Digital Health

Digital heart health tool linked to significant blood pressure reductions in women before and after menopause

Key Takeaways: 

• One of the largest studies of a digital heart health programme has found that Hello Heart significantly reduces blood pressure in women, particularly those in perimenopause and postmenopause.
• Blood pressure reductions were sustained over 12 months, even among women with severely elevated baseline levels.
• The findings highlight the potential of targeted digital tools to address a long-overlooked cardiovascular risk window in women’s health.
  

Introduction

A new peer-reviewed study published in the American Journal of Preventive Cardiology has found that the Hello Heart digital heart health programme is associated with meaningful reductions in blood pressure among people with hypertension, particularly among women undergoing perimenopause and postmenopause. These are life stages where cardiovascular risk rises sharply, yet care and prevention strategies are often lacking.

The study, titled “Blood Pressure Reduction by Gender and Menopause Status Among Hypertensive Participants of a Mobile Health Cardiovascular Risk Self-Management Program”, analysed data from over 48,000 participants enrolled in the Hello Heart programme between July 2015 and September 2023. Women comprised 55 percent of the study cohort, making this one of the largest evaluations to date of a digital intervention focused on cardiovascular health in women.

Results: Blood pressure reductions across sex and menopause status

The study found that participants using the Hello Heart app experienced statistically significant and sustained reductions in blood pressure. Although the programme was effective for all users, women achieved particularly notable results.

• Greater reductions in women: Women experienced greater improvements in blood pressure than men, even after adjusting for confounding factors such as age and medication use.
• Improvement during menopause: Women in postmenopause, despite having higher baseline blood pressure levels, achieved blood pressure reductions comparable to their premenopausal peers.
• Impact on those at highest risk: Among individuals with stage 2 hypertension (≥140 mmHg), reductions were sustained over 12 months, and women in this group achieved greater reductions than matched male participants.
  

Dr Jayne Morgan, the study’s lead author, cardiologist, and Vice President of Medical Affairs at Hello Heart, commented:

“This study is a crucial step forward in understanding how to support women at a time when they face heightened cardiovascular risk. We’ve known for years that menopause is a turning point for heart health, but until now, there’s been little data on scalable, effective ways to intervene during this phase. These findings show that digital tools can help close the gap.”

About the Hello Heart programme

Hello Heart is a mobile-based digital health platform designed to help individuals self-manage their cardiovascular risk. The programme combines several features:

• A Bluetooth-connected blood pressure monitor
• Reminders and tools to support medication adherence
• AI-driven personalised coaching
• Educational content tailored specifically to people in menopause
  

By integrating behavioural science and personalised insights, the programme aims to empower individuals to make sustained changes in cardiovascular health behaviours. The targeted educational content for menopausal women is a distinguishing feature and was cited in the study as a likely contributor to the improvements seen.

Experts call for proactive use of digital health in women’s heart care

Dr Erin Michos, Professor of Medicine and Director of Women’s Cardiovascular Health at Johns Hopkins University School of Medicine, and a co-author of the study, highlighted the significance of digital access:

“Menopause is a time of change, but it should not be the beginning of cardiovascular decline. Accessible digital health programmes like Hello Heart’s can empower women with actionable tools to reduce risk of a serious event.”

Dr Martha Gulati, Director of Cardiovascular Prevention and Associate Director of the Barbra Streisand Women’s Heart Center at Cedars-Sinai, echoed this sentiment:

“We need to think of menopause as a window of opportunity when it comes to heart health and get proactive about helping women understand their risk and take action. This is a strong example of how we can do that.”

A scalable solution for a neglected risk window

The study’s findings add weight to the growing call for tailored, sex-specific strategies in cardiovascular prevention. Menopause is now widely recognised as a critical juncture for cardiovascular health, yet many routine care models still overlook this stage.

By demonstrating that a scalable, app-based intervention can deliver measurable and lasting benefits for women during and after menopause, the study represents a promising step forward in closing the gender gap in heart health outcomes.

Read More

AI detects unseen diabetes risk in people with normal blood sugar levels

Key Takeaways:

• A large-scale, multimodal study found that many individuals with ‘normal’ glucose test results still experience harmful glucose spikes, revealing hidden diabetes risk.
• Using continuous glucose monitoring (CGM) and machine learning, researchers developed a personalised glycaemic risk model that outperformed standard tools such as HbA1c.
• The model revealed wide risk variability among people with prediabetes and highlighted the influence of microbiome diversity, diet, and lifestyle on glucose regulation.
  

Rethinking Glucose Testing in Diabetes and Prediabetes

A new study published in Nature Medicine has shown that current diagnostic methods for type 2 diabetes (T2D) and prediabetes may be missing key signals. Researchers analysed detailed health data from over 2,400 individuals, identifying significant differences in glucose spike patterns between those with normoglycaemia, prediabetes, and T2D — even when standard test results appeared normal.

The study’s central innovation is a multimodal risk model that integrates diverse data sources to build personalised glycaemic risk profiles. These profiles may help identify individuals with prediabetes who are at greater risk of progressing to T2D, offering a more precise and inclusive alternative to existing tools such as fasting glucose and glycated haemoglobin (HbA1c).

A Complex Picture of Blood Sugar Regulation

Despite being widely used, HbA1c and fasting glucose do not fully capture the complexity of glucose regulation. Blood glucose fluctuations can be influenced by numerous factors — including diet, genetics, sleep, stress, gut microbiome composition, physical activity, and age. Of particular concern are postprandial glucose spikes, defined as increases of at least 30 mg/dL within 90 minutes of eating, which have been documented in people who otherwise appear healthy.

“Many people have these hidden spikes in blood sugar that are not captured by the standard diagnostics,” said Eran Segal, senior author of the study and a computational biologist at the Weizmann Institute of Science in Israel.

The PROGRESS Study: A Multimodal Approach to Glycaemic Risk

To investigate glycaemic variability more comprehensively, the researchers conducted the PROGRESS study — a fully remote, U.S.-based digital clinical trial. A total of 1,137 participants were enrolled, 48.1% of whom identified as being from groups historically underrepresented in biomedical research. Participants spanned the full glycaemic spectrum from normoglycaemia to type 2 diabetes.

In addition to wearing continuous glucose monitors (CGMs) for 10 days, participants provided blood, stool, and saliva samples from home, submitted electronic health records, and logged food intake and lifestyle habits via mobile apps. The team excluded individuals with conditions likely to skew results, such as recent antibiotic use, pregnancy, or type 1 diabetes.

CGM data were processed into one-minute intervals, and six key glycaemic metrics were calculated:

• Average glucose
• Time spent in hyperglycaemia
• Spike frequency
• Spike duration
• Nocturnal hypoglycaemia
• Spike resolution time

Complementary data included heart rate, sleep patterns, diet, physical activity, genomic profiles (including polygenic risk scores), and gut microbiome diversity.

Building and Testing a Machine Learning Model

Using this rich dataset, the researchers developed a machine learning model that combined data on demographics, anthropometrics, diet, microbiome, and CGM output. The goal was to predict glycaemic status and identify hidden risks.

