<table style="border:1px solid #adadad; background-color: #F3F1EC; color: #666666; padding:8px; -webkit-border-radius:4px; border-radius:4px; -moz-border-radius:4px; line-height:16px; margin-bottom:6px;" width="100%">
<tbody>
<tr>
<td><span style="font-family:Helvetica, sans-serif; font-size:20px;font-weight:bold;">Health Tech | Fierce Healthcare</span></td>
</tr>
<tr>
<td> </td>
</tr>
</tbody>
</table>
<table style="font:13px Helvetica, sans-serif; border-radius:4px; -moz-border-radius:4px; -webkit-border-radius:4px; background-color:#fff; padding:8px; margin-bottom:6px; border:1px solid #adadad;" width="100%">
<tbody>
<tr>
<td><a href="https://www.fiercehealthcare.com/ai-and-machine-learning/gen-ai-chatbots-continually-struggle-differential-diagnoses-mass-general" style="font-family:Helvetica, sans-serif; letter-spacing:-1px;margin:0;padding:0 0 2px;font-weight: bold;font-size: 19px;line-height: 20px;color:#222;">LLMs still fall short in ‘clinical reasoning abilities’: study</a>
<div style="font-family:Helvetica, sans-serif; text-align:left;color:#999;font-size:11px;font-weight:bold;line-height:15px;">Apr 20th 2026, 13:45</div>
<div style="font-family:Helvetica, sans-serif; color:#494949;text-align:justify;font-size:13px;">
<p><div class="col content" morss_own_score="5.345029239766082" morss_score="51.5105969131964">
<p>Despite the increasing use of artificial intelligence in healthcare by <a href="https://www.fiercehealthcare.com/ai-and-machine-learning/ai-chatbot-use-health-information-16-2024-rock-health-survey">patients</a> and <a href="https://www.fiercehealthcare.com/ai-and-machine-learning/75-us-healthcare-systems-use-plan-use-ai-platform-2026">providers</a> alike, a <a href="https://www.massgeneralbrigham.org/en/about/newsroom/press-releases/ai-chatbot-lacks-clinical-reasoning">new study</a> from Mass General Brigham found that publicly available generative AI models often fail to properly navigate diagnostic situations.</p>
<p>The study, published April 13 in <a href="https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2847679?guestAccessKey=ef7d2c87-beca-4c8b-b6a2-85d11bcf8185&utm_source=for_the_media&utm_medium=referral&utm_campaign=ftm_links&utm_content=tfl&utm_term=041326">JAMA Network Open</a>, evaluated 21 different general-purpose large language models (LLMs) on 29 standardized clinical cases from January to December 2025. Models received sequential case transcripts which “preserved clinical context and maintained continuity” throughout the process of clinical reasoning. </p>
<p>Medical student evaluators then scored outputs at each stage against the MSD Manual. Researchers also developed a new measure, dubbed Proportional Index of Medical Evaluation for LLMs (PrIME-LLM), to determine accuracy across five clinical reasoning domains. </p>
<p>Among the LLMs tested by Mass General Brigham’s MESH Incubator researchers were GPT-5, Gemini 3.0 Flash and Grok 4. </p>
<p>While all LLMs achieved a correct final diagnosis more than 90% of the time, researchers found models “performed poorly in generating differential diagnoses and navigating uncertainty relative to other reasoning stages.” All models failed to produce an appropriate differential diagnosis more than 80% of the time.</p>
<p>"These models are great at naming a final diagnosis once the data is complete, but they struggle at the open-ended start of a case, when there isn't much information," said Arya Rao, lead author, MESH researcher and M.D.-Ph.D student at Harvard Medical School, in a statement. </p>
<p>MESH Incubator Executive Director Marc Succi, M.D., was one of the study’s corresponding authors. Succi said in a <a href="https://www.massgeneralbrigham.org/en/about/newsroom/press-releases/ai-chatbot-lacks-clinical-reasoning">statement</a> that off-the-shelf LLMs “are not ready for unsupervised clinical-grade deployment” despite continual improvements.</p>
<p>“Differential diagnoses are central to clinical reasoning and underlie the ‘art of medicine’ that AI cannot currently replicate,” Succi said. </p>
