Michelle M Mello’s research while affiliated with Stanford Medicine and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (290)


Representation of Women, Older Adults, and Racial and Ethnic Minoritized Patients in Pivotal Trials for US Food and Drug Administration Novel Oncology Therapeutic Approvals, 2012-2021: Bright Spot Trials and Trends Over Time
  • Article

June 2025

·

1 Read

JCO Oncology Practice

·

William Pelletiers

·

·

[...]

·

Joseph S Ross

PURPOSE To examine trends over time in clinical trial diversity for oncology therapeutics, identify bright spot trials adequately representing Black or Latino patients, and examine their associations with trial characteristics. METHODS This retrospective cross-sectional study of trials supporting US Food and Drug Administration (FDA) oncology therapeutic approvals (2012-2021) compared proportions of enrolled women, older adults, and racial and ethnic minoritized patients with US patient populations with approved indications and US census data, exploring trends over time. We estimated the likelihood that an organization sponsoring a trial adequately representing Black patients also sponsored a trial adequately representing Latino patients, and vice versa, and examined associations of bright spot trials, trials adequately representing Black or Latino patients, with trial characteristics. RESULTS The FDA approved 111 oncology therapeutics (2012-2021) from 70 sponsors, on 121 trials enrolling 50,871 patients, of which 22,155 (44%) were women, 9,442 (19%) older adults (>64 years), 5,319 (9.6%) Asian, 1,353 (3%) Black, and 1,511 (3%) Latino. Zero trials adequately represented all patient demographics with no improvements for any group over the past 10 years ( P values >.05), and 80% adequately represented women, 44% older adults (>64 years), and 2% racial and ethnic minoritized patients. We identified 33 bright spot trials, the likelihood of which was associated with individual sponsors ( P < .0001), but not sponsor size ( P = .22), indication ( P = .20), or trial phase ( P = .32). There were no associations between a company sponsoring a trial adequately representing Black patients and that company also sponsoring a trial adequately representing Latino patients, or vice versa ( P values >0.05). CONCLUSION Diverse trial enrollment is inadequate, without improvement over time. There are bright spot trials that may merit study for factors contributing to their successful recruitment of representative patients.


A Framework for Assessing the Permissibility of Academic Leaders' Outside Activities

June 2025

·

6 Reads

Milbank Quarterly

Policy Points Many have urged academic institutions to rethink conflict of interest policies governing leaders’ outside activities, which pose not only individual conflicts for leaders themselves but institutional conflicts for their academic employers. Although the American Association of Medical Colleges and Association of American Universities have provided guidance on managing such conflicts, neither offer a structured approach for determining when and under what conditions it is appropriate for a leader to engage in specific outside activities. To address this gap, this article develops a decision‐making framework that institutional oversight bodies can use to assess the permissibility of academic leaders’ proposed outside activities.




Estimated Savings From Extending Prescription Drug Inflationary Rebates To All Commercial Plans

March 2025

The inflationary rebate provisions of the Inflation Reduction Act (IRA) of 2022 require pharmaceutical manufacturers to pay money back to Medicare if they raise prices faster than the rate of inflation. Opportunities now exist for Congress and the states to extend inflationary rebates to commercial plans. We demonstrate the potential savings from applying inflationary rebates to prescriptions filled in commercial plans by simulating savings under several policy design options. We estimate that a policy mimicking the IRA's design could have saved as much as 8.1billionin2021byimposingrebateson1,100drugs.Tominimizetheadministrativeburdenofassessinginflationaryrebates,Congressorstatescouldrestrictrebatestocertainhighcostdrugsandstillgarnersubstantialsavings.Forexample,includingonlydrugscostingmorethan8.1 billion in 2021 by imposing rebates on 1,100 drugs. To minimize the administrative burden of assessing inflationary rebates, Congress or states could restrict rebates to certain high-cost drugs and still garner substantial savings. For example, including only drugs costing more than 830 per month or ranked in the top 300 drugs by total spending could halve the number of rebate-eligible drugs yet capture 96 percent of estimated savings that could be available under a policy that mimics the IRA design (7.7billion).Restrictingeligibilitytothetop300drugsbytotalspendingcouldcapture85percentofthosepotentialsavings(7.7 billion). Restricting eligibility to the top 300 drugs by total spending could capture 85 percent of those potential savings (6.9 billion), imposing rebates on just 194 drugs.


Ensuring Safe Practice by Late Career Physicians: Institutional Policies and Implementation Experiences

November 2024

·

17 Reads

·

1 Citation

Annals of Internal Medicine

Background: Late career physicians (LCPs; physicians working beyond age 65 to 75 years) may be at higher risk for delivering unsafe care. To oversee LCPs, some health care organizations (HCOs) have adopted LCP policies requiring cognitive, physical, and practice performance screening assessments. Despite recent controversies, little is known about the content and implementation of such policies. Objective: To characterize key features of LCP policies and the perspectives of medical leaders responsible for policy development and implementation. Design: Mixed-methods study using content analysis and key informant interviews. Setting: 29 U.S. HCOs with LCP policies active in 2020. Participants: 21 purposively sampled interviewees in physician leadership roles at 18 HCOs. Measurements: Descriptive statistics of policy features and content analysis of interviews. Results: Although policies had many commonalities-mandatory universal screening at a trigger age around 70 years, a strategy of screening followed by in-depth assessment of positive results, and commitment to patient safety as the key motive-they varied substantially in the testing required, funding, processes after a positive screening result, and decision making around concerning results. Policies prioritized institutional discretion in interpreting and responding to test results; many lacked clear language about appeals or other procedural protections for physicians. Leaders were generally satisfied with policies but reported preemptive retirements as physicians approached the screening age and cautioned that substantial investment in cultivating physicians' buy-in was required for successful rollout. Limitations: Sampled policies and interviews may not be representative of all HCOs. The analysis excluded the experiences of HCOs that tried and failed to implement LCP screening. Conclusion: Policies about LCPs are considered successful by institutional leaders. Policy variations and early adopters' implementation experiences highlight opportunities to improve physician acceptance and program rigor. Primary funding source: The Greenwall Foundation.