To validate the model, they applied it to a second dataset — the Israeli Health and Prevention Project (HPP) — a longitudinal observational study of over 1,300 individuals. The model achieved high accuracy in distinguishing normoglycaemia from T2D and, importantly, revealed substantial variability in risk levels among people with prediabetes, even when HbA1c values were identical.

What the Data Revealed

Of the 1,137 individuals enrolled in the PROGRESS study, 347 were included in the final analysis:

• 174 with normoglycaemia
• 79 with prediabetes
• 94 with T2D
  

Key findings included:

• Glucose spike metrics differed significantly across groups, especially between T2D and the others.
• Prediabetic individuals more closely resembled normoglycaemic individuals than those with T2D in terms of spike frequency and intensity.
• Gut microbiome diversity was inversely associated with glucose spikes — suggesting that a more diverse microbiome supports healthier glucose control.
• Higher body mass index (BMI), resting heart rate, and HbA1c were linked to worse glycaemic outcomes.
• Physical activity was associated with more stable glucose levels, while higher carbohydrate intake led to faster spike resolution but also more frequent and intense spikes.
  

Implications for Diabetes Prevention

This study offers several key insights:

• Standard diagnostic tools may miss important warning signs, particularly among people with prediabetes.
• Personalised, multimodal risk profiling can detect hidden glucose dysregulation and guide early interventions.
• The model’s success in a diverse, decentralised cohort also highlights the potential of remote, real-world clinical research to advance precision medicine.
  

“We see people with the same HbA1c level but very different glucose dynamics,” said Segal. “By using continuous glucose monitoring and other personalised data, we can find those at greater risk earlier.”

Limitations and Future Directions

While the study design and analysis were robust, several limitations must be acknowledged:

• CGM performance can vary by device.
• Self-reported data, such as food logs, may introduce errors.
• Some participants were using antihyperglycaemic medications, which could influence results.
  

Further research is needed to validate findings in broader populations and assess the long-term predictive value of these glycaemic risk profiles.

Towards More Precise and Inclusive Diabetes Care

The study underscores the importance of moving beyond a one-size-fits-all approach in diabetes diagnostics. By integrating CGM data with lifestyle, genetic, and microbiome information, healthcare professionals may one day tailor prevention strategies to individuals — catching hidden risk earlier and supporting more effective interventions.

“This is not just about identifying risk,” said Segal. “It is about offering people more accurate and personalised care.”

Read More

Twelve-month trial finds digital asthma programme significantly enhances symptom control in adults

Key Takeaways:

• A year-long decentralised trial found that a digital asthma self-management (DASM) programme significantly improved asthma control, particularly in individuals with uncontrolled asthma at baseline.
• Participants using the Apple Watch–integrated app saw greater improvements in Asthma Control Test (ACT) scores than those receiving usual care.
• The study highlights the importance of culturally responsive digital health solutions, as effectiveness varied across different ethnic groups.
  

A major decentralised clinical trial led by Elevance Health and the University of California, Irvine, in collaboration with Apple, Inc., has demonstrated that a digital asthma self-management (DASM) programme can meaningfully improve asthma symptom control in adults. The benefits were sustained over a 12-month period, making this one of the most substantial demonstrations to date of the potential for digital health technologies to enhance chronic disease management.

Trial Design and Participation

The randomised, controlled and virtual study enrolled 901 adult participants across 41 U.S. states. Participants included those with both commercial and Medicaid insurance, ensuring broad demographic representation. The intervention group used an iPhone application integrated with the Apple Watch, enabling them to:

• Track asthma-related health data
• Log symptoms
• Receive personalised alerts and reminders
  

This combination of continuous data monitoring and tailored prompts formed the core of the DASM approach, showcasing its scalability and potential for nationwide implementation.

Symptom Control Outcomes

The trial measured asthma control using the validated Asthma Control Test (ACT). Among individuals with uncontrolled asthma at baseline, those in the DASM group improved their ACT scores by an average of 4.6 points over 12 months, compared with only 1.8 points in the control group receiving standard care. Improvements of this magnitude typically indicate meaningful changes in symptoms—such as fewer episodes of breathlessness or reduced disruption to daily activities.

Even among those with controlled asthma at the outset, the DASM group showed favourable outcomes when compared with usual care, albeit with a smaller effect size.

Equity and Engagement Insights

Importantly, the study observed positive results across different insurance types and ethnicities, reinforcing the potential for DASM programmes to be deployed broadly. However, the improvements were less pronounced among African American participants, suggesting a need for culturally adaptive strategies in digital health design.

“These results support continued development of digital asthma self-management programmes,” said Dr Jordan Silberman, co-lead author and Director of Clinical Analytics at Elevance Health. “Findings also underscore the need for culturally adaptive strategies to ensure equitable engagement and impact across all communities affected by asthma.”

Secondary Benefits and Usage Patterns

Participants using the DASM intervention also experienced several secondary benefits, including:

• Improved medication adherence
• Greater confidence in managing their condition
• Reduced impact of asthma symptoms on work productivity
  

The researchers found that higher levels of app engagement—such as more frequent symptom logging—were strongly associated with better clinical outcomes.

“We are proud to support research that not only improves chronic disease management but prioritises underserved populations often left behind in digital innovation,” said Dr Richard A. Lee, co-author and Associate Clinical Professor of Pulmonary and Critical Care Medicine at UC Irvine.

Conclusion

This study presents compelling evidence for the role of digital technologies in long-term asthma management, especially when personalised, user-friendly tools are combined with wearable technology. It also highlights the critical importance of equity and inclusion in digital health to ensure that benefits are widely shared among all individuals living with asthma.

Read More

AI turns old diabetes drug Halicin into powerful antibiotic against superbugs

Key Takeaways:

• Halicin, originally developed for diabetes, significantly inhibited 17 out of 18 tested multidrug-resistant (MDR) bacterial strains, demonstrating potential as a broad-spectrum antibiotic.
• The study confirms Halicin’s unique mechanism of action and highlights its effectiveness against ESKAPE pathogens, with the exception of Pseudomonas aeruginosa.
• This work underscores the power of artificial intelligence in repurposing old pharmaceuticals to address urgent public health threats such as antimicrobial resistance.
  

Introduction

Artificial intelligence (AI) continues to redefine the frontiers of medicine. A new study published in Antibiotics demonstrates how AI can accelerate drug discovery by uncovering novel uses for discontinued or previously overlooked pharmaceuticals. The research focuses on Halicin – a drug initially designed to treat diabetes – and its newly revealed antibacterial properties. Specifically, the study investigates Halicin’s efficacy against 18 clinical strains of multidrug-resistant (MDR) bacteria, revealing promising results that may influence future antimicrobial strategies.

The Global Threat of Superbugs

Multidrug-resistant bacteria – often referred to as ‘superbugs’ – are one of the most pressing global health threats. Among these, the ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species) are of particular concern. These organisms are named for their ability to “escape” the effects of antibiotics and are repeatedly flagged by the World Health Organization (WHO) as urgent priorities.

