We analysed direct and indirect health costs in German outpatients starting AM therapies for chronic disease under routine conditions. Compared to the pre-study year, costs did not differ significantly in the first year after enrolment, whereas in the second year costs were significantly reduced by 13% (416 Euro per patient).
Strengths of this study include a large patient sample, a long follow-up period, high follow-up rates, and the participation of 37% of all AM-certified physicians in Germany. Participants resembled all eligible physicians with respect to socio-demographic characteristics, and included patients resembled not included, screened patients regarding baseline characteristics. These features suggest that the study to a high degree mirrors contemporary AM use in outpatient settings. Moreover, since patients with all diagnoses were included, our study offers a comprehensive picture of AM practice. Therefore, in the present early phase of economic AM evaluation, the inclusion of all diagnoses is an advantage. On the other hand, we did not attempt to separate disease-specific costs from overall health costs. Our analysis is comprehensive, including cost domains (physician and dentist services, psychotherapy, physiotherapy, ergotherapy, medication, inpatient treatment, sick-leave compensation) amounting to 87% of healthcare expenditures of the German Statutory Health System [8] (13% not analysed: dentures, medical appliances, nursing, patient transport, and health prevention programs).
A limitation of the study is the absence of a comparison group. We do not know if in similar patients in similar settings receiving conventional or no treatment, costs would have increased, been stable, or been reduced.
Another limitation is that cost analysis was not based on direct cost measurement but on patient self-reporting of resource utilisation, which can be affected by recall bias. In this pre-post analysis, however, any systematic recall bias would probably have been conservative, making results appear less favourable. The reason is: While at study entry patients were asked about therapies and health services during the preceding 12 months, these items were thereafter asked every six months (medicine use also after three months). Since patients' recall of resource utilisation declines over time with a net tendency towards under-reporting [17], under-reporting is more likely for the 12-month pre-study period than for the shorter periods after study entry.
Dropout is unlikely to have biased the analysis of resource utilisation: For this analysis, 88% of patients were evaluable. Moreover, there is no a priori reason to assume that therapies and health services are used differently by dropouts and respondents.
Since patients were treated by AM physicians who could possibly have an interest in AM therapies having favourable outcomes, study data were largely collected by patients and not physicians. For this analysis, any bias affecting physician's documentation would not affect outcomes (resource utilisation), since these outcomes were documented by patients. Also, physicians' documentation of baseline health status (main and comorbid diagnoses) did not affect patient recruitment, since patients were enrolled regardless of diagnoses.
Major determinants of cost changes were an increased use of AM therapies (corresponding to a cost increase from the pre-study year of 377 Euro and 116 Euro per patient in the first and second year, respectively) and a reduction of hospitalisation (corresponding to a cost reduction of 310 Euro and 513 Euro, respectively), whereas other costs differed by less than 50 Euro per year. The increase of AM therapies is a consequence of the study inclusion criterion of patients starting new AM therapies. The reduction in hospitalisation was paralleled by a reduction of disease severity and improvement in quality of life [5] and may thus be related to successful therapies or spontaneous improvement. Another possible cause is frequent or long hospitalisation early in the course of disease (diagnosis, therapy initiation) followed by a normalisation of hospitalisation rates. Sensitivity analysis, however, suggests that this factor could at maximum explain 37% of the hospitalisation reduction in the second year (primary analysis: decrease by average 1.78 days = 100%, compared to the pre-study year; patients with disease duration of at least one year: decrease by 1.13 days = 63%).
Moreover, changes in health-care implementation may affect the frequency and duration of hospital treatment. However, during this study, the average number of hospital days per person-year in Germany decreased by only 0.21 days (1999–2003: 2.07→1.86 days) [18]. This reduction of 0.11 days per two years corresponds to only 6% of the observed reduction of 1.78 days per two years in our study patients. Therefore, the reduced hospitalisation in our study cannot be explained by changes in health-care implementation. A possible setting-related cause of reduced hospitalisation is the policy of AM general practitioners to provide more comprehensive patient care and avoid unnecessary referrals to secondary care [3, 19]. Study implications: The reduction of hospital treatment in this cohort following AM therapies is in accordance with other findings: In two Dutch studies [20, 21] and a British NHS audit [22] patients of AM physicians had 10%–35% less hospital days than local or national averages.
In Germany, patients may use specialist health services without referral from a primary care physician, generating additional costs. Our study is the first economic analysis of AM therapies taking into account such direct health costs generated outside the AM setting, as well as indirect costs (sick-leave compensation). In the first study year, costs of AM therapies amounted to 15% of total health costs and were largely outweighed by the reduced hospital costs; therefore, total costs were not significantly increased from the pre-study-year (as found in our previous analysis of a smaller patient sample of this study [5]), and in the second year a cost decrease of 416 Euro per patient (bootstrap 95%-CI indicating a decrease of at least 264 Euro) was found.