Sophisticated software can help doctors reach more accurate diagnoses
A middle-aged patient recently visited the clinic of Dr Thomas Paine in Seattle when he complains of a significant and unexpected weight loss and a tingling sensation in his feet. The doctor wondered the meaning of the phenomenon: the symptoms could indicate any problem, starting with a simple inflammation and ending with any of dozens of more complicated and serious diseases, such as diabetes or cancer. Payne, who also serves as medical director of information technology services at Washington University School of Medicine, therefore decided to take an unconventional route. After conducting a routine physical examination and filling in the details in his patient's medical record, he enlisted the help of an online tool - Dxplain.
Payne entered the symptoms into the computer and the computerized support system suggested several possible conditions, including a rare disease called Amyloidosis, in which abnormal proteins accumulate in the body, disrupt the function of various organs, and damage the nervous system. Additional tests as well as a biopsy conducted at another medical institute confirmed the diagnosis suggested by the computerized system. The patient is one of the 4,000 people diagnosed with the disease in the US every year.
Until just five years ago, if he had encountered such an unsolved case, Payne would have first sought the help of a trusted colleague or spent hours reviewing the mountains of professional books and scientific research articles to arrive at such a rare diagnosis. The DXplain clinical decision support system draws on the same professional literature and peer-reviewed studies to offer its own assessments - but the computerized system does it in seconds. "Could I have come up with a similar list of possible situations? Maybe, if I had set aside enough time for it," says Payne, who usually has the chance to treat patients with the flu or arthritis rather than patients suffering from damage to the nervous system, the cause of which is unclear. However, Payne warns, the scientific literature shows that "when we are working under time pressure, we do not give sufficient thought to such and such cases, and those rare diagnoses may escape our notice."
Indeed, too often we miss the correct diagnosis, as is evident from the wide-ranging reports published by American National Academy of Sciences, Engineering, and Medicine which analyze the causes of medical errors that occurred in the USA in 1999 and 2015. These errors stem in part from careless management of medical records or poor communication. But often, the cause of the mistake is misdiagnosis. Reviews of Medical transcriptions It turns out that it is possible to attribute 6% to 17% ofThe negative events in which harm was caused to patients in hospitals due to errors in diagnosis. according to estimate 2015 National Academy of Sciences Report, 10% of the deaths of patients in the US are caused by incorrect diagnoses and as a result of the incorrect treatment given as a result.
As a solution to the problem, the academies recommended, among other things, that hospitals and attending physicians be assisted by additional tools - officially defined asClinical decision support systems - which may help improve their decision-making processes. At the most basic level, this is a checklist (checklist) according to which one must work to avoid skipping key steps in important medical procedures. And more than that, more and more medical schools, university hospitals and other medical centers are purchasing subscriptions to computer-based support systems such as Dxplain or competing systems, VisualDx and-Isabel. The most popular of them is the VisualDx medical diagnosis support system, which is currently installed in more than 1,600 hospitals and clinics across the US, according to the manufacturer's data.
The clinical decision support system industry markets its products as tools that can be used by treating physicians to confirm their own diagnoses or examine alternative diagnoses offered by these systems. But until now, doctors have not been in a hurry to adopt the new tools. The crucial question is whether software solutions of this type really have the power to significantly improve diagnosis and medical treatment, or whether they are nothing more than an unnecessary burden on the already busy agenda of doctors.
Processing power
Utilizing computing power to aid in medical diagnoses is not a new idea. Already in the 70s, first attempts were made to use a computer in order to prevent mistakes by doctors. In the following decade, in the mid-80s, Massachusetts General Hospital began developing the DXplain system in order to help improve diagnoses. This approach seemed promising, but in those years it was not very successful, partly because medical records were still written by hand and the use of a computer-based system was therefore another burdensome step in the doctors' work routine.
A lot has changed since then. Computers today are an integral part of routine medical work. Computerized records are replacing the manual records in most clinics, hospitals and private medical institutes, and all this, in no small part thanks to government incentives. These changes contribute to improving the quality, safety and efficiency of the healthcare system.
At the same time, there are also changes in the computerized systems that support clinical decision-making. Over time, these systems are refined and become much faster and often, they are linked online to the studies on which they are based, allowing treating physicians to evaluate the evidence in their hands on the spot and get more information about any possible diagnosis. The VisualDx system, for example, emphasizes the "visual" aspect - it presents diagrams of the body organs that may be affected in a given situation and images illustrating the diseases that may affect them, with the aim of simplifying the comparison between the data.
