Lecture Notes: Computers in Medicine (1985: Queen Mary, Osteopathic Internists Convention)
Please note that these are lecture notes likely extracted from a variety of sources and searches 20 years ago and my references are lost. Perhaps it may be of some historic interest to someone. If so, I might wite a then and now, 06-16-2005
Computers are in their infancy in medicine. Who knows what the future may bring.
The history of analog computers pre-dates Stonehenge. The Chinese abacus (digital) was developed in about 3000 BC and the slide rule (analog) in 1630. The marriage of engineers and slide rules has only recently been supplanted by sophisticated hand-held digital calculators. If one ignores the abacus, the first digital adding machine was developed by Blaise Pascal; a tax clerk, mathematician and philosopher; in 1642. His was an adding machine and Leibniz in 1670 extended this to a multiplication machine. In 1801 Jacquard used punched cards to instruct his looms in the weaving of various patterns on cloth.
Charles Babbage (1792-1871) is considered the father of the programmed computer. He developed the "Difference Engine" in 1822 which carried out sequences of computations by moving levers, gears and cams. He proposed but never developed an "analytical engine" to use punched cards both as input and output of data. This was the first computer in the modern sense.
Herman Hollerith developed a punch card method for the tabulation of the 1890 census. The 1880 census took 7 years to tabulate. Using electronically charged rods with a sorting mechanism data processing was accomplished. In 1896 Hoolerith founded the Tabulatory Machine Company which in 1924 became International Business Machines Corporation (IBM).
In 1939 George Stibitz and his colleagues at Bell Laboratories produced a computer using electromechanical relays to establish discrete settings. It contained 9000 relays, weighed 10 tons, and could add 2 numbers in .03 seconds but took an entire second to multiply 2 numbers. The IBM supported Mark 1 followed in 1944 spawning the Mark 2, 3, and 4. This device was controlled by paper tape with punched holes. In 1946 ENIAC (electronic numerical integrater and computer) was developed by Eckert and Mauchly using vacuum tubes and performing 300 multiplications per second. The UNIVAC 1 was installed at the Census department in 1951 containing 5000 vacuum tubes; adding 2 numbers in 2 millionths of a second and using magnetic tape for input of data.
The electromagnetic relay was supplanted by the vacuum tube for speed in the storage and processing of data. The vacuum tube with its huge power demands, dissipation of heat, and limited life was soon replaced by the transister (transfer resister) made origionally of germanium in 1947 at Bell Labs. Silicon was eventually substituted in the transister which served as a highly efficient on/off (eg. 0/1) switch. The application of this insurgent technology to military situations, Ballistic Missile Early Warning Systems (1958) and finally the space program ushered in a generation of smaller, faster and more reliable computers (second generation). Third generation generation computers eliminated the need for wiring between components with the advent of printed circuit boards. Transistors were etched into intrigate integrated circuits. An immense hardware technology followed.
A silicon chip, a quarter of an inch on a side, can hold one million electronic components which is 10 times more than the original 30 ton ENIAC digital computer. The microprocessor is a "computer on a chip" and is 30,000 times as cheap as ENIAC. It uses the power of a night-light instead of that of a hundred lighthouses. The microprocessor performs one million calculations per second which is 200 times as many as ENIAC ever could. Eventually one billion transistors will be on a single chip.
-By 1990 the state of the art memory devices will store the total information contained within the Library of Congress in a device the size of a 50 cent piece.
III. What is a Computer?
A computer is simply a calculator, like an adding machine. It has an accelerated operating system and the ability to store large quantities of information. In an anthropomorphic sense, it is an "electronic brain" with an immense "memory".
The actual methods of computer calculation are "moronically" simple. But an incredible operating speed allows the performance of a series of arithmetical operations that in a period of an hour might take a person months to years. The computer is the master of "dumb" repetition. It is most suited to massive population studies and statistical design; that is Data Processing.
There are two kinds of computers, analog and digital. Analog computers existed primarily before 1950 and displayed results of calculations on gauges and dials; like a watch, strain gauge, or speedometer. Analog measures are direct representations of mechanical or electrical events in a continuous display range.
Digital computers operate in a number system. Electronic digital computers coordinate groups of circuits and switches. The binary nuber system consists of two digits, 0 and 1. The computer recognizes 0 as a low voltage and 1 as a higher charge. By combinations of 0's and 1's, an entire memory map of the electrical circuitry can be represented and data can be manipulated both internally and externally.
