Limitations of the study
The sample was recruited from the in- and out-patients of three psychiatric university hospitals in Germany. A selection bias cannot be ruled out, as the in-patients were drawn from the PREDICT study , and thus had to adhere to this study's inclusion criteria and were treated according to its protocol, while out-patients did not follow this regime exactly, but were instead treated under similar conditions in one of the three university hospitals.
As we could not collect information from patients or family members who declined to participate, the question remains open as to what extent the study sample was representative of the participating centres’ patients. That said, a post-hoc comparison of the families of in- and out-patients does not suggest that there was a selection bias between the two. Additionally, the participating families’ gross family income (€3162 per month on average) differed very little from the total 2008 average for Germany (€3561), suggesting that the study participants’ socio-economic status was more or less typical for Germany . Finally, the fact that the difference between the alcoholism-related family burdens at baseline for study completers and for dropouts was very small does not suggest any systematic cause responsible for participants’ dropping out. The study was restricted to patients who were living together with spouses or families. Its findings should not be generalized to alcoholic patients who live alone, but who also receive some informal care.
Finally, it should be noted that the study did not include a control group, meaning that the findings suggest an influence, but cannot be interpreted as simply the effect of alcoholism treatment. The quasi-control group (family members of treated and non-relapsing patients) was not large enough for proper assessment. Further studies should adopt a controlled design as well as a large enough sample size to guarantee statistical power. Being part of the larger PREDICT study, the economic analyses did not include a separate statistical power analysis. Therefore, all results and conclusions must be considered as exploratory.
The lack of a control group also means that we were unable to draw conclusions about or compare our findings to the burden imposed on families of alcoholics who have not undergone treatment or who were sober for a significant period but then began drinking again due to some significant life event and had to return for treatment (‘regression to the mean’). However, our results for the subgroup for relapsing patients—although perhaps not representative of alcoholics who have never been treated—may provide some insights into this tendency towards regression to the mean.
Alcoholic patients’ actual health-service consumption levels may, in some cases, have differed from the levels indicated by the utilization data used in the study. In particular, some of the in-patients received an unknown amount of additional psychotherapeutic treatment, which was not represented in the cost estimates used in our analyses. For this reason, treatment costs may be slightly underestimated in cases that included hospital treatment.
Our study's major findings can be summarized as follows:
- The addiction-related financial burden on family members of patients suffering from alcoholism is very large, amounting to approximately 20% of family income before taxes.
- The financial burden decreases significantly when the addict receives treatment. When the patient manages to remain abstinent for a 12-month period, the financial burden may be reduced significantly more sharply than in cases of relapse. However, these savings—as is the case for all financial effects analysed in this study—cannot be considered as offsetting the costs of treatment.
- Family members’ quality of life increases significantly during the 12-month period after treatment is initiated.
- The cost per QALY) can be estimated for family members and relatives using the overall cost of treating the patients. Costs per QALY range between €5470 in cases of out-patient treatment and € 37 661 in cases of in-patient treatment.
This study represents an innovative methodological approach. Per-QALY cost estimates for family members of alcoholic patients associated to treatment costs have not yet been published in international addiction research journals. Our findings demonstrate that the financial and the psychosocial burdens on families (in terms of relatives’ quality of life) can be ameliorated by the treatment of the affected family member. Sustaining abstinence for 1 year is likely to reduce the families’ alcoholism-related expenditures to zero. The underlying assumption of our analysis is that any observed change in families’ expenditures occurs immediately following treatment initiation and persists for the entire follow-up period. In those cases where the family member did not relapse, these expenditures were reasonably considered to have been reduced to zero immediately after treatment initiation and to have stayed at zero thereafter. In the case of relapses, however, it is unclear when and for how long disorder-related expenditures rose again. In these cases, we estimated the level of expenditures recorded for the month immediately preceding their follow-up to have been the level for the entire period between treatment initiation and follow-up. It should be noted that this may slightly overestimate the actual increase, as there were probably also some periods of zero additional expenditures during the months before the relapse.
Whether our method over- or underestimates gains in quality of life, and therefore the cost per QALY, depends on exactly when and for how long the improvement in quality of life actually occurred. For example, if it were to emerge that quality of life improved linearly from baseline to follow-up, then our method of taking the difference between quality-of-life levels at baseline and follow-up would overestimate the actual increase by a factor of two. Conversely, our method may also underestimate quality-of-life increases by including only one of the patients’ family members in the analyses. If it were to come to light that the other household members also experienced a comparable increase in quality of life, then our estimates would need to be multiplied by a factor of 1.9 to account for the average household size of 2.9 persons.
The fact that there was an overall increase in quality of life suggests that the effect may be attributable to the simple fact that the subject has begun treatment—without regard for the outcome. However, the possibility that the increase in quality of life may not be attributable to treatment at all, but is due instead to an independent upwards trend in quality of life, as has been shown in other studies, cannot be ruled out [35, 36]. Similarly, regression to the mean may also be responsible for the quality-of-life increases.
Although the QALY concept is controversial , calculating the cost per QALY allowed us in this study to assess the relative value for money of addiction treatment in terms of its indirect effects on patients’ families. As the findings show, treatment yields substantial benefits additional to the effect on the patient.
What the threshold QALY value should be, below which medical treatment should be considered affordable, is a source of ongoing debate. In the United Kingdom in 2007, the National Institute for Health and Clinical Excellence (NICE) set £20 000 (approximately €29 500 at 2007 exchange rates) for one QALY gained as the upper limit above which treatment would not be considered cost-effective . Such limits are not fixed, however. In 2008, for example, NICE deemed a cost of £30 000 (approximately €37 500 at 2008 exchange rates) per QALY as acceptable for pharmaceutical treatment in terminal stages of cancer . In the context of preventive measures, thresholds as high as €40 000 have also been discussed.
Our findings lie within this range. It should be noted, however, that relatives’ QALY gains are a secondary, although additional, effect of addiction treatment. If we used the aforementioned QALY values thresholds as benchmarks, the QALY gains for alcoholics’ relatives—if they were to be confirmed in larger studies—must be added to the QALY gains generated for the patients themselves by their treatment (which we did not assess here). The sum of these two QALY gains is then what should be compared to the actual cost of treatment, which would substantially increase the cost-effectiveness ratio of alcoholism treatment.
A recent Australian study calculated the cost per QALY gained through treating alcoholism at AU$12 966, or approximately €8065 in 2005 prices, a figure that is more or less comparable to those we applied in this study [40, 41]. Such findings give an impression of the added value that may be generated when putting secondary effects on family members into the equation.
Until now, these effects have generally been neglected in health-care economic addiction research. This perspective is, however, a necessary one if we are to achieve a realistic understanding of the economic and psychosocial strain caused by alcoholism and its treatment.
Declarations of interest
HJS received financial support for oral presentations and advisory board membership from Eli Lilly, Pfizer and Janssen-Cilag, but not for this study. MF received financial support and sponsoring for oral presentations and symposia from Lundbeck, Janssen-Cilag, Novartis, Lilly, AstraZeneca, Pfizer, Servier, Bristol-Myers Squibb and GlaxoSmithKline, but not for this study.