Richard Smith: Computers take histories better than doctors – why don’t they do it more?

Richard Smith

Here’s a simulated doctor patient consultation that took place today at the Royal Society of Medicine. A 65 year old woman (cunningly disguised as a bald, male professor from the Mayo Clinic) who is known to be hypertensive and on treatment says that her blood pressure has gone up over the last 10 days. An Australian doctor plucked from the crowd starts the consultation.
“Why do you think your blood pressure might have gone up?”
“I don’t know.”
“Has anything changed in your life.”
“I don’t think so.”
“Are you under any stress?”
“Is everything OK at home?”
“Have you changed your diet?”
“Have you been taking any new medicines?”
“It may not mean anything. Your blood pressure isn’t that high. Keep measuring your blood pressure. Let me know if it increases suddenly. Otherwise, come back in two weeks.”
This case is based on a real one, and the patient had given her history to a computer before seeing the doctor. The computer – which never tires, has unlimited time, and can ask prepared questions in a way that doctors often don’t remember – had elicited all that the doctor elicited and much more. In particular it had asked about known but rare causes of raised blood pressure and had discovered that over the past 10 days the women had been eating imported licorice. She didn’t tell the doctor because it never occurred to her that it could be important.
In the real case the doctor knew that licorice could have this effect – and it leapt out at him when he scanned the summary of the history taken by the computer. Doctors are good and fast at scanning such histories to pick out the crucial facts, which is why a partnership of computers and doctors can be so potent.
One reason that computers take better medical histories than doctors is that doctors do it badly. Doctors, said Professor John Bachman from the Mayo Clinic at an RSM conference on Using the Internet to Practice Medicine, use jargon and miss 50% of psychoscial psychiatric problems. Studies at the Mayo Clinic, indisputably one of the world’s leading clinics, show that they miss 54% of patient problems and 45% of patient concerns. In half of consultations the doctors and patients did not agree on the presenting complaint. Doctors do, of course, have limited time, although the Mayo doctors consulted for 30 minutes.

Computer interviews are structured and provide more complete data than interviews by doctors. Many studies have shown this. Computer interviews can be done at the patient’s pace, and most patients like computer interviews. Patients will also tell computers things that they won’t tell doctors, even though they know that the doctors will see what they have told the computer. Plus the computer records exactly what the patient says, and it’s easy to incorporate validated scales into the questioning.

Dr Allen Wenner, a consultant physician from South Carolina and one of the pioneers of the use of computer interviews, told the story of a middle aged man who filled in the MAST questionnaire for alcohol problems and scored in the “alcoholic” range. Wenner was able to point this out to the patient, something less judgemental than him saying to the patient “I think that you may be an alcoholic.” Nevertheless, the patient was furious and told his wife that he would never consult Wenner again.

But another advantage of computer interviews is that the doctor can give a copy to the patient as well as keep a copy in the electronic patient record. The man took home his interview, and later his wife asked him about the consultation. She pointed out to her husband that it was him that had given the information that suggested he was an alcoholic. It wasn’t a judgement of Wenner. Given time the man came round to the idea that he needed to do something about his drinking, and returned to Wenner.

Computers can also interview family members, have no problems with different languages, don’t get impatient, can use multimedia, and provide marvellous data for research.

Almost everything I’ve written so far was known in the 60s, and it was all in an editorial in the BMJ in 1988 written by Mike Pringle, who subsequently became chairman of the council of the Royal College of General Practitioners.
Yet computers are hardly used in Britain for taking histories, although they do begin to be used more in the US.

Why? The answer seems to be that we don’t change until we have to, and we haven’t yet had to. It may be, however, that financial cuts in the NHS will force change in a way that generous funding of the NHS has not. And help is at hand – from people affected by thalidomide. The Thalidomide Trust helped organise the meeting – because people affected by thalidomide get a raw deal from the NHS and are desperate to find better ways. Consultations through the internet incorporating computer histories could be one of those ways.

Competing interests: Richard Smith is the chair of Patients Know Best, a start up business that aims to use information technology to enhance the clinician patient relationship. I am not paid but have some equity. The company has a contract with the Thalidomide Trust.

Richard Smith is a former editor of the BMJ.

  • amcunningham

    Thanks for this. The Pringle editorial has not been cited widely but I can see some more recent research. Which are the best pieces?

    Something incorporated in to the medical records sounds like a great idea. How different will our jobs be in 5 years?

  • Richard Smith

    John Bachman, who spoke at the meeting, wrote a review in the Mayo Clinic Proceedings in 2003; 78: 67-78.78.

    Ray Jones, who also spoke at the meeting, did a review for NHS Choices, which he said we could find by Googling “NHS Choices report Jones Plymouth.” I did that but couldn't find it. You could do better, I'm sure.

    Then Azeem Majeed, professor of primary care at Imperial, has emailed me to say that his team have written a review, which the BMJ rejected. He's now seeking a home elsewhere–but would probably be willing to share a copy.

    Finally, the whole meeting was videod and should be available soon.

    There is lots of evidence of benefit from computer administered records. the problem, as always, is getting people to change.

  • Ellen Grant

    Dear Richard,

    Computor histories are only as good as their questions. I found the history questionaire used in the 1970s at Charing Cross Hospital migraine clinic was brilliant for information about what was already known to relevant. This included present and past medications, smoking and hormone use, cheese, chocolate, oranges and alcohol. However, if pre-programmed questions do not include what is commonly eaten at every meal each day, the effect of a high intake of the commonest hidden allergen – wheat – will be missed.