Analog to digital conversion is the process of converting measured data (eg. serum Na+ concentration) into electronic voltage potentials which are stored within registers and stacks consisting of capacitors and transistors. Computer memory is either resident within the device (eg. RAM) or external (eg. disk drives; hard disk; RS232 and Modem sources; etc.)
Software is the programs and associated documentation which causes the hardware to accomplish a desired goal. The earliest programs were wired into plug boards or punched into cards and processed as a series of 1's (no hole) or 0's (a hole). For example, the instruction 001 might mean to add two numbers stored in separate registers together. This programming is in "machine language" and is a very difficult art to master. Assembly language substitutes mneumonics to represent machine codes in a symbolic fashion. Higher level programming languages ensued, such as FORTRAN (FORmula TRANslator), COBOL (COmmon Business-Oriented Language), BASIC (Beginner's All-purpose Symbolic Instruction Code), etc.
A computer contains "words", a central processing unit (CPU), and input/output devices. The CPU contains ROM (read only memory) which are permanent set instructions which can not be modified and allow the processing of data by preprogrammed rules. Other parts of memory can be read from or written to ( input or output ) which is RAM (random access memory.) Storage is measured in words and bytes and speed in microseconds and nanoseconds.
Understanding Computer Basics:
-The word computer comes from the Latin computare meaning "to reckon" or "to figure out." A computer is a device to help us figure things out, to help us solve problems.
-BInary digiT = 8 bit sequence = one byte
-Words are the the memory structure of the computer and consist of either 1, 2, or 4 bytes depending on the capacity of the microprocessor (computer of either 8, 16, or 32 bits.)
-Central Processing Unit (CPU) contains all the logic for processing instructions etc.
-Arithmetic Logic Unit (ALU)
-Control Unit (CU) is the computer inside the computer which responds to ROM program instructions.
-Buses are electrical pathways thru which the CPU interfaces with peripheral data storage or display devices
-Clock is part of the CPU which produces regular, periodic voltage pulses that initiate and synchronize all the actions of the CPU. The clock frequency or "machine time" is measured in millions of cycles/sec (MHz) and varies between 1-12 MHz. The faster the cycle time, the better the processing time.
-Memory: Working Memory is fixed (ROM)or read but not written to, and variable (RAM) or read & writeable to.
-Core Memory (based on the original permanent nonvolatile magnets.
-MOS (metallic oxide semiconductors), CMOS (complementary MOS)
-ROM, PROM, EPROM
-Interface, keyboard, data rate - Baud (bits/second rate)
Classification of the unknown:
-GIGO... Garbage in, Garbage out; or the answer is only as good as the question, is the first principle of computerization
-The human mind has a memory of 10^20 bits. The computer has 10^7 bits. Electronic circuits conduct 10 million impulses per second (neurons conduct 100) and the computer can subtract serial 7's 1 million times faster than the brain. The computer has small number of exceedingly reliable and fast circuits, and is designed to perform many simple tasks 1 to a few at a time. The brain has a huge number of relatively slow circuits and is better suited to cope with large amounts of information presented and processed simultaneously. The brain does poorly with serial information (eg. comparing all the FEV1's of the population of the United States) and the computer does poorly with the recognition of patterns.
IV. Medicial Perspective
Prescription for Computerphobia:
Physicians lag behind other professionals when it comes to using computers and computerphobia is the reason why.
-Improving technology, lower prices, peer pressure, better education
-17% of Dr. s surveyed in one study used computers in their offices in 1983.
-Medical profession's continued use of dangerously inadequate information-handling methods. This is the "Detroit Complex- that is relying heavily on previously useful but now outdated methods. Reluctance to keep pace with its competition and trends of modernization like the auto industry.
-Benefits of computer such as education, practice aid, data processing and economic savings do not allay the "fear of the machine"...Seminars and workshops (computer education firms) covering multiple topics in intensive sessions. Hands on experience is often the ticket to exorcising computerphobia. Fosters familiarity and comfort with computers and therefore overcoming fear.
-Training disks and tutorials; courseware...AMA/Net...AOA/Net with data banks...Club Mediterranee vacation seminars in a low stress environment
The Physician in the Information Age
-Information age's impact on medicine may indicate a fundamental change in the doctor patient relationship and even change the role of the specialist. It may have a dramatic effect on the income and employment status of physicians.
-Comparing the automobile to the computer industry; if they had developed at the same rate, a Rolls-Royce would sell today for $2.75 and would get 3 million miles to the gallon and it would drive the Queen Mary. You would be able to put 6 of them on the head of a pin.
