High Noise Healthcare Data Environments Are a Prime Candidate for Advanced Technologies

By: Dan Fisher, Principal, IPM.ai

For a number of years, the application of artificial intelligence in healthcare has held great promise. One of most obvious use cases is to identify patient populations that represent the proverbial needle in a haystack, such as undiagnosed rare disease patients, patients soon to fail a line of therapy, patients most at risk for non-adherence or patients who fit the profile for a clinical trial. However, as with many new technologies, the realization of this promise has been disappointingly slow.

The challenges that have hindered widespread adoption include: notoriously dirty data with duplicate entries, missing fields, incomplete coding, etc.; the complex interactions among multiple conditions that are also temporally sensitive; the identification and ranking of contributing conditions and the translation of results into real world business objectives. IPM.ai developed the industry’s first system of insight that solves these challenges, and delivers patient insights that are better, faster and in the format that answers your most critical questions aligned with your product lifecycle. Our system of insight reflects a unique combination of people, process and technologies designed to to help life sciences companies identify, engage and activate patients who could benefit from new therapies and modalities of care

Neural Nets and Deep Learning – These leading edge techniques uncover the deep underlying connections between any number of medical conditions, arrayed across any time span. It is these previously-unknown interactions that are most predictive in applications such as discovering undiagnosed patients.

Natural Language Processing – NLP helps solve the problem of incomplete, missing or unspecified coding. For example, it can let us “see” ICD-10 150.8, Unspecified Heart Failure when emphasizing the relationship extraction between comorbid or concomitant codes when correctly re-mapped and indexed for level of severity, thus greatly enhancing the specificity of the model.   

Ensemble of Algorithms – Models that are carefully engineered to be as independent as possible provide an average result more precise than any individual result. This approach identifies the key features that drive the model and their relative importance, providing insights on the most predictive triggers that can inform other business decisions around a therapy.

Non-Parametric Models Tuned to Business Objective – Traditional models measure the propensity of a population having the desired characteristic, with populations scored on a scale of 0 to 1. However, stating a population is 0.9 vs 0.8 provides no insight into the tradeoffs a company faces when trying to pick the correct population to target. IPM.ai’s models are always tuned to answer a business question, such as “What is the correct population to target that results in a number of patients where a 20% conversion rate would result in a program ROI of 100%?” 

To learn more about the how IPM.ai’s system of insight can help find, engage and activate your ideal patients and their healthcare ecosystems, contact us at hello@ipm.ai

About the Author

Dan Fisher

Principal, IPM.ai 

As Principal for IPM.ai, Dan leads a team that utilizes machine learning, artificial intelligence and advanced analytics to deliver valuable insights that guide and accelerate the clinical and commercial decisions of life sciences companies. With a focus on specialty markets, Dan’s deep expertise in rare disease and oncology disease states helps biopharma clients better understand and more effectively uncover ideal patients and their health care providers. Prior to joining IPM.ai, Dan led commercial operations and clinical analytics projects for ZS Associates. He holds a Master of Business Administration (MBA) from Vanderbilt University.