Conventional antibiotic pipelines are struggling to keep pace, constrained by labour-intensive discovery methods and rising rates of antimicrobial resistance. In this context, the use of AI and machine learning (ML) offers a transformative approach – enabling the rapid screening and simulation of existing compounds to uncover overlooked antibacterial effects.

From Diabetes to Antibacterial: The Case of Halicin

Halicin was originally developed as a c-Jun N-terminal kinase (JNK) inhibitor for managing diabetes-related pathways. However, in a breakthrough application of AI at the Massachusetts Institute of Technology (MIT), deep learning algorithms identified Halicin’s unique ability to disrupt the bacterial proton-motive force – a critical energy-generating mechanism. Unlike conventional antibiotics, Halicin does not target bacterial cell walls or protein synthesis, making it an attractive candidate for combatting antibiotic-resistant bacteria.

Despite its potential, limited studies had evaluated Halicin’s minimum inhibitory concentrations (MICs) across clinically relevant MDR strains, leaving significant gaps in the evidence base.

Study Aims and Methodology

This new study – the first of its kind in Morocco – sought to quantify Halicin’s MIC values against a wide panel of MDR bacterial isolates collected from Moroccan hospitals. A total of 18 clinical isolates were tested, including strains from the ESKAPE group. To ensure quality control, two standard reference strains – Staphylococcus aureus ATCC® 29213™ and Escherichia coli ATCC® 25922™ – were also included.

Initial testing involved agar disk diffusion assays to confirm multidrug resistance against 22 commonly used antibiotics. Subsequent MIC assays were conducted using broth microdilution methods in accordance with protocols established by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) and the Clinical and Laboratory Standards Institute (CLSI).

Dose-response curves were generated to determine the relationship between drug concentration and bacterial growth inhibition. Scanning electron microscopy (SEM) was used to visualise morphological changes in E. coli after Halicin treatment. The Kruskal–Wallis non-parametric test was employed to analyse differences in MIC distributions between species.

Key Findings

Halicin exhibited potent antibacterial effects against a majority of the tested strains:

• The MIC for E. coli ATCC® 25922™ was 16 μg/mL, and for S. aureus ATCC® 29213™, 32 μg/mL.
• Among the clinical MDR isolates, most MIC values ranged from 32 to 64 μg/mL.
• Pseudomonas aeruginosa was notably resistant to Halicin at all concentrations tested. Researchers attributed this to its highly selective and impermeable outer membrane, which likely hinders drug penetration.
  

The results suggest that Halicin maintains broad-spectrum antibacterial activity, even against highly drug-resistant organisms. Its distinct mode of action – targeting bacterial energy production – may also make it less susceptible to common resistance mechanisms.

Implications for Antimicrobial Research

This study represents an important step forward in antimicrobial development, particularly in the context of drug repurposing. As the authors note, “The present study validates the antibacterial efficacy of Halicin, a largely discontinued anti-diabetic relic, in significantly inhibiting the growth of 17 of the 18 (94%) clinical MDR bacterial isolates tested.”

Given its ability to evade resistance mechanisms by targeting an unconventional bacterial function, Halicin holds promise for future clinical application – pending further investigation into its pharmacokinetics, toxicity, and safety in human populations.

Moreover, the study highlights the transformative impact of AI and machine learning in drug discovery. By enabling the identification of novel therapeutic properties in existing compounds, these technologies could accelerate the development of urgently needed antimicrobials.

Future Directions and Cautions

While the findings are encouraging, the authors emphasise that further research is essential. Key next steps include:

• Evaluating Halicin’s pharmacological safety and tolerability in vivo
• Investigating synergistic effects with existing antibiotics, particularly against resistant organisms like P. aeruginosa
• Establishing bacterial resistance monitoring programmes to detect any emerging resistance trends
  

Importantly, the limited current use of Halicin means that no resistance has yet been observed – a favourable position that must be safeguarded through careful stewardship.

Conclusion

Halicin – once relegated to the pharmaceutical archives – has re-emerged as a compelling antimicrobial candidate through the application of AI-driven drug discovery. This study reinforces its efficacy against a wide range of multidrug-resistant bacteria, positioning it as a potential weapon in the global fight against antimicrobial resistance. As AI continues to evolve, its role in revolutionising medicine – particularly in addressing antibiotic scarcity – is only beginning to be realised.

Read More

AI-powered digital twin developed to anticipate individual health trajectories

Key Takeaways:

• Researchers at the Weizmann Institute of Science have developed a personalised “digital twin” using artificial intelligence to predict disease risk, guide prevention strategies, and simulate treatment responses.
• This innovation builds on data from the Human Phenotype Project—a 25-year longitudinal study collecting in-depth biological, genetic, and lifestyle data from tens of thousands of participants worldwide.
• The AI model can estimate a person’s biological age across 17 body systems and has already shown success in detecting pre-diabetes and menopause onset earlier than conventional tests.
  

Rethinking personal health decisions

Many individuals naturally simulate potential outcomes before making major life decisions. However, doing so for one’s health—such as selecting an appropriate treatment or diet—is often much more complex. Personal biological differences make it difficult to predict how a given choice will affect a specific individual.

In response to this challenge, researchers from Professor Eran Segal’s laboratory at the Weizmann Institute of Science have developed an artificial intelligence-powered “digital twin.” This digital twin is capable of identifying an individual’s risk for certain diseases, simulating how they might respond to different interventions, and guiding preventative care decisions. Their findings, recently published in Nature Medicine, are rooted in data collected through the Human Phenotype Project—a large-scale initiative that has gathered detailed medical information from over 13,000 individuals.

Beyond the genome: Capturing the full human picture

Although the Human Genome Project, launched in 1990, identified tens of thousands of genes linked to human traits and diseases, it quickly became clear that genetics alone could not fully explain health outcomes. Environmental factors, the microbiome, ageing processes, and lifestyle play crucial roles.

To capture this broader picture, Professor Segal initiated the Human Phenotype Project in 2018. This 25-year longitudinal study involves comprehensive testing of participants every two years across 17 different body systems. The assessments include:

• Anthropometric measurements
• Nutritional intake tracking
• Ultrasound imaging
• Bone mineral density scanning
• Voice recordings
• Home-based sleep studies
• Two-week continuous glucose monitoring
• Gene sequencing
• Protein profiling
• Microbiome analyses of gut, vaginal, and oral samples
  

“When we launched the project in Israel in 2018, our initial goal was 10,000 participants,” Segal stated. “Since then, more than 30,000 people have signed up, and we hope to reach 100,000 in the future.”

The project is now expanding internationally, with new branches in Japan and the United Arab Emirates, the latter in collaboration with Professor Eric Xing from the Mohamed bin Zayed University of Artificial Intelligence. The age range of participants is also expanding to include both younger and older individuals, increasing the diversity of the data collected.

“We recognised the importance of sharing this resource with the scientific community and have now made it accessible digitally to research groups worldwide, while maintaining the privacy of the participants,” Segal added. “We believe that the data we have compiled will profoundly affect the field of medicine.”