The potentials -and limitations- of computers in medical practice:
-Ideal system will be small, modular, stand alone, and affordable.
-Automated patient records (AMR: Automated Medical Record) = The "Living Record": improved legibility, accuracy and precision of data; standardization of terminology and increased availabilyt of patient data.
-Problems: tasking of the system, basic science lacking in the decision making process, a scarcity of medical people trained and educated in the computer sciences and computer scientists having problems solving clinical problems...the high cost of development... Potenetials and limitations of computers and answer whether a computerized system will actually benefit office practice.
-Data base may have too much information which may be irrevelent and lead to a data overflow situation.
-Physicians often ask for more information on the system then is needed Ground rules for AMR: identify the decision process and relationships.
-an adaptive person-machine sytem whose precision and efficiency increase with use.
Ethical and Legal Issues in Clinical Computing.
-Mechanism to evaluate whether a computer program is safe for human use.. eg. Laboratory Data transfer programs (PROMIS, HELP); Medical Record keeping (Regenstrief Medical Record System, COSTAR, TMR, ONCOCIN); Medical Information retrieval programs; and Medical Decision Making.
When to use a Medical Computer program?
-If it improves the quality of care at an acceptable cost in time or money or if it maintains the existing standardat a reduced cost. Improved quality is reflected in: improved diagnostic ability; improved therapeutic results; a better patient sense of well being; easier and more rapid access to patient information. If these precepts are present then consideration of computerization may be considered.
Who should use the systems and how?
-The user may be Dr's, students, nurses, PA's, paramedical, lab and office personel, government and insurance agencies, patients themselves.
-The user must be sufficiently trained to provide medically reliable information as input and be able to override a program's advise is it is in error!!!!
-Such demands for an educational background might make some physicians unqualified to use certain computer based aids. Only specialists might be expected to handle unforseen complications.
Specially trained paramedical people may replace the Dr. in certain areas where isolated conditions exist. In this setting a computer based aid system could substantially improve the standard of care.
-Another area of concern is the use of computers by third party to evaluate the performance of professionals in providing care; eg a clerk at an insurance company. There are to many variables in the realm of patient care to be handled by present technology. An untrained asshole could damage the reputation of a health care professional.
-Liability... intent, negligence, and strict liability; the latter 2 are only relevent to computers in medicine because intentional harm is unlikely in this scheme. Negligence occurs when one's conduct fails to conform to standards of care in the profession (failure to render reasonably prudent care.) In this sense it is the responsibility of the physician to either accept or reject the advise of the consultant according to the patient's best interests. The consultant's liability results from rendering substandard advise. The computer as consultant is more complicated liability and for it to be applicable to strict liability it would have to be considered as a product or a good produced by a company rather than a service. Many medical computer programs exhibit elements of both goods and services. It seems unlikely that liability will be found with programs persee by strict liability because physician discretion is applicable to independent judgement. There is also potential negligent liability for the physician if the computer program is not used especially if it becomes customary practice; that is failing to use available technology to the detriment ofthe patient.
-Strict-liability standards recognize errorless function only which is a defective-product oriented standard which makes its use "unreasonably dangerous". Physicians are judged liable on the basis of negligence which allows for the occassional occurence of mistakes. Computer programs must fall somewhere in the middle.
Problem of Confidentiality:
-Patient confidentiality and privacy are based on the fundamental principle of autonomy or respect for persons
-Legal protections...Privileged communications...
-Fiduciary nature of the patient-physcian relationship to act only in the patient's best interests. The law of torts confers a general right of privacy on all persons. If disclosure of health information violates the privacy and cause damage, suit can follow.
The problem of Validation:
-FDA sanctions the program
-How the systems will be sanctioned and continually updated.
-The need for accepted standardization of all aspects of medical care.
-System to allow modification and updates and correction as new information unfolds.
-Diagnoses are rarely made with absolute certainty but with reasonable certainty and considering the consequences of being wrong in the setting of the patient's unique situation with recommendations. Current computer technology is unable to make complex medical decisions.
-A large part of medicine involves human interactions. The judgement of the physician will always be important. It must be a human to decide to run the particular specialty diagnostic program.
-Automated Tower of Babel
-Resolving subtle discrepancies in the history such as when the patient smiles and says I'm having the worst pain of my entire life. Physicians hook up to high tech medicine:
-Business tool in billing
-Satisfying yourself that the computer is saving you a great deal of money.
-Staff training and turnaround is a real problem.
