Machine Learning as a Surrogate for Efficient Care - Valor HealthTech Corner

For quite some time, we at Valor, have you been bringing you game-changing candidates and great career opportunities. In 2016, we began our efforts to support that work with news, tips, blog perspectives and updates on M&A activity around the industry. Today, we unveil our bi-weekly Healthcare blog feature, where we’ll bring you the latest and greatest ponderings and ideas, straight from the minds of the top Health IT/Software Execs in the industry.

Our first featured blogger is Steve Curd https://www.linkedin.com/in/stevecurd/ , one of most successful and notable CEO’s our industry has known. Steve spent time as Technology and Information leader at organizations like CIGNA and United Health, before becoming COO at Healtheon/WebMD and leading the company through its IPO and 13 acquisition integrations. In recent years, Steve has been COO and CEO at leading innovators in and around the Healthcare Analytics and Precision Medicine world, including NantHealth, and most recently, Wanda/Oncoverse, where he commercialized and launched their Oncology Collaboration Platform.

Foreword by Drew Ramsey

Machine Learning as a Surrogate for Efficient Care

Among all advanced industrialized nations, the US stands alone. We have no uniform health system, no universal coverage, and have only recently legislated coverage for nearly all our citizens. Although with the new administration’s mantra of “repeal and replace,” even this is up in the air, leaving our Country’s amorphous hybrid healthcare system again in limbo. Perhaps this helps explain why the US spends over 16% of our GDP on healthcare ($8,700 per person), as compared to the industrialized nation average of 9%. Since decades of legislation have yet to close this $750 billion efficiency gap and with no systemic solution in sight, can machine learning help us to crack the efficiency code? Since most health information is now in digital form, and with significant recent advances in health analytics, our growing understanding of health and healthcare through the data will result in organic improvements in efficiency and in overall outcomes.  Let’s consider a few bellwether examples.

The Power of Health Analytics

At Wanda, we leveraged a decade of machine learning research from UCLA related to observing individuals with congestive heart failure, an extraordinarily expensive disease which has become the number one killer. By training the algorithm with minimal data from each of 1,500 individuals – blood pressure, heart rate, weight, and simple symptom information – a formal clinical trial demonstrated an ability to predict 90% of all hospitalizations days prior to the adverse event. This enables an unprecedented ability to focus resources just on those people for whom an intervention is highly likely to eliminate that expensive hospitalization.

Here are a few other examples where machine learning is meaningfully improving efficiency and outcomes:

  • Precision medicine bio-marker discovery: as genomic and proteomic analysis uncovers the mutations and genetic predisposition for disease by sifting through patterns encoded within the 3 billion base pairs of the human genome, machine learning is an essential tool to accelerate discovery and correlation across massive data sets
  • Real-time patient risk stratification: care teams must continue to do more with less; machine learning can help focus expensive resources just on those patients most likely to be headed for an exacerbation. It has been shown that even simple daily activity patterns can become strong predictors of emerging health issues
  • Effectively dealing with sparse, missing and “spammy” data: physiologic data collection is fraught with invalid data points for a variety of reasons. Machine learning can reduce false positives, and fill in the blanks when necessary
  • Automated assessment of medical imaging: pattern and anomaly detection are ideal applications for machine learning techniques, to amplify diagnostic accuracy
  • Efficient, scalable parsing of clinical data: much health care data exists in the form of human language text; machine learning can help extract actionable insights from natural language to better inform the clinical team. Likewise, a droid which classifies and mines clinical literature can enable physicians to remain current and informed

There are few silver bullets in healthcare, however there has never been a time where analytics can have such a profound impact, with new examples emerging daily. By working smarter as opposed to harder, and armed by actionable insight provided through advanced analytics, we are learning that it is indeed possible to improve efficiency and diagnostic accuracy even in an uncoordinated and chaotic health “system.”


Steve Curd

Steve's career-long passion has been to accelerate the implementation of emerging science to improve health and patient outcomes while eliminating inefficiencies and avoidable costs. He is named as inventor on patents related to advanced image processing techniques, applying commercial aviation flight deck safety technologies in the operating room, and an innovative real-time wireless care team collaboration platform. Previously he served as the CIO of UnitedHealthCare, as COO of Healtheon / WebMD during its period of extraordinary growth, and has been the chief executive of innovative healthcare technology companies in the areas of electronic health records, surgical information systems, care team coordination, and personalized medicine focused on managing cancer. Having participated in 15 M&A transitions and an IPO, he has experience and strong competence in business integration, as well as driving organic growth.