Measuring biological age: A more nuanced health indicator

Traditional medicine tends to benchmark health indicators against average values for one’s age and sex. However, these reference ranges can miss important individual variations in health and disease risk.

To address this, a team led by Drs Lee Reicher and Smadar Shilo within Segal’s lab developed an AI model to calculate a person’s biological age by analysing 17 physiological systems across the body. The technology is based on a platform developed by Pheno.AI, a health-focused artificial intelligence company.

“The model assigns scores to each body system and compares these values to the expected values for the participant’s chronological age, sex and body mass index,” explained Segal. “Based on the deviation from these predicted values, the model determines the participant’s biological age. The older the apparent age of a body system, the greater the risk of associated diseases.”

One of the early successes of this model has been in the detection of pre-diabetes. By analysing blood glucose patterns across the lifespan, the team identified pre-diabetes in 40% of participants who had been classified as healthy by conventional testing.

Gender-specific patterns have also emerged. “While men’s biological age generally increases relatively linearly, we observe an acceleration in women’s biological ageing during their fifth decade of life,” Segal noted.

“Menopause is a pivotal event in many medical respects, and it appears to reset the biological age clock. For example, we found that a decrease in bone density is more strongly correlated with the time that passed since the onset of menopause than with chronological age. Furthermore, our measurements make it possible to detect the start of menopause early, so that hormonal treatment can be planned accordingly.”

Microbiome signatures for early disease detection

The Human Phenotype Project has also revealed new paths for early diagnosis of conditions such as breast cancer, endometriosis, and inflammatory bowel disease. These illnesses are often preceded by detectable changes in the composition of the individual’s microbiome. Such microbiome alterations form unique “signatures” that can be identified well before symptoms emerge.

The rise of the digital twin: Personalised medicine in action

The most transformative potential of the project lies in its application of AI to create a digital twin for each participant—a sophisticated, personalised model that simulates future health trajectories.

This ongoing work, led by doctoral researcher Guy Lutsker, involves training an AI system on the full suite of medical data from each participant. The model learns by repeatedly being asked to predict missing pieces of information based on the rest of the record. Over time, this training allows the AI to anticipate likely future events, such as disease onset or treatment response.

Already, the system has successfully forecast which individuals with pre-diabetes are most likely to progress to type 2 diabetes within two years. It has also been used to simulate the effects of different dietary interventions and medications, helping to determine which strategies are most likely to benefit each person.

Eventually, the digital twin model is expected to incorporate all aspects of the dataset, enabling the prediction of a wide range of health events. This may significantly reduce the trial-and-error process that currently characterises treatment decisions in many areas of medicine.

Bringing AI-driven health insights directly to participants

Professor Segal emphasised the critical role of the participants in enabling this progress:
“This achievement is primarily made possible by the community of participants in the Human Phenotype Project. It is a dedicated group of individuals committed to advancing medicine and to the continuous monitoring of their health.”

To empower participants with direct access to their health data and insights, the team is now developing a mobile application. This app will provide each person with a dynamic view of their individualised “health trajectory”—a predictive map of their future health status.

“We are living in an era of incredibly rapid change. The realms of health and medicine will undergo dramatic transformations in the coming years, becoming increasingly AI-driven.

“Our project is poised to be a leading global source of information and innovation, and this is all thanks to our participants. I want to take this opportunity to express my sincere gratitude to each and every one of you—your exceptional collaboration is the true driving force behind this revolution in medicine,” Segal concluded.

Read More

Healthcare AI drives nearly $4 billion in venture funding, powering digital health’s strong momentum in 2025

Key Takeaways: 

• AI-enabled healthcare startups secured nearly $4 billion in venture capital in the first half of 2025, accounting for 62% of digital health investment.
• Adoption of generative AI, particularly ambient documentation tools, is accelerating across healthcare, with some hospitals reporting utilisation rates as high as 90%.
• Despite economic and policy uncertainties, the digital health sector is seeing robust M&A activity and larger average deal sizes, underscoring investor confidence.
  

Strong start for digital health in 2025

The digital health sector began 2025 with considerable strength, marked by major funding rounds for AI-focused startups, a revival in the IPO market, and increased uptake of AI tools by healthcare providers.

According to Rock Health’s mid-year funding report, venture capital investment in digital health reached $6.4 billion in the first six months of 2025, surpassing the $6 billion recorded in the first half of 2024 and the $6.2 billion from the same period in 2023.

In the second quarter alone, digital health companies raised $3.4 billion, significantly above the quarterly average of $2.6 billion since early 2023. By comparison, the sector raised $10.2 billion over all of 2024 and $10.9 billion in 2023.

Rock Health analysts noted that the sector’s resilience is notable given an uncertain policy and economic environment. “Big developments in the past six months—the thawing of the IPO freeze with Hinge Health and Omada Health going public and investor excitement for AI solutions—signals market maturity and momentum,” wrote Megan Zweig, Mihir Somaiya and Tiffany Marie Ramos.

Fewer deals but larger investments

Although the first half of 2025 saw fewer funding deals—245 compared to 273 in H1 2024—the average deal size rose to $26.1 million, up from $20.4 million last year. This growth was driven by larger investments in later-stage rounds (Series B through D) and the accelerating influence of AI across digital health.

AI-enabled startups received the bulk of venture funding, capturing 62% of capital deployed so far this year, amounting to $3.95 billion. On average, these companies raised $34.4 million per round—an 83% premium over the $18.8 million typical of their non-AI counterparts.

For Series A and B rounds specifically, AI-focused startups secured average investments of $24.4 million and $54.8 million, respectively, compared to $15.6 million and $39.6 million for non-AI digital health businesses.

Rock Health defines AI-enabled startups as those using artificial intelligence, machine learning, or deep learning as central elements of their products or services.

AI reshaping key areas of healthcare

The top three funded areas in the first half of 2025 were non-clinical workflow ($1.9 billion), clinical workflow ($1.9 billion) and data infrastructure ($893 million). These three areas combined accounted for 55% of total digital health investment, marking the first time such a concentration has been recorded since Rock Health began tracking funding in 2011. All are being fundamentally transformed by AI and automation.

AI-enabled startups also dominated the largest funding rounds. Of the 11 mega deals (over $100 million) in H1 2025—already on pace to surpass the 17 mega deals seen throughout 2024—nine went to AI-driven companies.

Notable transactions included:

• Abridge, an AI scribe company, which raised $300 million in a Series E round in June after securing $250 million in Series D funding just four months earlier.
• Innovaccer ($275 million Series F)
• Hippocratic AI ($141 million Series B)
• Qventus ($105 million Series D)
• Truveta ($320 million Series C)
• Commure ($200 million growth round)
• Persivia ($107 million growth round)
• Tennr ($101 million Series C)

OpenEvidence, which closed a $75 million Series A in February, is reported by Newcomer to be approaching a mega deal.

Rapid uptake of generative AI tools

Investor confidence in healthcare AI is matched by rapid adoption among providers. The Peterson Health Technology Institute (PHTI) observed that medical ambient documentation tools are spreading at an unprecedented rate—faster than any previous technology in healthcare.