-Source for drug information
Patient Confidentiality - on line
-25-40% of hospital activity is devoted to collecting, recording, communicating, and reporting medical information. Computerized records are more detailed, centralized, permanent, and easily transmittable than paper records. Its easy, speddy access to information is what make it both desireable and problemsom. In this respect they pose a threat to patient confidentiality. Computers are tools and physicians are sworn to ethical standards.
-Researchers may be tempted to make statisical studies from patient data bases. Eg. find names and addresses through the system...
-Blackmail regarding Herpes, etc
-Employers or insurance companies could use data out of context; eg. venereal diseases or psychiatric hx.
-Safegyuarding patient information is the job of the Dr. and not the computer.-Computers may become more and more useful in clinical practice, but in the endm it's the doctor's judgement that really counts.; at least legally.
Clinical Computing in a Teaching Hospital
-Potential encroachment of this technology on the practice of medicine and the doctor-patient relationship.
-"Any doctor who could be replaced by a computer deserved to be."
-The less experience people have with a machine, the more they tend to react to it with both fear and unreasonable admiration.
-In medicine, as fear of the computer wanes, expectations increased and prophesy becomes a substitute for accomplishment.
-1) Information on the system is immediately available. 2) Response time should be rapid. 3) Reliable with no loss of data and a minimal down time. 4) Confidentiality must be protected. 5) User friendly. 6) A common registry for all patient data.
-This study found a huge physician and medical magnitude of use associated with a highly efficient, centralized clinical computing system.
The Computer as a Consultant.
-Adjunct to patient care.
-It is immune from fatigue and carelessness, works day and night without break. They can not discuss a patient's condition with family, perform a physical exam, or verify the data input as reliable...Garbage in = Garbage out.
-The computer constitutes yet another blow to man's ego (Copernicus, Darwin, Freud) for it places him in a continuum with the machines that he builds. We must set about the task of developing humanitarian uses for these inanimate machines.
Clinical judgement AND Computers:
-Where can we use computers and where must we use human beings?
-Medicine is as much art as science and rules can not replace intuition and instinct. Although when we are teaching house staff, we project a set of rules depicting medical science.
-If clinical judgement can not be described by rules, though, it is magical.
-The computer performs more accurately, and rapidly than man.
-Physicians are in a quandry for they don't know what to expect from computers. A perceived threat exists.
-The judgemental process in medicine is a reduction of an enormous set of possibilities to a small number of probable states. The clinical judgement involves this differential diagnosis; the selection of appropriate historical questions; searching for specific signs; ordering the next test; and the choice among alternatives.
Protocol-Based Reminders, the quality of care and the no-perfectability of man:
-There are limits to man's capabilities as an information processor that assure the occurence of random errors in his activities... Sensory overload... Many of the physician's informational tasks are by rote and repetitive to which computer protocols can be better addressed. The amount of data per unit time is more than a physician can process without error. Physicians detected and responded to twice as many events when given computer recommendations as when not... Multiple short deterministic protocols... In order to process more information per patient for the individual physician it would require more time per patient which can not be given in an already saturated schedule. Housestaff are one means by which some physicians spend more time per patient without doing it. Computers will be just another and more efficient physician extender. The individual physician is not perfectable, but the system of care is. The computer will play a major part in the perfection of future care systems.
-Computer programs can organize medical data; permit statistical analysis and research; help physicians manage medical problems; take histories from patients and provide patients with health care instructions
Physicians tend to be skeptical of probabilities; "No disease is rare to the patient who has it." But knowing probabilities force the clinician to abandon imprecision.
The computer in Clinical Decision Making:
-By efficient and comprehensive data collection and organization and the computation of useful derrived functions normally calculated by the physician, the computer has facilitated the physician's traditional role without as yet drastically changing it... The computer as a coordinator, record keeper, and medical knowledge resource
-To sucessfully apply computers in medical practice, physicians and nurses, etc, must accept them, use them, and be convinced of their benefits in daily practice
-A hundred dollar machine now plays chess better than 95% of the population and interprets electrocardiograms and pft's better than most physicians.
-The rough calculations taht physicians must constantly make are performed automatically without possibility of mathematical error.
-Computerized information management giving the Dr. better access to data in his role of bedside caregiver. The usability of a system is ultimately determined by bedside users and not programmers.
Information, Computers, and Clinical Practice:
-The physician is an information manager who acquires, processes, stores, retrieves, and applies information related to patient's history, clinical course, diagnostic and therapeutic protocols, disease patterns, epidemiology, and publishe knowledge.
