Healthcare

Customer Opportunities & Challenges

Across the broad spectrum from the medical research front-end to the operational back-end of healthcare in hospitals and clinics, there is constant push to improve patient care and outcomes, at reduced cost, without violating patient privacy.  Data and analytics provide an opportunity to be more targeted and anticipative, and there are ever more sources of information available that can be exploited toward that end. For example, enhanced inpatient monitoring with video and other sensors combined with patient logs can help to predict risk of falling, stroke, bedsores, etc. so that preventative measures can be taken before an incident occurs. Medical researchers at the experimental front end can test hypotheses across diverse attributes such as phenotype, genotype, environmental and family histories to assess risk of diseases such as cancer, and determine which combinations of factors are most predictive. There is significant opportunity to improve the specificity and personalization of diagnostic models across the board.

However, some significant barriers to realizing these opportunities in the healthcare field are:

• Cost and complexity of the information systems to bring the data together, mediate formats, and provide governance and process around managing it
• Information privacy and security, complicated by the fact that the privacy issue is sometimes actually created when the data is brought together in one place
• Establishing the value of combining different sources of data for analysis; it can be expensive and time consuming to bring the data together to figure out whether or not it is worth bringing the data together
• Determining which test or diagnostic information is most useful to obtain next and combine in the context of specific diagnostic questions, and if tests are being accumulated, which test given current conditions and its cost profile would provide the most overall utility

Our Value Proposition for Healthcare

Collaborative Analytics (CA) provide predictive analysis of distributed multisource data across silos, without integrating the data. The approach overcomes some significant barriers to progress with multi-silo analysis to provide the following benefits to healthcare customers:

• Contains the complexity of the information architecture by providing a way to efficiently assess the value of combining different sources of information, and so can guide the design of a decision system with the highest performance at the lowest cost and complexity
• Mitigates privacy risks because patient data is never moved or centralized, but is combined at an abstract level in the form of obscure signaling statistics whose content cannot be interpreted in isolation
• Allows rapid exploration of factors and models across silos, and allows new silos (e.g., new kinds of tests) to be integrated incrementally into the system to establish benefit
• Quantifies the value of information provided by different sources, and can guide selection of which information to add next to best resolve ambiguity
• Provides real-time monitoring capability with streaming sources of data, and because the model learning and prediction are natively distribued, provides a natural way to deal with data evolving at different rates (addresses data turbulence)
• Sources can be mined locally and separately to obtain the highest performance with the most appropriate techniques, e.g., deep learning neural nets for image data, boosted trees for text documents, etc., and CA will compose them into a globally optimized and unified system

Collaborative Analytics lead naturally to information architectures for the healthcare space which address constraints on cost and concerns about patient privacy intrinsically and by design.