“The promise of these solutions to reduce burnout and improve workflows has driven an industry with notoriously long sales cycles and implementation timelines to adopt ambient scribes,” PHTI wrote in a recent report. “Ambient scribe represents the first large-scale application of generative AI in health systems.”

Adoption rates are now estimated at between 30% and 40% across physician groups, with some leading hospitals reporting usage levels of up to 90%.

M&A activity and new private equity strategies

Digital health is also experiencing significant merger and acquisition activity. There were 107 M&A transactions in the first half of 2025, putting this year on track to nearly double the 121 deals recorded in all of 2024.

“Digital health companies continue to be the most frequent acquirers of other digital health companies, accounting for 63% of all deals so far this half,” wrote Zweig and colleagues. They noted that many are pursuing what Rock Health terms a “tapestry weaving” strategy—acquiring diverse capabilities to build more comprehensive offerings. Examples include longevity-focused startup Superpower acquiring Base and Feminade.

Private equity firms are experimenting with a new playbook that pairs AI-native startups with established healthcare businesses. New Mountain Capital, for instance, combined Access Healthcare (a revenue cycle and business process outsourcing company) with the AI technologies of SmarterDx and Thoughtful.ai to form Smarter Technologies. Earlier this year, the same firm acquired Machinify and integrated its AI solutions into a combined entity with traditional payment integrity companies Apixio, Varis, and The Rawlings Group.

“The bet is that combining the established distribution networks and trusted services of legacy companies with cutting-edge technology will drive meaningful efficiency, margin, and scale gains. New Mountain Capital (NMC) has been putting this new playbook to work,” Rock Health researchers explained.

Navigating policy and economic uncertainty

Despite strong funding and operational trends, digital health startups face notable policy and economic headwinds. Broader economic uncertainty, trade tariffs, and the implications of President Donald Trump’s wide-ranging healthcare legislation all loom large.

The bill’s provisions—including Medicaid work requirements and changes to the Affordable Care Act marketplace—are projected to leave millions uninsured. This could shrink addressable markets and exacerbate losses from uncompensated care. Although the legislation offers a modest temporary rise in Medicare reimbursement rates, it does not incorporate long-term physician payment reforms found in earlier drafts.

Rock Health advised digital health startups to align their solutions with federal priorities, such as chronic disease management and food-as-medicine initiatives. Federal agencies and legislative committees have indicated a desire for a more technology-driven healthcare system. The Department of Health and Human Services recently sought input on the use of AI in clinical decision support and new care delivery models.

“Participating can both steer federal digital health initiatives and strengthen future lobbying positioning. All-in-all, early engagement during this first-year window could shape the contours of policies that will determine how healthcare is paid for, regulated, and accessed—all of which set innovation trajectories in motion,” Zweig wrote.

Read More

Smartwatch data shown to spot early Parkinson’s signs more accurately than traditional tests

Key Takeaways: 

• A UK-led study found that data from ordinary smartwatches can detect early biological changes linked to Parkinson’s disease more sensitively than established clinical risk scores.
• The research demonstrated that a digital risk score derived from wearable data outperformed traditional assessments and closely aligned with gold-standard brain scans and spinal fluid tests.
• This approach could provide a simpler, non-invasive screening method, making early Parkinson’s detection more accessible and cost-effective.
  

Wearable technology uncovers early Parkinson’s signs

Data collected from smartwatches have been shown to detect early brain changes associated with Parkinson’s disease with greater sensitivity than widely used clinical risk assessments, according to new research.

The study, led by Dr Cynthia Sandor at the UK Dementia Research Institute at Imperial College London, highlights the potential of wearable devices to identify subtle, early signs of Parkinson’s long before a formal diagnosis. The findings, published in eBioMedicine, suggest that everyday technology could offer a more accessible and less invasive way to screen for Parkinson’s risk.

Comparing smartwatch data with traditional clinical and biological tests

Previous work from Dr Sandor’s team had already demonstrated that wearable devices could predict Parkinson’s years ahead of diagnosis by analysing subtle changes in movement. However, this latest study is the first to directly compare smartwatch-derived data with gold-standard biological markers – including specialist brain imaging and cerebrospinal fluid tests – as well as with established clinical scoring systems.

The researchers used data from the Parkinson’s Progression Markers Initiative (PPMI), a major international study led by the Michael J. Fox Foundation. Participants in this project wore Verily smartwatches for an average of 16 months, during which the devices continuously and passively tracked their sleep patterns, heart rate and physical activity.

These biological markers included dopamine transporter imaging (DaTscan), a type of brain scan that highlights loss of dopamine function, and cerebrospinal fluid tests looking for misfolded α-synuclein – both considered key hallmarks of early Parkinson’s.

Developing a more sensitive digital risk score

Drawing on the extensive smartwatch data, the research team developed a digital risk score designed to differentiate people living with Parkinson’s from healthy individuals. They compared this digital score with the Movement Disorder Society (MDS) research criteria, a widely used clinical risk score for Parkinson’s, and also assessed it in a separate group of individuals who were considered at increased risk due to either genetic variants or early symptoms.

The digital score not only correlated strongly with both the clinical and biological markers but also demonstrated higher sensitivity than the MDS criteria in detecting early Parkinson’s-related changes. Notably, when the digital risk score was combined with a smell test (to detect hyposmia, a common early sign of Parkinson’s), it identified over 80% of people who had abnormal brain scans or spinal fluid markers.

Towards simpler, earlier detection

The study’s authors believe that this digital approach could pave the way for a straightforward, non-invasive screening tool to help identify individuals most likely to benefit from more detailed neurological assessments. This could potentially make early diagnosis of Parkinson’s both more widely available and more affordable.

Dr Cynthia Sandor, Edmond & Lily Safra Assistant Professor in Parkinson’s Disease at Imperial’s Department of Brain Sciences, said:

“Our findings suggest that everyday smartwatch data could help flag early signs of Parkinson’s long before a clinical diagnosis is made. The accuracy of this data is on par with current standard tests, which can be expensive and invasive. This kind of digital monitoring could be a game-changer – offering a simple, non-invasive way to screen those most at risk and helping to guide who should receive more definitive testing.”

Read More

NHS to become ‘digital by default’ under landmark 10-year plan to transform care delivery

Key Takeaways:

• The UK government’s new 10-year health plan aims to make the NHS ‘digital by default’, moving most outpatient care out of hospitals by 2035.
• Local neighbourhood health centres will house integrated services, from diagnostics to social support, fundamentally changing where and how people receive care.
• The ambitious strategy requires significant investment in technology, estates, and local leadership to succeed, with experts warning sustained support is essential.
  
  

Prime Minister Keir Starmer has unveiled the government’s new 10-year health plan, setting a bold ambition for the NHS to become ‘digital by default’. The strategy, launched on 3 July 2025, outlines how the majority of outpatient care will be delivered outside of traditional hospital settings by 2035, reshaping the health system to focus on community-based, technology-enabled care.

