We found no statistically significant differences between the three treatments in total societal costs, direct costs, or the cost of production losses. The overall results are consistent with other recently published studies.5,15
However, as with clinical outcomes,16
our finding of no difference in costs must be interpreted with caution. As is usual, cost data were highly variable, and our study may have been underpowered to detect differences in costs that would be considered important by decision makers. However, in our comparisons of patients allocated to the psychological therapies by either of the randomisation procedures we effectively doubled the sample size available and still found no significant differences. Furthermore, none of the sensitivity analyses significantly influenced the results. The validity of our finding of “no difference” may be strengthened through further original research with larger samples of patients, although recruitment in primary care is often difficult17
and the costs of such studies may prove problematic. Meta-analysis of cost data from similar studies may be another method of overcoming this problem.
Our three treatment groups showed little variation in the use of specialist services for non-psychiatric problems (table ). The differences between the groups were in the use of primary care services, psychotropic drugs, and mental health facilities. To assess the impact of the psychological therapies on the use of all other healthcare services, we removed the cost of the two therapies from the analysis. At four months, there was a significant difference in direct treatment costs between the group given usual general practitioner care and those given cognitive-behaviour therapy (mean difference £163 (95% confidence interval £12 to £313); P=0.031). This suggests that in the short term the costs of providing cognitive-behaviour therapy were recouped through reduced use of other healthcare services. There were no such significant differences between general practitioner care and non-directive counselling at four months, nor between general practitioner care and either therapy at 12 months.
None of the three treatments seemed to be associated with markedly lower rates of time off work or lost production costs. The inclusion of production losses in economic evaluation is still a matter for debate, mainly because of criticisms of the valuation methods used.18
The valuation of production losses on the basis of earnings, as used in this study, ignores the fact that the existence of unemployment allows the replacement of workers who leave the labour force at little cost. Hence, attention has recently turned to the friction cost method of calculation, which attempts to account for the level of scarcity in the labour market.19
Although we did not try the friction cost method, we know that these costs would lie somewhere between the human capital valuations we reported and zero. Since the conclusions of our study were not altered by inclusion or exclusion of productivity costs from the analysis, friction cost valuations would not affect the results. Equally controversial is the method by which zero value was placed on productivity losses for patients not in paid employment. However, the more equitable analysis (using an average wage rate for all patients) did not significantly influence the results.
We excluded various cost elements from the analysis—such as the travel costs associated with specialist referrals and the costs of non-psychotropic drugs—and assumed missing data on referrals and drugs to be zero because of the considerable resources that would be required to collect such data. Thus the calculated total direct costs are probably lower than the actual costs incurred. The impact of these exclusions, however, is likely to be small. Travel costs were a relatively small proportion of total costs and differed little between the three groups at the final follow up. The number of patients with missing data was relatively low, and the sensitivity analysis provided no evidence that this was a significant influence on the results. It is unlikely that the inclusion of such costs would significantly change our results.
In conclusion, the use of psychological therapies in general practice was associated with short term benefits in the mental health of depressed patients compared with usual general practitioner care. Since our study failed to find a significant difference in total costs between the three interventions it is possible that the psychological therapies were also more cost effective than usual care in the short term. However, this finding must be considered preliminary, given the low power of the cost calculations. At 12 months, we found no significant differences between the three treatments in outcomes or total costs, and thus there was no evidence that psychological therapies were more cost effective than usual care in the long term. Given such equivalence, commissioners of services are in a position to decide on services based on factors other than outcomes and costs, such as staff and patient preferences or staff availability.
What is already known on this topic
The cost effectiveness of psychological therapies in general practice for depression is not always measured in randomised clinical trials
A small number of published trials have reported that the costs of psychological therapy and general practitioner care are similar
What this study adds
The data suggest that both brief psychological therapies may be significantly more cost effective than usual general practitioner care in the short term, as benefit was gained with no significant difference in cost
This finding must be considered preliminary, given the low power of the cost calculations
There was no evidence that psychological therapies were more cost effective than usual care in the long term