Class 10 — Physician agency

Tuesday, September 29

This recap comes from a recording of class, transcribed with Whisper and summarized with Claude, with a light edit from me. It’s meant as a convenience rather than a careful document, so the phrasing will sometimes be a little off. If something here looks wrong, please tell me and I’ll fix it.

What we covered

Homework 1 first, then the start of module 2. Why treatment varies across physicians, what physician agency means, and the Dartmouth evidence on geographic variation, ending on why variation isn’t the same thing as waste. We ran the patient-side simulation at the end.

Key points

  • Medicine is learned by watching people work rather than from a rulebook, and there’s real gray area in what a patient needs. Some variation in treatment across physicians is therefore what you’d expect. The question for this module is how much of it comes from something other than patient need.
  • Physician agency doesn’t require anyone behaving badly. The case that matters is the one where the care genuinely helps, just not by much, and costs more than you’d have agreed to if anyone had told you the price.
  • The Dartmouth map shows utilization, not prices. It’s Medicare, which pays a formula adjusted for local wages, so adding up payments tells you how much care was delivered rather than what it cost. It’s also adjusted for race, age, and sex, which is why “sicker people live there” doesn’t account for the spread.
  • Variation is not automatically waste. Waste is easy to name afterward and hard to identify in advance, since you’d have to know which patients would get no benefit before treating them. So the goal isn’t identical care for everyone. It’s understanding where the variation comes from, and payment incentives are where we go next.
  • Not-for-profit status means one thing for certain, which is that revenue above costs doesn’t go to owners. Tax exemption and cheaper capital follow from it. What’s also supposed to follow, community benefit, isn’t well measured or well regulated, so in observable behavior a not-for-profit can look much like a for-profit. Either way it still needs a margin.

Thursday

Model day, and the math is where this module trips people up. Read the slides and the chapter before you come so the notation isn’t new. Be comfortable with derivatives. We’ll do a constrained maximization, but in a simplified way by substitution rather than with a Lagrangian.

Homework

Grades and comments for homework 1 are in Canvas. Email me if something looks wrong, since we could have missed something in reviewing your repositories, and an answer key is coming.

Three things to carry into homework 2. Merge on identifiers rather than names, since county names repeat across states and merging on the name duplicates rows. A national rate isn’t the average of the county rates, so weight by counts, or sum the numerator and the denominator and then divide. And use git properly, working locally, committing, then pushing; uploading files through the GitHub website cost nobody points this time and will cost points next time.