Healthcare has a bad track record of resisting innovation. For example, we’re all aware of uncontrollable costs, and yet, although this would normally be a homing beacon ushering in a flurry of funding and technology solutions, we’ve seen plenty of the former, but very little of the latter.
One of the primary problems is that decision-making sits in the hands of the licensed provider and can’t be shifted to anyone else, not just for obvious legal reasons around who is qualified to deliver medical advice and prescribe, but because of strong patient and referring provider expectations. So, although, as in other fields, we have lower-cost ancillary personnel who can be helpful for coaching, nudging, and carrying out administrative tasks, they can’t fundamentally replace the provider in decision-making. So, substantial task-shifting to clinical staff is out of the question. And in my 30 years in medicine, I haven’t seen any sign that clinical decision-making will shift to the patient, other than perhaps a little more web searching followed by buying over-the-counter remedies at CVS, or coming to their doctor with a list of ideas they found on chat forums.
So if we’re going to scale in healthcare, it’s going to have to be at the provider level. And given problems in access and costs, the gains will have to be dramatic – 50% time-saving or more in terms of carrying out their tasks. And given the shortage of physicians, we’re going to need to rely heavily on advanced practice providers, even for complex patients, while somehow addressing significant existing differences in training quality and intensity. These enormous problems of scaling and expertise should, in principle, be addressable by technology – but they’re not going to be solved by our current toolkit of AI scribes, medical reference search tools, or healthcare chatbots. Although minimally disruptive and thus readily adopted, I don’t anticipate this batch of innovation having any meaningful impact on our most pressing healthcare challenges. A different model is needed.