Abstract
Popper's point of view that a hypothesis can only be proven wrong by finding one counter-instance and that confirmation is unscientific is widely acknowledged. In medicine, however, we can only optimise the probability that a therapy works. For this we have to search for counter-instances for our existing ideas. The inclusion of a medicine in a repertory symptom-rubric when the symptom is seen in a cured case is wrong. This biased idea could be replaced by the Bayesian likelihood ratio (LR).
LR can be assessed by prognostic factor research (PFR) using data collection by homeopathic practitioners. In the past 25 years, several PFR projects have been performed. It appeared that practitioners correct for biased entries in the repertory by finding practical solutions. Such solutions are keynote prescribing, selecting sub-rubrics that semantically confirm the preferred medicine and lowering threshold values for symptoms that confirm the preferred medicine. There is also variation between practitioners in confirming medicine selections by repertorisation. This way the bias of the repertory results in biased data collection. Statistical analysis of biased data results in biased conclusions.
Research protocols should deal with bias in data collection and statistical analysis. Practitioners and researchers involved in data collection should be thoroughly trained. Incoming data should be monitored from the start resulting in adequate feedback to observers.
Keywords
data collection - bias - repertory - PFR - homeopathy
