Precision Medicine, AI and the New Frontier

If the current buzz is to be believed, we stand at the precipice of a new dawn in healthcare. Let’s put aside the enthusiasm and provide a dose of reality.

Achieving the objective of delivering AI-driven healthcare can be a risky path. Embarking on this pathway with data sets driving diagnosis must be tempered with clinical and analytical governance and oversight. In this blog post, we discuss:

• The challenges and opportunities of applying AI to precision medicine, which aims to provide personalized and effective healthcare based on data and evidence.
• The need for data quality, clinical governance and ethical oversight to ensure AI solutions are reliable, safe and beneficial for patients and providers.
• A proposed five-step process for healthcare providers to explore and implement AI solutions in their organizations, involving data analysis, clinical identification, ROI estimation, MVP modelling and pilot testing.

If artificial intelligence solutions are driven by data, clean non-biased data is required to avoid potential pitfalls in a data-driven outcome. What is “clean non-biased data?” Examples include demographics, labs, appointment schedules, prescriptions, patient reported measures and presenting complaints. Given the current powerful algorithms AI uses, all data points need to be reassessed and algorithms adjusted to weight the data appropriately. For precision medicine to be effective, it must view the patient entirely based on trusted clinical markers.

There is strong evidence the standardization of care is part of an evidence-based clinical journey. Currently, the clinical pathway begins at the point of diagnosis. However, in the near future, it will start even earlier in the process. I believe this will be precision medicine 2.0, where treatment is determined by the patient’s presentation and associated information.

Driving precision medicine across the ecosystem is no easy task. Healthcare providers need solutions that can be easily deployed and positively affect patient outcomes while at the same time understanding how clinical risk is mitigated in the ecosystem.

Managing the transformation process required to deploy AI in healthcare is complex and needs coordinated action. Before deploying AI into healthcare, organizations should establish appropriate governance teams, including clinical ethical and operational teams.

Healthcare providers exploring AI in their operations can follow this five-step process as they embark on this path.

1. Review the historical data set and define what intelligence you can obtain by leveraging this existing data. Most organizations do not have this expertise in-house and would need to liaise with a specialist team of clinical informaticians and data experts to understand the current environment.
2. Work with the clinical team to identify key client groups or system workflows where care (defined in terms of the IHI quadruple aim) can be improved by leveraging data more effectively.
3. Research the potential return on investment by incorporating the data analysis in step one with the clinical target identified in step two.
4. Finally, the solution will be modeled (MVP), and a pilot will be conducted with a view to a fast deployment into the live clinical environment.
5. Continue back-testing and working with the clinical teams to ensure relevance and impact.

While many institutions exemplify this approach, others try to chase “the next shiny thing” without a coherent plan. Without a dedicated team driving precision medicine through the organization, the project is set up for failure. This approach supports the careful and coordinated development of AI solutions across the ecosystem.

This is indeed a new frontier. A close partnership between healthcare providers, clinical teams and technology vendors is essential for a positive healthcare outcome for the patient.

Source: dell.com