-Optimal performance is obviously in excess of cognitive human capability. The history is limited by patient's inadequate recall, the degree of physician intuitive problem solving strategies, the time and emotional constraints of the interview, and an archaic record system. Diagnosis is likewise an uncertain science and treatment is often base on habit or familiarity. Therapeutic regimens are rarely suited to each individuals unique circumstance.
-Physican burnout and patient dissatisfaction are inevitable results.
-Information technology must assist the physician in this context. It can not perfect him or replace him. It must become a clinical partnership; one which the medical profession has been slow to accept. This has resulted in the medical profession's continued use of dangerously inadequate information-handling methods.
-Reluctance to accept it is related to habitual behavior; power and ego problems... the Detroit Complex... Self deception.
-"Sell no software before it's time"
V. Pulmonary and CCU Applications
Computers in Critical Care and Pulmonary Medicine:
Computer applications in the healing arts have been slow to develop. There usage in the research setting has revolutionized statistical design and application; but in patient care, computers have had little place. In evaluating the lung (and any other organ system), computerized radiographic techniques (eg. CT scanning, MRI, Positron-Emmision Tomography, Radionuclide studies) have become accepted modalities. Noninvasive assessment of the lung parenchyma and vasculature is now readily available.
Chest Computerized Axial Tomography has revolutionized many aspects of pulmonary evaluation. It is the most accurate means of assessing pleural abnormalities and assists greatly in differentiating an empyema from a lung abscess. It is highly sensitive for the presense of lung nodules and when coupled with a phantom of known density it may be able to differentiate benign from malignant intraparenchymal lesions. This is an area of controversy; especially in evaluation of the SPN. Furthermore, CT-guided fine needle aspiration of SPN is highly accurate in obtaining a tissue diagnosis of malignancy in clinical context. Mediastinal visualization is facilitated by this computerized technique. By identifying mediastinal adenopathy, various staging techniques for bronchogenic CA can be better applied. The chest CT-scan also is applied to therapy for it is a most accurate means of planning radiation ports. A further computer application in this same field are brachytherapeutic approaches delivering high dose, local radiation therapy endobronchially. The chest CT-scan is also of value in determining the presence of pulmonary arterial hypertension and can identify pulmonary emboli in certain instances. Another recent study found chest CT-scanning to be better than pulmonary function testing in diagnosing emphysema by the presence of low density and identifying bullous disease.
In addition to better intrathoracic diagnosis and therapy of bronchogenic CA, computers have been applied in identifying high risk occupational settings and therefore to prevention. By analysis of Tumor registry data the occupational risks of Bronchogenic CA of Asbestos, diesel fuel, etc. have been realized. A recent review used decision analysis to approach the solitary pulmonary nodule. By using Bayesian techniques coupled with retrospective probabilities of mortality, a risk benefit analysis was applied using computerized simulation.
The acceptance of computerized assisted diagnostic and monitoring devices is established. There is no question that these approaches have enhanced this realm of medical endeavor. The application of computer techniques to the bedside and to clinical caregiving has been slow and met with reluctance by both physicians and patients.
Pulmonary Function Lab:
-generation of Flow Volume Loops
-presenting data is an easily interpretable format
-calculating and customizing various staistical normals and comparison with an individual.
-performing tedious calculations.
-data processing and storage for rapid comparison in the same patient.
-portable and bedside techniques using microprocessor technology.
-telecommunication and transmission of spirometry data to central, networked laboratories.
-rapid interpretation and quality control.
-expired gas analysis and exercise physiology; second by second analysis
-expert artificial inteligence programs to interpet, give differential diagnosis and recommend therapy.
-Data management in a lung function unit...Computerized lung function lab
-Sulphur Hexafluoride Washout
-Oxygen washout on mechanical ventilation for FRC
-Sequential lung ventilation with dual ventilators...2 synchronized servo 900 ventilators for differential lung ventilation
-feedback by end-tidal CO2 and regulation.
-reverse I:E ratio ventilation
-compliance curves, airways resistance
-On-line Graphic Presentation of Lung Mechanics during Mechanical Ventilation...pressure, flow, volume data; breath by breath
-computerized ventilator which allows sine, square, accelerating, and decellerating wave forms...matching the ventilator's flow pattern to the natural flow pattern of the individual patient.....alarms and monitoring systems (neonatal)
-an objective system allowing customized and individualized therapy by having a huge memory and open to random functional application. In this sense computers may offer a more humanistic approach than an individual with his particular biases is able.
