The method of conjoint analysis (CA) is in principle suitable for finding out what preferences patients have when it comes to therapy goals. However, in order to use them widely in cost-benefit assessments, some questions, particularly methodological ones, still need to be clarified. This is the result of a recently published pilot project (almost 2.5 years since the commissioning) by the Institute for Quality and Efficiency in Health Care (IQWiG). After the Analytic Hierarchy Process (AHP), CA is the second method that the institute has had external experts - Prof. Jizermann (AHP) and Prof. Axel Mühlbacher (CA) - tested.
In two pilot projects and in two different indications, IQWiG tested the two most widely used methods internationally that can be used to determine patient preferences. The IQWiG had already presented the report on the Analytic Hierarchy Process (AHP), presented by Prof. Maarten IJzerman (University of Twente) in June 2013, and now the long-awaited working paper on CA is also available, with Prof. Axel Mühlbacher from the Neubrandenburg University had been commissioned. This had submitted its draft to IQWiG around 1.5 years ago, but changes were repeatedly requested.
Of course, IQWiG did not award the two pilot contracts without reason. But he knows full well that his efficiency limit concept, which was presented unsolicited at the beginning of the AMNOG process, has reached limits that have so far been insoluble. Until now, efficiency limits could only be mapped for an aggregated endpoint or for a single target criterion such as mortality (mortality), morbidity (symptoms and complaints) or quality of life. What makes matters worse is that data is often only available for individual target criteria. What was previously not possible was the creation of efficiency limits for various patient-relevant endpoints and their merging into an overall assessment, what experts call “aggregating”. To do this, however, the individual results must be weighted.
This is also done in the AMNOG process, but so far it has been purely paternalistic or based on eminence by the members of the G-BA's Medicines Subcommittee, who each make value decisions for the entire group of GKV insured people individually without any algorithm. But this could also be done differently by taking into account the preferences of patients. Because they are, so to speak, the “end consumers” of medical measures. That's why they have been included in assessments of benefits and costs internationally for a long time, just not yet in Germany.
In Germany, the entire AMNOG process is transparent; almost everything, except for “secret module 5” of each drug dossier, is published online. The only exception is the value decision made by the “benches” of the G-BA in the G-BA Subcommittee on Medicines, which has so far been anything but systematic, transparent and comprehensible. This could change with quantitative approaches such as CA or AHP if they were recognized and used by the G-BA.
Both those who are sick and those treating them are interviewed
Using a choice-based variant of CA (discrete choice experiment), the commissioned scientists working with Mühlbacher interviewed both people suffering from chronic hepatitis C (HCV) and experts involved in their care - first in separate focus groups, then using questionnaires. A total of 326 patients and 21 doctors took part.
The questions addressed various dimensions of benefit and harm: effectiveness (e.g. freedom from viruses), avoidance of side effects (e.g. gastrointestinal problems) as well as effort (e.g. frequency of injections) and duration of therapy.
The participants were asked to decide 18 times between two fictitious therapy alternatives, which were made up of different therapy properties (attributes) and differentiated according to characteristics (levels).
The characteristics of the total of seven attributes were varied many times and combined again and again: While in such a scenario therapy A lasted 48 weeks and therapy B 24 weeks, the probability of gastrointestinal problems was 35% (A) and 25% (B). Therapy A in the next scenario with 12 weeks only took half as long as B (24 weeks), but this time the risk of gastrointestinal side effects with 25% was significantly lower for A than for B with 45%.
Patient preferences versus expert judgment
By analyzing the results of all of these choices using logistic regression models, it is possible to derive the relative importance (weighting) of the individual therapy attributes. For example, it can be calculated how much higher the chance of recovery (virus-free) must be for patients to accept a certain higher risk of more or more severe side effects.
If you compare the evaluation for patients and doctors, it becomes clear that the ranking of the therapy goals is largely the same. However, there are differences in the strength of the weighting: “sustained virus freedom” is in first place for both groups. But when it comes to choosing a therapy, being virus-free is even more important for doctors than for patients.
Patients can manage CA
After this pilot project, IQWiG judges CA – like the AHP before it – to be a fundamentally suitable and manageable method. “Patients can cope with it and the procedure delivers useful results. When it comes to weighting endpoints, CA could be used,” says Andreas Gerber-Grote, head of the health economics department at IQWiG.
While the CA works with multidimensional scenarios, the AHP makes pairwise comparisons (e.g. duration of therapy versus side effects). “For CA, you could say that the process is closer to the decision-making situations that occur in the real world. The attributes are always evaluated in a bundle,” explains Gerber-Grote. However, the complexity of the decision increases with the number of attributes. If - as with the AHP - you only look at two attributes, it is more transparent for the respondents.
Before widespread use: clarify methodological questions
“Both pilot projects were successful. But before we can widely use CA or AHP in assessment, there are still a whole series of challenges, and they are primarily of a methodological nature,” points out the deputy director of the institute, Stefan Lange. It would have to be clarified in advance who should be surveyed: patients, doctors or – as in Great Britain – a sample of the general population? In any case, there would have to be an agreement on how representative the selection of respondents must be. And you would have to determine what level of accuracy is required, i.e. how reliable the results would have to be. “The higher the demands here, the more people you have to interview, and the greater the effort. Unlike clinical studies, there are currently no standards for how to plan the number of cases,” Lange continued.
