Applying Artificial Intelligence to Predictive Analytics Can Save Lives

By Shalini Sahai for For the Record

What better time in history to challenge the medical community to find better ways to manage life-threatening diseases than during a pandemic? My recent work in this area focuses on leveraging artificial intelligence (AI) across the continuum of hospital admissions, laboratory tests, diagnosis, pharmacy, therapeutics, and suggested treatments stored in patients’ EHRs. For the most part, the scientific, medical, and technology communities agree on the value of using AI to diagnose diseases based on observed algorithms. In their June 2019 Future Healthcare Journal article, “The Potential for Artificial Intelligence in Healthcare,” Thomas Davenport and Ravi Kalakota posit that with the increase and complexity of health care data, AI would be a logical addition to the predictive analytics platform. The World Economic Forum suggests rates of diabetes, congestive heart failure, and COPD would be positively affected by AI-assisted inclusion of social determinants of health (eg, birthplace, heredity, and air quality) in managing preventive health care initiatives. In my work to date, I have found both benefits and challenges.

Benefits of Using AI in Predictive Analytics

  • Speed. AI is being used predominantly in multispecialty and emergency care centers where quick data processing and suggestions are necessary to save a patient’s life. Here’s one application: Think about the value of early infection recognition to start treatments sooner in critical intensive care settings.
  • The ability to detect and treat disease. For example, AI can often detect early-stage lung cancer based on a patient’s CT scan, recognize sequences in their EHR, and then weigh and suggest patient-centric treatment options. A better, more informed diagnosis facilitates superior treatment choices to the medical team and ultimately the patient.
  • AI provides a common language. Through AI’s profound application of machine learning, doctors can decide the type of treatment based on the computer’s thorough and quick examination and synthesis of a patient’s EHR. Through natural language processing, computers can handle speech recognition, text detection, and other language-related processes, making the human-data interface more effectual. Surgical robots and chatbots converse and assist during the patient’s visit. Predictive analytics become “post-dictive” in monitoring patients after hospitalizations and preventing instrument failures. Doctors and hospitals can collaborate globally on treatment plans through cloud-based data integrating software.
  • Health care accuracy improves. My studies have found that researchers using the appropriate algorithms and training sets can detect cardiovascular diseases with 80% to 90% accuracy.

Failures of Old Technology

In our inquiries to date, we have found the rule-based Environment, Health, and Safety systems used in health care centers today have failed to match the precision of results delivered by machine learning and other AI methods. To overcome this gap, scientists have advised manufacturers to focus more on the use cases where they can learn the most suitable, practically adaptable AI techniques. Science needs to be accessible and appropriate before it can help humankind.

The second practice focuses more on the operational state of technology and regulating the update period since symptoms and diagnostic procedures are subject to change. Finally, more importance should be placed on outcomes. In the case of predictive analytics, however, the data themselves play a more vital role. Because the patient’s data are used to diagnose the disease and suggest suitable treatment, the data’s integrity is paramount.

AI Challenges

Delicate, profoundly significant, life-giving patient data are being lifted from analog storage onto a digital platform. This is a problem that needs to be addressed as we go forward with this exciting intersection of science, medicine, and technology. We need protocols for anticipating and dealing with the following issues:

  • Data corruption or algorithm dysfunction may result in inaccurate and misleading reports. This, in turn, can endanger the patient’s life if unnoticed. What predefined regulations will streamline data collection and their usage? Can AI be developed to police itself? Can data integrity be left to humans? Who is at fault if something fails?
  • People’s privacy can be heightened by limiting their health data to the treatment point only. But what of database breaches? What are the ethics surrounding those?
  • Proper data collection is a time-consuming process and requires manual inputs. How can the health care industry manage and pay for collection of accurate data? What happens if we don’t?

A Connected World

The global digitalization of EHRs allows developing areas and nations a better opportunity for survival. But what if hospital “A” has AI-assisted diagnostics and prevention and hospital “B” in another part of the world does not? When will everything be connected? AI-assisted predictive analytics platforms serving our medically challenged communities can save lives before, during, and after the pandemic but also generally push for a higher good. IBM’s Watson computer has been recruited in this endeavor, but we’re also seeing open source (free) programs challenging Big Blue’s mainframe to make “him” work a little harder on our behalf.

I believe in AI’s predictive analytics can manage life-threatening diseases. We need to seek ways to protect data while we develop the best way of implementing every available technology. By doing so, we will save lives and enhance the quality of lives. By using AI in detecting and treating—if not curing—arthritis, Parkinson’s disease, and Alzheimer’s disease, We. Make. Life. Better. That’s why I do my research and my work.

Share Article: