Removing bias from health care AI tools

by University of Oxford

Credit: Pixabay/CC0 Public Domain

Rapid advances in artificial intelligence (AI) have opened the way for the creation of a huge range of new health care tools, but to ensure that these tools do not exacerbate preexisting health inequities, researchers urge the use of more representative data in their development.

11 april 2024--Researchers from Oxford University's Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences (NDORMS), University College London and the Center for Ethnic Health Research, supported by Health Data Research UK, have for the first time studied the full detail of ethnicity data in the NHS. They outline the importance of using representative data in health care provision and have compiled this information into a research-ready database.

The new study, published in Scientific Data, is the first part of a three-phase project that aims to reduce bias in AI health prediction models which are trained on real-world patient data. The project, which addresses ethnicity disparities that were highlighted during the pandemic, is part of the UK Government's COVID-19 Data and Connectivity National Core Study led by Health Data Research UK.

The researchers used de-identified data on ethnicity and other characteristics from general practice and hospital health records, accessed safely within NHS England's Secure Data Environment (SDE) service, via the British Heart Foundation Data Science Center's CVD-COVID-UK/COVID-IMPACT Consortium.

This is the first time that patient ethnicity data has been studied at this depth and breadth for the whole population of England. The researchers were able to combine records to analyze patient self-identified ethnicity recorded through over 489 potential codes.

Researchers analyzed how more than 61 million people in England identified their ethnicity in over 250 different groups. They also looked at the characteristics of those with no record of their ethnicity, and how conflicts in patient ethnicity data can arise. The data, now available for other researchers to use, shows that 1/10 patients lack ethnicity records, and around 12% of patients had conflicting ethnicity codes in their patient records.

Sara Khalid, Associate Professor of Health Informatics and Biomedical Data Science at NDORMS, explained, "Health inequity was highlighted during the COVID19 pandemic, where individuals from ethnically diverse backgrounds were disproportionately affected, but the issue is long-standing and multi-faceted.

"Because AI-based health care technology depends on the data that is fed into it, a lack of representative data can lead to biased models that ultimately produce incorrect health assessments. Better data from real-world settings, such as the data we have collected, can lead to better technology and ultimately better health for all."

Professor Cathie Sudlow, Chief Scientist at Health Data Research UK and Director of its BHF Data Science Center said, "We are delighted to be supporting hundreds of researchers to harness the power of the UK's rich health data. This study on ethnicity recording highlights how different sources of health data from the whole English population can be accessed and analyzed in a safe and secure way, providing insights that are relevant to everyone.

"The findings will empower health professionals, patients, carers and policymakers to make better decisions that will benefit people of all ages, ethnic groups, and social backgrounds across the country."

The study assessed the available detail of ethnicity data in NHS England, including across different types of ethnicity codes. For example, NHS hospitals record patient data via 19 ethnicity codes, while GPs use the globally recognized SNOMED-CT Codes, of which there are 489. However, health researchers lose the finer detail from these recording systems as they typically collapse these groups into just five or six, potentially leading to less accurate research.

The researchers plan to demonstrate the value of these findings in the subsequent phases of the project, which will first focus on using these detailed results on ethnicity data to better describe how different ethnicities were impacted by the COVID-19 pandemic, and then feed into more equitable artificial intelligence and machine learning tools suitable for use by diverse patient groups.

More information: Marta Pineda-Moncusí et al, Ethnicity data resource in population-wide health records: completeness, coverage and granularity of diversity, Scientific Data (2024). DOI: 10.1038/s41597-024-02958-1

 

Study finds daily fiber supplement improves older adults' brain function in just 12 weeks

by King's College London

Credit: Pixabay/CC0 Public Domain

A daily fiber supplement improved brain function in people over 60 in just 12 weeks. The study, published recently in Nature Communications by researchers from the School of Life Course & Population Sciences showed that this simple and cheap addition to diet can improve performance in memory tests associated with early signs of Alzheimer's disease.

11 april 2024--However, the prebiotic supplements inulin and FOS were found to have no effect on muscle strength over this period.

"We are excited to see these changes in just 12 weeks. This holds huge promise for enhancing brain health and memory in our aging population. Unlocking the secrets of the gut-brain axis could offer new approaches for living more healthily for longer," says first author Dr. Mary Ni Lochlainn from the Department of Twin Research.

As populations age globally, the prevalence of age-related conditions such as cognitive decline and muscle loss is on the rise. Researchers at TwinsUK, the U.K.'s largest adult twin registry based at King's College London, sought to understand how targeting the microbiota, the diverse community of microorganisms residing in our intestines, using two cheap, commercially available plant fiber supplements inulin and FOS, could impact both muscle health and brain function.

