The future of artificial intelligence in health care

The benefits of AI in health care

From patient self-service to chat bots, computer-aided detection (CAD) systems for diagnosis, and image data analysis to identify candidate molecules in drug discovery, AI is already at work increasing convenience and efficiency, reducing costs and errors, and generally making it easier for more patients to receive the health care they need.

While NLP and ML are already being used in health care, they will become increasingly important for their potential to:

1. Improve provider and clinician productivity and quality of care 
2. Enhance patient engagement in their own care and streamline patient access to care 
3. Accelerate the speed and reduce the cost to develop new pharmaceutical treatments 
4. Personalize medical treatments by leveraging analytics to mine significant, previously untapped stores of non-codified clinical data

While each AI technology can contribute significant value alone, the larger potential lies in the synergies generated by using them together across the entire patient journey, from diagnoses, to treatment, to ongoing health maintenance.

Implementing AI in health care: Lessons learned

Based on our work with clients on applications of AI in health care, we offer these insights:

1. Factor in extra time and cost for early adoption: even relatively small projects require additional time and effort up front performing business case validations and proof of concept. 
2. Reduce cost and complexity by leveraging open-source technologies and limiting customization. 
3. Build solutions for average transaction length and volumes but with capacity for longer transactions and peak volumes.
4. Involve personnel with a combination of technology and health care backgrounds who have a clearer understanding of end users’ needs and preferences, as well as options for technology solutions. 
5. Carefully select the data used to “train” any AI/ML model: Make sure it accurately represents the production data and does not incorrectly train and bias the model.
6. Since training of models is an ongoing process, expected return on investment (ROI) should include the time period and time frame.