Last week, I had a 43-year-old patient on 16! medications burst into tears at the idea of increasing the dose of one of her existing medications, fully aware that she represents a payer-identified care gap.
Many people are pointing to the deplorable rates of guideline-directed medical therapy in conditions like hypertension and heart failure and primarily framing it as an access problem. The guidelines have been written, the meds exist, but patients just can’t find anyone to prescribe them. Technology, and soon autonomous prescribing AI, will be the savior.
Having managed thousands of complex patients of this sort, many in D-SNPs with multiple comorbidities, I am confident that improved access will only be a part of the solution. Many of these patients abhor medications. They are already on 10-15 of them, they feel awful, and the idea that someone is proposing to add more is a non-starter, no matter what the guidelines say. And they’re not wrong to be upset. The typical ischemic cardiomyopathy diagnosis comes with 6-8 cardiac meds. Their diabetes diagnosis comes with 3-4 meds. And the high rates of migraines, obesity, urinary incontinence, anxiety and depression, and chronic pain round it out with another 5-7.
And yet we’ve been very successful in managing this exact group. What has worked (our NPS rating is 95) is having patients engage with a highly credentialed expert who spends short, frequent intervals with them (2-5 minutes) explaining, anticipating, responding to data, negotiating, willing to combine meds, and above all, listening. That person soon becomes their most trusted advisor, and they are willing to accept almost any recommendation. The total amount of time spent longitudinally is probably no more than 20-25 minutes across 4-5 virtual encounters. But the results are incredible. Patients who have been refractory for decades are at goal for the first time.
The technology in this case, structuring and automation across every part of care that’s amenable, enables this model to work. But the fundamental engagement with the patient has to be human, and in most cases a perceived human expert. (I’ve written separately on the role of LLMs in our direct-to-consumer practice, and on how these patients are now arriving for care.)
All of this will make the uptake of models like ACCESS with the inevitable AI agents coming to the rescue an interesting experiment. The framing of many of these initiatives is there is primarily an access (pun-intended) problem. But from my experience, the model that gets patients to enroll and more importantly achieve successful outcomes will need to be much more. There is no doubt that technology and AI will be essential to make the unit economics work. And scaling “highly credentialed” individuals is not a trivial problem. But designing a solution that realistically overcomes the obstacles I’ve presented is far less straightforward than simply throwing an LLM at it.