Health insurance expenditures for CAM were the subject of a long debate about the economics of CAM, which was characterized as much by ideological and political controversy as it was by a lack of valid data on costs and benefits in terms of patient-centred outcomes of CAM within Swiss primary care [3–5]. A further comparison with the international literature aimed at the economics of CAM therapies shows that research in this field is almost entirely restricted to specific indications and procedures, whereas studies adopting health system perspective are currently of low quality if not lacking at all. The goal of this study was therefore to provide accurate information on the nature of CAM practice and its cost to social health insurance within the system of ambulatory care in Switzerland.
It is acknowledged in this context, that reimbursed cost do not necessarily reflect actual resource cost of care. However, the amount of socialized cost is a crucial component in maintaining and improving cost-efficiency and equity of health systems. The study focused on costs arising from interpersonal and medical-technical care, considered as proxies for consultation related priorities of patients and physicians . Average costs per patient were placed at the centre of this study.
Processes of care
The results confirm other observations that patients seeking CAM treatments are younger and more often tend to be female [9, 10]. Furthermore, the study provides evidence of differences in process and management of care between CAM and COM. Certified CAM physicians treat fewer patients but spend more consultations and also more consultation time with those patients. They also treat fewer accident-related patients and perform fewer home visits than COM physicians. These differences, along with the low self declared activity in primary care of certified CAM physicians, have important implications for understanding CAM in Swiss primary care. CAM physicians care for a particular, distinctive selection of patients, and their consultation patterns may therefore not be fully in line with the formal definition of general practice/family medicine. This obvious mismatch of defined and observed practice may adversely affect decisions on resource allocation and reimbursement policy for CAM.
Extent of health insurance expenditures
Data on health care costs were obtained from the large data pool of all Swiss health insurers (santésuisse). These data are categorized into costs related to consultations, and referrals. Consultation-related costs cover almost 100% of all expenditures accounted for by the basic Swiss health insurance. Data on referral costs are restricted to prescriptions, lab analyses, and physiotherapy. The availability of cost data for other referrals – including hospitalisations, specialist treatments, and diagnostic or therapeutic procedures – is limited as only a minority of primary care providers act as gatekeepers. Hence respective data structures of insurers and health care providers do not allow a match for the remainders. Accordingly, expenditures on these procedures could not be compared in this study.
The modeling procedures of this study were based on a behavioural model in which predisposing factors such as beliefs and socio-demographic and behavioural attributes of patients indirectly influence health care use via patients' expectations and direct medical needs. Health of patients was therefore regarded as an intrinsic component of providing and consuming care within a specific treatment philosophy, i.e. particularly CAM patients have specific procedures in mind when they decide to consult a physician. It was therefore not deemed appropriate to model resource utilisation as a function of CAM or COM by additionally controlling for health status of patients, although such data would have been available for a subsample of physicians.
Regardless of these limitations, our data on total annual treatment costs per patient show substantial differences between CAM and COM. Certified CAM physicians appeared to generate the lowest and COM physicians the highest costs to social health insurance. However, modelling procedures indicate significant confounding effects of some health system and patient-related cofactors. Differences between groups decreased considerably after incorporating these factors into the analyses. Consequently, total treatment costs per patient do not significantly differ, statistically, between groups. Statistical tests and effect sizes indicated significant and particularly large effects of patient age, patient gender, and the frequency of accident-related consultations as a proxy indicator for a more somatically oriented consultation style (Table 4). Hence these three factors are important predictors of providing and consuming CAM in primary care. Our study therefore provides evidence that apparent lower treatment costs of CAM in Swiss primary care are mainly related to structural attributes of care and to a more cost effective patient population; in other words, CAM physicians treat younger patients and have a larger proportion of less costly consultations with female patients. We have to assume in this context that female patients are seeking more often specialist care for cost intensive health problems than male patients and have therefore less costly individual consultations in primary care. This phenomenon, that younger, better-educated patients who have diseases of longer duration but slightly better overall health status are more prevalent in specialist practices than in generalist practices is well know in the literature. Our finding are also in line with other studies within the same project evaluating CAM in Swiss primary care indicating that CAM patients utilize more frequent and more diverse medical services than COM patients. It therefore remains doubtful that including CAM in basic health coverage would have had a long term, cost containing effect, on overall expenditures in Swiss health care. We also doubt that an inclusion of CAM in basic health insurance will be cost neutral. Firstly because every new procedure in the catalogue of reimbursed health services will add to the overall cost of the system per se. Additionally, the literature and earlier work in the context of this project suggest that CAM is not always a substitute for orthodox care and may be an additional expense.
