Dr. Albert Isaac (AI) Will See You Now

Imagine a world where human error was been virtually eradicated from hospitals. The process of patient diagnosis and treatment is completely automated, with a certified doctor being the final gatekeeper to sign off on the recommendations wholly generated by a machine. This is the world envisioned by DeepMind, the AI division of Google that is seeing new applications in the medical field.

7934619-a-doctor-banging-his-head-realizing-a-mistake-he-forgot-isolated-on-white-background.jpg

Human error in medicine is one of the leading causes of death in America – #3, in fact. The introduction of AI into the patient care process has the potential to drastically cut erroneous diagnoses as a result of the fact that we’re simply human beings. Regardless of how well-trained, diligent, and keen those who care for us are, how many times have you been told your nagging sore throat “is just a cold” when it eventually developed into strep throat or the flu? With the assistance of DeepMind-like AI, your medical history and physical information can be computed just as any variable would. By handing the reins of medicine to a specialized computer, small details and seemingly innocuous symptoms would be immediately flagged for further observation. This could also become crucial during moments when serious and life-threatening diseases share the same initial symptoms as more benign illnesses.

ICU_SciSource_DC0018-2048x1152.jpg

Today, patients staying for a particular duration of time may find themselves surrounded by machines such as the ones featured in the photo above. Although they may seem comprehensive and diligent in the care they’re administering, they are, in fact, simply “dumb” machines that are only showing outputs and administering minimal inputs. Nurses are the ones tasked with making routine rounds between each patient ward, during which they interpret biosigns, answer patients’ questions, and administer care in the form of medicines or treatments. In an AI-enabled health care facility, the machines would be able to actively interpret the outputs and administer care to the patient with minimal input from a caregiver. In the event of an emergency or post-op, a patient is extremely vulnerable. In such situations, milliseconds could mean the difference between life or death, time that is easily wasted for a nurse to arrive at the patient’s ward, interpret the data, and administer a solution.

the-slippery-slope-sm.png

I’ve included this graph to better illustrate the rate of patient deterioration and why it is such a problem in hospitals. A patient in the stages of deterioration will quickly slide towards death if proper care isn’t promptly administered. During a nurse’s rounds, and depending on the ratio of nurses to patients, a patient that is the last to be seen each round may have already entered into an irreparable state before a nurse can be in attendance. In the best case scenario, this patient may yet be saved if high-cost lifesaving techniques are used such as those found in intensive care units. With the shortage of nurses and doctors becoming an ever-increasing problem in the near-future, AI may be the answer to this critical problem that would otherwise take a generation of schooling and certifications to resolve.

What do you think? Even with the current abilities of AI, diagnosis and treatment would rarely be erroneous. But how comfortable would you feel if AI took over day-to-day caregiving in our hospitals?

8 comments

  1. Hi Mike, I had no idea that human error was such a leading cause of death in the US at #3. I think that AI absolute has a place in the hospital. The value of it being correct most of the time is clear. However at the same time, I don’t think that human doctors would be obsolete entirely. I see AI as being more of a safety net assistant type thing for hospitals. I feel that doctors make errors due to carelessness rather than an actually deficiency in knowledge. I think AI and doctors would agree most of the time and for the times that they disagree, it’s always good to have a second opinion.

  2. Good timing, Wired just posted an article this morning on using AI to more rapidly identify brain blockages in CT scans of potential stroke victims. As you described, this is an example of when fast treatment is crucial and could be the difference in maintaining brain function.
    https://www.wired.com/story/using-ai-to-help-stroke-victims-when-time-is-brain/
    I agree with @jamessenwei that I don’t think AI is going to replace many jobs in healthcare, but rather it will be a tool that doctors and nurses use to provide better and faster care. AI can assist in setting priorities, diagnostics, and searching through electronic health records to connect dots that may have missed. Exciting times for healthcare.

