How a Little Bit of Fear Breeds Innovation

Krupa Srinivas of Owned Outcomes solving healthcare data puzzles

Co-founder of Owned Outcomes Krupa Srinivas explores the value of fear in an entrepreneur’s journey as she describes partnering with a US healthcare intermediary to solve the problem of cataloguing hospital supplies.

The US hospital supply chain is one of the largest supply chains in need of an overhaul. Market participants struggle with the lack of standardization, interoperability and transparency on quantities and prices of devices and supplies they purchase.

Without industry-wide master unique identifiers for medical-surgical (med-surg) items, hospitals cannot easily compare products by their attributes to identify cost-savings opportunities, or map product selection to the best patient outcomes. In other words, an for hospital supply products is needed.

The typical hospital spends about 30% of its annual budget on four to seven million items ranging from gloves and catheters to robotic surgery systems, totaling a staggering US$50 to US$250 million. Ordering is decentralized, based on personal preferences in both products and vendors, and at variable prices without strong ties to clinical outcomes.

In 2015, one of the largest healthcare intermediaries in the country presented us these challenges. The organization asked our team of data scientists and engineers to build the most comprehensive med-surg catalog the healthcare industry had ever seen.

This challenge spelled fear and opportunity in equal parts.

We didn’t know the domain

While we take pride in writing algorithms and building models to uncover the truth in healthcare data, we did not comprehend supply chains. We had to quickly understand the nuances in manufacturers’ product characterization, as well as how each purchaser (hospital) translated that into its own ordering systems. To build intuition, we turned to human experts. Observing and interviewing, we learned how to categorize a product, how to assign attributes to that product, and how to find values for those attributes.

For instance, SKU #L122UV from Bausch+Lomb was categorized as Intraocular Lens (also called “IOL,” one of the test categories our customer assigned us) and had the following attributes: Optic Shape, Optic Material, Haptic Material, Pieces, Optic Diameter, Overall Diameter, Power, Haptic Design, A-Constant, AC Depth and Haptic Angle. We used this type of expert knowledge to power our semantic engines.

LESSON: Watch an expert solve a problem before you build a machine to attempt it.

Other people had struggled on this journey

Incumbent human processes were cumbersome, laborious, costly, slow and demoralizing. Having worse odds than a coin toss, the results were often obsolete before they could be compiled. New product introductions, product repackaging, rebranding, bundling and misaligned incentives were to blame.

We started with the easiest problem that technology could attempt to solve. We wrote up a small list of words related to IOL: eye, ocular, lens, optic, power, etc., and had our scripts collate vocabulary (bag of words) from published literature. In a matter of hours, we had programmatically parsed through 1.1 million published articles, isolated around 2,500 of those that were related to IOL, generated about 1,200 attributes and identified roughly 3,300 possible values for them.

Turning our attention then to the process issues, we whittled down to a manageable number of attributes (<25). Finally, we used human intelligence as the adjudication engine.

LESSON: Technology solves silicon-intensive problems and humans solve judgment-intensive ones.

Tight timeline, undefined goals

Our partner assigned a 60-day clock, provided 3TB of data spanning multiple years to build our proof of concept and asked us to do our best—and then wished us luck. Unsure of the bar for acceptance, we dove in to the data to discover that it contained over 20,000 items and unstructured text descriptions from thousands of hospitals.

Escalating heart rates, sleep deprived eyes and much anxiety later, we delivered our first category 45 days later. Our attribution rate of 91% was a roaring success—we’d created a scalable technological solution in the process!

LESSON: When solving large-scale problems, chase the art of the possible instead of settling for a pre-agreed target.

If you are interested in entrepreneurship, innovation and accelerating your growth, then you will want to register for EO24/7, EO’s annual virtual-learning event. This year’s event will feature profound insights from today’s top teachers in the fields of business growth and innovationLearn more now. 

Ultimately, a little bit of fear breeds innovation

We love placing technology in service of tough problems. We celebrate the scrappiness, grit and risk tolerance. It turns our fears and applies our energy into building quick and iterating even quicker. Dreaming gets us started, creativity gets us excited, fear keeps us humble and relentless iteration gets us to the finish line.  

Krupa Srinivas is the co-founder of Owned Outcomes, a software company that enables data-driven decision making for healthcare providers and payers as they seek financial sustainability alongside clinical outcomes in patient care. She is currently serving a governor-appointed position on the Information Technology Advisory Board for the State of Nevada. Krupa joined EO Las Vegas in 2018. 

Categories: Best Practices Coaching Entrepreneurial Journey FINANCES general Technology


Leave a Comment

  • (will not be published)