Analysis of diabetic patients through their examination history

Abstract

The analysis of medical data is a challenging task for health care systems since a huge amount of interesting knowledge can be automatically mined to effectively support both physicians and health care organizations. This paper proposes a data analysis framework based on a multiple-level clustering technique to identify the examination pathways commonly followed by patients with a given disease. This knowledge can support health care organizations in evaluating the medical treatments usually adopted, and thus the incurred costs. The proposed multiple-level strategy allows clustering patient examination datasets with a variable distribution. To measure the relevance of specific examinations for a given disease complication, patient examination data has been represented in the Vector Space Model using the TF-IDF method. As a case study, the proposed approach has been applied to the diabetic care scenario. The experimental validation, performed on a real collection of diabetic patients, demonstrates the effectiveness of the approach in identifying groups of patients with a similar examination history and increasing severity in diabetes complications.

MeTA

“…Both drugs are likely to be prescribed to patients with cardiovascular diseases in conjunction with specific examinations. Rosuvastatin is mainly used to treat patients with primary hypercholesterolemia, whereas ramipril (code: C09AA05) is commonly prescribed to reduce blood pressure [ATC 2013]. The confidence values of rules (2) and (3) indicate that approximately 11% of middle-aged patients actually take the specific drugs.…”

“…The drug generalization hierarchy contains as leaves the drugs encoded using the fifth level of the ATC classification system [ATC 2013]. Drugs are aggregated into the corresponding drug category according to the first level of the standard ATC classification system.…”

“…It contains 26 examinations clustered into seven examination categories. The selected examination categories are based on the expert-driven classification reported in Antonelli et al [2013].…”

“…The diagnostic and therapeutic procedures were defined using the ICD 9-CM (International Classification of Diseases, 9th revision, Clinical Modification) [ICD-9-CM 2011]. Drugs were identified by the pharmaceutical coding system adopted by the Anatomical Therapeutic Chemical (ATC) Classification System [ATC 2013]. …”

“…Item generalization is driven by a taxonomy, which consists of a set of is-a hierarchies built over the analyzed data. For example, drugs can be generalized based on the addressed pathology [ATC 2013], whereas examinations are clustered based on the focused area (e.g., liver or cardiovascular system). Aggregating items into higher level concepts (e.g., examinations into the corresponding category) prevents the discarding of potentially useful knowledge and thus counteracts the issue of data sparseness.…”