Software can help make better diagnoses

I don't have the best memory, and often rely on my PDA to find out what dosages to recommend or what esoteric disease i'm leaving off a differential diagnosis. And I wonder how doctors, so far removed from their training years, can remember what they learned -- especially the rare diseases -- all the while keep up with new diseases and treatments.

Apparently, this is a problem. It certain was for one young girl who's father decided to do something about it. This, from the 23 May 2005 WSJ article "Software for Symptoms":

To the several doctors who examined three-year-old Isabel Maude in 1999, her malady at first seemed to be a routine case of chicken pox. But the fever wouldn't drop, her skin had developed a bluish discoloring and she complained of increasing pain.

Eventually rushed to the emergency room, the toddler spent two months in the hospital -- half that time in intensive care -- suffering from multiple organ failure and cardiac arrest brought on by toxic shock syndrome and necrotizing fasciitis (sometimes called the "flesh eating" disease).

Isabel's brush with death was a life-changing experience for her father, Jason Maude. He eventually left his career as an insurance company's investment manager in London to become one of the founders of a venture that makes software designed to help physicians make diagnoses. Mr. Maude believes that "diagnostic-decision support software," which aims to reduce misdiagnoses by presenting physicians with a full array of possible conditions, might have made all the difference for his daughter.

"We all know computers are better at remembering things than we are," he says.

He started a company named Isabel (after his daughter -- who survived).

Most of today's DDS programs work in a similar fashion: The physician enters basic data about the patient, such as age and sex, along with the patient's symptoms. The program then lists various diseases and conditions to consider, sometimes ranking them by likelihood or organizing them by medical category -- gastroenterology, cardiology, oncology, etc. The systems also enable doctors to look up background material, such as articles from medical journals with the latest disease research.

These programs often suggest maladies that aren't always obvious even to experienced physicians. Consider the case of a 10-year-old boy who came to an emergency room after suffering from nausea and dizziness for two weeks. Harold Cross of Beaufort, S.C., the attending emergency-room physician, says the case was puzzling because the boy had a good appetite, no abdominal pain and only one headache over the two-week period. Dr. Cross found no other physical or neurological problems.

To aid in the diagnosis, Dr. Cross turned to software from Problem-Knowledge Coupler Corp., based in Burlington, Vt. A common thread emerged among the possible causes the program suggested for each of the boy's symptoms: trouble in the back portion of the brain. Dr. Cross ordered an MRI scan, which revealed a tumor in the back of the boy's brain. "My personal knowledge of the literature and physical findings would not have prompted me to suspect a brain tumor," Dr. Cross says. The tumor was removed two days later.

A study published in the February 2005 issue of Casebook, a journal published by the United Kingdom's Medical Protection Society, tested the diagnostic software called Isabel -- Mr. Maude's project -- against 88 cases where doctors had missed the correct diagnosis or had made a delayed diagnosis.

It found that the choices offered by the software included the correct diagnosis for 69% of those cases. (The software is sold by Isabel Healthcare Ltd. and Isabel Healthcare Inc., the two commercial subsidiaries of the U.K.-based Isabel Medical Charity.)

So why hasn't diagnostic software found more of a market? Some doctors say it takes too much time to enter extensive patient data into some systems. "If your HMO allows you 10 and a half minutes to see a patient, how are you going to do this?" asks David Goldmann, a physician who serves as vice president and editor in chief of the Physicians' Information and Education Resource, a guide to clinical care published by the American College of Physicians in Philadelphia.

Others worry about the potential for mistakes whenever computers are involved in medicine. For example, the March 9 issue of the Journal of the American Medical Association is filled with articles about errors in the computerized ordering of drugs and medical tests by physicians. Isabel Healthcare seeks to soothe such concerns by calling its program "diagnosis reminder" software -- to emphasize that diagnostic software leaves the final decision up to the doctor.

There are financial considerations as well. A doctor pays $750 a year to subscribe to Isabel, for example, while a hospital pays $180 per bed. But the majority of the roughly 700,000 practicing physicians in the U.S. are self-employed.

Despite these obstacles, DDS programs will become more popular many in the medical community believe.

Still, there are signs that these diagnostic systems may gradually gain a wider following. The increased use of personal digital assistants, or PDAs, and other handheld computers by doctors could boost the market for diagnostic software. More than half of U.S. doctors use one of these devices regularly, according to a study by the American Medical Association and Forrester Research Inc. of Cambridge, Mass.

