income shifts. For example, reductions in government expenditure have been associated with reductions in taxation. Reductions in injury compensation for work-related injuries could result in reductions of employer levies/premiums for workers’ compensation and consequential increases in dividends to the owners of businesses.
The statements have also been presented with the intent to influence public policy.3,6,14
Others have critically noticed the statements—describing them as surprising and most ludicrous—noting they are claiming that time off work causes time off work.15 They point out that association is not causation, but did not note the statement’s erroneous nature.
Not only are the statements an incorrect interpretation of a graph, the data in the graphs are also being generalised to all persons off work, whereas the graph was for a subset of males who have had more than 10 days off work after an injury. The data are censored at 520 days, and hence extrapolation beyond one year five months to ever getting back to work would be unlikely to be reasonable. The data came from 1993–1998 and due to changing circumstances may not be applicable to inform today’s policies.
The statement that, if a person is off work for 70 days, the chance of ever getting back to work is 35%, is not justified by the study the position statement references. The statements appear to be based on an incorrect interpretation of a graph in the referenced study.
A reasonable summary of the graph would be that after more than two weeks off work due to an injury, most people return to work within a year. However, even that summary only applies to males with certain sets of characteristics, at a certain time.
It is not known what effect the statements are having, but they are based on an incorrect interpretation of the referenced study and should be corrected.