spsMOSAICS has been developed to estimate latent cohort heterogeneity whilst also accounting for the impact of informative censoring from competing risks in survival estimates. Applicable to both time-to-event or ordinal class type data, spsMOSAICS estimates relative frailties, baseline hazard rates, covariate associations and hazard ratios, 95% confidence intervals and p-values for each modelled risk and each latent class. Benefitting from both Bayesian regression and Bayesian model selection, the pipeline optimally estimates a cohort’s substructure with little risk of overfitting. With the addition of both retrospective and prospective subgroup assignment, spsMOSAICS is ideally suited to robustly identifying differing disease trajectories and, potentially, those patients most likely to benefit from a given treatment.

Other features include:

• Determination of the number of statistically significant patient subgroups in a clinical data set

• The sizes and detailed characteristics of each subgroup, including differences in frailties, risk associations, or base hazard rates

• Probabilities for individual patients to belong to each detected subgroup

• Whether and how subgroup membership can be predicted from the data

• Simultaneous analysis of multiple risks, enabling survival estimates to be cleaned of informative censoring

• Fully automated analysis reports, including survival curves