Ending ‘hospital by default’: a preventative, digital-first NHS

A press release issued by the Department of Health and Social Care (DHSC) on the same day stated that the longstanding status quo of ‘hospital by default’ would come to an end. Instead, the plan establishes a new preventative principle: care should be ‘digital by default’.

The DHSC noted:

“The government’s plan will bring it into the digital age, making sure staff benefit from the advantages and efficiencies available from new technology.”

This includes a pledge to roll out new digital tools for general practitioners over the next two years. These tools will support quicker referrals and introduce AI scribes, which are expected to “end the need for clinical note taking, letter drafting, and manual data entry to free up clinicians’ time.”

Additionally, digital telephony systems will be introduced so that calls to GP practices can be answered promptly, enabling same-day digital or telephone consultations when required.

Care on the doorstep: introducing the Neighbourhood Health Service

As part of this transformation, Starmer announced the creation of a Neighbourhood Health Service, designed to ensure that people receive care closer to their homes. Outlining his vision, he said:

“Our 10 year health plan will fundamentally rewire and future-proof our NHS so that it puts care on people’s doorsteps, harnesses game-changing tech and prevents illness in the first place.

That means giving everyone access to GPs, nurses, and wider support all under one roof in their neighbourhood – rebalancing our health system so that it fits around patients’ lives, not the other way round.”

Neighbourhood teams will be based in new health centres, intended to operate 12 hours a day, six days a week once fully established. These centres aim to bring traditionally hospital-based services – such as diagnostics, post-operative care and rehabilitation – into the community. They will also incorporate broader services, including debt advice, employment support, and stop smoking or weight management programmes.

A shift that could ‘turn the NHS on its head’

Health secretary Wes Streeting described the scale of the proposed changes as one of the most significant in NHS history:

“Our 10 year health plan will turn the NHS on its head, delivering one of the most fundamental changes in the way we receive our healthcare in history.

By shifting from hospital to community, we will finally bring down devastating hospital waiting lists and stop patients going from pillar to post to get treated.”

Healthcare leaders have largely welcomed the vision. Daniel Elkeles, chief executive of NHS Providers, called the plan potentially a “gamechanger” for the NHS, adding that it is:

“a win for patients who will be better informed and empowered to direct their care as never before.”

Matthew Taylor, chief executive of the NHS Confederation, also praised the announcement, describing it as:

“a vital step towards a more preventative, community-based NHS.”

However, he cautioned that realising this ambition would hinge on continued investment and local empowerment:

“Delivering on this ambition will require sustained investment in digital and estates, support for the NHS’s workforce, and a commitment to decentralise national control by empowering local leaders to do what is best for their populations.”

Underpinned by major investment in NHS technology

The plan builds on an earlier pledge of an additional £10 billion investment in NHS IT. Other recent measures include the introduction of ‘innovator passports’ to accelerate the rollout of new technologies, updates to the NHS App, and the implementation of an AI system designed to detect unsafe care.

Taken together, these initiatives represent a decisive shift towards a digitally enabled, locally delivered model of care – one intended to reshape the NHS for the next generation.

Read More

New study shows digital tests can uncover Alzheimer’s signs well ahead of clinical diagnosis

Key Takeaways:

• A recent review published in the Journal of Alzheimer’s Disease demonstrates that digital cognitive tests can detect signs of Alzheimer’s and other dementias long before traditional methods.
• Researchers identified that tiny delays in response times — even when answers are fully correct — serve as early biomarkers of cognitive decline.
• The approach builds on the Boston Process Approach, emphasising the value of analysing how individuals arrive at answers, not merely whether they are correct.
  

Digital Testing Shown to Identify Dementia Risk Well Before Symptoms Appear

A new study featured in the Journal of Alzheimer’s Disease has provided compelling evidence that digital cognitive assessments can uncover the earliest, often imperceptible, indications of cognitive decline — potentially years before conventional diagnostic methods or overt symptoms would allow detection. This advancement could transform how clinicians identify people at risk of Alzheimer’s disease and other dementias.

The paper, titled “Precision neurocognition: An emerging diagnostic paradigm leveraging digital cognitive assessment technology,” was co-authored by David J. Libon, PhD, Professor at the New Jersey Institute for Successful Aging at Rowan University, and Rod Swenson, PhD, Clinical Professor of Psychiatry and Behavioural Science at the University of North Dakota School of Medicine and Health Sciences. Both authors are also scientific advisors to Linus Health, a company specialising in digital brain health innovations.

Going Beyond Correctness: The Critical Role of Response Latency

The researchers focused on a key insight: that the true diagnostic power of digital assessments lies not just in whether a person answers correctly, but how quickly they do so. Their findings arose from analysing a brief, 7-minute speech-based digital cognitive test developed by Linus Health, which captures rich, quantifiable data related to neurodegenerative risk.

In their peer-reviewed synthesis, Drs. Libon and Swenson introduce new digital biomarkers based on what they term “latency” — the subtle pauses and reaction times between hearing or seeing a prompt and giving a response. They discovered that even when an individual achieves 100% accuracy on these tests, small delays in reaction times may signal early disruptions in the brain’s cognitive processing.

Dr. Libon underscored this point:

“As our review demonstrates, just because an individual answers a question on a digital assessment 100% correctly does not mean a 0% risk of cognitive impairment. We have shown here that the time it takes to correctly respond or recruit the necessary brain regions or strategies for efficient correct responding likely provides rich information regarding the probability of the eventual emergence of serious cognitive decline.”

This approach enables clinicians and researchers to detect vulnerabilities in brain health that would be invisible using traditional paper-based or purely accuracy-focused tests.

Anchored in the Boston Process Approach

The emerging model described in the study builds upon principles from the Boston Process Approach (BPA), a respected framework within neuropsychology. Developed by Dr. Edith Kaplan — under whom both Drs. Libon and Swenson trained and collaborated over many years — the BPA emphasises evaluating the process by which a person completes cognitive tasks, rather than focusing solely on the final score or outcome.

By drawing on the BPA, Linus Health’s digital assessments can uncover nuanced cognitive patterns. This allows clinicians to identify meaningful changes in the brain’s function long before such shifts would be apparent in standard assessments. As a result, this digital methodology could play a crucial role in guiding early interventions and planning for those at heightened risk of Alzheimer’s or related dementias.

Read More

Microsoft claims new AI tool can diagnose complex cases better than doctors

 Key Takeaways:

• Microsoft’s new AI “Diagnostic Orchestrator” outperformed doctors by a factor of four when diagnosing intricate medical cases, achieving 85.5% accuracy versus around 20% for clinicians under restricted conditions.
• The system employs an innovative method where multiple AI agents debate and collaborate, offering a transparent “chain of debate” reasoning process.
• Though promising in simulated trials, experts caution the technology remains at an early stage, untested in real-world clinical practice and not yet peer reviewed.
  

Microsoft unveils AI tool it says could revolutionise medical diagnosis

Microsoft has revealed an artificial intelligence–driven medical tool that it claims is four times more successful than human doctors at diagnosing complex health conditions. The development, which the company hopes will accelerate treatment pathways, was announced alongside research it believes could signal a major step towards so-called “medical superintelligence”.

