Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Compensation status and surgical outcome

Systematic review

The history of worker compensation, or other compensation, and medical outcomes is a difficult and tortuous one, which Bandolier has no hope of reviewing. For those with an interest in its history, though, there is at least one good reference [1].

What about outcome after surgery and compensation status? A new meta-analysis suggests that poor outcomes may be twice as likely in compensated versus non-compensated individuals [2].

Systematic review

The study sought published papers of any trial of surgical intervention in which compensation status was reported, and compared results according to that status. A compensated patient was defined as one who received workers' compensation payments for their condition, or who experienced litigation as a result of their preoperative condition. The minimum size was set at a single compensated and a single non-compensated patient.

Any surgery was included, but not rehabilitation, injections, or similar interventions. Region-specific outcome scores were the main outcome of interest, but generalised function scores, health outcome scores like SF-36, patient satisfaction score, or a pain score were also used. Satisfactory or unsatisfactory outcomes were as reported in the publications, with general descriptions like fair, poor or failure included in unsatisfactory.


The search identified 211 studies, of which 175 described a worse outcome in the compensation group, 30 described no difference, and one described better outcomes in the compensation group. Five did not comment.

Dichotomous scores were available in 129 studies with over 20,000 patients, ranging in size from nine to 4,200 patients. The median size was 74 patients, and all but seven of the studies had fewer than 500 patients. The results differed somewhat depending on the size of the study (Table 1, Figures 1-3).

Table 1: Results analysed according to size of study

Number of
Pecent unsatisfactory outcome
Size of study
Relative risk
(95% CI)
(95% CI)
Below median (≤74)
1.3 (0.8 to 2.3)
3.3 (3.0 to 3.7)
Moderate (77-499)
2.0 (1.8 to 2.3)
3.8 (3.5 to 4.1)
Largest (≥500)
2.0 (1.8 to 2.1)
10 (8.8 to 13)
All studies
2.0 (1.9 to 2.1)
4.8 (4.5 to 5.1)

Figures 1-3: Unsatisfactory surgical outcome in smallest, larger, and largest trials

The smallest studies, those below the median size, had the highest proportion (51%) of unsatisfactory outcomes with surgery (Figure 1). The NNT in this group of patients was 3.3 (3.0 to 3.7).

Larger studies between 75 and 499 patients in size had a lower, but still high, proportion (45%) of unsatisfactory outcomes with surgery (Figure 2). The NNT in this group of patients was higher at 3.8 (3.5 to 4.1).

The seven largest studies with more than 500 patients in each had the lowest proportion (22%) of unsatisfactory outcomes with surgery (Figure 3). The NNT in this group of patients was much higher at 10 (9 to 13).

Other sensitivity analyses had limited value because of the propensity of smaller studies to have bigger effects, especially when there were so many small studies. But study design, follow up time, type of procedure, or type of compensation seemed to make little difference to the overall conclusion.


A number of lessons are to be learned here. First is that of size. We forget at our peril that small studies tend to over-estimate effects, and the smallest allowable trial here was of two (2) patients. Some trials had only two or three patients in one of the groups. In this example only 14% of all patients were in half of all trials (the smallest), but 41% were in the largest seven trials.

Both large and small studies reported a similar direction of result, the magnitude of which was larger in the smallest trials of below median size. The magnitude of the difference declined with increasing size. So the large difference in small trials (1 more unsatisfactory outcome for every 3 compensated versus non-compensated patients) became less important in the largest trials (1 more unsatisfactory outcome for every 10 compensated versus non-compensated patients).

We know from other circumstances that small studies often lack the rigour of large studies, which is why we see a bias towards greater effects in small studies. The implication is that we should disregard them in favour of larger studies. But how large is large enough? As we have no quality filter for the studies in the review, this leaves us in a bit of a pickle. We cannot rule out that the difference between compensated and non-compensated is due to bias from poor study quality.

The largest trials demonstrated only half the differential effect of compensation status compared with the whole sample of trials. Moreover, as none of the trials was randomised by compensation status, we do not know whether patients who were being compensated had the same degree of severity as those who were not being compensated. So while we do not know it to be the case, we cannot rule out that the compensated were initially worse than non-compensated, which is why there is a difference.

All in all, a bit of analysis takes what is on the face of it a straightforward result that fits our prejudices nicely, and challenges whether there is any effect at all. But that probably takes it too far, and concerns about bias and baseline differences are not justified. The best we can say is that it is likely that unsatisfactory results from surgery are more common in people who are being compensated, but that we have no good idea just how much more common.


  1. G Mendelson. Compensation and chronic pain. Pain 1992 48: 121-123.
  2. I Harris et al. Association between compensation status and outcome after surgery. JAMA 2005 293: 1644-1652.

previous or next story