Researchers assigned 36 twin pairs—72 individuals—over 60 years old to receive either a placebo or the supplement every day for 12 weeks. Neither the analysis team, nor the participants knew which they received until the analysis was complete (double-blind). Alongside this, all study participants did resistance exercises and ate a protein supplement which was aimed at improving muscle function.

Researchers monitored participants remotely via video, online questionnaires and cognitive tests. They found the fiber supplement led to significant changes in the participants' gut microbiome composition, particularly an increase in the numbers of beneficial bacteria such as Bifidobacterium.

While there was no significant difference in muscle strength between the groups, the group receiving the fiber supplement performed better in tests assessing brain function, including the Paired Associates Learning test which is an early marker for Alzheimer's disease, together with tests of reaction time and processing speed. These measures are important for daily living—for example reacting to traffic or stopping a simple trip-up turning into a fall.

"These plant fibers, which are cheap and available over the counter, could benefit a wide group of people in these cash-strapped times. They are safe and acceptable too. Our next task is to see whether these effects are sustained over longer periods and in larger groups of people," says senior author Professor Claire Steves, professor of aging and health.

Another novel aspect of the study was its remote design which demonstrated the feasibility of conducting trials in older adults without the need for extensive travel or hospital visits, which could be delivered in many settings globally. Challenges such as digital literacy and access to the necessary technology were acknowledged and will be addressed in future larger scale projects with the aim to enhance the quality of life for aging populations worldwide.

More information: Mary Ni Lochlainn et al, Effect of gut microbiome modulation on muscle function and cognition: the PROMOTe randomised controlled trial, Nature Communications (2024). DOI: 10.1038/s41467-024-46116-y

 

More schooling is linked to slowed aging and increased longevity

by Columbia University's Mailman School of Public Health

Credit: Pixabay/CC0 Public Domain

Participants in the Framingham Heart Study who achieved higher levels of education tended to age more slowly and went on to live longer lives as compared to those who did not achieve upward educational mobility, according to a new study at Columbia University Mailman School of Public Health and The Robert N. Butler Columbia Aging Center. Upward educational mobility was significantly associated with a slower pace of aging and lower risk of death. The results are published online in JAMA Network Open.

11 april 2024--The Framingham Heart Study is an ongoing observational study first initiated in 1948 that currently spans three generations.

The Columbia analysis is the first to connect educational mobility with pace of biological aging and mortality.

"We've known for a long time that people who have higher levels of education tend to live longer lives. But there are a bunch of challenges in figuring out how that happens and, critically, whether interventions to promote educational attainment will contribute to healthy longevity," said Daniel Belsky, Ph.D., associate professor of Epidemiology at Columbia Mailman School and the Aging Center and senior author of the paper.

To measure pace of aging, the researchers applied an algorithm known as the DunedinPACE epigenetic clock to genomic data collected by the Framingham Heart Study.

The latest findings showed that, according to the yardstick of the DunedinPACE epigenetic clock, two years of additional schooling translated to a 2–3% slower pace of aging. This slowing in the pace of aging corresponds to a roughly 10% reduction in risk of mortality in the Framingham Heart Study, according to previous research by Belsky on the association of DunedinPACE with risk of death.

DunedinPACE was developed by the Columbia researchers and colleagues and reported in January 2022. Based on an analysis of chemical tags on the DNA contained in white blood cells, or DNA methylation marks, DunedinPACE is named after the Dunedin Study birth cohort used to develop it.DunedinPACE (stands for Pace of Aging Computed from the Epigenome) is measured from a blood test and functions like a speedometer for the aging process, measuring how fast or slow a person's body is changing as they grow older.

Biological aging refers to the accumulation of molecular changes that progressively undermine the integrity and resilience capacity of our cells, tissues and organs as we grow older.

The Columbia researchers used data from 14,106 Framingham Heart Study spanning three generations to link children's educational attainment data with that of their parents. They then used data from a subset of participants who provided blood samples during data collection to calculate the pace of biological aging using the DunedinPACE epigenetic clock.

In primary analysis, the researchers tested associations between educational mobility, aging, and mortality in a subset of 3,101 participants for whom educational mobility and pace of aging measures could be calculated.

For 2,437 participants with a sibling, the researchers also tested whether differences in educational attainment between siblings was associated with a difference in the pace of aging.

