Ziad Obermeyer (@oziadias) / X

Ziad Obermeyer

2,448 posts

Ziad Obermeyer

@oziadias

Physician + prof @UCBerkeley. Co-founder, @NightingaleOS, @Dandelion_AI4H. Machine learning, health, emergencies, espresso.

Berkeley, CA

Joined March 2013

Ziad Obermeyer
@oziadias
Jul 6, 2020
Imagine testing everyone for COVID, every day. Sure it’d be expensive—but we’d diagnose infections in real time, and keep everyone else safe. We think we have a way to make this possible… at <$5/person/day. A thread on our new working paper: tinyurl.com/ydz664ju
Ziad Obermeyer
@oziadias
Jan 15, 2021
Medical puzzle: Take Black vs. White patients with similar knee x-rays Why do Black patients have more pain? Our paper in @NatureMedicine provides one answer: go.nature.com/3i6zAuc Algorithms see causes of knee pain in Black patients, that human radiologists miss. 🧵
An algorithmic approach to reducing unexplained pain disparities in underserved populations
From nature.com
Ziad Obermeyer
@oziadias
May 4, 2024
Ever wonder why your insurer charges a copay for statins (and other high-value meds)? Don’t they want to prevent heart attacks, strokes, etc? Copays never made sense to me. In this new paper in @QJEHarvard, we show copays kill. 🧵 tinyurl.com/nhk4v93f
320K
Ziad Obermeyer
@oziadias
Apr 11, 2022
Is this patient having a heart attack? After 10+ years in the ER, it’s still an agonizing question for me I turned the angst into a long paper with @m_sendhil in @QJEHarvard—doi.org/10.1093/qje/qj… Tl;dr—machine learning can have *huge* benefit, and help fix docs' errors 🧵
Diagnosing Physician Error: A Machine Learning Approach to Low-Value Health Care*
From academic.oup.com
Ziad Obermeyer
@oziadias
Dec 13, 2021
I spend a lot of “research” time pleading for access to health data @m_sendhil and I co-founded a new nonprofit, @NightingaleOS, so you don’t have to We link medical imaging data + ground truth outcomes, and make it available—free! Launching tomorrow at #NeurIPS2021
Ziad Obermeyer
@oziadias
Jan 26, 2020
You can’t pick your doctor in the ER. But your doctor can pick you. Two implications: 1) A lot of "practice variation" may be *preference* variation 2) Don't assume "as-if-random assignment" to docs Thread on a new, short @JAMANetworkOpen paper tinyurl.com/w8bqqkj
Ziad Obermeyer
@oziadias
Jun 22, 2023
Those of us building health AI products have a problem: Our algorithms look great—in our own data But how will they perform elsewhere? On other machines? On diverse patients? We’ve built a way to find out, at @Dandelion_AI4H: A free, public service for algorithm audits
59K
Ziad Obermeyer
@oziadias
Jan 15, 2021
Replying to @oziadias
What if there were a different way to train the algorithm? L̵e̵a̵r̵n̵ ̵f̵r̵o̵m̵ ̵t̵h̵e̵ ̵r̵a̵d̵i̵o̵l̵o̵g̵i̵s̵t̵ Listen to the patient We train the algorithm to predict the *patient’s pain*, not the radiologist’s read
Ziad Obermeyer
@oziadias
Dec 10, 2021
Your doctor sends you for a lab test. Ever wondered if the outside temperature on the day of the test changes its result? Neither did @devingpope or I… until we did this study! Turns out, temperature distorts results Thread—new paper in @MedCellPress
Cell Press
@CellPressNews
Dec 10, 2021
Ambient temperature influences the results of some of the most used laboratory tests, and these distortions likely affect medical decision-making, including whether to prescribe medications. Read more in @MedCellPress here: cell.com/med/fulltext/S… @oziadias, @Devin_G_Pope
Ziad Obermeyer
@oziadias
Jun 21, 2021
For the past 2 years, we’ve worked to diagnose and fix bias in dozens of algorithms. @caseymross wrote a wonderful piece about it today: statnews.com/2021/06/21/alg… And our bias ‘playbook’ distills the process into 4 simple steps: tinyurl.com/rnnhckz6 🧵
‘Nobody is catching it’: Algorithms used in health care nationwide are rife with bias
From statnews.com
Ziad Obermeyer
@oziadias
Oct 7, 2021
Just amazing news. And almost as amazing, Glaxo will sell it at just 5% above cost—huge value for the world. Science!
Gabriele Steinhauser
@gksteinhauser
Oct 6, 2021
WHO endorses world's 1st malaria vaccine. “I longed for the day that we would have an effective vaccine against this ancient and terrible disease,” said @DrTedros. “Today is that day. An historic day.” W @deniseroland wsj.com/articles/world… via @WSJ
Ziad Obermeyer
@oziadias
Aug 12, 2020
We know communities of color suffer most from COVID-19. The way we’re allocating $175b in relief funding is also biased against them. Thread on our new paper, led by superstar @pragyakakani (supporting roles: @amitabhchandra2, @m_sendhil, me) jamanetwork.com/journals/jama/…
Ziad Obermeyer
@oziadias
Nov 9, 2019
Like the rest of the scientific community, I am profoundly shocked and saddened by this news of another epidemiological association driven by unmeasured confounding. Thoughts and prayers and also we’ll keep running the same regressions (Yes this is a joke)
JAMA
@JAMA_current
Nov 9, 2019
It now seems safe to conclude that many prior epidemiological associations between #vitaminD deficiency and adverse health outcomes were driven by unmeasured residual confounding or reverse causality ja.ma/33vpaMz #KidneyWk
Ziad Obermeyer
@oziadias
May 4, 2024
Replying to @oziadias
3) Patients at the *highest risk* of drug-preventable events (e.g. heart attack)… …are *most* likely to stop the drug when they run out of money. This is not what simplistic economic models would predict—these patients should be the *most* willing to pay for the drugs.
19K