Data mining for on-line support of general practice

Statistical relationships among symptoms, diagnoses and treatments can be inferred from large data-bases of health records. We investigate how these "empirical norms" can be utilized to improve the efficiency and quality assurance capability of on-line systems in General Practice medicine. Using a survey-based database of General Practice records, we assess hotlists (case sensitive menus) of diagnoses to speed data entry. We also explore norm violation as an indicator of poor quality in practice or data recording. We find that efficient hotlists of diagnoses can be generated from symptoms. Also, we find that less frequently used hypertension treatments are assessed as of a lower quality than more common ones. The results support the hypothesis that empirical norms have a role to play in improved General Practice systems.