The new system, named the Microsoft AI Diagnostic Orchestrator (MAI-DxO), represents the first major output from Microsoft’s dedicated AI health unit. This division was established last year under the leadership of Mustafa Suleyman, who co-founded DeepMind, the AI research lab now owned by Google. Microsoft attracted staff from DeepMind to join this effort.

Towards medical superintelligence

In an interview with the Financial Times, Mustafa Suleyman, now chief executive of Microsoft AI, described the results as an early milestone on the road to vastly more capable AI systems that could help ease staffing shortages and cut waiting times across pressured healthcare systems.

Suleyman said:

“Microsoft is nearing AI models that are not just a little bit better, but dramatically better, than human performance: faster, cheaper and four times more accurate. That is going to be truly transformative.”

How the “orchestrator” works

The technology’s backbone is an AI “orchestrator” that assembles virtual panels comprising five distinct AI agents, each acting like a specialised doctor. These agents take on roles such as formulating diagnostic hypotheses or selecting appropriate tests. They then interact and effectively “debate” among themselves before reaching a collective diagnostic decision.

This innovative framework employs a method called “chain of debate”, which compels the AI models to articulate step by step how they arrive at conclusions, offering unprecedented transparency into machine reasoning.

Testing against medical cases

To assess the system’s abilities, Microsoft researchers used 304 documented case studies drawn from the New England Journal of Medicine (NEJM). These cases represent some of the most challenging diagnostic puzzles solved by doctors in practice.

The orchestrator integrated leading large language models (LLMs) from OpenAI, Meta, Anthropic, Google, xAI and DeepSeek. Although the orchestrator improved the performance of all these models, it achieved the highest success rate when paired with OpenAI’s “o3” reasoning model – correctly diagnosing 85.5% of the NEJM cases.

For comparison, experienced physicians – who in this test were not allowed to consult textbooks or colleagues, factors that would likely have raised their success rates – solved around 20% of these same cases.

Potential integration and cost efficiencies

Microsoft indicated that elements of this diagnostic technology could soon be incorporated into its Copilot AI chatbot and Bing search engine, which currently process approximately 50 million health-related queries daily.

Dominic King, formerly the head of DeepMind’s health team and who joined Microsoft late last year, noted:

“The programme had performed better than anything we’ve ever seen before. There is an opportunity here today to act almost as a new front door to healthcare.”

King also highlighted that the AI agents were prompted to consider costs, which led to fewer diagnostic tests being ordered. In some scenarios, this resulted in hundreds of thousands of dollars in theoretical savings during the trial simulations.

Caution from experts and early-stage caveats

However, King was careful to stress that the technology remains in its infancy. The work has not yet undergone peer review and is not ready for deployment in live clinical environments.

Eric Topol, a cardiologist and director of the Scripps Research Translational Institute, commented:

“This is a landmark study. While this work was not done in the setting of real world medical practice, it is the first to provide evidence for the efficiency potential of generative AI in medicine – accuracy and cost savings.”

Microsoft’s broader AI strategy and industry dynamics

This effort also arrives as Microsoft continues to deepen its investment in AI. The company has put nearly $14 billion into OpenAI and holds exclusive rights to use and sell its technology. Nonetheless, the partnership faces tensions as OpenAI seeks to shift towards a for-profit model, with negotiations ongoing about future terms.

Suleyman remarked that although OpenAI’s model delivered the strongest performance, Microsoft remains flexible regarding which of the major models underpin the orchestrator:

“We have long believed that they’ll become commodities … it’s the aggregate orchestrator which I think is the differentiator.”

A sign of things to come?

The launch of MAI-DxO follows in the wake of other AI–healthcare breakthroughs, notably from Suleyman’s former organisation, DeepMind. Last year, the lab’s chief Sir Demis Hassabis was jointly awarded a Nobel Prize in chemistry for work using AI to unravel the structures of proteins fundamental to life.

For now, Microsoft’s technology stands as a striking example of AI’s growing capacity to tackle some of healthcare’s most demanding challenges. Yet its real-world impact will depend on rigorous validation and careful integration into the delicate fabric of clinical care.

Read More

New algorithm improves fitness tracker accuracy for people with obesity

Key Takeaways: 

• Researchers at Northwestern University have developed a validated, open-source algorithm that significantly improves the accuracy of fitness trackers for people with obesity.
• The new model achieves over 95% accuracy in estimating energy expenditure, addressing long-standing issues in existing algorithms that fail to reflect the lived experience of individuals with higher body weight.
• The technology is informed by both clinical testing and real-world scenarios, aiming to redefine how effort and activity are recognised across diverse body types.
  

A Breakthrough in Inclusive Fitness Technology

Fitness trackers have become an essential tool for many in monitoring daily physical activity and energy expenditure. However, for people living with obesity – whose energy usage, walking gait, and movement dynamics often differ from those without obesity – standard tracking algorithms have repeatedly fallen short.

Now, scientists at Northwestern University have introduced a new algorithm designed specifically to bridge this gap. The model enables wearable fitness devices, particularly wrist-worn smartwatches, to more accurately estimate the number of kilocalories burned during various forms of physical activity among people with obesity.

“People with obesity could gain major health insights from activity trackers, but most current devices miss the mark,” said Dr Nabil Alshurafa, Associate Professor of Behavioural Medicine at Northwestern University Feinberg School of Medicine. His laboratory – the HABits Lab – developed and validated the algorithm, which is open-source, rigorously testable, and ready for others to build upon. A companion activity-monitoring app, compatible with iOS and Android, is due for release later this year.

Why Current Trackers Fall Short

Most existing activity-monitoring algorithms are optimised for individuals without obesity, resulting in inaccurate data when used by people with higher body mass. Hip-worn trackers in particular are prone to errors due to variations in walking patterns and tilt angles, often underestimating energy expenditure.

While wrist-worn models offer better comfort and potential for consistent wear across body types, Alshurafa notes a key flaw: “Without a validated algorithm for wrist devices, we’re still in the dark about exactly how much activity and energy people with obesity really get each day – slowing our ability to tailor interventions and improve health outcomes.”

To evaluate performance, the Northwestern team compared their new model against 11 leading algorithms used in research-grade devices. Their method was not only more inclusive but also demonstrated superior accuracy. They even employed wearable cameras to validate each instance where wrist sensors failed to accurately capture calorie burn.

The team’s findings will be published on 19 June in Nature Scientific Reports.

A Personal Motivation: Fitness Should Not Exclude

Dr Alshurafa’s motivation for the project came from a personal experience while attending a fitness class alongside his mother-in-law, who lives with obesity.

“She worked harder than anyone else, yet when we glanced at the leaderboard, her numbers barely registered,” he recalled. “That moment hit me: fitness shouldn’t feel like a trap for the people who need it most.”

This disconnect between perceived effort and measured output led Alshurafa to rethink how fitness technology serves people across a range of body types.

Matching Gold-Standard Methods

The newly developed algorithm is trained using data from commercial fitness trackers and achieves accuracy comparable to gold-standard methods for measuring energy burn. It estimates how much energy a person with obesity expends per minute, even during complex or low-movement tasks, and consistently surpasses 95% accuracy in real-world settings.