"A key confound in studies like these is that people with different levels of education tend to come from families with different educational backgrounds and different levels of other resources," explained Gloria Graf, a Ph.D. candidate in the Department of Epidemiology supervised by Belsky, and first author of the study.

"To address these confounds, we focused on educational mobility, how much more (or less) education a person completed relative to their parents, and sibling differences in educational attainment—how much more (or less) education a person completed relative to their siblings. These study designs control for differences between families and allow us to isolate the effects of education."

By combining these study designs with the new DunedinPACE epigenetic clock, the researchers were able to test how education affects the pace of aging. Then, by linking the education and pace of aging data with longitudinal records of how long participants lived, the team was able to determine if a slower pace of aging accounted for increased longevity in people with more education.

"Our findings support the hypothesis that interventions to promote educational attainment will slow the pace of biological aging and promote longevity," noted Graf.

"Ultimately, experimental evidence is needed to confirm our findings," added Belsky. "Epigenetic clocks like DunedinPace have potential to enhance such experimental studies by providing an outcome that can reflect impacts of education on healthy aging well before the onset of disease and disability in later life."

"We found that upward educational mobility was associated both with a slower pace of aging and decreased risk of death," said Graf.

"In fact, up to half of the educational gradient in mortality we observed was explained by healthier aging trajectories among better-educated participants." This pattern of association was similar across generations and held within family sibling comparisons: siblings with higher educational mobility tended to have a slower pace of aging as compared with their less educated siblings.

More information: Educational Mobility, Pace of Aging, and Lifespan Among Participants in the Framingham Heart Study, JAMA Network Open (2024). DOI: 10.1001/jamanetworkopen.2024.0655

 

Loneliness worse than smoking, alcoholism, obesity: Study suggests primary care clinicians can offer solutions

by Regenstrief Institute

Credit: Pixabay/CC0 Public Domain

Loneliness is a significant biopsychosocial stressor with a mortality risk comparable to smoking more than 15 cigarettes a day and more harmful than alcoholism, obesity, and lack of physical activity.

11 april 2024--Despite its harmful effects, interventions to address the discrepancy between desired and actual social interaction are few and limited.

In a new study, Regenstrief Institute and Indiana University School of Medicine research scientists Monica Williams-Farrelly, Ph.D., Malaz Boustani, M.D., MPH, and Nicole Fowler, Ph.D., MHSA, identified evidence suggesting primary care clinicians can play an important role in developing and maintaining personal connections for patients experiencing loneliness.

The study found that 53 percent of older adults in the primary care population experience loneliness. Evidence also suggests that when older adults experience loneliness, their physical and mental health related to quality of life are reduced significantly.

"The first and obvious answer for loneliness is for primary care physicians to screen their patients," said Dr. Williams-Farrelly, the study's first author, a Regenstrief research scientist and an assistant research professor at IU School of Medicine.

"Based on the literature and research, loneliness has influences on health that are quite significant and quite strong, so in the same way that we ask older adults: Do you smoke? Or do you measure your blood sugar? We should be inquiring about and measuring loneliness and offering solutions."

Dr. Williams-Farrelly suggests that it is imperative for primary care physicians, nurse practitioners and other clinicians also to provide resources to patients to help address this significant issue.

"The topic of loneliness is more relevant now than ever given the May 2023 U.S. Surgeon General's call to action to tackle the loneliness epidemic," said Dr. Fowler, principal researcher and senior author.

"This research is important because it identifies and suggests evidence for interventions that are necessary for older adults in primary care who experience loneliness. Primary care clinicians should discuss loneliness with their older adult patients and provide resources to help them create meaningful social relationships." Dr. Fowler is also a Regenstrief research scientist, an associate professor, and a director of research at the IU School of Medicine.

An effective intervention, the researchers suggest, is the Circle of Friends concept, which consists of a three-month, group-based, psychosocial rehabilitation model aimed at enhancing interaction and friendships between participants. The model has shown effectiveness in both reducing loneliness and improving health outcomes, including subjective health, cognition, mortality, and lower health care costs.

"As older adults age, they have a lot of changes in their life due to a lot of circumstances—retirement, divorce, or the death of family and friends—making it a little more difficult for them to maintain social relationships. When connections are lost with coworkers or loved ones, it can be jarring," said Dr. Williams-Farrelly. "Older adults need their primary care physicians to screen and suggest effective resources that can allow them to maintain, foster and develop social relationships."

Data was gathered during the COVID-19 pandemic, but the researchers identified a steadily increasing trend in loneliness in this population prior to the global pandemic. The numbers are still increasing today.

