Healthcare: 3 Ways to Merge AI with Human Expertise

Table of Contents

Population Healthcare Data Automation

“What we’re really talking about is taking advantage of advances in things like computing and AI and merging that with the work of human beings in order to make people more productive and also get a better result. So it’s really people plus machines, as opposed to people being replaced by machines.” Dr. Kaveh Safavi, Senior Global Managing Director for Accenture Health

Analysts predict that machines will take up to half of our jobs within the next two decades. Automation is currently affecting a broad range of jobs, from bank tellers to telemarketers…and it’s starting to be used for more critical-thinking processes, such as news writing.

Automation holds perhaps the greatest promise in improving the quality of health care delivery: using health data for more precise segmentation and engagement that takes each user through highly personalized treatment plans. In fact, the “population health” model is based on this vision, but healthcare automation is hard. In addition to structural limitations (siloed, incompatible health data, etc.) there are analytic limitations (the ability to automate the development, delivery and ongoing management of meaningful patient-treatment interactions).Healthcare automation is particularly challenging because it’s not just about automating manual processes or crunching numbers, or setting triggers…it’s about human nature.

The goal of any population health or corporate wellness platform that embeds some degree of automation is the creation of actionable data that promotes user engagement, utilization, adherence and track-able, positive outcomes. Which is why healthcare automation is not automatic – once you get beyond the clinical analytics, and what processes you want to automate, you need to account for individual behavior – for instance, what mode of communication will increase the likelihood of user engagement?

If you’re contemplating a major automation project – health-related or otherwise — the odds are against success without taking the time to properly plan: establishing the vision and goal, allocating the right resources, asking the right questions, setting realistic expectations, and understanding that all automation projects are inherently works in progress needing constant evolution in order to align with the ever-changing needs and utilization habits of the user population.

Planning begins by precisely defining what to automate, taking into account where you’re most likely to get the greatest return.
Set success criteria linked to the goal – but be ready to adjust: unrealistic projections, sloppy execution and bad design can derail the project. Expect that you’ll need to make periodic modifications
Establish testing plans early: once the system has been designed it has to be tested multiple times. Set aside sufficient time to run the necessary tests
Communicate often