“This advancement makes it easier for more people with obesity to track their daily activities and energy use,” said Alshurafa, underscoring its potential to support healthier outcomes through more reliable data.

Rigorous Testing in Lab and Life

To ensure accuracy, the research included two distinct study groups:

• Controlled Lab Testing:
  A group of 27 participants wore both a wrist-worn fitness tracker and a metabolic cart – a face mask device that measures oxygen intake and carbon dioxide output – to calculate kilocalorie burn and resting metabolic rate. Each participant performed a range of physical activities while data were collected and compared.
• Free-Living Evaluation:
  Another group of 25 participants wore both a fitness tracker and a body camera while going about their daily lives. The camera provided visual confirmation of physical activity, helping the team identify instances where the algorithm over- or under-estimated energy expenditure.

The researchers also explored less conventional forms of exercise. “Many couldn’t drop to the floor, but each one crushed wall-pushups, their arms shaking with effort,” said Alshurafa. “We celebrate ‘standard’ workouts as the ultimate test, but those standards leave out so many people. These experiences showed me we must rethink how gyms, trackers and exercise programmes measure success – so no one’s hard work goes unseen.”

Study Details and Contributors

The study, titled “Developing and comparing a new BMI inclusive energy burn algorithm on wrist-worn wearables,” represents a collaborative effort within and beyond Northwestern University.

In addition to lead author Boyang Wei, the research team included Christopher Romano and Bonnie Nolan. External collaborators were Mahdi Pedram and Whitney A. Morelli, both formerly affiliated with Northwestern.

Read More

Stanford clinicians trial new AI tool that lets you ‘chat’ with electronic medical records

Key Takeaways:

• Stanford Health Care has developed ChatEHR, a secure AI-powered assistant that enables clinicians to interact directly with electronic health records via natural language queries.
• ChatEHR can summarise complex patient histories, retrieve key data points, and assist with administrative tasks, freeing clinicians to focus more on patient care.
• The tool remains in pilot testing with 33 users, and its developers are actively building new features including automated clinical evaluations and in-record source citations.
  

Transforming Clinical Workflows with AI

Clinicians at Stanford Health Care are now piloting a new artificial intelligence (AI) tool, ChatEHR, which enables them to interact with patients’ electronic medical records (EMRs) in a conversational, intuitive manner – similar to how one might “chat” with a large language model such as GPT-4.

Currently in its pilot phase, ChatEHR draws upon information from a patient’s health record to respond to queries, generate summaries, and support routine clinical tasks. It is designed to be secure, context-aware, and seamlessly integrated into clinicians’ existing digital workflows.

“AI can augment the practice of physicians and other health care providers, but it’s not helpful unless it’s embedded in their workflow and the information the algorithm is using is in a medical context,” explained Nigam Shah, MBBS, PhD, Chief Data Science Officer at Stanford Health Care, who spearheaded the development of the system.

“ChatEHR is secure; it’s pulling directly from relevant medical data; and it’s built into the electronic medical record system, making it easy and accurate for clinical use.”

Origins and Development

The idea for ChatEHR took shape in 2023, when Dr Shah, Anurang Revri, Vice President and Chief Enterprise Architect for Stanford Health Care’s Technology and Digital Services, and a team of Stanford Medicine researchers recognised the clinical potential of large language models. Their aim was to develop a tool that could streamline interactions with complex patient data.

“ChatEHR opens up a new way for clinicians to interact with electronic health records in a more streamlined and efficient manner,” said Michael Pfeffer, MD, Chief Information and Digital Officer for Stanford Health Care and the School of Medicine.

“Whether that’s asking for a summary of the entire chart or retrieving specific data points relevant to the patient’s care, this is a unique instance of integrating LLM capabilities directly into clinicians’ practice and workflow. We’re thrilled to bring this to the workforce at Stanford Health Care.”

Enhancing Information Retrieval and Workflow Efficiency

Currently, the pilot programme involves 33 clinicians – including doctors, nurses, physician assistants, and nurse practitioners – who are testing the software’s performance, identifying areas for improvement, and contributing to feature development.

The interface welcomes users with:
“Hi, 👋 I’m ChatEHR! Here to help you securely chat with the patient’s medical record.”

Clinicians can then type natural-language queries such as:

• Does this person have any allergies?
• What were the results of their most recent cholesterol test?
• Have they undergone a colonoscopy? If so, were the findings normal?
  

“Making the electronic medical record more user friendly means physicians can spend less time scouring every nook and cranny of it for the information they need,” noted Dr Sneha Jain, Clinical Assistant Professor of Medicine and an early adopter of the tool.

“ChatEHR can help them get that information up front so they can spend time on what matters – talking to patients and figuring out what’s going on.”

Supporting Emergency and Transfer Care

The tool also has potential to ease the burden of clinical information gathering in emergency and transfer cases.

“It’s not just the chest pain they’re having in that moment that matters – it’s their whole story, what led up to this moment,” said Dr Jonathan Chen, MD, PhD, a hospital physician and Assistant Professor of Medicine and of Biomedical Data Sciences.

“All their prior history is relevant. What medications were they on, what side effects did they have, what surgeries took place and how did that affect them? It’s a ton of work to go back and find all of that information during a time-sensitive case, so speeding up that process would be a big help.”

In transfer cases, patients often arrive with vast documentation – sometimes hundreds of pages long. ChatEHR offers a rapid summarisation feature to distil these into concise, relevant overviews. Importantly, users can also ask follow-up questions for greater context.

Moving Beyond Queries: Automating Evaluative Tasks

The development team is expanding ChatEHR’s functionality to include “automations” – algorithmically-driven evaluations based on a patient’s clinical record.

For example, one automation helps determine whether a patient is eligible for transfer to the Sequoia Hospital patient care unit, a Stanford Medicine-affiliated facility with additional room capacity.

“That automated evaluation saves us the administrative burden of sifting through patient information and helps us quickly determine if a patient can be transferred, opening access to care here at Stanford Hospital,” Dr Shah explained.

Other automations under development aim to assist with hospice eligibility assessments and post-surgical follow-up recommendations.

Evaluating and Expanding ChatEHR

Stanford researchers are using MedHELM, an open-source framework for real-world medical AI evaluation, to systematically assess the accuracy and effectiveness of ChatEHR. Among upcoming features are citation tools, which will allow clinicians to see exactly where within a patient’s record specific pieces of information originate.

“We’re rolling this out in accordance with our responsible AI guidelines,” said Dr Shah.

“Not only ensuring accuracy and performance, but making sure we have the educational resources and technical support available to make ChatEHR usable and useful to our workforce.”

The eventual goal is to make ChatEHR available to all clinicians at Stanford who interact with patient charts, improving the speed, clarity, and usability of electronic medical records across the board.

The development of ChatEHR has been supported by Stanford’s Department of Medicine and the Center for Biomedical Informatics Research, reflecting a broader institutional commitment to integrating trustworthy AI into frontline care.

Read More