Michael D Abramoff

Father of three men. Physician-scientist founder & builder. Digital Diagnostics & University of Iowa. AI the Right Way. Metrics for ethics.

[Michael D Abramoff @MichaelAbramoff]

Trust, reimbursement & AI
Or: the three stages of trust

What can physicians and other healthcare providers learn from AI creators — and what can AI creators learn from the provider experience?
There is a lot of debate about AI in healthcare and how it relates to physicians and other providers. Too much of it is physicians telling AI creators how it should be, and too little is about listening to how it can be.
It is interesting to look at this from a financial perspective. In this thread I assume the US third-party payer system as given, and I primarily use Medicare data, since commercial payers often follow Medicare and there is substantial documentation on the why and how.

[Michael D Abramoff @MichaelAbramoff]

It is indeed unfortunate that autonomous driving is so far behind acceptance of autonomous AI in medicine.

Where I disagree however, is that is due to process measures. In fact, process measures, like Press-Ganey scores have a **negative** -0.9 correlation with patient outcome.

IMO the primary reason for this strong support from all stakeholders - illustrated for example by increasing reimbursement - is because we are able to show improved patient outcomes in RCTs with autonomous AI.

And as you argue, evidence for such autonomous AI being safe had been around for decades - but while this led to FDA approval, it did not lead to widespread acceptance yet. Hard evidence for better outcomes was needed.

[The change in medical ethics this requires was a whole separate chapter.]

More here (Including refs):

https://t.co/dqdyxHgmLJ

[Michael D Abramoff @MichaelAbramoff]

Trust, reimbursement & AI
Or: the three stages of trust

What can physicians and other healthcare providers learn from AI creators — and what can AI creators learn from the provider experience?
There is a lot of debate about AI in healthcare and how it relates to physicians and other providers. Too much of it is physicians telling AI creators how it should be, and too little is about listening to how it can be.
It is interesting to look at this from a financial perspective. In this thread I assume the US third-party payer system as given, and I primarily use Medicare data, since commercial payers often follow Medicare and there is substantial documentation on the why and how.

[Michael D Abramoff @MichaelAbramoff]

BLS just helped me find the most recent data (URL below), and...

ambulatory healthcare total factor productivity in 2024 was 97.96, and continues to be below that in 1987 (100)

So over almost 40 years now, we have managed to decrease US healthcare productivity.

https://t.co/w8DewhIeve
(NAICS 621)

[Michael D Abramoff @MichaelAbramoff]

More terrifying than sad, but yeah. I should probably pull updated BLS data since I published it. Many ascribe declining productivity to the Baumol effect, but that only explains why it doesn't rise, not a decrease.

X doing its thing with figures from nature papers, but the key figure is this, showing the 40% clinic productivity when autonomous AI is used vs control:

[Michael D Abramoff @MichaelAbramoff]

Disagree with where this is going.
I have been developing an ethical framework for AI, not based on utilitarianism, starting from beneficence, justice and autonomy, and the need to optimize between them.

It has been great being able to discuss potential metrics / constraints for these principles with Yochai.

I learnt that there is absolutely a structural difference between Effective Altruism and Halacha for such metrics. The latter presupposes a relational hierarchy between subjects - Yochai keeps citing Bava Metzia 62a to me - i.e. a relative comes before a stranger.

Major differences for the trolley problem, and if an AI company uses EA ethics, its AI is more likely to operationalize the right than the left side of Fig 5.

https://t.co/TBWsIQE9wb

[Michael D Abramoff @MichaelAbramoff]

@PierreEliasMD For how I created the autonomous AI charge, see:
https://t.co/WBQZ3rS3PL

and
https://t.co/JcxSjG9KQU

[Michael D Abramoff @MichaelAbramoff]

Trust, reimbursement & AI
Or: the three stages of trust

What can physicians and other healthcare providers learn from AI creators — and what can AI creators learn from the provider experience?
There is a lot of debate about AI in healthcare and how it relates to physicians and other providers. Too much of it is physicians telling AI creators how it should be, and too little is about listening to how it can be.
It is interesting to look at this from a financial perspective. In this thread I assume the US third-party payer system as given, and I primarily use Medicare data, since commercial payers often follow Medicare and there is substantial documentation on the why and how.