And some DDS programs, including Isabel, can be used on PDAs as well as desktops. Doctors may be more favorably inclined toward these programs if they don't have to be tied to their desks to use the software. Meanwhile, some makers of diagnostic programs are trying to address doctors' financial concerns. The home page for Isabel Healthcare, for example, has an "ROI calculator" that doctors can click on to consider the potential return on an investment in Isabel -- including the possibility of avoiding malpractice cases.

The growing adoption of electronic medical records also presents an opportunity for makers of diagnostic software. The DXplain software developed by Massachusetts General Hospital in Boston can be linked to electronic medical records, automatically prompting doctors, for example, about potential causes for abnormal results in a patient's latest lab tests.

"Diagnostic support systems will be more widely used when we can link [more of] them" with electronic medical records, says Octo Barnett, a professor of medicine at the Harvard Medical School and senior scientific director of the Laboratory of Computer Science at Massachusetts General.

Also, Dr. Barnett suggests, as the profession becomes more accustomed to using other forms of health-care information technology, like electronic medical records and computerized ordering of medications and tests, some of the resistance to diagnostic software will fade.

Meanwhile, diagnostic-support programs have benefited from greater input from doctors in their development, as well as from the latest advances in software. The Isabel diagnostic tool, for instance, uses powerful software from U.K.-based Autonomy Corp. to analyze the vast amount of medical information that provides the basis for diagnoses of multiple symptoms. Isabel's use of Autonomy's pattern-recognition technology is a key reason why the program is able to find the most likely matches between symptoms and diseases, says a recent report by Forrester.

"Hopefully, someday these systems won't be a burden to be borne by the physician community, but a tool to be embraced," says Eric Brown, a vice president at Forrester.

According to the review article "New online diagnostic tool launched to help doctors" that ran in the British Medical Journal in 2002, "Trials in four hospitals have found that in 95 out of 100 paediatric cases, the Isabel tool came up with the correct diagnosis. More extensive trials are planned for August of this year." This means Isabel is quite accurate. However the power remains with the human professional. "Lord Hunt, health minister for IT and clinical quality, who attended the launch, emphasised that the software was a support system but it was the doctor who must use his or her clinical training to decide on a diagnosis."

The human brain is limited, and there is no surprise if a few diagnosises are wrong or missed. But according to a 22 February 2006 NY Times article "Why Doctors So Often Get It Wrong," it isn't just a few.

With all the tools available to modern medicine — the blood tests and M.R.I.'s and endoscopes — you might think that misdiagnosis has become a rare thing. But you would be wrong. Studies of autopsies have shown that doctors seriously misdiagnose fatal illnesses about 20 percent of the time. So millions of patients are being treated for the wrong disease.

As shocking as that is, the more astonishing fact may be that the rate has not really changed since the 1930's. "No improvement!" was how an article in the normally exclamation-free Journal of the American Medical Association summarized the situation.

The writer of this article poses a question to end his piece, interestingly. "Clearly, misdiagnosis costs far more than that. But in the current health care system, hospitals have no way to recoup money they spend on programs like Isabel. We patients, on the other hand, foot the bill for all those wasted procedures and pointless drugs. So we keep getting them. Does that make any sense?"

Isabel isn't the only DSS program in the market. Here's one highlighted by Microsoft. These are but two of many.

One interesting question that takes me back to an old post is can computers become better diagnosticians than seasoned doctors? Not better doctors, but better diagnositicians.

Speaking of the VA team (which is still the best team of residents, sub-i's and attending I've been on), I remember a woman who had sudden-onset substernal chest pain mimicking a heart attack. However, the EKG and cardiac enzymes were negative for it. I had just watched a nighttime show on Broken Heart Syndome, and this woman fit the typical picture: middle-aged woman who has chest pain because of a stressor which mimics but isn't a myocardial infarction.

I suggested this disease to my sub-i (the smartest guy in his class according to his peers) and my upper-level resident, but they didn't consider it at all. In fact, my sub-i didn't even heed any attention to BHS until I showed him the Up-to-Date article on it. Why were they so dismissive of BHS? Because they never heard of it. And I think this is another place where these DSS programs can help -- a computer program can keep track of new diseases better than a doctor, especially in this age where information has gone from a trickle to a rushing river.