Human medical diagnose utilizes a logical, rational thought process that is well understood. There are knowledge, models, and techniques to teach people how to become experts at diagnosing human medical conditions. Because of this deep understanding regarding how doctors learn and process medical information and the high value of medical personal in society, creating a computerized medical diagnostic system is both practical and valuable. Over the past 40 years many attempts at designing a computerized medical diagnostic system have been attempted; however to date no such system is being used on a wide scale. It is clear from the demand for such a system that computerized medical diagnostic systems are the future of the human health process, but it is only through the examination of the past designs of medical diagnostic systems and the systems placement and effect on society that a widely used computerized medical diagnostic system will be conceived and developed.
A computerized medical diagnostic system is a system that given data about the current and past medical state of a patient will accurately assess medical problems that a patient may have. Medical diagnostic systems have the potential to predict a patient’s future risk of developing various medical conditions and suggest preventive measures for the patient. Typically, medical diagnostic systems are implemented as expert systems borrowing from the ideas and technology produced by research in artificial intelligence.
One of the first medical diagnostic systems, named Mycin, was inspired by artificial intelligence technology and was developed as an illustration of how artificial intelligence techniques could be used to solve problems involving uncertain and incomplete knowledge.8
The Mycin expert system, which was written in LISP by Edward Shortlife in around the early 1970s,7 focused on the diagnosis and the treatment recommendation of bacterial infections. Since rapid response times are critical when working with medical conditions the Mycin system was constructed to work on incomplete data if definitive test data was not available.9 When recommending treatment to a patient if there were more than one possible diagnosis due to incomplete data on hand then the Mycin system was designed to give a detailed drug treatment plan that would cover all the possible infections.9 Mycin operated by asking a series of Boolean or textual questions from which the diagnostics system would produce a list of possible bacteria that was mostly ailing the patient along with a confidence rating for each diagnosis.7 Furthermore, each of the diagnosis could be queried and reasoning that lead to their prediction could be exhibited allowing a physician to easily verify that the diagnosis followed logically.7
The knowledge base of the Mycin system consists of only around five hundred7 IF-THEN rules and uncertainty factors for each rule.9 Also included as part of the knowledge representation aspect of the Mycin system are knowledge tables which are used to contain basic facts that are needed by the system.9 The Mycin program proved to be correct around 65% of the time--better than most physicians, excluding bacterial infection specialists.2 However, even when compared to specialists, whom have an average correct diagnosis rate of about 80%, Mycin’s diagnoses were only slightly less accurate than the specialists.7 Although Mycin was successful, the noise introduced by Mycin’s certainty factor system caused it to have a limited depth to its reasoning hierarchy.7 This problem lead to it being used as a cautionary tale for the generation of ad hoc probability frameworks. However, it has been found that using more rigorous probabilistic frameworks, such as Bayesian statistics, work better. Even with its high accuracy rate, Mycin was never moved out of the lab into popular use because "no one knew who to sue when [Mycin] was wrong."2 So the legal and ethical implications were found to overshadow the benefits of the system.7
Internist, a medical diagnostic system, which uses a more complex problem solving strategy than MYCIN, where the system attempts to capture the way human medical experts learn and make diagnoses.9 The Internist system diagnosis method utilizes a strategy where “ the direct evocation of differential diagnostic task structures, and outlined methods for managing the size and complexity of the space of task definitions result in the course of analyzing a complex clinical problem.”10 This strategy works well when there are a large number of possible hypotheses to analyze—in this case diseases.9 Internist uses three basic steps to carry out its diagnostics. First it uses symptoms to trigger or suggest likely diseases, then it determines expected symptoms given the suggested diseases, finally it gathers specific data based on the expected symptoms to distinguish between the diseases and then it repeats the process until the system eliminates enough hypotheses to make a diagnosis.9
The knowledge base that powers Internist stores profiles for each disease that it contains. The disease profile consists of symptoms and evidence for a given disease and associates two numbers having a range between zero and five for each piece of evidence.9 The first number is an “evoking strength” value which is the likelihood of the given disease given the correlated evidence and the second number is a “frequency” which tells the likelihood of the symptom given the disease.9 Although techniques similar to traditional conditional probabilities are used to manipulate the symptom data, ad hoc techniques are used instead.9 It is important to note that using these ad hoc statistical method Internist is able to adjust its own disease profile figures and in this way has the ability to learn and adjust out errors, becoming more accurate as it works.10
In the end Internist consisted of around 600 diseases.10 All this knowledge was accumulated by a team of physicians who used many years of careful literature reviews and case discussions to compose the knowledge base.9 But the internist system never caught on. The system was designed to be run on a large mainframe computer, needing a large amount of computational resources; this limitation made it not suitable for practitioners.9 Furthermore the accuracy of the system was limited because the system used time-independent profiles, while many health condition symptoms change over time, and the system did not take into account a patient having multiple health conditions as often one condition affects another’s presence.14 Lastly, Internist, unlike Mycin did not give an explanation why it predicted a patient has a certain condition, making it hard for medical personal to discern if the assumptions Internist made were valid. Due to the limitations of Internist, the system was adapted into a new system called, QMR (Quick Medical Reference), where its powerful, well structured database was put to use in there different modes: diagnostics, an electronic textbook, and symptoms-disease correlation.9
