Friday, April 24, 2020

Reinventing Clinical Decision Support

In our latest book, Reinventing Clinical Decision Support: Data Analytics, Artificial Intelligence, and Diagnostic Reasoning, Paul Cerrato and I explore the promise of artificial intelligence and machine learning for improving clinicians’ ability to make more informed diagnostic and therapeutic decisions. Here’s an excerpt from Chapter 2:

“AI is a once-in-a-generation transformative technology. As such, expect its impact to be on the scale of the advent of electricity or the Internet,” says Jean-Claude Saghbini, Wolters Kluwer Health.(1)

“Artificial intelligence and machine learning are set to transform healthcare. From front line care delivery, including triage, clinical decision support and patient experience to back office operations, such as billing and revenue cycle, algorithms and emerging technologies are already proving their value,” according to a recent report from Healthcare Information Management Services Society (HIMSS). (1)

Both enthusiastic visions suggest that artificial intelligence (AI) and machine learning (ML) are poised to transform medicine and bring in an era of cost effective patient care. But these predictions have to be weighed against less optimistic views, including those that suggest AI will disrupt the workforce in healthcare and other industries, causing many to lose their jobs to soulless
algorithms and robots.

Israeli historian Yuval Noah Harari, for example, believes that: “For now, most of the skills that demand a combination between the cognitive and manual are beyond AI’s reach. Take medicine . . . ; if you compare a doctor with a nurse, it’s easier for AI to replace a doctor—who basically just analyzes data for diagnoses and suggests treatments. But replacing a nurse, who injects medications and bandages, is far more difficult. But this will change; we are really at the beginning of AI’s full potential.” (2)

There are futurists who are far more optimistic, however. They imagine a scenario in which every patient gets the same quality of care afforded presidents in affluent countries or billionaire CEOs at major technology companies. With the assistance of AI, machine learning, and massive databases, they envision a world in which we each have the electronic equivalent of a personal physician who has access to the very latest research, the best medical facilities that specialize in each individual’s health problems, access to cutting-edge data sets, predictive analytics, testing options, clinical trials currently enrolling new patients, and much more. For example, Alvin Rajkomar, MD; Jeffrey Dean, MD, of Google; and Isaac Kohane, MD, PhD, of Harvard Medical School, describe a possible future in which:

A 49-year-old patient takes a picture of a rash on his shoulder with a smartphone app that recommends an immediate appointment with a dermatologist. His insurance company automatically approves the direct referral, and the app schedules an appointment with an experienced nearby dermatologist in 2 days. This appointment is automatically cross-checked with the patient’s personal calendar. The dermatologist performs a biopsy of the lesion, and a pathologist reviews the computer-assisted diagnosis of stage I melanoma, which is then excised by the dermatologist.(3)

This scenario stands in stark contrast to the current state of affairs that often transpires in today’s broken healthcare ecosystem. As Rajkomar et al.3 point out, in today’s ecosystem, this patient is more likely to ignore his skin lesion for far too long; his primary care physician may misdiagnose the melanoma because of its atypical appearance, and the delay may result in a metastatic malignancy that requires systemic chemotherapy.

With such contrasting views, clinicians have to wonder: What precisely will the future look like? Our purpose in Chapter 2 is to explore the strengths and weaknesses of AI and ML and to help clinicians and technologists gain a realistic view of the near future—a future that promises to deliver more cost effective, more personalized care but also one that faces numerous challenges. We will explore basic terminology and concepts and discuss AI/Ml solutions in

a variety of medical specialties. In Chapter 3, we will outline the many challenges that stand in the way of the full implementation of these solutions. 

More details from Chapter 2 will appear in a subsequent blog. A full discussion of AI/ML is available in our book.


1. “AI and Machine Learning: What Cuts Hype from Reality.” Healthcare IT News. Retrieved on April 8, 2019.

2. Kaufman D. (2018, October 19). “Workers Beware: Algorithms Could Replace You—Someday.” The New York Times, p. F2.

3. Rajkomar, A., Dean, J., and Kohane, I. (2019). “Machine Learning in Medicine.” New England Journal of Medicine, vol. 380, pp. 1347–1354.

Saturday, April 18, 2020

A COVID update

I realize that my blog post frequency has diminished during COVID.   Writing time has been redirected to the national COVID-19 coalition and its 14 workgroups:

Modeling & Simulation
Health Systems and Clinical SME
Supply Chain
ICU and Mechanical Ventilation
COVID-19 Data Standardization (mCoVD = Minimal COVID-19 Viable Dataset)
Non-Pharmaceutical Interventions
Optimization of Clinical Therapeutics/Protocols
Data Storage and Source
Privacy Advisory
Contact Tracing

Below is a description of this week's highlights

a. Critical Shortages of Personal Protective Equipment (PPE)

Worked with international suppliers to deliver 580,000 FDA-certified respirators to New York City hospitals, working through Governor Cuomo’s procurement office.

Connected 100,000 FDA-certified respirators to Masks for America, a volunteer coalition of everyday Americans joining forces to deliver protective masks directly to frontline healthcare workers in COVID-19 hotspots.

Prepared guidelines, assisted purchases, and piloted novel solutions to decontaminate respirators so they can be reused multiple times, extending their lifecycle.

b. Implementing Social Policies

Launched a Non-Pharmaceutical Intervention  (NPI) dashboard that provides real-time tracking of state-wide NPI implementation.  As States begin relaxing NPIs, we’ll be able to see the impact. 

c. Achieving Data-Driven Clinical Care Outcomes

Coalition members, including numerous electronic health record vendors, are developing a minimal common data set for COVID-19.  This is now being used to research the outcomes of clinical outcomes on treatments, such as hydroxychloroquine, Remdesivir, and others.   We have designed a federated query approach so that any institution can participate in studies by running queries locally in their EHR. 

Supporting the rapid scale-up of telemedicine as part of a fully integrated healthcare delivery system.  With Mayo's help we are running multiple studies about the adoption of telemedicine during COVID.   See a starting set of best practices in the Resource Library

Detecting Dis/misinformation Related to COVID-19.   Dis/misinformation or outright fraud can affect patients' ability to understand and adhere to non-pharmaceutical interventions and, most importantly, their health.  A new dashboard will be published soon.

Supporting the Mayo Clinic-led national convalescent plasma trial, which is attracting thousands of health systems and patients.  We have worked closely with EHR vendors to develop order sets and a data analysis approach.

Supporting the Contact Tracing activities of Apple, Google, MIT and others with connections to the Association of State and Territory Health Officers, Bluetooth experts at Lincoln Labs, and leveraging an Amber alert like technology called Sara alert for infected patient followup. 

d. International Regulatory Standards and Comparison for N95 Respirators

Due to the shortage of N95 respirators in the U.S. during the pandemic, organizations are ordering similar respirators from international companies. We’ve listed what’s available from 7 countries and which products are FDA-authorized.  

e. Ventilator Training App

This multi-vendor library of training and product materials for medical professionals was created through a partnership of leading ventilator manufacturers and Allego, Inc. You’ll find free mobile access to video overviews, instruction manuals, and other training materials for equipment that is critical to treating patients suffering from COVID-19-related respiratory distress.   See

f. Coronavirus Scientific Literature Topic Browser

Researchers looking for COVID-19-related articles of clinical or scientific interest should try this interactive tool. The COVID-19 Open Research Dataset (CORD-19) is presented in several views, allowing users to quickly find clusters of papers on a desired topic.