(21) Applying Sequence Clustering Methods to Characterize Healthcare Pathways of Patients at Different Prostate Cancer Stages in The French Nationwide Healthcare Database (SNDS)

Publication Number: 695

Background: There is a great heterogeneity of prostate cancer patient journeys for a same stage of prostate cancer, which makes their evaluation complex using descriptive statistics. Unsupervised machine learning, has the potential to reveal patterns within heterogeneous data. However, impact of such methods in real-world studies is not yet clear.

Objectives: To illustrate how clustering and visualization of healthcare pathways can enhance the characterization of patients with prostate cancer, at all disease stages.

Methods: Patients with prostate cancer in 2014 were identified in SNDS and their healthcare data were extracted with up to 5 years of history and 4 of follow-up. Fifty-one specific healthcare encounters constitutive of prostate cancer management were synthetized into 4 macro-variables using clustering of variables approach. Values of these macro-variables over patient follow-ups constituted healthcare pathways. Optimal matching using TRATE substitution method was applied to calculate distances between pathways. Partitioning around medoids algorithm was then used to define consistent groups of similar pathways across four exclusive cohorts of incident prostate cancer patients: hormone-sensitive (HSPC), metastatic hormone-sensitive (mHSPC), castration-resistant (CRPC), and metastatic castration-resistant (mCRPC). Index plots were used to represent pathway clusters.

Results: The repartition of macro-variables values – surveillance, local treatment, androgenic deprivation, and advanced treatment – appeared to be consistent with HSPC, mHSPC, CRPC and mCRPC status. Two to five clusters of healthcare pathways were observed in each of the different cohorts, corresponding for most of them to relevant clinical patterns. For instance, clustering allowed to distinguish patients undergoing local treatment with or without androgen deprivation therapy in HSPC; patients with rapid and slow disease progression in CRPC; patients with curative, non-curative treatment, or treatment-free interval in mHSPC.

Conclusions: Visualization methods combined to clustering approach enabled the identification of clinically relevant patterns of prostate cancer management. This experience also highlighted potential improvement to leverage the impact of results. Characterization of these care pathways is an essential element for the robust assessment of healthcare technologies effectiveness.