"Loneliness may seem simple, but it can be complex to identify and address. It started to become a problem before COVID-19, and then with the national stay-at-home order caused by the pandemic, social contact was being prevented, which exacerbated the problem," said Dr. Williams-Farrelly.

This study used baseline data from the Caregiver Outcomes of Alzheimer's Disease Screening (COADS) clinical trial, supported by the National Institutes of Health's National Institute on Aging grant R01AG056325.

"Loneliness in older primary care patients and its relationship to physical and mental health-related quality of life," is published in the Journal of the American Geriatrics Society.

More information: Monica M. Williams‐Farrelly et al, Loneliness in older primary care patients and its relationship to physical and mental health‐related quality of life, Journal of the American Geriatrics Society (2024). DOI: 10.1111/jgs.18762

 

Four in 10 adults choose telemedicine visits

by Lori Solomon

Many patients, including those with the greatest care needs, choose telemedicine even when in-person visits are available, according to a study published online March 22 in JAMA Network Open.

11 april 2024--Eva Chang, Ph.D., M.P.H., from Advocate Health in Milwaukee, Wisconsin, and colleagues assessed patient characteristics associated with telemedicine use and telemedicine mode. The analysis included data from 5,437 U.S. adults with health care visits in 2022 participating in the Health Information National Trends Survey.

The researchers discovered that 43 percent of patients had a telemedicine appointment, with 70 percent choosing a video consultation and 30 percent opting for an audio-only visit. Telemedicine visits were less frequent among older patients (75 years and older), those who did not have internet access, and people living in the Midwest.

Telemedicine visits were positively associated with female sex (adjusted odds ratio [aOR], 1.43), chronic conditions (aOR, 2.13), and multiple health care visits (two to four visits: aOR, 1.77; at least five visits: aOR, 3.29).

Audio-only visits were associated with older age (65 to 74 years: aOR, 2.13; 75 years and older: aOR, 3.58), no health insurance (aOR, 2.84), and no internet use (aOR, 2.11). There were no significant differences observed in telemedicine use or mode by education, race and ethnicity, or income.

"These findings suggest that while many patients chose to use telemedicine when in-person visits are available, telemedicine access, particularly video visits, was less likely among key populations who experience barriers to accessing care," the authors write.

More information: Eva Chang et al, Patient Characteristics and Telemedicine Use in the US, 2022, JAMA Network Open (2024). DOI: 10.1001/jamanetworkopen.2024.3354

 

Study shows ChatGPT can produce medical record notes 10 times faster than doctors without compromising quality

by Uppsala University

Credit: CC0 Public Domain

The AI model ChatGPT can write administrative medical notes up to 10 times faster than doctors without compromising quality. This is according to a study conducted by researchers at Uppsala University Hospital and Uppsala University in collaboration with Danderyd Hospital and the University Hospital of Basel, Switzerland. The research is published in the journal Acta Orthopaedica.

11 april 2024--They conducted a pilot study of just six virtual patient cases, which will now be followed up with an in-depth study of 1,000 authentic patient medical records.

"For years, the debate has centered on how to improve the efficiency of health care. Thanks to advances in generative AI and language modeling, there are now opportunities to reduce the administrative burden on health care professionals. This will allow doctors to spend more time with their patients," explains Cyrus Brodén, an orthopedic physician and researcher at Uppsala University Hospital and Uppsala University.

Administrative tasks take up a large share of a doctor's working hours, reducing the time for patient contact and contributing to a stressful work situation.

Researchers at Uppsala University Hospital and Uppsala University, in collaboration with Danderyd Hospital and the University Hospital of Basel, Switzerland, have shown in a new study that the AI model ChatGPT can write administrative medical notes up to 10 times faster than doctors without compromising quality.

The aim of the study was to assess the quality and effectiveness of the ChatGPT tool when producing medical record notes. The researchers used six virtual patient cases that mimicked real cases in both structure and content. Discharge documents for each case were generated by orthopaedic physicians. ChatGPT-4 was then asked to generate the same notes. The quality assessment was carried out by an expert panel of 15 people who were unaware of the source of the documents. As a secondary metric, the time required to create the documents was compared.

"The results show that ChatGPT-4 and human-generated notes are comparable in quality overall, but ChatGPT-4 produced discharge documents ten times faster than the doctors," notes Brodén.

"Our interpretation is that advanced large language models like ChatGPT-4 have the potential to change the way we work with administrative tasks in health care. I believe that generative AI will have a major impact on health care and that this could be the beginning of a very exciting development," he maintains.