-Computerized Ventilator for the critically ill.
-continuous MVO2 sats
-end-tidal CO2 curves
-early warning systems
-esophageal ph testing
-bedside monitors and cardiac output computers
-kinetic exponential algorithms
-drug interactions both known and potential
History and physical data acquistion
-pulmonary sounds: crackles, wheezing by spectral analysis and Fournier, etc
-acoustic detection of sleep apnea
-biofeedback and intervention on psychological dependency (cigarrets, ELISA)
-interstitial lung disease
-Swan Ganz Catheter
-rural programs for paramedical
-health care general public software
-Automated Physiologic profiles...Organization of data in a creative, customized format.
-Data Management...ICU system acquisition monitoring and alarms
-reference manual, drug interactions, continuing medical education
-Interactive computer during consultation...patient anxiety, satisfaction; stress vs Intrusiveness on Dr. / Pt. relationship.
-Computer for business accounting; collections; administration...not for patient records, longitudinal data collection, clinical decision making and educational merits. The computer can revolutionize the interpretation and value of medical information. The CT scanner manipulates conventional radiographic and electromagnetic data to generate something new, different and better; that is a cross sectional representation of anatomy. -The computer processes raw data information such as H&P findings, laboratory, longitudinal follow up. It may make data collection standardized to allow large population studies to be generated from the community setting. The office computer may expand our currently narrow base of information on common diseases. -Cookbook medicine and the specialist... vs the insensitive automaton mechanically following the commands of an insensitive computer. Systematic approach to clinical applications and treatment allowing meticulous attention to detail... eg. diabetic care... more consistent and more detailed and more uniform data about patients. The computer can assist the Dr. in the limitless recall of details.
-Using networking to access similar patients in a community setting to customize the paradigm textbook approach to the unique case. -Risk profiling... Health Hazard Appraisals... DRG's
Computers and Behavioral Medicine: Automating Lifestyle Changes
-Computers can individually tailor instructions specifically to the individual and in so doing explicitly teach people new health related behaviours. In smoking cessation, some patients prefer computerized to conventional treatment.
The Acid Test:
-Computer based ambulatory monitoring that measures esophageal pH levels
Estimating the probability of Malignancy in SPN
-The likelihood of cancer is calculated from the literature based on likelihood ratios from the general population, age, size of the lesion and smoking history.-eg. In a 55 y.o. man with a 1.5 pack/day smoking habit and a 2.5cm nodule: The probability for CA in a clinical setting is 40% and the likelihood ratio is (.4)/(1-.4):1... For size the probability of CA is 34% and benign is 20% with LR=1.7 (34/20)... For age 20% w/ CA and 14% benign; LR=1.5... Smoking shows 38% CA and 19% Benign w/ LR=2... Therefore Odds CA= .7*1.7*1.5*2= 3.57:1; which is 78% probability of cancer ((3.57)/(1+3.57))*100
-The management of SPN demands decision between IMMEDIATE SURGERY, BIOPSY and OBSERVATION. This decision is influenced by the consideration of the probability that the nodule is malignant, risks of surgery, accuracy of biopsy technique, fear that delay in resection will forfeit curability, and the patient's attitude to these options. The results of a complex decision analysis found a striking similarity between average life expectancy predicted for each of the management strategies. One method was only determined better than another in terms of weeks added to survival. At very low probabilities of cancer, OBSERVATION gave a slightly better average life expectancy, 38.27 yr (oposed to IMMEDIATE SX @ 38.04yr)
-They conclude that when alternative strategies produce such small differences in survival, then other considerations besides life expectancy should be given relatively more weight in deciding how to manage the nodule such as personal preferences. The decision analysis showed that decisions between strategies were a very close call. Something clinicians have known already for some time.
Computer Guidelines for Pulmonary Laboratories
Advantages of PFT Lab Computers:
1) Automation may reduce cost in time and money and increase accuracy
2) Assurance that standardized procedures are followed
3) Significant reduction in major measurement errors
4) Storage and retrieval of information quickly and efficiently
5) Implementation of automated calibration and system check procedures within the instrument
6) Standardized and consistent interpretation of results
1) Incremental increase in initial cost of equipment
2) Requires more careful training of personnel
3) May limit the flexibility of some testing procedures
4) Inability of the user to update and correct the software
-The analog spirometer's accuracy and the microprocessor's must now be considered in overall accuracy.