    Dr Ellen Grant

  • Dear Richard

    If ever we “had to”, it’s now. Consultations need to be liberated from the endless drudgery of single-finger typing and data entry (by the majority). We could have head-up displays instead of the present PC screens that require gaze avoidance and keyboards similar. iPads and pipeline products, including AI evidence-based decision-making assistive software, can bring touch back into the equation, as well as face-face engagement -> death to the keyboard, using increasingly sophisticated voice recognition technologies. These will also enable far more accurate (and cheaper to execute long-term) qualy-quanty research through the use
    of different software products that can already analyse and score/rate both visual and verbal inputs.

    Is it too much to hope for that such technologies will also show, through research, that the most important ‘technologies’ in the consulting room are those of the participants in conversation, whilst they are using their brainminds? Einstein should know – he said it first.

    Kind regards

    A loyal follower of the Smiths

    (Dr) Chris Manning

    We could also have far more bored and univolved people in ‘communities’ and lowered skilled (and therefore paid) staff entering data (where required), instead of paying GPs a flying fortune to do it. I’m thinking espy here of the QoF – most of which could be delivered by trained pigeons tapping suitably organised and arranged
    sphygs and blood-test form fillers – of course suitably incentivised with Trill?

  • Tim Benson

    Delighted to see this raising its head again.

    More than 30 years ago the Department of Health funded a trial of six computer-assisted history-taking projects using software developed at the National Physical Laboratory (MICKIE). This showed that on one hand these systems did everything that their supporters claimed and on the other hand that developing such questionnaires is non-trivial. Much of this work is referenced in Pringle's editorial.

    Given 30 years of Moore's law and the wide availability of inexpensive touch-screen tablets, it is clear that technology is not one of the issues that has held back adoption.

    Tim Benson

  • I have long held, in jest, that in order to get good service from a doctor (or accountant or mechanic…) I have to know at least as much about their field as they do.

    As a hard-core geek and veteran computer programmer, this story has helped me understand why. Better yet, I now know a way forward.

    Thanks for this story!

  • Pawan Randev

    Hi Richard

    I believe Ray Jones paper is at this link

    I wonder if this methodology may be applied to help in earlier diagnosis of cancer, given the recent paper in the BJGP about key cancer symptoms.

  • Richardswsmith

    You can learn more from what was an interesting and well done meeting (which most aren't) at:

  • This is spot on. I am already making sure that the symptoms from the paper (which are already in the questionnaire) are flagged with a reference to the paper, in the output from Instant Medical History – the system that Professor Bachman uses. I am the UK representative for this and have anglisied the question set. There is already 20 years of medical knowledge built in and it is being continually updated. The lastest work on Ovarian Cancer is another good example. Did you see

    Do get in touch if you wish, vbw Richard

  • Cdupont

    I do not believe that problem orientated records are a great help in managing patients. We accumulate a huge mass of largely irrelevant detail. I am sorry that Dr. Richard Smith exhumes the Weed method initially advocated around 1972.

    Competing Interest: If computers take better histories than doctors and “Patients Know Best” – I am redundant and so is the rest of the medical profession!

  • Ady

    I attended this meeting as a patient and hoped it would help bring a system like IMH into service with the NHS as soon as possile.

    Being a man and ex services going to the Doctor is an alien act it really goes against the grain but playing with a compiter does not. I rang NHS direct while having a heart attack and was told after just over an hour to ring my GP in the morning, The reason I mentioned IMH is that about 6 months later I became freinds with Dr Richard Sills who showed me IMH and allowed me to have a go with it. if i could have accessed that software when I was suffering my heart problems i would have been in hospital at least 3 days sooner and would not have had such a serious event.

    People are not embarrased answering questions with a computer as far as they are concered a PC can't TUT and give a disaproving look no matter how hard a doctor tries they may so people tell the truth more often. This gives a GP more accurate information to work on.

    The time saved and the accuracy gained by using this software will remove a load of pressure from over burdened medical professionals, But the biggest thing i do not understand is why every doctor in the country is not screaming out for this to be used if a patient withholds information to the GP and something goes wrong all the GP has is the notes from his examination with the person to prove he did his job properly. with IMH a record is produced that gives all of the questiones asked if the patient provided missinformation to the PC it will be recorded and can be used in court, How much wil this reduce parctice insurance? Tie that in with the Talking Point software that was at the same meeting and GP'S will be able to take up golf.

    Looking from the outside in this seems a no brainer to me.

    about 4 years ago after attending another meeting where IMH was discussed Muir Grey and other senior NHS officials said it was Imorral to delay the implimentation of a working software package any longer!!!!


  • Sarah Bruml

    I was at the meeting and have just signed my practice up to PKB. At the moment a number of our patients have chosen to use freestyle email to me as their GP thro the practice website; this raises issues about confidentiality – a patient emailed me yesterday from her work email address asking about counselling services for depression and suicidal feelings – I dont think she really understood the consequences of putting this sort of personal detail on a work email address. PKB could have allowed her to communicate privately with me. Adding an 'instant medical history' on line would not have made a diagnosis or treated her – but it hopefully will allow me as her GP to get a full history – gathered with various validated questions eg PHQ9 – completed before she came to see me and she could complete it in her own time; at the consultaion we could then focus on important issues and make an agreed plan. I believe the internet offers one way forward for improved and accurate history taking in ever more squeezed clinical time. Let us see commissioners in Primary Care support pilots of this technology in the UK – it may well be effective.

  • ady

    Hi Cdupont,

    a computer based interview will not replace any medical professional but will give him more information to do a better job with less stress.

    The reason you all have to deal with patients that if you could have seen a couple of days before would have been easier to sort is that most of us patients know you are busy so dont want to bother you.

    If a simple computer programme can save you time it can save lives but only with the professional that understands the output.

  • Guest

    Computor histories are only as good as their questions.