[Michael D Abramoff @MichaelAbramoff]

Especially for something as controversial - as autonomous AI was 6 years ago - for CMS was all about clinical utility. We had already published the ethical framework for what to charge for the AI, so from was clear what the cost savings would be (see refs below).

Adding the continuous stream of RCTs and other studies showing better outcomes was the background for a very public debate about the introduction of autonomous and its potential risks and benefits - which all showed up 4 years of CMS's MPFS proposed and final rules on 92229.

Even now, looking back it is very interesting to read the pages and pages of proposed / final rules and the public comments from patients, professional societies etc all weighing in on autonomous AI. Here are some key moments:

Key CMS references for CPT 92229 (autonomous AI retinal imaging): • CY 2021 Final Rule (created the code): https://t.co/DjzpiZtEQ6... (pp. 84629–84630) • CY 2022 Proposed Rule: https://t.co/cXRtJ59Ojm... • CY 2022 Final Rule: https://t.co/vSIYsNZ5MW...

Ultimately it the mechanics are often easier for MACs to go first so that is what happened, with CMS creating national coverage a year later.

[Michael D Abramoff @MichaelAbramoff]

@signulll Agreed, but the need to build trust is unavoidable.
In healthcare, autonomous AI has been improving outcomes for a while now. Second order effects of the trust that has been built are kind of interesting, see my post here:

https://t.co/mmAFViUcSd

[Michael D Abramoff @MichaelAbramoff]

Trust, reimbursement & AI
Or: the three stages of trust

What can physicians and other healthcare providers learn from AI creators — and what can AI creators learn from the provider experience?
There is a lot of debate about AI in healthcare and how it relates to physicians and other providers. Too much of it is physicians telling AI creators how it should be, and too little is about listening to how it can be.
It is interesting to look at this from a financial perspective. In this thread I assume the US third-party payer system as given, and I primarily use Medicare data, since commercial payers often follow Medicare and there is substantial documentation on the why and how.

[Michael D Abramoff @MichaelAbramoff]

13. Goodhart CAE. Problems of Monetary Management: The UK Experience. In: Monetary Theory and Practice. Macmillan; 1984:91–121.
14. Olivero WC et al. Correlation Between Press Ganey Scores and Quality Outcomes From The National Neurosurgery Quality and Outcomes Database. Neurosurgery. 2018;65(CN_suppl_1):34–36.
15. Husten L. Medicine Or Mass Murder? Guideline Based on Discredited Research May Have Caused 800,000 Deaths In Europe Over The Last 5 Years. Forbes. January 15, 2014.
16. Wolf RM et al. Autonomous artificial intelligence increases screening and follow-up for diabetic retinopathy in youth: the ACCESS randomized control trial. Nature Communications. 2024;
15(1):421.
17. Abràmoff MD et al. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. npj Digital Medicine. 2018;1(1):39.
18. Channa R et al. Effectiveness of artificial intelligence screening in preventing vision loss from diabetes: a policy model. npj Digital Medicine. 2023;6(1):53.
19. Verbraak FD et al. Diagnostic Accuracy of a Device for the Automated Detection of Diabetic Retinopathy in a Primary Care Setting. Diabetes Care. 2019;42(4):651–656.
20. Grzybowski A, Brona P. Analysis and Comparison of Two Artificial Intelligence Diabetic Retinopathy Screening Algorithms in a Pilot Study: IDx-DR and Retinalyze. Journal of Clinical Medicine. 2021;10(11):2352.
21. Sedova A et al. Comparison of early diabetic retinopathy staging in asymptomatic patients between autonomous AI-based screening and human-graded ultra-widefield colour fundus images. Eye. 2022;36(3):510–516.
22. Mehra AA et al. Diabetic Retinopathy Telemedicine Outcomes With Artificial Intelligence-Based Image Analysis, Reflex Dilation, and Image Overread. American Journal of Ophthalmology. 2022;244:125–132.
23. Dow ER et al. AI-Human Hybrid Workflow Enhances Teleophthalmology for the Detection of Diabetic Retinopathy. Ophthalmology Science. 2023;3(4):100330.
24. American Diabetes Association. Standards of Care in Diabetes—2026. Section 12: Retinopathy, Neuropathy, and Foot Care. Diabetes Care. 2026;49(Suppl 1):S261–S276.
25. American Academy of Ophthalmology. Can AI Close the Diabetic Retinopathy Screening Gap? EyeNet Magazine. 2026.
26. Centers for Medicare & Medicaid Services. Medicare Program; CY 2022 Payment Policies Under the Physician Fee Schedule. Federal Register. 2021;86(221):64996.
27. Centers for Medicare & Medicaid Services. Medicare Program; Contract Year 2027 Policy and Technical Changes. Federal Register. 2026.
28. Abràmoff MD, Lavin PT, Birch M, et al. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. npj Digital Medicine. 2018;1(1):39.
Productivity Paper (Scaling / Reimbursement & Productivity):
29. Abramoff MD, Dai T, Zou J. Scaling Adoption of Medical AI — Reimbursement from Value-Based Care and Fee-for-Service Perspectives. NEJM AI. 2024;1(5):AIpc2400083.