The plan is to launch an in-depth study shortly, with researchers collecting 1,000 medical patient records. Again, the aim is to use ChatGPT to produce similar administrative notes in the patient records.

"This will be an interesting and resource-intensive project involving many partners. We are already working actively to fulfill all data management and confidentiality requirements to get the study underway," concludes Brodén.

More information: Guillermo Sanchez Rosenberg et al, ChatGPT-4 generates orthopedic discharge documents faster than humans maintaining comparable quality: a pilot study of 6 cases, Acta Orthopaedica (2024). DOI: 10.2340/17453674.2024.40182

 

CBD products don't ease pain and are potentially harmful, study finds

by Vittoria D'Alessio, University of Bath

Credit: Unsplash/CC0 Public Domain

There is no evidence that CBD products reduce chronic pain, and taking them is a waste of money and potentially harmful to health, according to recent research led by the University of Bath in the UK.

11 april 2024--CBD (short for cannabidiol) is one of many chemicals found naturally in the cannabis plant. It's a popular alternative medicine to treat pain and is readily available in shops and online in the form of oils, tinctures, vapes, topical creams, edibles (such as gummy bears), and soft drinks.

However, consumers would do well to steer clear of these products, according to the study.

"CBD presents consumers with a big problem," said Professor Chris Eccleston, who led the research from the Centre for Pain Research at Bath. "It's touted as a cure for all pain, but there's a complete lack of quality evidence that it has any positive effects."

He added, "It's almost as if chronic pain patients don't matter and that we're happy for people to trade on hope and despair."

For their study, published in The Journal of Pain, the team—which included researchers from the Universities of Bath, Oxford and Alberta in Canada—examined research relevant to using CBD to treat pain and published in scientific journals up to late 2023.

They found:

• CBD products sold directly to consumers contain varying amounts of CBD, from none to much more than advertised.
• CBD products sold directly to consumers may contain chemicals other than CBD, some of which may be harmful and some illegal in some jurisdictions. Such chemicals include THC (tetrahydrocannabinol), the main psychoactive component of the cannabis plant.
• Of the 16 randomized controlled trials that have explored the link between pain and pharmaceutical-grade CBD, 15 have shown no positive results, with CBD being no better than placebo at relieving pain.
• A meta-analysis (which combines data from multiple studies and plays a fundamental role in evidence-based health care) links CBD to increased rates of serious adverse events, including liver toxicity.

Medical vs. non-medical CBD

In the UK, medical cannabis is the only CBD product that is subject to regulatory approval. It's occasionally prescribed for people with severe forms of epilepsy, adults with chemotherapy-related nausea, and people with multiple sclerosis.

Non-medical CBD is freely available in the UK (as well as in the US and many European countries) as long as it contains negligible quantities of THC or none at all. However, CBD products sold on the retail market are not covered by trade standards, meaning there is no requirement for them to be consistent in content or quality.

Most CBD products bought online—including popular CBD oils—are known to contain very small amounts of CBD. Moreover, any given product may be illegal to possess or supply, as there's a good chance it will contain forbidden quantities of THC.

Chronic pain

An estimated 20% of the adult population lives with chronic pain, and sufferers are often desperate for help to alleviate their symptoms. It's no surprise then that many people reach for CBD products despite their high price tag and the lack of evidence of their effectiveness or safety.

Dr. Andrew Moore, study co-author and former senior pain researcher in the Nuffield Division of Anaesthetics at the University of Oxford, said, "For too many people with chronic pain, there's no medicine that manages their pain. Chronic pain can be awful, so people are very motivated to find pain relief by any means. This makes them vulnerable to the wild promises made about CBD."

He added that health care regulators appear reluctant to act against the spurious claims made by some manufacturers of CBD products, possibly because they don't want to interfere in a booming market (the global CBD product market was estimated at US$3 billion in 2021 or £2.4 billion and is anticipated to reach US$60 billion by 2030 or £48 billion) especially when the product on sale is widely regarded as harmless.

"What this means is that there are no consumer protections," said Dr. Moore. "And without a countervailing body to keep the CBD sellers in check, it's unlikely that the false promises being made about the analgesic effects of CBD will slow down in the years ahead."

The study's authors are calling for chronic pain to be taken more seriously, with consumer protection becoming a priority.

"Untreated chronic pain is known to seriously damage quality of life, and many people live with pain every day and for the rest of their lives," said Professor Eccleston. "Pain deserves investment in serious science to find serious solutions."

