gms | German Medical Science

Background and aim: There are yearly "goal agreements" on regional levels between Statutory Health Insurances and the National Association of Statutory Health Insurance Accredited Physicians orientated on principals given by national law. Additionally or sometimes excluded are checking’s on total budget level (so called “Richtgrößenprüfungen”) dependent on the specialization of the physicians.

Material and method: We use cluster methods with Mathematica 6.0 to determine specializations on the basis of ATC codes on different levels using prescription data of one year. The results are compared with historically determined groups of specialization. If given groups (in some context) are fixed, one has to determine the individual position of physicians. So one can transfer a grouping with special properties to other regions. We propose a new maximum method for he mentioned decision process. It shell not depend on the scale of a special group. It is based on the amount of an ATC fraction measured by costs or by daily drug doses for he considered specialization in comparison with others.

Results: Cluster methods offer the possibility of further differentiation without the danger of getting to small groups. Further on they show similarities of groups, as Urologists and Gynecologists. There are difficulties to separate Surgeons from a lot of other groups on this level. The determination of Neurological, Oncologists, Gastroenterologists, Rheumatics and Nephrologists can be well done; separation of Cardiologists from General Practitioners is more difficult by the used procedure. One can imagine, that Ophthalmologists with an ATC code S01 (ophthalmologic) with prescription fraction value of 96% can easily determined (but mention, that this is the value of the group, the individual values of the physicians may differ much so it is not clear at all if the positions are not such extremely). The top position G03 (sex hormones and modulators of the genital system) of Gynecologists is 38% in comparison with 24% of Urologists, their top position is L02 (endocrine therapy) with 49%. The top position of Neurological is R03 (pneumologicals) with 74% of costs. Cardiologists have a top position of only 18% in costs: B01 (antitrombotic agents). The Oncologists have a top position of 44% drugs without ATC code (special preparations). General Practitioners have top positions of 10% for both A10 (drugs used in diabetics) and C09 (agents acting on the renin-angiotensin system). The level of consideration the ATC code is important for the classification of groups of physicians (in the examples we used the ATC 3 level). For some groups there are separation problems on a global level. Then one can use hierarchical or local methods that shall mean that one has to separate only subgroups of physicians and ATC codes. The determination of subgroups shall be done by algorithms based on ATC-information and not by individual decisions using additional and personal determined information.

Conclusions: It is possible to group the Physicians using ATC codes for the prescriptions. As a first step one can use global procedures. For some details one has to use local procedures and the ATC code in deeper positions. The ATC code in comparison with the prescription behavior of physicians and groups of physicians lead to useful benchmark results. In order to ensure therapeutically and pharmacologically high quality of prescriptions while providing patients with drugs under budget conditions one has to use classifications as a result of a combination of pharmacological data bank information, market information and actual therapeutic and pharmacological standards.