[Michael D Abramoff @MichaelAbramoff]

References
1. Neprash HT et al. Association of Evaluation and Management Payment Policy Changes With Medicare Payment to Physicians by Specialty. JAMA. 2023;329(8):662–669.
2. MedPAC. Physician and other health professional services: Assessing payment adequacy and updating payments. Report to the Congress: Medicare Payment Policy. March 2025.
3. Centers for Medicare & Medicaid Services. Medicare Program; CY 2026 Payment Policies Under the Physician Fee Schedule. Federal Register. 2025.
4. Li ML, Dai T. The Future in Sight: LumineticsCore and the First Autonomous AI for Diagnostics. Harvard Business School Case Collection. 2025.
5. Abràmoff MD et al. A reimbursement framework for artificial intelligence in healthcare. npj Digital Medicine. 2022;5(1):72.
6. Abramoff MD, Dai T, Zou J. Scaling Adoption of Medical AI — Reimbursement from Value-Based Care and Fee-for-Service Perspectives. NEJM AI. 2024;1(5):AIpc2400083.
7. Social Security Act, Title 42, § 1395. Prohibition against any Federal interference.
8. Perlis RH et al. Trust in Physicians and Hospitals During the COVID-19 Pandemic in a 50-State Survey of US Adults. JAMA Network Open. 2024;7(7):e2424984. 9. Elder RJ, Allen RD. A Longitudinal Field Investigation of Auditor Risk Assessments and Sample Size Decisions. The Accounting Review. 2003;78(4):983–1002.
10. Katz R, O’Brien D. A Complete Guide to MIPS Quality Measures. Health Catalyst. 2026.
11. Hsiao WC et al. Resource-based relative values: An overview. JAMA. 1988;260(16):2347–2353.
12. Kantner AC. Understanding and Improving Your Work RVUs. Family Practice Management. 2023;30(2):4–8.

[Michael D Abramoff @MichaelAbramoff]

