Patient-relevant endpoints from the perspective of the GKV

The benefits of statutory health insurance include the prevention, early detection and treatment of illnesses as well as aftercare. All services provided at the expense of the GKV are subject to the cost-effectiveness requirement according to §12 SGB V. The population-oriented distribution of morbidity (approximately 20% of the insured persons tie up around 80% of the services) means, conversely, that approximately 80% of the insured persons actually receive no or only very few services claim. A financing model must take this into account (false price signals) as well as the politically responsible design of an effectiveness-based performance framework. This highlights the need to examine which services should be collectively financed. Medical procedures, products and medicines must be evaluated with regard to their effect (ideally the patient-relevant benefit) in order to include them in the GKV's catalog of services or, if necessary, exclude them.

>> To assess the effectiveness of medical procedures or products, they are tested under study conditions and the resulting effect is measured. Defined endpoints are used to measure these effects. In the area of cost-benefit analysis of pharmaceuticals, the legislator has for the first time specified which aspects endpoints must cover to measure the effects (Section 35b of the German Social Code Book V). These are:
• Improvement of health status
• Shortening the duration of illness
• Extended service life
• Reduction of side effects
• Improvement of quality of life

When designing studies, it is important to carefully distinguish between the primary endpoint, for which the required number of participants was calculated, and other secondary endpoints that were defined and established before the start of the study. There is no objection to considering additional endpoints that only emerged during the course of the study if it is made clear that these are purely exploratory analyses that require confirmation in further studies. Furthermore, endpoints should always be chosen that measure a patient-relevant benefit. For example, when evaluating coronary stents, it is legitimate to determine flow rates a few weeks or months after implantation. However, this is merely a surrogate parameter. The patient-relevant endpoints here would be survival time or the absence of further deterioration in health. This could be reflected or measured, for example, by a reduced re-infarction rate or a reduced need for inpatient treatment (fewer hospital days).
A study investigating the use of beta-interferons in multiple sclerosis illustrates how misleading the above-mentioned surrogate parameters can be. The number of disease flares and the number of radiologically detectable lesions were measured, and the reduction in both measurements was used as a surrogate for improved quality of life or change in health status. However, a later retrospective analysis showed that the improvement in quality of life assumed based on the surrogate parameters could not be demonstrated with the disability scores, which measured the patients' actual abilities in everyday life (Shirani 2012). However, even if the endpoints were defined prospectively, measured in a patient-relevant, and valid manner, the respective assessment perspective is crucial for the selection of the endpoints and the measurement method.
Early benefit assessment - provision of a
Benefits paid by the statutory health insurance
There are various perspectives in the context of statutory health insurance (GKV) care. One of these is the view of GKV as a collective. This focuses on the question of whether services should be included in the universally applicable service catalog of the entire GKV. The services under examination are usually at a very early stage of their product life cycle.
In this phase, the data and findings available for evaluation are primarily from the relevant approval studies. These are usually designed as randomized controlled trials (RCTs). Such studies are initiated by the manufacturer to demonstrate the approval criteria, efficacy, and safety of their product. The manufacturer is obligated to select a study design that reflects the maximum internal evidence. Maximum internal validity is usually achieved at the cost of weak external validity.
The inclusion criteria for these studies must be selected so that the experimental and control groups are highly homogeneous and comparable. Ideally, these are subjects who, apart from an isolated disease under investigation, are relatively "healthy," i.e., exhibit little to no concomitant disease. The selected treatment protocols are meticulously adhered to in these complex, personnel-intensive, and expensive studies, and even the smallest deviations are documented.
The selection or definition of success parameters or endpoints in these studies is limited only by the resulting effort. However, each study is conceptually focused on only one primary endpoint, for which the required number of participants must be recruited. In addition, additional, so-called secondary endpoints are possible. Anything measurable can be specified in the corresponding study protocols and subsequently queried or determined. This applies in particular to parameters directly perceived by the patient, such as pain intensity, quality of life, and ability scores. Since the data is primarily collected to measure benefit, the time of collection can also be freely chosen. In particular, the collection is not tied to billing periods or invoicing. All of these characteristics make the RCT a useful, valuable, and indispensable decision-making aid at this early stage of product marketing. The validity of the conclusions reached regarding the individual parameters is extremely high, given the highly selected and homogeneous group of study participants. As a result, the internal validity of such studies is optimized, but the transferability of the results to other collectives, i.e., the external validity, is low. When selecting endpoints, the service provider will prefer those that can demonstrate maximum effectiveness. The extent to which this reflects the actual patient interests can certainly be debated in each individual case. But there are also differences between patients and medical experts regarding the weighting of endpoints. For example, the IQWiG (Institute for Quality and Quality in Health) found in a pilot project on depression: "For patients suffering from depression, it is more important that they noticeably respond to therapy than that they recover completely. For physicians who treat people with this illness, the opposite is true: they prioritize remission over response." (IQWiG 2013). To increase patient-relevant benefit, greater inclusion of the patient perspective in study planning is essential in the future.
Differentiation of the RCT from everyday care
At the time of the decision regarding the inclusion of a service in the statutory health insurance benefits catalogue, primarily data from RCTs are available, and these therefore often form the basis for the decision. Once a service can be provided to statutory health insurance policyholders at the expense of the statutory health insurance, the assessment perspective in everyday care changes. Often, it is no longer strictly defined, uniform patient groups that are treated, but rather "normal" statutory health insurance policyholders with a diverse range of comorbidities in the "normal" treatment process, which in individual cases will differ considerably from the study conditions of the RCT. For example, it has been shown that in the care of oncology patients, the group treated in the "real" world is significantly older than the subjects in most approval studies (Hutchins: 1999). An analysis of patients suffering from rheumatoid arthritis treated with biologics shows that only approximately one-third of the patients treated would have met the inclusion criteria of the respective RCTs. The patient population covered by statutory health insurance (SHI) therefore generally differs significantly in composition from the population in which efficacy and safety were demonstrated. Consequently, the efficacy found in this patient group is also lower than that determined in clinical trials. However, these effects, found under real-world conditions, are the ones that are relevant for long-term evaluation by payers.