-Errors may go undetected for months
-Suggested guideline for quality control of pulmonary lab computers was porposed.
Computerized Reference Management: Searching the Literature
-Addition of 225,000 articles yearly to the medical literature... 6,000 to 7,000 scientific articles are written each day and the number increases by 13% per year. If the physician reads and average of 3 hours per week, he must 13,000 article an hour to keep up. Over one third of surveyed internist were dissatisfied with their ability to keep up with the medical literature. The Index Medicus is frustrating and time comsuming. Medical librarians searches are expensive and take days to complete and can only be on specified topics.
-MEDLINE contains 4 million references from more than 3200 journals covering 1966 to the present. When using Index Medicus, the article must be pulled and reviewed; on MEDLINE abstracts can be quickly scanned and selected faster.
-MEDLINE; MEDLARS; COLLEAGUE; DIALOG; COMPUSERVE; PAPERCHASE--- All systems are comparable and capable in retrieving pertinent articles addressing a given clinical problem. Differences existed in the ease of use, cost, searching time, and the proportion of articles relevent to the topic.
VII. Artificial Intelligence
Computer AS Colleage:
-Puff an artificially intelligent program with expert reasoning to interpret pulmonary function tests.
-includes a health questionnare.
-Physicians agreed with Puff in 95% of cases. Physicians agree with each other in one study 92% of the time. The computer is strictly consistent and won't miss certain subtleties.
Artificial Intelligence: Competitor or Helpmate?
-Machine could ape man.
-Rob people of their functions
-Fears of computers as enormous, heartless, autonomous breathing brains leads to an instinctual rejection of computer assistance by physicians.
-Argument that computers will help and not replace physicians.
-Health Hazard Appraisals
-The general internist encountering a complex diagnostic problem must have more than 300,00 items of medical knowledge in memory to do the problem justice. The average Dr. can not consider more than 5 clinical hypotheses at a time; the computer can consider virtually all possibilities.
-Compensating for human inadequacies, the computer extends the physician's ability to perform well, rather than diminishes it with rivalry.
-All any computer does is apply the rules and criteria established by human beings. The apply the same rules rigidly, every time without variance; something people do not do well. Consistency!!!
-At times an expert computer program may outperform its inventor; indeed it was written when he was in his lucid moments and fresh with the material at hand.
-The physician is the source of medical decisions. He can employ laboratory tests, textbook reference, articles in the literature, conversations with colleagues and computer programs; but only he is to blame if his decision proves unsatisfactory.
-Funnel representing clinical judgement and the shrinking cognitive span. At the beginning the physician's comprehension must include the totality of everyday world. Common sense well acquainted with the world. Computers can take nothing for granted.
-Point B where alternative have been narrowed down, is the most applicable for specialized computer application such as in the application of formulas or kinetic modeling.
-Expert systems are better than specialists in many instances.
-Expert systems operate in a small and well structured task domains that were initially organized by human judgement. None is functioning at point A. The human being must decide to enact the use of an expert system in the first place
-About 2 million "facts" would comprise great intelligence in a computer system encompassing all of internal medicine with its subspecialties.
-Once point B has been reached in clinical judgement; the irrelevant facts have been filtered off and the medical problem has become well structured. History taking is the most crucial step in the entire diagnostic process and separates the exceptional dr from the less able. There are about 3000 diseases and medical conditions. The value of a positive response that a patient offers is much more powerful than a negative one (the former would include a limited number of conditions in which that complaint occurs while the latter would exclude the few conditions while included still a huge differential diagnosis.) The diagnostic or selection power of a positive response is 100 times greater than a negative one. The amount of information obtained from a random inquirey about symptoms (mostly negative responses) would be small. A computer is faulted by this problem. Irrelevent information is frequent. Common sense and general medical knowledge is required to determine what is relevent in a clinical situation. Computer systems to begin at point A must represent the entire world ("you can't have everything; where would you put it.")
-Artificial intelligence system function reasonably well in microworlds which are carefully structured
-Inorder to function close to point A
Computer-Assisted Medical Decision Making.
-Computers may never replace physicians, but they can and will assist their decisions.
-Assist, not replace.
-The ultimate decision which requires personal and occasionally subjective considerations, still will rest with the practitioner.
-"Clinical judgement must take precedence; these are only suggestions."
-Physician often makes the crucial decision before turning the machine on
-Computers don't tire, overlook, forget, or equivocate
-Physicians should be able to evaluate the computer's reasoning process and make conclusions as to the validity of its conclusions... How human cognition works.
