Hospital and health system CIOs and innovation leaders are often faced with the tough question of build versus buy; when a problem arises, is the correct strategy to build internally or buy an existing platform.
A panel at the Becker's Hospital Review 5th Annual Health IT + Revenue Cycle Conference in Chicago in October 2019 briefly addressed the issue. The panel included Liz Popwell, CSO of Ascension Michigan; Ben Patel, CIO of Cone Health in Greensboro, N.C.; Viraj Patwardhan, vice president of digital design and consumer experience at Thomas Jefferson University and Jefferson Health in Philadelphia; and Paul Black, CEO of Allscripts.
Here are five key takeaways from the panel:
1. Consider what the problem is, and whether the solution is a commodity or a transformational technology that you cannot buy. If the issue is complex and would require your data science team to build algorithms that already exist, it may make more sense to buy and iterate from there.
2. Where you are in your transformational journey matters. If you want to differentiate yourself from others, build may be the way to go.
3. Look around and see what existing platforms offer. If you can't build something better with your current team that will solve the problem faster, then buy is the way to go. But if no technology exists that will solve the problem, then build makes the most sense.
4. Once you decide to either build or buy, the next step is to think about how the system will run the platform and continue iterating on that run. Conduct onsite sprints to help make those decisions and understand how you will conduct future iterations to avoid leaving the platform stagnant.
5. If you decide to partner with outside companies on projects, select partners carefully.
While members of the panel have gone the partnership route, CIO of Cedars-Sinai in Los Angeles Darren Dworkin said in an interview with Becker's that his system preferred to build and has been successful in that.
"We have operationalized over a dozen machine learning models with clinical and operational impact; we have automated machine learning for greater than 700,000 to 800,000 'what if' scenarios on inpatient days; and we are creating new inpatient dashboards at two per month," he said. "We continue to find all sorts of amazing things we can do and add them to our repository of more than 1.5 billion livestream datapoints so we can inform our machine learning models."
While the health system doesn't have as many data scientists as Google, it does have the right number to solve real-world challenges associated with data flow, admissions, follow-ups and gaps in care. "A lot of that low-hanging or medium hanging fruit meets our needs without going to the most advanced corners of data science," he said.