We have seen this before, artificial neural networks, analytics, big data analytics, and now AI. These are not all the same terms, but the message has been the same over the past 10-20 years. The use of data and algorithms is going to transform medical devices and healthcare, more broadly. To some degree it has been true, but more often than not, we see MedTech companies investing in these trends and ultimately coming up short of finding the promise of a significant return on investment. These technologies are unquestionably valuable, but why do so many organizations fail to realize the value?
Any one of these initiatives is a significant undertaking; many organizations jump into the initiative without identifying the full cost or precisely recognizing how their business will translate the investment into revenue.
AI with an undetermined cost and no business model is not going to add value to your business. Assessing the full impact of any new technology is a rough exercise and delaying participation in the next “gold rush” could leave your business behind the competition. But, spending potentially millions or tens of millions of dollars without a return is likely a larger risk. R&D dollars are precious and represent the fuel for your organization’s future growth.
Some common mistake patterns are:
- Scatter shot – Attempting to apply a new technology in too many places, lacking focus.
- Experiment to victory – Informally evolving a proof of concept into a formal product that is difficult to control.
- Limited data infrastructure – In AI, data has a lifecycle, just like software code. The infrastructure to support it is required to be efficient and compliant.
- No financial model – Not having a defined connection between your AI initiative and your P&L.
- No product strategy – What is your product strategy, and does AI align to that strategy?
Going through a few basic stages will help prevent poor use of precious R&D dollars, missed market opportunities, and lost organizational focus; increasing the likelihood of AI transforming your business.
Find the right problem to solve
Start with your customers’ existing unsolved problems – you likely already know them. Which of these problems would result in cost (or time) savings for your customers? Typically, AI would be most likely to help a healthcare customer make a decision quicker and/or identify an opportunity to avoid an extra step.
Establish Success Criteria
What does your business need to declare the initiative a success? Most of the time success means a positive impact on the P&L, but are you going to get there with a higher average selling price, increased market share, or is there an option to lower your costs? Ultimately, if there is no business model, then there is no “gold rush”.
Reimbursement in healthcare is highly controlled, so if the target problem does not yield time savings, additional patient volumes (e.g. new indications), or measurably improved outcomes, it is unlikely that the clinician will pay for the solution. Focusing on problems that fit these criteria will prevent organizational fatigue.
Assess & Experiment
Where does AI logically fit into your product portfolio and existing organizational capabilities? Any product that is disconnected and/or limited in its ability to generate data is not likely a good landing spot for data driven AI. Before pursuing AI, it is critical to understand if your product has the right ingredients.
If the data source and connectivity are there, it’s time to experiment. At this stage, you have made little to no investment in infrastructure, so it is about using data to confirm whether or not the problem can be solved with AI. You may learn that the data signals are not present, or a critical piece of data is missing.
Assessing your organization’s capabilities can be difficult. It is not just “can we do it”, the assessment needs to address whether resources will be able to be applied efficiently. As with any technology trend, it is essential to be quick, so injecting external help is an opportunity to shorten learning cycles.
Plan & Commit
Getting to this stage implies that you have a strong business opportunity, a logical landing spot in your product portfolio, and evidence that AI can solve the target problem(s). It is tempting to blur experimentation with formal development but, particularly in healthcare, the ability to control development efforts is critical. Now it is time to plan the infrastructure and process.
The infrastructure can be as complex as the product itself; the data use cases must be defined in support of the developers, data scientists, as well as compliance concerns such as privacy. For example, how will you remove data from a series of models of a patient invokes their right to be forgotten? Failure to account for these types of use cases may result in a painful and expensive mitigation later.
Processes needed to support AI are going to have a similar level of intricacy to software processes. Version control, provenance tracking, and deidentification all must be controlled and repeatable.
Conclusion
The promise of AI is to transform more than just healthcare, it has the ability to transform how we work every day. However, to participate in this transformation medical device organizations must be hyper focused on opportunities that have a precise value for your business.
Full Spectrum has a suite of services to help get your AI initiative off the ground. Give us a call, we are confident that we can ensure practical AI is applied to your business.
References:
| Title | Notes | Link |
| Trend 12 is analytics
Trend 15 = better data leads to better outcomes |
https://hitconsultant.net/2014/01/31/best-healthcare-technology-infographics-of-2013/3/ | |
| Big Data trend | https://www.forbes.com/sites/johnnosta/2013/01/02/2013-the-year-of-digital-health/ | |
| Big Data = New Value | https://www.industryweek.com/technology-and-iiot/emerging-technologies/article/21959051/six-tech-trends-that-will-shape-the-pharmaceutical-industry-in-2013 | |
| IBM Watson | https://www.informationweek.com/it-sectors/9-digital-health-trends-for-2014#close-modal | |
| Great info graphic | https://www2.deloitte.com/us/en/insights/focus/tech-trends/2014/2014-tech-trends-cognitive-analytics.html | |