Table 1 Summary of the five steps of network construction and the informed decisions made by the expert panel in the ACD project
Step | Guiding study objectives | Informed decisions and explanations | Technical operationalization in the dataset | |
|---|---|---|---|---|
1. Units for network construction | A | We chose individual ambulatory physicians (and not practices) in order to identify all physicians involved in the patients' treatment Physicians from the same practice may also be interested in patients' pathways within their practice but also in other practices | Vertices are identified through the unique physician ID | |
2. Health care providers | B | Exclusion of physicians of predefined specializations and authorized physicians Reasons for exclusion: - Specialized in care of patients who are not included (e.g., children and adolescent specialists) - Specialists with no direct contact with patients do not play an active role but are mainly conducting contracted services (e.g., laboratory medicine or pathology) - Specialists providing mainly contracted services treat a large number of patients and would affect network construction (e.g., radiologists) | Excluded physician specializations: Children and adolescent specialists Laboratory medicine Microbiology Oral and maxillofacial surgery Pathology Radiology Radiation therapy Transfusion medicine | |
3. Patient population | C | Patients with one of 14 selected ambulatory care-sensitive conditions. Conditions are chronic or acute, have a high prevalence, and need continuous and/or interdisciplinary treatment We assigned patients to every physician who they consulted in a presumably "face-to-face" consultation and excluded selected types of billed services | Detailed information on operationalization is available in Table 1 Excluded types of billed services: Referral or billed service for a laboratory service as contract service; Request for laboratory service in a laboratory community Medical emergency service; Replacement during holiday or illness; Emergency; Emergency service with taxi; Rescue service; Central emergency service | |
4. Network identification | B and D | The minimum number of shared patients to define a connection between two physicians was set to 20 to ensure data protection. The 20 patients needed to account for at least 5% or more of the total patient population for at least one of the two physicians to ensure relevance The multilevel algorithm of the package igraph is used in an iterative manner to obtain networks of a size between 20 and 120 physicians | Â | |
5. Patient allocation | C and D | We defined a usual provider network for each patient who is mainly responsible for his or her care | We allocated patients to the network in which they had the most frequent physician consultations (in days). The number of consultations to one network had to exceed 50% of all physician consultations per patient | |