DATA MINING BASED ON REAL WORLD DATA IN CHRONIC KIDNEY DISEASE PATIENTS NOT ON DIALYSIS- THE KEY ROLE OF EARLY HEMOGLOBIN LEVELS CONTROL
Author(s)
Frimat L1, Pau D2, Sinnasse-Raymond G2, Choukroun G3, Magrez D2
1INSERM CIC-EC CIE6, Vandoeuvre, France, 2Roche, Boulogne-Billancourt, France, 3CHU Amiens, Amiens, France
OBJECTIVES: The aim of the OCEANE non interventional study was to describe in real life conditions the management of anaemia with C.E.R.A. in patients with chronic kidney disease not on dialysis. We used data from this study to perform exploratory analysis to evaluate factors influencing haemoglobin levels. METHODS: To identify these factors, supervised and unsupervised data mining models and statistical approaches such as Random forest, hypercube analysis, Bayesian networks and mixed model for repeated measures were used. For supervised analysis, the targeted outcome measure was the haemoglobin level around 6 month of treatment, using EMA guidelines (haemoglobin level between [10-12] g/dL). As treatment patterns are very different, analyses have been performed by subgroup of patient naïve or not of ESA. Patients were followed up every 3 months during 1 year. RESULTS: CONCLUSIONS: These techniques on real world data seems to be a way to broaden the pathology, compound and practice patterns interactions. Dose adjustment around 3 months of treatment is a key factor for achieving the recommended haemoglobin target after 6 months. Our study confirms the importance of personalized anaemia management based on the patient’s profile.
Conference/Value in Health Info
2015-11, ISPOR Europe 2015, Milan, Italy
Value in Health, Vol. 18, No. 7 (November 2015)
Code
PUK4
Topic
Clinical Outcomes
Topic Subcategory
Comparative Effectiveness or Efficacy
Disease
Urinary/Kidney Disorders