A flexible mixed-data model applied to claims data for post-market surveillance of prescription drug safety behavior.

We develop a new modeling framework for jointly modeling first prescription times and the presence of risk-mitigating behavior for prescription drugs using real-world data. We are interested in active surveillance of clinical quality improvement programs, especially for drugs which enter the market under an FDA-mandated Risk Evaluation and Mitigation Strategy (REMS). Our modeling framework attempts to jointly model two important aspects of prescribing, the time between a drug's initial marketing and a patient's first prescription of that drug, and the presence of risk-mitigating behavior at the first prescription. First prescription times can be flexibly modeled as a mixture of component distributions to accommodate different subpopulations and allow the proportion of prescriptions that exhibit risk-mitigating behavior to change for each component. Risk-mitigating behavior is defined in the context of each drug. We develop a joint model using a mixture of positive unimodal distributions to model first prescription times, and a logistic regression model conditioned on component membership to model the presence of risk-mitigating behavior. We apply our model to two recently approved extended release/long-acting (ER/LA) opioids, which have an FDA-approved blueprint for best prescribing practices to inform our definition of risk-mitigating behavior. We also apply our methods to simulated data to evaluate their performance under various conditions such as clustering.

Cluster activation mapping with application to computed tomography scans of the lung.

Purpose: An open question in deep clustering is how to explain what in the image is driving the cluster assignments. This is especially important for applications in medical imaging when the derived cluster assignments may inform decision-making or create new disease subtypes. We develop cluster activation mapping (CLAM), which is methodology to create localization maps highlighting the image regions important for cluster assignment. Approach: Our approach uses a linear combination of the activation channels from the last layer of the encoder within a pretrained autoencoder. The activation channels are weighted by a channelwise confidence measure, which is a modification of score-CAM. Results: Our approach performs well under medical imaging-based simulation experiments, when the image clusters differ based on size, location, and intensity of abnormalities. Under simulation, the cluster assignments were predicted with 100% accuracy when the number of clusters was set at the true value. In addition, applied to computed tomography scans from a sarcoidosis population, CLAM identified two subtypes of sarcoidosis based purely on CT scan presentation, which were significantly associated with pulmonary function tests and visual assessment scores, such as ground-glass, fibrosis, and honeycombing. Conclusions: CLAM is a transparent methodology for identifying explainable groupings of medical imaging data. As deep learning networks are often criticized and not trusted due to their lack of interpretability, our contribution of CLAM to deep clustering architectures is critical to our understanding of cluster assignments, which can ultimately lead to new subtypes of diseases.