  3. Read an article about integrating more AI machine learning in diabetes medication that created software and hardware that help predict and at the same time manage insulin for patients, so I think this is very relevant in health care today! I am really onboard with this, and IBM’s Watson computer is a great example of how great this is. As human being our brains just cant compute and sift through the same amount of data that computers can at such high speeds.
    I think the human aspect of nurses and doctors in hospitals is necessary to make patients feel more comfortable, especially in a place that already makes people feel uncomfortable and scared, but as
    @mpduplesmba mentioned above, AI and machine learning can be great tools to properly treat and predict illnesses and diseases that might be missed or misdiagnosed.

    Links:
    https://www.wired.com/story/with-ai-your-apple-watch-could-flag-signs-of-diabetes/
    http://www.nydailynews.com/news/world/ibm-watson-proper-diagnosis-doctors-stumped-article-1.2741857

  4. I feel that AI most likely, like @jamessenwei said, would be an aid. Think about the number of times people go to “check on a patient” on what seems to be an “arbitrary schedule”. I also feel like AI needs to be trained to know though that more information doesn’t necessarily mean that a patient is hurting more. How often pain is put into the machine or when certain symptoms are started with a timestamp may be hard to pinpoint for patients just as they are right now. It will take time to understand how patients interact with AI or how doctors interact with it. If some of the factors are based on quantities, it is extremely hard to compare one’s pain to someone else’s quantitatively.

    This is why I feel that AI for healthcare may actually be further off than we think. I feel that doctors are great interpreters of data, and suggestion to them would be helpful, but to think that diagnostics taken by a machine could be properly organized and point towards when someone needs help is difficult to imagine. This also worries me about a supersaturation of hospitals or care facilities because machines would most likely need to ere on the side of safety and going to see a doctor rather than not.

  5. I agree with both what @jamessenwei and @mpduplesmba said regarding AI taking over doctors positions. I do not think that AI should completely take over doctors positions in healthcare because I would rathe be taken care of by an actual human being rather then a machine. I would feel more comfortable with a doctor instead of AI because I would be able to interact with the doctor. Doctors are trained to interact with their patients, and they know how to deliver news in the most positive way possible. Also, being able to have a human-to-human conversation is something that people value when they are in the ER. Humans want to have interactions with other humans, and replacing doctors with AI eliminates that experience. I do, however, believe that AI systems could be used in healthcare to assist doctors or take over if their is no immediate doctor to respond. I think that hospitals should use AI systems, but they shouldn’t start them at the top right away. AI should slowly be incorporated into the system, and once patients start feeling more comfortable around AI, the more it can be integrated in the workplace.

  6. I worry about the lack of human care in the healthcare system already. We always try and put a number or a label to diseases. Less often doctors are having full conversations with their patients. AI is just going to exacerbate this system. If AI is to be used I think it needs to be in the background, with the doctor heading up the operation. Personally, as a patient, I want that personal connection with doctors. Without this personal connection, it feels like you are just a number in the healthcare system, and you risk patients not feeling cared for. I wonder if maybe a better solution is shorter working hours for doctors so that they aren’t careless or more apt to make mistakes with patients. Hospitals are the one place that I would be very scared to replace humans with AI. With human lives at stake, I would hate to have a computer crash or glitch harm a patient. Great post, very thought provoking!

  7. I totally agree with @tylercook95 here. There already isn’t nearly enough communication/facetime with Doctors in patients. As a son of a physician, I am well aware of how outside actors influence healthcare in a variety of ways. I can imagine some person giving a TED talk about how great everything will be when AI makes its way into health care when in reality its a mixed bag. There will be windfall gains in efficiency but making the doctor less important in the healthcare system isn’t necessarily all good.

    Though you have outlined some of the good things that AI in medicine will achieve, I hope we as a society can ask some of the tough questions about our healthcare before things change too much.

  8. Great post, Mike. I’d have to agree with Tyler and Mike here – many aspiring and current doctors I know have said medicine is becoming more and more dominated by insurance and limited, efficient, and business-like approaches to patient care – resulting in far less facetime with patients. I could see AI being useful to nurses in the rounds cycle, but I will be interested to see how it might integrate with checkups, emergency room visits, and more up-front medical treatments.

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

w

Connecting to %s

%d bloggers like this: