1. Read and learn about AI.
Executives can take quick courses from sources such as Coursera or Udacity, and there are free, very up-to-date textbooks, such as DeepLearningBook.org, said Katherine Chou, head of product, health, at Google Research. There also are some informative lectures on YouTube, and Quora has some solid suggestions, as well.
2. Start transforming your team’s mindset now.
Organizations of the future will have a mindset that embraces digital transformation completely, said Peter Lee, corporate vice president at Microsoft Research. Healthcare providers who make the same mistakes newspaper companies made in the 1990s will face the same fate, he added. Just like a winning strategy didn’t mean posting your newspaper online, expecting AI as a drop-in replacement for human expertise would be naive. Successful digital transformation demands radical ground-up reinvention of technology, process and skills, Lee advised.
3. Educate the leadership team.
Educate executives on where these tools and technologies are being implemented today in other industries, and help them shape the organization’s vision on how they could be translated to healthcare, said Brett MacLaren, vice president of enterprise analytics at Sharp HealthCare.
4. Remember AI is a tool.
Consider what problem you want to solve and determine if AI is the right technical solution. Rich, deep and longitudinal data sets will make the possibility of machine learning being able to find meaningful patterns much more likely; for example, medical imaging, genomics, medical records and waveforms all are highly dense data sets where the machine is likely to extract insights that humans require a lot of training and time to detect, Chou said. Where there’s a consistent pattern, machine learning can often pick up on it automatically so you can multiply the power of your clinicians and researchers.
5. Ask technical partners if they’re getting up to speed.
Have they taken the free online courses or deeper curricula? Are they starting to experiment? Are they learning to use the free, open source tools that are already available and ubiquitous among machine learning practitioners, such as TensorFlow? A quick investment can pay off big time, Chou said.
6. Identify low-hanging fruit.
Identify easy gets for areas where critical thinking is minimal and tasks are ripe for process automation and AI to assist and replace certain tasks or workflows, MacLaren said.
7. Develop a roadmap.
Create a roadmap for your data and analytics infrastructure, which should be tightly aligned with the objectives of your strategic plan, to see what kind of data will be needed to drive the strategic outcomes in the plan, MacLaren said.
8. Be strategic when thinking about the cloud.
Today providers are moving data to the cloud as storage for enhanced security, but it’s equally important to think about privacy management and computing tools available in the cloud environment selected, Lee said. Leveraging healthcare data to improve medical outcomes will play a key role in the way healthcare providers operate in the future. Providers should also be evaluating their cloud platform for tools that can adhere to privacy, consent and compliance when applying AI, Lee said.
9. Technology won’t drive change, patient outcomes will.
To achieve success with machine learning in healthcare, providers should be starting with focused outcomes they want to achieve versus starting with the technology, Lee said. The machine learning approach for identifying improved operational efficiencies in a provider network may have a different path for success than one that focuses on reducing patient readmissions to the emergency room. It’s critical for providers to start by evaluating their goals and the patient outcomes they want to deliver in their organizations first, Lee said. This approach directs the right datasets and technology needed to achieve the outcomes they want.