And just as importantly, scientists are learning more and more about the causes of human errors and the ways to prevent them. Researchers have identified several thinking traps, or cognitive biases, in the diagnostic processes of doctors. One of these, easily preventable with the help of the computer, is known as anchor tilt. Studies show that often, doctors tend to adopt the first diagnosis that comes to mind - the anchor - even when it is wrong. Captive of this concept, they may unconsciously attach more importance to any piece of information that confirms their diagnosis and ignore other relevant data - and sometimes, they don't even bother to look for additional data.
Human errors
Anchoring biases can disrupt medical diagnoses in overcrowded hospital wards or clinics and other medical centers for many reasons. A busy attending physician may forget to ask a patient whether he has recently visited a foreign country, even when the answer to the question can completely change his initial, seemingly reasonable diagnosis. In such cases, for example, a doctor may send a patient infected withEbola Returned to his home with instructions to take simple capsules to reduce fever and relieve pain instead of putting him in isolation and giving him medical treatment from appropriate hands. The origin of other medical errors in the teaching method in medical schools. Often, presented in front of the students test cases which reflect a classification of typical symptoms according to pre-defined parent categories, and not complicated and complex cases like those they will encounter in their work in the field. But these test cases described in the textbooks are not exactly the common cases, contrary to what one might think.
Errors of this type are errors for which the computerized support systems may prove to be most effective. Each of the utilities uses its own proprietary algorithms to link symptoms to diagnoses and draw attention to the most likely and most dangerous situations and, accordingly, to situations that should be ruled out outright. Some of the computerized support systems are even able to automatically retrieve information from the patient's electronic records and thus free the doctors from the need to go back and enter the information.
It is still difficult to assess to what extent the systems that support medical decision-making will be able to prevent errors. Anyway, the data looks promising. As part of a study that tested the performance of the VisualDx system in 2011, a comparison was made between the degree of success of doctors in emergency medicine departments at two different institutions in diagnosing a certain skin infection with the help of the computerized support system and without it. Attending physicians who used the VisualDx system reached a correct diagnosis in 64% of the cases. In contrast, the doctors who did not use the system were able to correctly diagnose the infection in only 14% of the cases. According to the findings of a preliminary study that examined the performance of the Isabel system and was presented at a conference held in 2014, the system improved the diagnostic ability of 40 medical students by approximately one third. In a study that examined the performance of the DXplain system and was published in 2010, it was found that when doctors specialize inMayo Clinic The system was used in complicated and difficult-to-diagnose cases, there was a dramatic decrease in medical costs, as the system contributed to the shortening of hospitalization times in the hospital and the optimization of treatment during hospitalization.
Obstacles on the way to implementation
However, often, changes for the better do not take place at the desired rate. In July 2017, the American National Academy of Sciences held a one-day conference to review progress in reducing diagnostic errors. John Ball, the doctor who chaired the committee that drafted the 2015 report of the Academy, said even before the conference that he expected "disappointing" results, since many of the recommendations to reduce the number of errors - including increased use of computer aids for decision-making - have not yet been widely adopted . Ball says that the network of seven hospitals in North Carolina where he works is still far from fully integrating these systems into its routine care.
The problem in North Carolina, Ball says, at least in part, is that different hospitals and different doctors in the network work with different electronic systems and different medical record protocols, making it impossible to standardize the changes. Another problem, he says further, is that doctors are not enthusiastic about devoting time to learning the computerized support systems until they are convinced of their inherent usefulness.
Institutional conservatism is a problem all over the US, he says Mark Graber, president and co-founder of The Association for the Improvement of Diagnostics in Medicine. "The health organizations do not, in fact, recognize the problem of errors in diagnosis and do not see it as an issue on which they should focus," says Graber. "For the most part, doctors think they are doing a good job and they think they have no reason to be upset."
Besides that, some of the experts, including them Sandra Freihofer, the former president of Association of Internal Medicine Physicians in the USA, who currently works as an internist in Atlanta, fear that widespread adoption of these utilities could have unintended consequences. She is troubled by the possibility that as these programs become more accessible to the general public, patients will use them for self-diagnosis and forgo a visit to the doctor or, alternatively, will be anxious, without justification, due to a terrifying diagnosis that the software will offer - a phenomenon that, according to doctors, they are already witnessing today , when people search for an explanation of their symptoms on the Internet.
On the other hand, other doctors, like Payne, are not afraid that the computer will replace them. They predict that the computerized support systems will ensure an advanced, safer and smarter approach to medical diagnosis, similar to the complex and sophisticated backup systems in the cockpit. They hope that with the help of such systems, which prevent redundancies and indicate possibilities for a solution, they will be able to chart a more reliable and smoother path for the benefit of all of us.
About the writers
Dina Payne Maron - Award-winning journalist and editor at Scientific American who deals with medical and health issues.
See more on the subject on the science website:
