Better Medical Diagnoses Through Software
from the doctoring-2.0 dept
We’ve written in the past about how doctors and hospitals can sometimes be slow to adopt new technologies. That isn’t always true, of course. Medical equipment is quite advanced, obviously. However, at the end of the day, a diagnosis of the problem is left up to the knowledge of the doctor on hand — often leading to big mistakes. Doctors have little on the line when making a wrong diagnosis (potentially malpractice, but only in the worst cases), and so the incentive to them to absolutely get it right isn’t quite as high. Obviously, most doctors have their reputations to worry about, but a mis-diagnosis here and there doesn’t kill a reputation — though it may kill a patient (which isn’t a pleasant thought if you happen to be that patient). One company is trying to change that, by creating a software product that will help doctors better diagnose health issues, based on a bunch of different symptoms. Especially when it comes to rare diseases, a doctor may not be able to easily put all the symptoms together in a way to recognize what’s really happening. The software tries to help. However, the software is pretty expensive for hospitals, and without that incentive to more accurately diagnose patients, hospitals may not feel it’s worth it. Of course, plenty of patients who feel misdiagnosed might like to get their hands on it, but that might just lead to more cases of cyberchondria — the belief that you have a disease because you read about it on the internet. Still, there are cases where such internet research helped figure out where the problem was, while doctors completely missed out.
Comments on “Better Medical Diagnoses Through Software”
IF it helps
Software is only as good as the data it is fed. A machine that just cranks out a “diagnosis” based on a fill-out form is worthless. Clinical diagnoses rely on a great many intangible factors that cannot be easily quantified, or filled out on a data entry form. If clinical diagonoses were that easy, the job would have been automated a long time ago.
Re: IF it helps
It isn’t worthless if it turns up a list of potential conditions that the doctor might not have been aware of. According to TFA that is exactly what happened.
Re: Re: IF it helps
Could the software know about all the obscure medical problems out there, though? It will only be as good as the people who wrote the software. How conservative or liberal should the software be? Would you want the software to be giving out diagnoses of “rare leukemia” often, or not often enough? Rare diseases, by their very nature, have highly variable symptoms that can masquerade as other illnesses. Also, software written in, say, Silicon Valley will not be very good at spotting “rare” regional illnesses that are common in other parts of the country.
Re: Re: Re: IF it helps
Any expert system is only as good as the experts (doctors) it was modeled on. You can put a lot more rules in a DB than in a doctor’s head and have them applied more consistently.
Re: Re: Re:2 IF it helps
How do you consistently define a disease that takes many different forms, though? If you ask 10 experts, you’ll get 10 different answers. Even the definition of “hypertension” remains in open dispute among medical researchers, for example. A patient that does have an unexplained disease would need to go to a major medical center anyway, before anyone is satisfied with the judgement a piece of software gives. You would not want to pump your 5-y.o. child full of chemotherapy drugs that makes his/her hairs fall out, because a computer said to do so.
Re: Re: Re:3 IF it helps
My opinion is that software like this is usefull in triggering alarms, certainly not to replace a doctor’s diagnosis. What is startling is that, as noted, software like this is so expensive. Where are the government subsidies when you really need them?
“Still, there are cases where such internet research helped figure out where the problem was, while doctors completely missed out.”
20 yrs of the same exact tests being done still no concrete diagnosis and it took me maybe a week of study on-line to narrow it down to the chromosome. I’m thinking of asking directly for that sort of test at my next visit just because I’m sick of no one caring what’s wrong with people. It needs to change and not like tomorrow. You have to take then into effect that only 40% of American doctors even have basic computer skills. What is this telling you it would take to get this into play. About a generation or sos.
I’m sure they are working closely with E Medicine and other governmental sites like these. National Institute of Neurological Disorders and Stroke.
>20 yrs of the same exact tests being done still no concrete diagnosis and it took me maybe a week of study on-line to narrow it down to the chromosome.
May I ask if this was a microdeletion/insertion? Many of these diseases are newly discovered, and there is no shortage of unreliable web sites out there.
Im sure some people misdiagnose themselves, but the last time I went to a Dr because I didnt feel well (thought I was getting sick, wanted to cut it short), they gave me Amoxicillin after I explained the symptoms (though it didnt seem like she was listening that closely), and didnt really tell me much about the drug.
The night I got REALLY sick, and I stayed really sick for 3 days. Then I went online and read the symptoms for both viral and bacterial infections, and all my symptoms sounded a lot like viral and a lot like the opposite of bacterial.
Anti-bacterial drugs dont help with viral infections at all, the Dr didnt do a throat culture so now any evidence of bacteria would have been destroyed, and I was allergic to Amoxicillin.
So because she didnt listen to me, I didnt know better, she didnt test, and she didnt tell me what to expect with the drug, I had a really miserable 3 days with probably no benefit to my health at all.
Only being able to look things up online clued me into the fact that this was all wrong. Patients having more infromation is a GOOD thing, that some people fuck that up is an education problem. One that wont go away because we dont educate people about health in any significant way, but I still want the information to save myself from doctors uncaring careless mistakes.
Imagine if the hospital itself was aware of your presence while you were admited. Every heart beat, every breath, every test… Then was able to help to diagnose your health issues. Over time (and considerable input) the software would learn and adopt giving ever more precise diagnoses; We are talking A.I. that does not require much beyond setting it up and turning it loose. The only problem I see is privacy issues, and what happens if someone hacks the system. Just imagine if it runs on windows…lol
Re: learning hospital
Sounds nice, but software that “learns” as such is prone to make false conclusions, finding statistical associations where there aren’t any meaningful links. For the foreseeable future, computers should remain data repositories, and judgement is best left to well-trained humans. We’re not talking about the stock market or a car factory here — lives are involved, the stakes are too high for social experimentation.
Re: Re: learning hospital
Good point dorpus; I would like to amend my previous post and say that all final decisions will of course be made by the doctors. Maybe include a kind of expert colaboration feature that allows your local doctor to send the information to experts for opinions; the more experts the better…
Re: Re: Re: learning hospital
I laud your good intentions. The expert collaboration you describe is otherwise known as “medical research”, which gets bogged down in debates, clash of egos, what have you. If a patient so desires, s/he could talk to an authority in the field about the problem — but that also comes with the risk that the patient will become the researcher’s guinea pig, and the researcher may not necessarily be good at caring for a patient. A patient’s problem may not necessarily be what they think it is, or want it to be. The best thing a patient can do is try the opinion of experienced practitioners.
Re: Re: Re:2 learning hospital
Do the doctors not “learn” by statistical approach. What I mean is doctors analyze and compare just like a computer A.I. approach would. If you include the opinions and expert analysis while retaining your local “known” doctor; is he/she not able to make more informed intelligent decisions. Information overload is one of the main issues facing doctors; no one doctor can know it all so retain technology to help with situational awareness and get correct diagnoses more of the time. I keep coming back in my mind to the times when HIV infection was considered by many to be a cancer. I wonder how long it took for the medical establishment to overcome sterotypes and really apply the knowledge of the day. I feel A.I./Data repositories would be much more impartical forcing sterotypes to the background and bringing root causes to the fore.
Re: Re: Re:3 learning hospital
The thing is that it’s good that there are ways by which doctors could build on the information they have in order to care for their patients better. I guess that even though there are some security issues that might arise because of such technologies – that is a trade off. Unless, of course, there would also be better ways of implementing security.
I agree that information overload can be daunting but at least the way I see it, the potential of such repositories to help narrow things down would be great.
Re: Re: Re:3 learning hospital
Doctors do certainly learn how to accurately diagnose problems from clinical experience. However, by agreed-to conventions in the medical field, statistical generalizations leading to new theories are best left to medical researchers who can conduct their research (observational, prospective, or experimental) under carefully controlled circumstances. A clinician’s perspective is certainly welcome for aiding researchers, but what s/he observes anecdotally in the clinic may contain biases that s/he may not be aware of. I’m in the field of biostatistics, which is all about (purposefully) looking at the numbers from an arm’s length. When the data is analyzed objectively, there often arise patterns that defy a clinician’s intuition. Of course, the statistical associations may be coincidental, so there is a complex art of looking at the numbers and discussing the findings with researchers from multiple fields, to arrive at the most scientifically plausible conclusion.
E.g. a project I’m working on right now says that older people are at lower risk of severe melanomas (a type of skin cancer). This would seem to defy the intuition that older people are more prone to cancers. But it is quite plausible, when one considers sociological factors that 1. old people did not get suntans in their youth, and 2. old people tend to go to the doctor more often, so their melanomas are treated before they get too bad.
I’m also finding that melanomas on the lower half of the body tend to be worse than those on the upper body. Is this because the lower body has higher lympathic pressure that pushes the cancer cells out more, or is it because melanomas on the lower body are harder to spot? A black spot on your butt will tend to be discovered later than a black spot on your arm, for example. I’ll be running more sophisticated statistical tests to derive a conclusion, which will then have to be repeated on other clinical data to verify the findings.
Re: Re: Re:4 learning hospital
I would love to see a medical analysis system that incorporates all known medical conditions with frontline research to give my local doctor expert analysis capabilites. Is such a system even possible? I guess only time will tell; would I participate in such a program regardless of the privacy issues? I might if I was facing a life ending illness/condition…
Another hurdle I just thought of is the people/corporations paying for the pure research are going to want some ROI (unless it is some government grant). If this research/knowledge can save a life, is it in the best intrest of the guy who you could save to withhold it just because its new and not proven. I guess I am moving in a ethical direction with all of this but I want to read the responses.