Citations (74)


... 25 Given that those tailoring efforts could increase the cost of LLM use in healthcare and frequently require significant expertise that is not always widely available, the superlative performance of the zero-shot approach was particularly encouraging. 26 Moreover, while previous studies using AI/LLMs heavily focused on vignette-based diagnostic performance, our findings are applicable to real-world screening scenarios. 27,28 The LLMs' ability to assist in screening depression or anxiety through text-based communications could open novel opportunities for timely detection of symptoms of comorbid depression or anxiety among those with various chronic diseases. ...

Reference:

Optimizing Large Language Models for Detecting Symptoms of Comorbid Depression or Anxiety in Chronic Diseases: Insights from Patient Messages
Avoiding Financial Toxicity for Patients from Clinicians’ Use of AI
  • Citing Article
  • September 2024

The New-England Medical Review and Journal

... 4) Bias (Does the LLM response contain content that perpetuates identity-based discrimination or false stereotypes?). This framework, developed prior to the red teaming event, has been subsequently published 29 . ...

Standing on FURM Ground: A Framework for Evaluating Fair, Useful, and Reliable AI Models in Health Care Systems
  • Citing Article
  • September 2024

NEJM Catalyst

... 204 This was followed by a landmark ruling from the Chilean Supreme Court in 2023, which further clarified the protection of mental integrity in the context of neurotechnology and AI. 205 Furthermore, existing legal structures, including anti-discrimination laws 206 , consumer protections, and medical device regulations, are increasingly adapted to mitigate AI-related risks. Concrete steps include for instance, updates of EU Medical Device Regulation, guidance of the U.S. Federal Drug Administration on AI, new requirements for health care organizations and insurers to address algorithmic impacts or efforts to anchor algorithmic discrimination specifically in anti-discrimination legislation. ...

Antidiscrimination Law Meets Artificial Intelligence—New Requirements for Health Care Organizations and Insurers
  • Citing Article
  • August 2024

JAMA Health Forum

... 10,11 While the perioperative field has seen exciting progress, they have also been met with caution as AI requires human oversight to ensure safety and reliability. 12 The American Medical Association reinforces this perspective, defining augmented intelligence as a tool designed to enhance, rather than replace, human judgement in clinical decision-making. 13 Despite the growing adoption of LLMs, their role in perioperative medicine remains underexplored. ...

Denial-Artificial Intelligence Tools and Health Insurance Coverage Decisions
  • Citing Article
  • March 2024

JAMA Health Forum

... Decisions that occurred in response to COVID-19 were not merely actions motivated by public health objectives, but also actions that were influenced by political affiliations [1,6,9,20]. Our research aligns with the existing literature on the distinct approaches taken by Democratic and Republican governors during COVID-19, and the damaging consequences of such divergence [1,6,[20][21][22]68]. While Democratic party leaders were typically in favor of stronger pandemic mitigation efforts in place for longer periods, Republican leaders generally prioritized the preservation of economic activity and a quicker return to a sense of normalcy [1,6]. ...

Legal infrastructure for pandemic response: lessons not learnt in the US
  • Citing Article
  • February 2024

The BMJ

... Another aspect hindering optimal AI use is an overreliance on AI-generated predictions, thereby diminishing their critical evaluation of diagnostic probability [27][28][29]. This was noticeable in the 16.4% of incorrect decision changes according to respective incorrect AI indications. ...

Understanding Liability Risk from Using Health Care Artificial Intelligence Tools
  • Citing Article
  • January 2024

The New-England Medical Review and Journal

... The recent U.S. Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence emphasized the importance of assuring the quality of medical AI. 25 While debate continues over the most valuable form of validation, 26 it is clear that independent, predeployment assessments are a crucial way to evaluate algorithm generalizability and reliability. In addition, regulatory systems In Panel A, the confusion matrix is normalized by actual diagnosis to allow for better visualization of clinician performance. ...

President Biden's Executive Order on Artificial Intelligence-Implications for Health Care Organizations
  • Citing Article
  • November 2023

JAMA The Journal of the American Medical Association

... Mandatory vaccination policies have also proven to be only partially effective: while increasing vaccination coverage, they are currently met with a significant degree of scepticism, sometimes evoking conspiracy sentiments (6,7). People subjected to mandatory vaccination were found to fight it by pseudoscientific arguments (8), and even healthcare workers were observed to strongly oppose such measures (9). ...

Vaccination Mandates-An Old Public Health Tool Faces New Challenges
  • Citing Article
  • July 2023

JAMA The Journal of the American Medical Association

... To ensure the accuracy of the information, consult multiple sources before choosing a DHT. There are discrepancies in how certain devices are marketed compared to their submitted 501(k) applications 28,29 . Neurologic DHTs may be reported using several CPT® code families (Box 2). ...

Discrepancies Between Clearance Summaries and Marketing Materials of Software-Enabled Medical Devices Cleared by the US Food and Drug Administration
  • Citing Article
  • Full-text available
  • July 2023

JAMA Network Open