Both Internist and Mycin were successful diagnostics systems from the technical standpoint and there are other such successful systems; however, the ethical, social, and legal issues that surround diagnostics systems have been the leading cause of failure of the systems to become popular and widely used in the medical industry. From an ethical standpoint it is important to question whether a new technology is beneficial to society and to weight the negative and positive impacts the technology will have. In the Canadian context it is easy to understand the need for computerized medical diagnostic systems. The Canadian health care system which is publicly supported is currently being stretched to its limits. Doctors are in high demand and consequently their salaries are climbing higher and higher, reducing the available funds in the system for equipment and building space. Currently the totally amount of federal money budgeted on Canadian health care is $27.2 billion which is roughly 33% of provincial spending.6
A computerized diagnostics system could alleviate this financial strain on the Canadian medical system by reducing the demand for highly trained medical personnel such as general practitioners. Instead a nurse practitioner, whom has less medical training than a doctor, could be used to evaluate a patient and then enter their evaluation into a computerized diagnostics system. The system then would diagnose any medical problems the patient is suffering from and recommend treatment as well as spot medical conditions that the patient has a high potential of developing and recommend preventative treatment. The nurse practitioner would then relay this information to the patient. If the computer diagnostics system could not make a diagnosis it would then calculate which information will most improve the accuracy of its diagnosis and request for this information. Furthermore, if the required information has been supplied and the computer diagnostics system was unable to determine with a high percentage of accuracy which ailment the patient had it would request that they were sent off to a specialist to be examined. In the case where the information requested was a medical test or in the case where the diagnostics system recommends that the patient visit a specialist the computerized diagnostics system would be able to locate a local and available specialist and book the appointment, further automating the medical system.
Because nurse practitioners would require much less training they could be trained quicker than doctors are now reducing the demand for doctors and consequently lowering federal spending on medical personnel spending. This will increase the quality of healthcare in Canada as with an increased number of medical personnel, waiting times decrease and more staff would increase the amount of personal care. Furthermore, it has already been shown that computerized diagnostics systems have proven to be more accurate than current medical professionals in diagnosing disease. With a central knowledge system the volume of diseases information and medical health information the diagnostics engine would have access to would be greater than any doctor could study in a life time. This means rare diseases that would most likely be missed by a typical doctor because they do not have knowledge of the disease could be diagnosed by the computer system. In a similar way, new research could be added to the system allowing it to always have access to the cutting edge advancements in medicine. No human doctor can possibly keep up to date on all the latest advancements in medicine.
It is important to ensure the clear understanding that a medical diagnostic system would not put doctors out of work, but instead just require them to be specialists or surgeons. The medical diagnostic system partnered with nurse practitioners would fill the role of the general practitioner medical system, but they would not be able to perform surgeries or remove the need for medical specialists.
The transition to a medical logistics system which relies heavily on a computerized medical diagnostic system will most likely be gradual. The first step, which is already being realized, is having medical personal assisted by computer technology. This hardware aspect of this technology is already making its way into the hands of doctors as doctors are the highest user group of PDA technology; however only a third of them used their device for viewing and manipulating mobile electronic medical records.18 There are disease-symptom search database which correlates a disease to symptoms and vice versa for PDA’s, which are used by a small number of doctors. Once there is a reliance on technology in this manner, doctors will be more willing to utilize artificial intelligence medical diagnostic software to assist them in their daily tasks. Even in this transitional state, medical diagnostic systems will make it easier for doctors to go about their work. Similar to the way they aided psychiatrists where “computers are commonly used in psychological assessments, to determine, for example, the potential of a patient for suicide, depression and anxiety, or substance abuse.”1 These assessment programs also calculate scores and write interpretive reports, freeing up time for the psychiatrist to do more clinical work.1 Doctors could also see this benefit from using computerized medical systems allowing them more time with the patient and less time filling out reports and other paper work. In these early transitional steps doctors are not displaced from their work, nor does the computerized diagnostics system have any real authority or effect without the doctors consent. Leaving the control in the physicians’ hands reduces the number or ethical, legal, and logistical problems of computerized diagnostics systems, allowing medical personnel to get comfortable using such system. However, the full advantage of these systems can only be realized if the computerized medical diagnostic systems are used as the door openers to the medical system, displacing the general practitioners that reside in the gate keeper positions at present.