More information: Andrew Moore et al, Cannabidiol (CBD) Products for Pain: Ineffective, Expensive, and With Potential Harms, The Journal of Pain (2023). DOI: 10.1016/j.jpain.2023.10.009

 

Study: Life expectancy increased as world addressed major killers, though poor pandemic management slowed progress

by Institute for Health Metrics and Evaluation

Credit: Unsplash/CC0 Public Domain

Global life expectancy increased by 6.2 years since 1990 according to a new study published in The Lancet. Over the past three decades, reductions in death from leading killers have fueled this progress, including diarrhea and lower respiratory infections, as well as stroke and ischemic heart disease.

11 april 2024--When the COVID-19 pandemic arrived in 2020, however, it derailed progress in many locations. This is the first study to compare deaths from COVID-19 to deaths from other causes globally.

Despite the challenges presented by the COVID-19 pandemic, the researchers found that the super-region of Southeast Asia, East Asia, and Oceania had the largest net gain in life expectancy between 1990 and 2021 (8.3 years), largely due to reductions in mortality from chronic respiratory diseases, stroke, lower respiratory infections, and cancer.

The super-region's strong management of the COVID-19 pandemic helped preserve these gains. South Asia had the second-largest net increase in life expectancy among super-regions between 1990 and 2021 (7.8 years), primarily thanks to steep declines in deaths from diarrheal diseases.

"Our study presents a nuanced picture of the world's health," said Dr. Liane Ong, co-first author of the study and a Lead Research Scientist at the Institute for Health Metrics and Evaluation (IHME). "On one hand, we see countries' monumental achievements in preventing deaths from diarrhea and stroke," she said. "At the same time, we see how much the COVID-19 pandemic has set us back."

The study also highlights how COVID-19 radically altered the top five causes of death for the first time in 30 years. COVID-19 displaced a long-dominant killer—stroke—to become the second-leading cause of death globally. The research presents updated estimates from the Global Burden of Disease Study (GBD) 2021.

The authors found that the super-regions where the COVID-19 pandemic hit hardest were Latin America and the Caribbean and sub-Saharan Africa, which lost the most years of life expectancy due to COVID-19 in 2021. While documenting the enormous loss of life caused by the COVID-19 pandemic, the researchers also pinpointed the reasons behind the improvements in life expectancy in every super-region.

Looking across different causes of death, the study reveals sharp drops in deaths from enteric diseases—a class of diseases that includes diarrhea and typhoid. These improvements increased life expectancy worldwide by 1.1 years between 1990 and 2021.

Reductions in deaths from lower respiratory infections added 0.9 years to global life expectancy during this period. Progress in preventing deaths from other causes also drove up life expectancy around the world, including stroke, neonatal disorders, ischemic heart disease, and cancer. For each disease, reductions in deaths were most pronounced between 1990 and 2019.

At the regional level, Eastern sub-Saharan Africa experienced the largest increase in life expectancy, which jumped by 10.7 years between 1990 and 2021. Control of diarrheal diseases was the leading force behind improvements in this region. East Asia had the second-largest gain in life expectancy; the region's success in slashing deaths from chronic obstructive pulmonary disease played a key role.

The GBD 2021 study measures mortality by cause of death and years of life lost at global, regional, national, and subnational levels. The analysis links specific causes of death to changes in life expectancy.

The study illuminates not only the diseases that have driven increases and decreases in life expectancy, but also looks at how patterns of disease have shifted across locations over time, presenting, as the authors write, an "opportunity to deepen our understanding of mortality-reduction strategies…[which] might reveal areas where successful public health interventions have been implemented."

GBD 2021 highlights places that have made huge strides in preventing deaths from major diseases and injuries. It also emphasizes how some of the most burdensome diseases are now concentrated in certain locations, underscoring opportunities for intervention. For example, in 2021, deaths from enteric diseases were largely concentrated in sub-Saharan Africa and South Asia.

For another disease, malaria, the researchers found that 90% of deaths occurred in an area inhabited by just 12% of the world's population in a stretch of land ranging from Western sub-Saharan Africa through Central Africa to Mozambique.

"We already know how to save children from dying from enteric infections including diarrheal diseases, and progress in fighting this disease has been tremendous," said Professor Mohsen Naghavi, the study's co-first author and the Director of Subnational Burden of Disease Estimation at IHME.

"Now, we need to focus on preventing and treating these diseases, strengthening and expanding immunization programs, and developing brand-new vaccines against E. coli, norovirus, and Shigella," he added.