Measuring effects in everyday care -
Non-interventional studies
Assessing effects under everyday conditions corresponds to the perspective that a health insurer must consider when selectively concluding contracts for its insured population. This is because the service being contracted must prove its worth and demonstrate its efficiency under these conditions. Measuring these effects requires non-interventional studies, i.e., measuring the form of care in routine care.
The design here is geared to measuring effects in everyday life. The inclusion criteria for care are very broad, and treatment is delivered under everyday conditions, with all the usual deviations from the intended standard treatment protocol, as well as the parallel treatments necessitated by the numerous comorbidities. The resulting inhomogeneity of the treatment group makes defining a control group challenging.
A further complication in the case of selective contracts between individual health insurance funds and manufacturers or providers is that randomization to a treatment group or control group for measuring effects is prohibited by the German Social Code (SGB V). According to the principle of equal treatment, it is simply not permissible to withhold effective treatment from a patient. Furthermore, for cost reasons, health insurance funds are often forced to base their analyses on secondary data. This data is sent by service providers to the health insurance fund for billing purposes. This primarily includes diagnosis and procedure data, but also prescription data (coded descriptions of the service), times of provision, and the respective billing rates applied. If additional, cost-increasing collection of primary data is to be avoided, the available data set is severely limited, and the collection times are also defined by the service utilization or billing. However, patient-relevant endpoints can also be defined on this severely limited data basis (Glaeske 2010). These include, for example, the number of hospital readmissions after a procedure, the number of prescriptions for medication, prescribed aids, the number of patients who have used rehabilitation measures, or the proportion of insured persons who have remained continent after a procedure.
When selecting the endpoints relevant for evaluation, health insurance companies often prioritize cost data, not least due to considerable economic competition. Here, too, the question must be asked whether this always adequately reflects the patient's interests.
Practical application of NIS in everyday cash register operations
When measuring effects under real-world conditions, the focus is less on the isolated effect of a service or prescription and more on the effect of this specific service in the overall context of all treatments for the patient. In concrete terms, this means that even a brilliantly performed surgical procedure cannot demonstrate its patient-related outcome if, for example, partial withholding or failure of services at subsequent levels of care are involved.
Such an effect is preferably measured in comparison to a suitable control group. Since, for the reasons stated above, randomizing insured persons into a treatment group and a control group is not permitted, DAK-Gesundheit uses an analysis tool based on the propensity score matching model. This involves retrospectively matching each insured person from a suitable target population who received a contractually agreed service with a mathematical/statistical twin who was equally likely to receive this service but who did not choose it. This provides a control group from the health insurance company's administrative data for evaluation that is similar to the contractual collective with regard to the parameters used to calculate the propensity score. DAK-Gesundheit pursues this approach for all contract models, but also to measure the impact of its structured treatment programs. This approach is widely used and accepted in health services research, as well as in empirical social research, and has already been defined as the "current gold standard" for the analysis of care processes in everyday population-oriented care.
An example of such a risk-adjusted outcome measurement using propensity score matching is shown in Figure 6: Risk-adjusted outcome measurement (illustrative example). Starting with a retrospective review before the program launch to demonstrate homogeneity and thus the quality of the matching (ideally along the O-line, narrow confidence interval), the outcome after the intervention (here, active care management) is analyzed over the quarters.
At the start of the program, significant additional expenditures were observed for the contract group, which could be explained by intensified targeted treatment in this very group at the start of the program. These significant additional expenditures are offset by savings in the second and third subsequent quarters (in this fictitious course), which, on the one hand, do not offset the additional expenditures and, on the other hand, are most likely due to chance (confidence interval above and below the O line).
When interpreting the results, it is always important to consider whether differences between the contract and control groups are due to the treatment provided or whether differences are caused by unaccounted confounders that are not evenly distributed between the groups. This may require further morbidity analyses or individual service analyses (DRG, ICD, OPS, EBM, etc.) to identify skewedly distributed influencing factors.
Within DAK-Gesundheit, this approach has made it possible to quickly evaluate even complex contractual constructs in the real world of healthcare and, without randomizing the provision of benefits, to assess the causal effects of new or alternative forms of care based on their impact on the insured and the health insurance fund. Currently, retrospective propensity score matching represents the "gold standard" for DAK-Gesundheit in the evaluation of new healthcare offerings, and this approach must be continually improved until new analytical approaches necessitate a different approach.
Summary
When defining patient-relevant endpoints, the timing of their determination in relation to the product life cycle must be considered. Early on, endpoints that allow the manufacturer or provider to demonstrate the maximum effectiveness of their service will predominate. Later, when it comes to determining the effect of the service under everyday conditions, the choice is often limited by the free availability of administrative health insurance data and the focus of health insurance funds on cost-effectiveness.
In both cases, it remains to be stated that in the majority of cases, the primary focus is not on the patient, but rather others define what is patient-relevant. In this regard, the patient should be better involved in the future. A challenge from the perspective of health insurance will be to resolve the tension between patient preference (in the sense of "sick benefit recipients") and insured preference (in the sense of "healthy financiers" of the services) in a generally acceptable manner.