It is clear that medical diagnostic systems would benefit society at any stage in their integration into the medical system, but along with these advantages come disadvantages, which ethically must be weighed against the benefits of the system. As technology has done much in the past, computerized diagnostics systems threaten to destroy general practitioner’s jobs in the medical system. The disadvantage of lossing doctors’ jobs can easily be remedied because general practitioners are generally well educated members of society and could be trained as specialists or medical researchers, areas where computer diagnostics systems are less likely to achieve success. Furthermore, the loss of highly trained medical personnel jobs opens up jobs for nurse practitioners, who need far less training. Due to the reduced cost of nurse practitioners more of them can be hired, providing more jobs for society.
The human species prefers interaction with other human beings and medical diagnostic systems threaten to further decrease the amount of human-human interaction and increase the amount of machine-human interaction. If computerized medical diagnostic systems are implemented in the medical system where patients are interacting directly with the computer system then this decrease of contact with human beings progresses even more. Furthermore, other ethical and legal issues arise if patients interface directly with the diagnostics machine, such as what if the patient lies about a symptom or evaluates their symptoms incorrectly? In order to eliminated these ethical and legal issues as well as increase human interaction in the medical care process the use of nurse practitioners which act as mediators between the computerized diagnostics system and the patient is recommended. Since the nurse practitioner is responsible for the physical analysis of the patients’ symptoms, while the computer is responsible for the interpretation of the data, the nurse practitioner should be held legally responsible if the diagnostics system is misused or some of the patients symptoms are missed by the practitioner; however if there is a mistake in the diagnosis of a patient and the correct information was provided to the diagnostics system the company that produces the diagnostics system should be liable. In the transitional stage before the general practitioners are removed from the medical system the legal liability remains on the doctor not the computerized system to insure the proper advice and care of the patient is given. Having a nurse practitioner as an interface to a computerized system also insures that care is taken when bad news and information has to be imparted to the patient or their family. Computer programs in their present state are unable to effectively deliver highly emotional news to the same degree a human is able to. For this reason, legal and ethical issues, and to increase of human to human interaction in a technical society, the logistics of a computerized medical system requires the use of nurse practitioners or other medical personnel to act as liaison between the patient and the diagnostics machine.
Privacy can not be ignored when it comes to assessing the ethical logistics of computerized systems, especially systems which store personal information relating to specific people. Canada legislation has provided the Personal Information Protection and Electronic Documents Act to help outline rules and regulations regarding the use and disclose of personal information including information collected, stored, or distributed in digital format.20 The act states the organization that is holding the private data is responsible for taking appropriate security measures, ensuring the information is accurate, and the information is used only for the purpose intended by the individual (patient) submitting the information.20 The current medical system already relies on computerized medical records these records are similar to what would be stored by any artificial intelligence computerized diagnostics software and need to be treated in a similar way. However, the current medical system, abiding by the government regulations of medical and private data, has proved successful in keeping personal medical information private. Following the same procedures computerized diagnostics system should experience similar success in this area. The current medical record system logs every change and view to a medical record, in this way patients know who has seen there record.19 The system also allows doctors to tag sensitive medical data so that only they can view it and if another doctor wants to be able to view it they need to have permission from the “host doctor” first.19 Finally encryption techniques are currently employed to ensure stored and transferred data do not get intercepted. All three of these security techniques should be mimicked in any widely used medical diagnostic system as well as care that the local laws of the region are followed; in this way medical diagnostic systems can limit the negative ethical and legal impact caused by security issues and insure that the implementation of a computerized medical diagnostic system would keep the security level of personal medical information on par to what it is currently.
Widespread use of medical diagnostic systems has not occurred yet mainly due to a slow acceptance by the medical community and legal and ethical issues. From the Mycin project it can be seen that even an accurate system is not enough to be a successful medical diagnostic system, but operation time and legality play a major factor as well. The Internist project shows that the system must be easily accessible to practitioners, having system requirements below the mainframe level, and maintain a high degree of accuracy while still maintaining the ability to produce results quickly. Both these systems fail at diagnosing patients with more than one illness accurately. In the diagnostics systems community there has been a shift away form general diagnostics systems and a progression towards systems that diagnose only a specific disease or at minimum a subset of the medical ontology. From systems such as the Cycorp project it can be seen that in order for an artificial intelligence system to be able to have real intelligence that is similar to a human one, having common sense and being able to rationalize ideas, such a system must have a large ontology. As medical diagnostic systems move away from this large ontology toward a smaller one they lose the real power that artificial intelligence could bring to diagnosing human ailments and it is mostly only with this power of knowledge of a full medical ontology that an artificial intelligence computerized medical diagnostic system would be able to perform on a consistent basis as general practitioners. The accuracy of medical diagnostic systems is going to be the most important factor to have them accepted by the medical community and then society. Many of the ethical issues surrounding medical diagnostic systems can be eliminated with care in system implementation, ensuring data security and a human friendly interface. It has been seen that society will not embrace all beneficial technology but often needs transitional stages to ease into the idea of it, clearly this is the case with medical diagnostic systems. However at the end of the day well-designed medical diagnostic systems will provide many benefits to society and will likely eventually surpass general practitioners in diagnosis accuracy. After all, given the same information as a human, computers always produce more accurate results.