<p>The new study builds on previous work from Succi and the MESH group. Researchers <a href="https://www.massgeneralbrigham.org/en/about/newsroom/press-releases/chatgpt-shows-impressive-accuracy-in-clinical-decision-making">evaluated</a> the clinical abilities of ChatGPT 3.5 in August 2023, which found the chatbot was about 72% accurate in overall clinical decision-making.</p>
<p>Researchers in the present study said most models <a href="https://www.massgeneralbrigham.org/en/about/newsroom/press-releases/ai-chatbot-lacks-clinical-reasoning">demonstrated</a> improved accuracy when provided with lab results and imaging in addition to text, with the most recently released models performing better than older versions. </p>
<p>Noted limitations included disabled web search and reasoning, inability to fully exclude prior exposure to standardized cases and evaluation does not incorporate model augmentations. </p>
<p>The study emphasized the potential of LLMs to “augment—not replace—physician reasoning.” </p>
<p>“The consistent gap between differential diagnosis and final diagnosis highlights how differently these systems process information compared with physicians,” researchers wrote. “Clinicians preserve uncertainty and iteratively refine differential diagnoses, whereas LLMs collapse prematurely onto single answers, a limitation that persists across model generations.” </p>
</div></p>
</div>
<div style="font-family:Helvetica, sans-serif; font-size:13px; text-align: center; color: #666666; padding:4px; margin-bottom:2px;"></div>
</td>
</tr>
</tbody>
</table>
<table style="font:13px Helvetica, sans-serif; border-radius:4px; -moz-border-radius:4px; -webkit-border-radius:4px; background-color:#fff; padding:8px; margin-bottom:6px; border:1px solid #adadad;" width="100%">
<tbody>
<tr>
<td><a href="https://www.fiercehealthcare.com/health-tech/doj-seeks-immediate-asset-freeze-receivership-against-telehealth-company-zealthy" style="font-family:Helvetica, sans-serif; letter-spacing:-1px;margin:0;padding:0 0 2px;font-weight: bold;font-size: 19px;line-height: 20px;color:#222;">DOJ accuses telehealth company Zealthy of fraud</a>
<div style="font-family:Helvetica, sans-serif; text-align:left;color:#999;font-size:11px;font-weight:bold;line-height:15px;">Apr 20th 2026, 13:45</div>
<div style="font-family:Helvetica, sans-serif; color:#494949;text-align:justify;font-size:13px;">
<p><div class="col content" morss_own_score="5.582608695652174" morss_score="60.30754108239997">
<p>The Department of Justice filed a motion seeking an immediate asset freeze and receivership for telehealth company Zealthy and its CEO, Kyle Robertson, over alleged fraudulent prescribing and other deceptive practices, according to court documents.</p>
<p>The DOJ filed an amended complaint last week against Zealthy and its CEO and founder, Kyle Robertson, the co-founder and former top executive at mental health startup Cerebral.</p>
<p>"At two consecutive telehealth companies that he founded and has led during the past six years—Cerebral, Inc., and then Zealthy, Inc.—defendant Kyle Robertson has engaged in wide-ranging lawbreaking that has harmed tens of thousands of telehealth patients," the DOJ wrote in its amended complaint.</p>
<p>Prosecutors said Zealthy presents itself as providing legitimate telehealth services but is engaged in "systemic improper and dangerous telemedical practices," including "the routine ordering of prescriptions by foreign call-center contractors and other non-clinicians with no medical license and the systemic misuse of the name and National Provider Identifier (“NPI”) number of doctors to order many thousands of prescriptions for patients they did not actually treat or order a prescription for, and without their knowledge or clinical supervision—and even when they have not even been employed by Zealthy," according to the court documents.</p>
<p>The DOJ filed the motion seeking an asset freeze after it obtained new documents during the discovery phase of a 2024 lawsuit against three telehealth companies.</p>
<p>Prosecutors say Zealthy lost its certification from a pivotal credentialing organization, LegitScript, in 2025.</p>
