Algorithmic Control of Care

I am an avid reader of @stifel’s @tim opler’s reports. His most recent (https://www.stifel.com/newsletters/investmentbanking/bal/marketing/healthcare/biopharma_timopler/2025/BiopharmaMarketUpdate_053025.pdf) is looking at how AI may change medicine, how the roles of physicians and clinicians may be affected, how tech moats compare to healthcare and biotech moats, all interesting topics.

The one that struck me, though, is something that should be somewhat trivial but is not in practice. Opler calls this Algorithmic Control of Care and, while perhaps a little scary-sounding (think ChatGPT deciding your fate), the underlying idea is super simple: “algorithms can be far more complex than today’s care guidelines that are administered by physicians from memory” (slide 56). I would say, more sophisticated rather than more complex, and the concept here is that of the physician as a pilot overseeing a plane run by instruments.

The example of diabetes prescriptions provides a simple illustration of that (slide 62): GLP-1s/incretins are top-of-mind but are not the right solution for all cases (obvious, right?).

“One-size” prescribing misses big cost and outcome gains. Customizing therapy can for example divert expensive injectables to those who benefit most (OII/MOD, SIRD) and keep low-risk MARD patients on inexpensive oral agents. GLP-1s are in the avoid zone for pure insulin-deficiency states and low-risk elderly patients, while they are indispensable where obesity, CV risk or severe IR dominate.

SGLT2 inhibitors double as a renal & cardio protective class. Even when they are only “yellow” for glycaemic efficacy (e.g., NOIS/MARD), kidney or heart overlay columns can bump them up to mandatory first-line.

Cheap legacy drugs still matter. Metformin, sulfonylureas, thiazolidinediones (PPARγ) and lifestyle (“Diet”) remain green across many boxes – they are the foundation that keeps total-therapy cost manageable.

Comorbidities often outrank the core phenotype. A MOD patient with eGFR 55 mL/min may need to forego yellow-flag PPARγ despite being ideal otherwise, in favour of kidney-protective SGLT2 + ACE/ARB.