As for dorpus’ current research project does any of the melanomas found really lower life expencency of someone who is “old”, What i mean is how old is old and is the body not running out of steam at this point anyway?
Also in the course of your research how do you model your findings, are you using technology to help with the research? Is your “frontline research information” available to the local doctor? And if so in what format is it available. e.g. Say I am a dermatologist specializing in melanomas, do you include my ideas and findings in anyway, is there a feedback loop between the pure researcher and the frontline doctor/practictioner???
Re: Re: Re:5 learning hospital
That gets into the issue of the data’s reliability. Data that is collected haphazardly from clinics will contain unknown biases and is not looked upon favorably. (It will be used sometimes.) There is the issue of HIPPAA compliance, a new law which protects the privacy of patient data. In most cases today, the patient must agree in advance to have their data used for the purposes of medical research.
Most medical research is funded by grants. Pharmaceutical companies do pay for their own (expensive) clinical trials, but they do not usually pay for basic research.
Melanomas are a treatable condition when detected early. I am modelling my findings using well-established mathematical techniques from the 20th century. The computer will speed up my computations, but all meaningful inferences must pass through human filters.
is UK based decision support software that I was a beta tester for. It’s similar to DxPlain and suggests diagnoses based on physical findings and lab results. It’s only as good as what you feed it, and that’s where it falls down. You have to know what few data points it really needs, because if you feed it too much, it gets confused and suggests many other diagnoses that aren’t relevant. What’s the use if you need to know which data points to input initially, anyways. It’s likely you already know the diagnosis at that point.
preventive medical records proposal
4447, Chemin du Lac Sept-Iles, St-Raymond, (Québec), Canada, G3L 2S6
To whom it may concern…
Médick which main activity was physical evaluation, has a large bank of statistical data that could be a real source of information for any research unit on health and fitness.
This valuable information has been adding up for thirteen(13)years of professional activity. Some thirty(30) physicians, and fifty(50)licensed nurses, have contributed to thirty six thousands(36,000) files of tests on patients, in which we find in each file :
Complete medical examination:
Electrocardiogram(E.C.G) in two(2) phases at rest and exercising.
(including the reading paper)
A fully detailed report we find:
The information showing the cardiac frequency evolution on different scales of efforts.
A rating of blood pressure at different stage of physical exercises.
A recuperation factor of each individual and percentage of fat found on each person.
These files provide hundreds of data of each individual who has been thoroughly examined.
A clientele of people, aged from 19 to 68 years old (averaging 35 to 40)
Computerization of these statistics, should be of great value to a prospective buyer. This a paramount to a research program in very field of activity related to the health and fitness of any population.
Those exclusive and confidential files and statistics are for sale, and can be
obtained by quantity of 2000 units, and we guarantee his exclusivity to the
For any other inquiry, you are welcome refer to the person mentioned below.
M. Claude Audibert , président
4447 Chemin du Lac-sept-îles
Quebec, Canada, G3L 2S6
4447, Chemin du Lac Sept-Iles, St-Raymond, (Québec), Canada, G3L 2S6
Objet : recherche médicale.
A qui de droit,
Ayant dirigé pendant plusieurs années les cliniques médicales Médick, dont la principale activité fut l’évaluation médicale de la condition physique; nous disposons d’une banque de données et de renseignements pouvant se transformer en une importante recherche sur la santé.
Il aura fallu plus de treize années de travail professionnel pour y arriver.
Ont collaboré, une trentaine de médecins, une cinquantaine d’infirmières autorisées, pour établir 36,000 dossiers/patients du même type.
Nous y retrouvons dans chacun des dossiers :
Un complet majeur, l’examen le plus complet .
Un électrocardiogramme au repos et un autre à l’effort .
Un rapport détaillé et chiffré laissant voir l’évolution de la fréquence cardiaque à plusieurs niveaux d’effort.
Une réponse de la tension artérielle à différents paliers d’effort .
Le facteur de récupération de chacun, le pourcentage d’adiposité des individus.
Ces documents donnent accès à des centaines de renseignements sur chacun des participants.(environ 8,000,000) d’informations confidentiel.
La clientèle varie entre 19 et 68 ans pour une moyenne d’âge d’environ 35 à 40 ans.
L’informatisation d’une telle banque de données, donnera à son acquéreur, des milliers de renseignements permettant une recherche poussée dans différents champs d’action, hypertension, cardiovasculaire, recherche sur les populations etc…
Voici un atout supplémentaire susceptible de révolutionner votre recherche.
Nous vous proposons un instrument de connaissances incomparable, cette banque de données est à vendre ,et peut être obtenu par quantité de 2000 unités, et nous garantissons son exclusivité à l’acquéreur.
Pour tous renseignements supplémentaires et professionnels, vous pouvez nous rejoindre
M. Claude Audibert , président
4447 Chemin du Lac-sept-îles
St-Raymond , Portneuf
Québec, Canada, G3L 2S6