Schmerztypische Diagnosemuster in Routinedaten – Identifikation mittels Classification and Regression Trees (CART)

Abstract

The identification of beneficiaries with persistent, recurrent or chronic pain in claims data by means of individual diagnoses or analgesic prescription is not sufficient and reliable. By using CLASSIFICATION AND REGRESSION TREES (CART) it was possible to identify specific diagnosis patterns for patients suffering from pain. Diagnosis patterns are considered as specific if they occur more frequently among beneficiaries with at least two opioid prescriptions within one year compared with beneficiaries who did not receive any analgesic therapy. Diagnosis and prescription data of 2006 were provided by the German sickness fund DAK. As a result, 65 diagnosis patterns occurred more frequently among beneficiaries treated with opioids than among the control group. These 65 patterns can be classified as follows: cancer-related pain (4), specific back pain/osteoporosis (8), spine-related pain (6), arthritis-related pain/rheumatoid arthritis (22), pain after traumatic fractures (5), pain in multimorbid, dependent patients (3), neuropathic pain (7), headache (5), non-specific back pain (5). The derived diagnosis patterns showed high predictive values (sensitivity: 78%, specificity: 66%) and are suitable for the identification of beneficiaries suffering from pain - the first step towards health services research in pain-based on claims data.