Applications of AI in Healthcare
We are passionate about helping people improve their health and are fortunate to work with amazing healthcare clients that have technology products serving providers, patients, and manufacturers. A decade ago it would have been unthinkable that any commercial organization outside of an insurance company could unlock the value of their health data, but it's clear we're rapidly climbing up to the tip of the iceberg.
There are five key trends driving this accelerating change:
Healthcare costs are skyrocketing -- approximately 6 - 7% annually
Enormous amounts of health data is being collected every day in hospitals, clinics, labs and from wearable devices
Health data silos are starting to converge as healthcare companies partner or merge
Large technology companies and healthcare startups are reimagining the overall healthcare experience and beginning to put patients first
Cost of compute time and data storage is dropping rapidly
As a data science organization, it's exciting to be at the forefront of this revolution. We see a broad range of use cases where data-driven applications are being applied or will be applied in the next few years.
Enhancing Disease Identification & Diagnosis
Doctors are already using AI applications to help uncover diseases using models trained by others' health data. For example, last year Stanford created a machine vision model for detecting diseases in chest X-rays. Now there are many machine vision applications that recommend diagnoses from medical images. These applications help support Radiologists and give them more time to spend with the patients.
With the vast amount of patient health information being collected and connected, personalized medicine is becoming real. As we've learned working with clients in the chronic illness space, the same expression of a disease may have a different root cause for a given patient. Weeding through population data manually is nearly impossible for a doctor. However, as health data sets become more and more connected, statistics and machine learning models are uncovering novel findings that allow doctors to pinpoint more effective, personalized remedies faster.
A recent example of personalized medicine is Eric Dishman's story. Eric was a health executive at Intel who was diagnosed with a rare kidney cancer and a very poor prognosis. By his early 40s and many costly dead ends, he provided his whole genome (3 terabytes) to Intel's clinical research group. Using massive compute power, they found insights that saved his life.
Improving the Patient Experience
There are huge inefficiencies in the workflow of the healthcare system that result in a poor patient experience. We feel this each time we go to a new doctor: Redundant forms and different procedures and costs. This poor experience is compounded as we interact with different players in the healthcare system -- insurance companies, hospitals, pharmacies, etc.
Innovative companies are starting to tackle these challenges by rethinking the patient experience. For example, in the Pacific Northwest a number of startups have solutions solving challenges in this space:
- Phreesia - providers create more capacity so they can spend more time with patients
- Bright.md - your virtual physician assistant
- Medsavvy - patients learn more about medications
- Healthsparq - helps people make smarter healthcare choices
- 98point6 - on-demand primary care when you need it
Cost inefficiencies are pervasive in the healthcare system. Many of these inefficiencies relate to process problems or human errors. For example, hospital readmission is a big process problem in Cardiac Care programs. One study found it costs a hospital $7,000 a year when a patient returns to the hospital after a procedure. Likewise, if a patient returns to a hospital within 30 days of a procedure, the hospital may be heavily fined by the federal government as part of the Affordable Care Act.
Innovative organizations like Hitachi, Ltd and Partners Connected Health and Cloudera and Intel are using machine learning models to predict readmissions. Using these predictions they are targeting likely readmission candidates with preventative programs to reduce the impact.
Medical human-errors are another enormous sinkhole of costs in the healthcare system. A recent estimate claims these types of errors cost the US billions of dollars per year and millions of deaths. Unfortunately, many medical errors are understood after the fact and result in heavy fines and lawsuits. Products like MedAware (focused on prescription errors) are attempting to address this use case by mining patient records and alerting clinicians before mistakes are made.
As we approach the tip of the iceberg, there are more and more organizations ready to unlock the power of their health data for improved outcomes and more efficient operations. We have only scratched the surface on the number of healthcare use cases where machine learning can make a big difference. Unfortunately, there are very few experienced data scientists to make this dream a reality. Given our passion and experience in healthcare, we hope to help more organizations realize their dreams and help them grow their own data science teams long term.
To learn more about how Synaptiq can help you apply AI and machine learning to your healthcare use case, please contact us.