In 1889 Jerome K. Jerome published a hilarious book called “Three Men in a Boat” about a trip on the Thames River. Jerome, a hypochondriac in his fictional account, decided to see his doctor first to find out if there was anything wrong with him. He had read a book in the British Museum that convinced him that he must be suffering from any of a thousand different maladies. He went to his medical man because he thought that “what a doctor wants is practice.” The doctor’s prescription was that the patient should not stuff his head “with things he didn’t understand.”
Who does understand the true nature of illness? Even the best of physicians can get stumped from time to time. Wired magazine reported that, according to Memorial Sloan Kettering Cancer Center, “it would take at least 160 hours of reading a week just to keep up with new medical knowledge as it’s published.” For this reason, Sloan Kettering teamed up with the healthcare company Wellpoint to see if IBM’s Watson could do more than play “Jeopardy!” Samuel Nussbaum of Wellpoint has claimed that Watson’s successful diagnosis rate for cancer is 90 percent, while human doctors come in at only 50 percent. (To learn more about Watson and artificial intelligence, see Don’t Look Back, Here They Come! The Advance of Artificial Intelligence.)
Isabel, IBM Watson and McKesson InterQual
In an article called “For Second Opinion, Consult a Computer?,” The New York Times told about Dr. Gurpreet Dhaliwal of the University of California, San Francisco, whose 45-minute demonstrations amazed crowds of admiring physicians. Dr. Dhaliwal would be given a series of symptoms and, one by one, he would discuss and rule out potential diagnoses until he arrived at the right one (to rousing applause). Dr. Dhaliwal believes that in medicine “thinking is our most important procedure.” But even Dr. Dhaliwal, who is an insatiable reader of medical journals, turns to the web-based Diagnostic Checklist System called Isabel for what he calls “a second check.” Whether you use a computer or your brain, said the doctor, the challenge is “deciding what’s signal and what’s noise.”
IBM’s foray into medicine appears to be initially focused on its own employees, but they plan to expand it to the public in the next five years. Deborah DiSanzio, the head and general manager of IBM Watson Health, told the Chicago Tribune in December, 2015, that “Watson Health is only seven months old, and it has close to a million electronic health records, 30 billion images, 100 ecosystem partners. It’s tremendous.” So we may see IBM as a big player in the industry within the next few years. (For more on electronic health records, see Electronic Health Records: Here’s What’s at Stake.)
Recently I spoke with Lisa Smith, the Director of Product Development and Care Management at DST Health Solutions, about the role of IT in medical diagnosis. While the layman may think of IBM’s Watson, Ms. Smith pointed me toward McKesson, a Fortune 500 company and “the healthcare tech giant you’ve probably never heard of,” and their InterQual product solutions. McKesson’s CEO John Hammergren described the evolution from paper to IT in an interview with Fortune magazine: “Health records, before we started to automate them, were all handwritten in paper files. Today, they are being automated very rapidly. I think the positive is that the records are available for things like data and analytics, and for sharing.” He explained that some 70 percent of hospital IT vendors are now working together in an initiative called the CommonWell Health Alliance.
Ms. Smith, an experienced nurse (and friendly to boot), also educated me about the industry’s use of ICD-10 to translate patient conditions from words into alphanumeric code. The International Classification of Diseases (ICD) is a joint effort of the member states of the World Health Organization (WHO). ICD-9 came out in 1976, but the adoption of ICD-10 has been staggered among the various countries. The U.K. adopted ICD-10 in 1995, and France did it in 2005. But the U.S. only implemented the standard in October, 2015. It is now part of the U.S. medical community’s HIPAA law.
Coding is something that we IT professionals know something about, so let’s take a closer look. The 68,000 codes of the Clinical Modification (CM) portion of the ICD library are to be used by all health care providers. The Procedure Coding System (PCS) contains some 76,000 codes that are targeted toward hospitals. While ICD-9 codes had only five characters, ICD-10 codes are seven characters long:
Format: _ _ _ . _ _ _ _
The first three characters make up the category. The fourth slot gives the subcategory. The fifth and sixth places are for specificity, like location or laterality (left of right). And the seventh character is an extender to offer more detail. The character “X” can be used as a place holder.
Short Description: Superficial frostbite of left arm, sequela
Code Classification XIX: Injury, poisoning and certain other consequences of external causes
(You can browse to this particular code here to see how the database is set up.)
Knowledge Bases and Decision Trees
Now it’s clear that we have moved from the realm of verbal descriptions to language that computers can use. While it may be meant for all health care professionals, in some settings the coding is left with the medical billing personnel. But there are more uses than financial reporting. Lisa Smith told me about the employment of ICD-10 in knowledge base and decision tree applications. The McKesson website calls their InterQual products “the undisputed gold standard in evidence-based clinical decision support.”
IBM Health and their hero Watson may be doing a lot, but so is McKesson. They are in 300 health plans, 3,700 hospitals, and 175 veterans’ facilities. Ms. Smith described to me how she works with medical professionals every day to educate them about using artificial intelligence tools to refine their diagnoses. Medical classification is used for more than statistical analysis and reimbursement.
Along with the written histories taken by doctors and nurses, diagnostic test results are included in a patient’s record. Advancements in imaging procedures like computerized tomography (CT), positron emission tomography (PET), magnetic resonance imaging (MRI), ultrasound and standard X-rays take advantage of built-in information technology that facilitates data transfer to medical documentation. My first job out of high school was working in the radiology department of my local hospital. I remember the tedium of manual film processing – and the pandemonium that ensued when a young orderly (not me!) had left the dark room door open and exposed thousands of dollars’ worth of X-ray film. Now all such images are easily attached as digital files to be viewed by physicians at will.
We are still not to the point that a holographic doctor can take over the duties of a deceased physician, as on the science-fiction television show “Star Trek: Voyager.” But even the best diagnosticians, like Dr. Dhaliwal, have seen the value of digital knowledge and computer-aided decision making. Such computer assistants are smart for a reason. Ms. DiSanzo said that for Memorial Sloan Kettering, “Watson ingested something like 15 million pages of medical content, looked at 200 medical textbooks, [and] read 300 medical journals.”
But computers can’t do it alone. The Isabel website says that their diagnostic tool is “the culmination of over 10 years and over 100,000 man-hours of continual development.” And you still need clever physicians to make final decisions about the health of the patients in their care. This sounds a lot like the man-computer symbiosis that J.C.R. Licklider envisioned at the dawn of the computer age.