In addition to providing new insights on COVID-19, the study reveals growing threats from non-communicable diseases, such as diabetes and kidney diseases, which are increasing in every country. The researchers also point to uneven progress against conditions such as ischemic heart disease, stroke, and cancer. High-income countries have driven down deaths from many types of non-communicable diseases, but many low-income countries have not.

"The global community must ensure that the lifesaving tools that have cut deaths from ischemic heart disease, stroke, and other non-communicable diseases in most high-income countries are available to people in all countries, even where resources are limited," said Eve Wool, senior author of the study and a Senior Research Manager at IHME.

More information: Mohsen Naghavi et al, Global burden of 288 causes of death and life expectancy decomposition in 204 countries and territories and 811 subnational locations, 1990–2021: a systematic analysis for the Global Burden of Disease Study 2021, The Lancet (2024). DOI: 10.1016/S0140-6736(24)00367-2 , www.thelancet.com/journals/lan … (24)00367-2/fulltext

 

Good evidence confuses ChatGPT when used for health information, study finds

by CSIRO

Credit: Pixabay/CC0 Public Domain

A world-first study has found that when asked a health-related question, the more evidence that is given to ChatGPT, the less reliable it becomes—reducing the accuracy of its responses to as low as 28%.

11 april 2024--The study was recently presented at Empirical Methods in Natural Language Processing (EMNLP), a Natural Language Processing conference in the field. The findings are published in Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing.

As large language models (LLMs) like ChatGPT explode in popularity, they pose a potential risk to the growing number of people using online tools for key health information.

Scientists from CSIRO, Australia's national science agency, and The University of Queensland (UQ) explored a hypothetical scenario of an average person (non-professional health consumer) asking ChatGPT if "X" treatment has a positive effect on condition "Y."

The 100 questions presented ranged from "Can zinc help treat the common cold?" to "Will drinking vinegar dissolve a stuck fish bone?"

ChatGPT's response was compared to the known correct response, or "ground truth," based on existing medical knowledge.

CSIRO Principal Research Scientist and Associate Professor at UQ Dr. Bevan Koopman said that even though the risks of searching for health information online are well documented, people continue to seek health information online, and increasingly via tools such as ChatGPT.

"The widespread popularity of using LLMs online for answers on people's health is why we need continued research to inform the public about risks and to help them optimize the accuracy of their answers," Dr. Koopman said. "While LLMs have the potential to greatly improve the way people access information, we need more research to understand where they are effective and where they are not."

The study looked at two question formats. The first was a question only. The second was a question biased with supporting or contrary evidence.

Results revealed that ChatGPT was quite good at giving accurate answers in a question-only format, with an 80% accuracy in this scenario.

However, when the language model was given an evidence-biased prompt, accuracy reduced to 63%. Accuracy was reduced again to 28% when an "unsure" answer was allowed. This finding is contrary to popular belief that prompting with evidence improves accuracy.

"We're not sure why this happens. But given this occurs whether the evidence given is correct or not, perhaps the evidence adds too much noise, thus lowering accuracy," Dr. Koopman said.

ChatGPT launched on November 30, 2022, and has quickly become one of the most widely used large language models (LLMs). LLMs are a form of artificial intelligence that recognize, translate, summarize, predict, and generate text.

Study co-author UQ Professor Guido Zuccon, Director of AI for the Queensland Digital Health Centre (QDHeC), said that major search engines are now integrating LLMs and search technologies in a process called Retrieval Augmented Generation.

"We demonstrate that the interaction between the LLM and the search component is still poorly understood and controllable, resulting in the generation of inaccurate health information," said Professor Zuccon.

Next steps for the research are to investigate how the public uses the health information generated by LLMs.

More information: Bevan Koopman et al, Dr ChatGPT tell me what I want to hear: How different prompts impact health answer correctness, Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (2023). DOI: 10.18653/v1/2023.emnlp-main.928

 

Long COVID leaves telltale traces in the blood, finds new study

by Andrew Czyzewski and Ryan O'Hare, Imperial College London

Graphical abstract. Credit: Nature Immunology (2024). DOI: 10.1038/s41590-024-01778-0

Findings from the largest UK study of patients hospitalized with SARS-CoV-2 infection show that long COVID leads to ongoing inflammation which can be detected in the blood.

11 april 2024--In an analysis of more than 650 people who had been hospitalized with severe COVID-19, patients with prolonged symptoms showed evidence of immune system activation.