References

1. Neprash HT et al. Association of Evaluation and Management Payment Policy Changes With Medicare Payment to Physicians by Specialty. JAMA. 2023;329(8):662–669.
2. MedPAC. Physician and other health professional services: Assessing payment adequacy and updating payments. Report to the Congress: Medicare Payment Policy. March 2025.
3. Centers for Medicare & Medicaid Services. Medicare Program; CY 2026 Payment Policies Under the Physician Fee Schedule. Federal Register. 2025.
4. Li ML, Dai T. The Future in Sight: LumineticsCore and the First Autonomous AI for Diagnostics. Harvard Business School Case Collection. 2025.
5. Abràmoff MD et al. A reimbursement framework for artificial intelligence in healthcare. npj Digital Medicine. 2022;5(1):72.
6. Abramoff MD, Dai T, Zou J. Scaling Adoption of Medical AI — Reimbursement from Value-Based Care and Fee-for-Service Perspectives. NEJM AI. 2024;1(5):AIpc2400083.
7. Social Security Act, Title 42, § 1395. Prohibition against any Federal interference.
8. Perlis RH et al. Trust in Physicians and Hospitals During the COVID-19 Pandemic in a 50-State Survey of US Adults. JAMA Network Open. 2024;7(7):e2424984.
9. Elder RJ, Allen RD. A Longitudinal Field Investigation of Auditor Risk Assessments and Sample Size Decisions. The Accounting Review. 2003;78(4):983–1002.
10. Katz R, O’Brien D. A Complete Guide to MIPS Quality Measures. Health Catalyst. 2026.
11. Hsiao WC et al. Resource-based relative values: An overview. JAMA. 1988;260(16):2347–2353.
12. Kantner AC. Understanding and Improving Your Work RVUs. Family Practice Management. 2023;30(2):4–8.
13. Goodhart CAE. Problems of Monetary Management: The UK Experience. In: Monetary Theory and Practice. Macmillan; 1984:91–121.
14. Olivero WC et al. Correlation Between Press Ganey Scores and Quality Outcomes From The National Neurosurgery Quality and Outcomes Database. Neurosurgery. 2018;65(CN_suppl_1):34–36.
15. Husten L. Medicine Or Mass Murder? Guideline Based on Discredited Research May Have Caused 800,000 Deaths In Europe Over The Last 5 Years. Forbes. January 15, 2014.
16. Wolf RM et al. Autonomous artificial intelligence increases screening and follow-up for diabetic retinopathy in youth: the ACCESS randomized control trial. Nature Communications. 2024;15(1):421.
17. Abràmoff MD et al. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. npj Digital Medicine. 2018;1(1):39.
18. Channa R et al. Effectiveness of artificial intelligence screening in preventing vision loss from diabetes: a policy model. npj Digital Medicine. 2023;6(1):53.
19. Verbraak FD et al. Diagnostic Accuracy of a Device for the Automated Detection of Diabetic Retinopathy in a Primary Care Setting. Diabetes Care. 2019;42(4):651–656.
20. Grzybowski A, Brona P. Analysis and Comparison of Two Artificial Intelligence Diabetic Retinopathy Screening Algorithms in a Pilot Study: IDx-DR and Retinalyze. Journal of Clinical Medicine. 2021;10(11):2352.
21. Sedova A et al. Comparison of early diabetic retinopathy staging in asymptomatic patients between autonomous AI-based screening and human-graded ultra-widefield colour fundus images. Eye. 2022;36(3):510–516.
22. Mehra AA et al. Diabetic Retinopathy Telemedicine Outcomes With Artificial Intelligence-Based Image Analysis, Reflex Dilation, and Image Overread. American Journal of Ophthalmology. 2022;244:125–132.
23. Dow ER et al. AI-Human Hybrid Workflow Enhances Teleophthalmology for the Detection of Diabetic Retinopathy. Ophthalmology Science. 2023;3(4):100330.
24. American Diabetes Association. Standards of Care in Diabetes—2026. Section 12: Retinopathy, Neuropathy, and Foot Care. Diabetes Care. 2026;49(Suppl 1):S261–S276.
25. American Academy of Ophthalmology. Can AI Close the Diabetic Retinopathy Screening Gap? EyeNet Magazine. 2026.
26. Centers for Medicare & Medicaid Services. Medicare Program; CY 2022 Payment Policies Under the Physician Fee Schedule. Federal Register. 2021;86(221):64996.
27. Centers for Medicare & Medicaid Services. Medicare Program; Contract Year 2027 Policy and Technical Changes. Federal Register. 2026.