-Algorithmic, heuristic and payoff approaches.
-There are very few difficult decisions in medicine. Most programs end up telling the physician what he knew to begin with.
-If the physician doesn't think of the correct diagnosis; he can not make it. A computer could generate a list or a menu to aid in this process.
-Usefullness also depends on the demographic eg if experts are available who are more competent that an expert system.
-Physicians to structure and program their own sytems.
-Physicians are not stick in the muds and when decision systems can do something that doctors can't do, every doctor will have one.
Expert Systems: robot physicians of the future?
-Conventional computer roles involve repetitive tasks such as word processing, electronic filing. automated billing and lab information.
-The science of imitating human intelligence with machines is called artificial intelligence (AI).
-Simulation of human intelligence is currently unrealistic, since machines are not able to match the "fringe consciousness" of human (the ability to focus on aspects of a problem without losing sight of the total picture.)
-Productivity tool...Decision support system...Expert System...
-Heuristics is the "rules of thumb" approach to solve a problem, limiting the deductions and inferences needed to solve a problem.
-Decision support systems will permeate the medical profession such as database managers, diagnostic and therapeutic algorithms, etc.
-An expert system is hampered by the same criticism that a subspecialist is; and that is the details of a specific medical problem obscures a global overview of the patient's complete situation. The expert system can not frequently deal with the global situation adequately. Therefore the physician will remain the principle decision maker.
PUFF: An Expert System for Interpretation of Pulmonary Function Data
-Artificial Intelligence system which captures the specialized knowledge of experts and is applied to perform difficult tasks.
These are tools for the future.
-PUFF interprets PFTS automatically without need to enter data in or user interaction. It represents domain-specific knowledge in the form of production rules, and performs consultation in that domain... Written in INTERLISP
-Puff has been operational since 1979. Some of its limitation regard basic computer inabilities which regard pattern recognition.
-The overall rate of agreement between two pulmonary physiologists on the diagnosis of PFT data was 92%. The agreement between PUFF and its expert designer was 96% and 89% with an outside expert.
-If the task, domain, and researchers are carefully matched in a medical situation, a system can be designed to perform reasonably well in a complicated skill.
-Artificial Intelligence is "using the computer to perform tasks that are considered to require intelligence when performed by humans".
-The problem oriented approach to medical decision making and diagnosis demands a rigorous and complete problem list approach. The unaided mind is unable to handle this volume of data as well as a computer could potentially. The computer can consider 40 causes or management strategies as easy as 5. It will never forget one or leave out a pertinent question. In this sense it can render a unique representation (or more so than a human with his limited recall and built in biases) of each patient.
-The computer's output is only as reliable as the bits of data input received.
VII. Programming Exercise
IX. Pocket Computer Applications
X. Teaching with Computers:
-Teaching files for interesting cases, research and publication and CPC's.
-Computer assisted teaching offers a means of tailoring the education to the individual. They set their pace and the computer keeps score; also noting progress and weak areas which need more work.
-This entails more preparation time for the teacher and also demands access to computer terminals for the students. Also problems in updating information are important.
-High-Quality video images are needed to teach certain pattern recognition as with pathology and with radiography. Interactive Video systems may be the answer. Video still frames are controlled by a computer... Videodisc is much more interactive with a faster response time and a higher quality of still images. The lase discs are very expensive to produce. One hour production costs $20,000.00 - 50,000.00
-Audiotape and slide projection can likewise be integrated into this schema. The student population must be selected carefully. Medical students, house-staff; CME programs.
-Bibliographic Searches access on line indices referenced by author, abstract...
-Integrated Academic Information Management Systems Program of the National Library of Medicine
-Memory holds 150 megabytes (or 4 gigabyte = 4 billion bytes) or 54,000 still video images
Education on a Silver Platter
-The student's own reactions and knowledge control the pace and amount of drilling involved.
-National Library of Medicine has radiology case studies, histology and pathology... CPR... Miles Laboraries... Calcium Channel Blocker... Arrhthmias
Computer-Aided Learning using a Laser Videodisc Player:
-PDX is a program for the diagnosis of pleural effusions (Elizabeth Rich of Case Western Reserve)
-University faculty which exercise monopoly over medical education are reluctant to admit the need because they have housestaff toiling ling machines to get comparable results. Medical students are bombarded with lecture and quickly forgotten facts rather than being trained in becoming problem solvers, self learners, and information managers.
-Medical schools imperitive to make physicians computerwise and not fall behind every other educational field.