Structure of health insurance expenditures
The observed differences in cost structure between CAM and COM are a direct reflection of different philosophies of care in complementary and alternative, and conventional medicine. CAM physicians claim to pursue a more patient-centred and holistic approach that focuses on patient empowerment and self-healing, rather than just applying the biomedical model to cope with or to cure a specific somatic disease. The concept of patient centredness is attributed to the work of Michael Balint, who used the concept as related to illness-centredness. The concept is, however, closely related to humanistic psychology and the person-centred therapy originally developed by Carl Rogers in the 1940-ies. Although patient centredness, despite being described as clinical method by Levenstein et al. , still remains difficult to define , the quality of the interaction between patients and their physicians is considered to be a major component of patient centredness . Patient-centred care also entails partnership and a focus beyond specific disease conditions [20, 21], which are key elements of care in the philosophy of both family medicine  and CAM. Furthermore, the literature indicates that the relationship between prescribing and direct consultation costs is associated with various structural and physician-related factors [23, 24]. Prescription patterns are also related to the duration of consultations [23, 25]: longer consultation time is associated with higher patient satisfaction, better patient enablement, and fewer prescriptions . Our data on cost structure document the direct financial consequences of differences in the way care is provided by CAM and COM physicians. The data also support allegations made by CAM advocates that conventional physicians mostly rely, irrespective of supporting evidence, on medications provided directly by the pharmaceutical industry. But it is not possible, at least based on these analyses, to conclude that this particular split of health care spending is associated with better outcomes in primary care.
Limitations and strengths
CAM procedures were defined within the legal framework of the Swiss health care system that included only homeopathy, anthroposophical medicine, neural therapy and traditional Chinese herbal medicine provided by physicians trained in both conventional and complementary primary care. As part of a project to evaluate the entire system of CAM provision in primary care in Switzerland, this study was not a controlled experiment. Selection bias and systematic differences that are not related to specific treatment philosophies were therefore unavoidable. It can be assumed that the motivation differed between participating physicians, since CAM physicians were under more pressure to demonstrate effective methods – which was not the case for COM physicians. It can only be speculated that the motivation of COM physicians is more attributable to a general interest in primary care research. In a strict sense, the generalisability of our results is therefore reduced to physicians with these distinct motivations. However, a comparison of the sample population with the general population of all Swiss primary care providers indicated no difference with reference to geographic location of practices and gender of physicians, clinical data of the project including patient perceived health status with regard to other recent research in Swiss primary care showed also no difference. Therefore, and regardless of the low sampling fraction, we consider the study sample as a reasonable representation of Swiss primary care. Further problems are associated with aggregated patient attributes at the practice level that were used to adjust for effects of cofactors. The data provide therefore no possibility to track consumption patterns of individual patients within the entire health system. The most severe limitations are related to the different case mix of patients treated by CAM and COM physicians that could not be accounted for in the analysis. Other data within the project show that our sample of COM physicians treated more cardiovascular conditions entailing higher treatment costs, whereas CAM physicians treated more psychiatric conditions  that are generally associated with lower costs per consultation. While there is no indication for bias in favour of COM, it remains; however, open to which extent these findings result in biased pro CAM estimates of annual patient costs because psychiatric conditions tend to require more consultations than cardiovascular problems which may partly or fully compensate differences of consultation cost. Nevertheless, caution is advised when interpreting differences of average reimbursements per patient between CAM and COM. Further limitations apply also to the fact that only costs for medication, referred laboratory analyses and physiotherapies are included in the health insurers data pool. No data are therefore available for costs of hospitalisations, of expensive diagnostic procedures such as MRI- or CT-scans and for referrals to other physicians and specialists.
Among the strengths of the study are the complete billing data of participating physicians, and, in contrast to other research , a considerable amount of variance in cost outcomes could be explained by statistical models used in the study. Most models had enough power to identify various statistically significant differences, or their lack, in health care expenditures between CAM and COM.
Additional research within the scope of this project will provide more information on health status of patients as seen by physicians and fulfilment of patient expectations. Further analyses will also investigate the relationships between use of resources, patient satisfaction, and treatment cost.