Issue:
A major pharmaceutical client sought to better understand distinct patient phenotypes that had a high likelihood of experiencing treatment initiation, treatment discontinuation, treatment switching and high total disease specific costs of their target product as well as comparator therapies among patients treated for chronic myeloid leukemia (CML).
Solution:
- We conducted a Classification and Regression Tree (CART) analysis of a sample of CML patients to identify patients that would have the greatest and lowest propensity to experience the above outcomes.
- Using claims data, we identified variables of greatest importance in predicting the outcomes described above. Both high and low probability trees were identified in the analysis. A number of follow-up stepwise logistic regression models were constructed to determine consistency in the key findings from the CART modeling.
- The findings from this analysis were disseminated internally to key internal decision makers.