The exact pattern of this activation varied depending on the sort of symptoms that they predominantly had—for example, mainly fatigue or cognitive impairment.

The research, led by Imperial College London, suggests that existing drugs which modulate the body's immune system could be helpful in treating long COVID and should be investigated in future clinical trials.

The study, published in the journal Nature Immunology, is the latest research from two collaborative UK-wide consortia, PHOSP-COVID and ISARIC-4C.

These involve scientists and clinicians from Imperial alongside collaborators from the Universities of Leicester, Edinburgh, and Liverpool, among others.

Professor Peter Openshaw, from Imperial's National Heart & Lung Institute and an ISARIC-4C lead investigator, said, "With one in ten SARS-CoV-2 infections leading to long COVID and an estimated 65 million people around the world suffering from ongoing symptoms, we urgently need more research to understand this condition. At the moment, it's very hard to diagnose and treat."

"This study, which includes detailed clinical data on symptoms and a raft of inflammatory blood plasma markers, is an important step forward and provides crucial insights into what causes long COVID."

Runaway inflammation

In the latest study, researchers included a total of 426 people who were experiencing symptoms consistent with long-term COVID-19 and had been admitted to hospital with COVID-19 infection at least six months prior to the study.

They were compared with 233 people who were also hospitalized for COVID-19 but who had fully recovered. The researchers took samples of blood plasma and measured a total of 368 proteins known to be involved in inflammation and immune system modulation.

They found that relative to patients who had fully recovered, those with long COVID showed a pattern of immune system activation, indicating inflammation of myeloid cells and activation of a family of immune system proteins called the complement system.

Myeloid cells are formed in the bone marrow and produce various types of white blood cells that circulate in the blood and migrate into organs and tissues, where they respond to damage and infection.

The complement system consists of a cascade of linked proteins that are activated in response to infection or tissue damage. Notably, overactivation of the complement system is known to be associated with many autoimmune and inflammatory conditions.

Dr. Felicity Liew, from Imperial's National Heart & Lung Institute, said, "Our findings indicate that complement activation and myeloid inflammation could be a common feature of long COVID after hospitalization, regardless of symptom type."

"It is unusual to find evidence of ongoing complement activation several months after acute infection has resolved, suggesting that long COVID symptoms are a result of active inflammation."

"However, we can't be sure that this is applicable to all types of long COVID, especially if symptoms occur after non-hospitalized infection."

Sub-types of long COVID

The researchers were also able to obtain comprehensive information about the range of symptoms that patients were experiencing and which ones were most common.

They found that certain groups of symptoms appeared to be associated with specific proteins. For example, people with gastrointestinal symptoms had increased levels of a marker called SCG3, which has previously been linked to impaired communication between the gut and the brain.

Overall, there were five overlapping subtypes of long-term COVID with different immune signatures, despite some commonalities, namely fatigue, cognitive impairment, anxiety and depression, cardiorespiratory, and gastrointestinal.

The researchers stress, however, that these groups are not mutually exclusive, and people can fall between groups depending on their symptoms.

Nevertheless, these long COVID subtypes seem to represent clear biological mechanisms of disease and highlight that different symptoms may have different underlying causes. The researchers suggest this could be useful in designing clinical trials, especially for treatments that target immune responses and inflammation.

One such treatment could include drugs called IL-1 antagonists, such as anakinra, which is commonly used to treat rheumatoid arthritis, as well as another drug class called JAK inhibitors, used to treat some types of cancers and severe forms of rheumatoid arthritis. Both drug types work by targeting components of the immune system that might be activated in long COVID.

The researchers highlight that one limitation of their study was that it only included people who had severe SARS-CoV-2 infections and who were hospitalized as a result. Yet a sizeable proportion of people who develop long COVID in the wider population only report mild initial SARS-CoV-2 infection, and it's unclear if the same immune mechanisms are at work.

Professor Openshaw concludes, "This work provides strong evidence that long COVID is caused by post-viral inflammation but shows layers of complexity."

"We hope that our work opens the way to the development of specific tests and treatments for the various types of long COVID and believe that a 'one size fits all' approach to treatment may not work."

"COVID-19 will continue to have far-reaching effects long after the initial infection has passed, impacting many lives. Understanding what's happening in the body, and how the immune system responds, is key to helping those affected."

More information: Peter Openshaw, Large-scale phenotyping of patients with long COVID post-hospitalization reveals mechanistic subtypes of disease, Nature Immunology (2024). DOI: 10.1038/s41590-024-01778-0. www.nature.com/articles/s41590-024-01778-0