<p>"Robertson has falsely touted low costs, misrepresented the nature of telemedicine services, and made hollow assurances that consumers! will “Pay $0 today” and can “Cancel anytime”—all to bait consumers into hard-to-cancel subscriptions while often immediately initiating unauthorized charges to their cards," the DOJ alleges.</p>
<p>The DOJ alleges that Robertson, while leading both Cerebral and Zealthy, "charged tens of thousands of consumers without their consent while deliberately stymying their subscription cancellation attempts in violation of the Restore Online Shoppers’ Confidence Act (ROSCA)."</p>
<p>The DOJ also accuses Robertson of directing the misuse of patients’ private health information without their knowledge, including data on their medical conditions, symptoms, and treatments, "to drive millions of dollars’ worth of social media marketing campaigns."</p>
<p>In its motion, the DOJ argues that a receivership and asset freeze are necessary to stop "Robertson and Zealthy’s runaway campaign of lawbreaking, which is actively deceiving telehealth patients, endangering their safety, and raiding their bank accounts," the DOJ said in the complaint.</p>
<p>The filing said the consumer redress and FTC rule violation penalty in the case may bankrupt Zealthy.</p>
<p>Zealthy launched in early 2023 and offers telehealth services for GLP-1 weight loss programs, mental health care, skincare treatments, hair loss solutions, erectile dysfunction support, birth control, and hormone-optimization therapies.</p>
<p>In November 2024, Cerebral agreed to <a href="https://www.justice.gov/usao-edny/pr/telehealth-company-cerebral-agrees-pay-over-36-million-connection-business-practices?bm-verify=AAQAAAAN_____yseBRHa2xYSzsnQWNjRM9KWzrUUy-qIy_hLlk1OW8q_hSZ71UuVwavCNxqSh6SRgKMiXJPssVdSHl6voGkpN6bUVY2wahISZhw32WeZ-SbC5pgdegZ-may2h4uCYdAXa-eNxsJPDXAzT9o4yd1mO4522F51mCPiVhNMwYLu_raFzHHc-WVUB5Z4xBvKWB_DnQr9Zh3RrUgTxib-TzqUZRSQC5JIEYKjBNRHd_zhPKfvSmcsQMAzebg-z4g8nz-DjrQrz0yIHOiuMwaFITH-LqoflGY-6oCectwjIDvcDCP25KGzN8RkKmflukCSGh0yREbGn0sryEzmqUbcEx7NojPg12GCNQZ0usf-3SNWmO_M3sb7DclqTE_LxSWGWkpMaIAzAyHiygpf7-mC-g">pay a fine </a>of more than $3.6 million for allegedly engaging in practices that encouraged the unauthorized distribution of controlled substances from 2019 to 2022. The fine was part of a non-prosecution agreement with the U.S. Attorney’s Office for the Eastern District of New York. An additional fine against Cerebral was deferred "in light of the company’s current financial condition," the federal government said when the agreement was announced.</p>
<p>In April 2024, the digital behavioral health company <a href="https://www.fiercehealthcare.com/regulatory/cerebral-agrees-7m-settlement-ftc-doj-over-user-data-sharing-cancellation-practices">agreed to pay the government</a> more than $7 million to resolve allegations that it disclosed consumers’ sensitive personal health information and other sensitive data to third parties for advertising purposes and failed to honor its easy cancellation promises.</p>
<p>Robertson <a href="https://www.fiercehealthcare.com/health-tech/online-mental-health-startup-cerebral-will-stop-prescribing-controlled-substances">stepped down</a> from Cerebral in May 2022 as the company faced a DOJ investigation for its prescribing practices as "possible violations" of the Controlled Substances Act. </p>
</div></p>
</div>
<div style="font-family:Helvetica, sans-serif; font-size:13px; text-align: center; color: #666666; padding:4px; margin-bottom:2px;"></div>
</td>
</tr>
</tbody>
</table>
<p><strong>Forwarded by:<br />
Michael Reeder LCPC<br />
Baltimore, MD</strong></p>
<p><strong>This information is taken from free public RSS feeds published by each organization for the purpose of public distribution. Readers are linked back to the article content on each organization's website. This email is an unaffiliated unofficial redistribution of this freely provided content from the publishers. </strong></p>
<p> </p>
<p><s><small><a href="#" style="color:#ffffff;"><a href='https://blogtrottr.com/unsubscribe/55wMX0?signature=d3c0935c0cdbd702ef36555e19dabf249df69613966bbb1ace8ed255419b1ff3'>unsubscribe from this feed</a></a></small></s></p>