Ancient Maya reservoirs offer lessons for today’s water crises

Peer-Reviewed Publication

UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN, NEWS BUREAU

 

IMAGE: 

THE MAYA BUILT AND MAINTAINED SELF-CLEANING WETLAND RESERVOIRS THAT SERVED URBAN POPULATIONS OVER MILLENNIA. U. OF I. ANTHROPOLOGY PROFESSOR LISA LUCERO WRITES THAT THE WATER-RELATED CRISES THEY FACED HOLD LESSONS FOR TODAY.

view more 

CREDIT: PHOTO BY FRED ZWICKY

CHAMPAIGN, Ill. — According to a new paper, ancient Maya reservoirs, which used aquatic plants to filter and clean the water, “can serve as archetypes for natural, sustainable water systems to address future water needs.”

The Maya built and maintained reservoirs that were in use for more than 1,000 years, wrote University of Illinois Urbana-Champaign anthropology professor Lisa Lucero in a perspective in the Proceedings of the National Academy of Sciences. These reservoirs provided potable water for thousands to tens of thousands of people in cities during the annual, five-month dry season and in periods of prolonged drought.

“Most major southern lowland Maya cities emerged in areas that lacked surface water but had great agricultural soils,” Lucero said. “They compensated by constructing reservoir systems that started small and grew in size and complexity.”

Over time, the Maya built canals, dams, sluices and berms to direct, store and transport water. They used quartz sand for water filtration, sometimes importing it from great distances to massive cities like Tikal in what is now northern Guatemala. A sediment core from one of Tikal’s reservoirs also found that zeolite sand had been used in its construction. Previous studies have shown that this volcanic sand can filter impurities and disease-causing microbes from water. The zeolite also would have been imported from sources about 18 miles (30 kilometers) away.

“Tikal’s reservoirs could hold more than 900,000 cubic meters of water,” Lucero wrote. Estimates suggest that up to 80,000 people lived in the city and its environs in the Late Classic period, roughly 600 to 800 C.E. The reservoirs kept people and crops hydrated during the dry season, Lucero said.

Maya royalty got much of their status from their ability to provide water to the populace.

“Clean water and political power were inextricably linked – as demonstrated by the fact that the largest reservoirs were built near palaces and temples,” Lucero wrote. The kings also performed ceremonies to gain the favor of ancestors and the rain god, Chahk.

A key challenge was to keep standing water in reservoirs from becoming stagnant and undrinkable, and for that the Maya likely relied on aquatic plants, many of which still populate Central American wetlands today, Lucero said. These include cattails, sedges, reeds and others. Some of these plants have been identified in sediment cores from Maya reservoirs.

These plants filtered the water, reducing murkiness and absorbing nitrogen and phosphorous, Lucero said.

“The Maya would have had to dredge every several years… (and) harvest and replenish aquatic plants,” she wrote. The nutrient-laden soils and plants removed from reservoirs could then be used to fertilize urban fields and gardens.

The most iconic aquatic plant associated with the ancient Maya is the water lily, Nymphaea ampla, which thrives only in clean water, Lucero said. Its pollen has been found in sediment cores from several Maya reservoirs. Water lilies symbolized “Classic Maya kingship,” Lucero wrote.

“The kings even donned headdresses adorned with the flowers and are depicted with water lilies in Maya art,” Lucero said.

“Water lilies do not tolerate acidic conditions or too much calcium such as limestone or high concentrations of certain minerals like iron and manganese,” she wrote.

To keep water lilies alive, water managers would have had to line the reservoirs with clay, Lucero said. A layer of sediment would be needed for plants’ roots. In turn, the water lilies and trees and shrubs planted near the reservoirs shaded the water, cooling it and inhibiting the growth of algae.

“The Maya generally did not build residences near reservoir edges, so contamination seeping through the karstic terrain would not have been an issue,” Lucero wrote.

The evidence gathered from several southern lowland cities indicates that, as constructed wetlands, Maya reservoirs supplied potable water to people for more than 1,000 years, failing only when the severest droughts took hold in the region between 800 and 900 C.E., Lucero said. She notes that current climate trends will require many of the same approaches the Maya employed, including the use of aquatic plants to improve and maintain water quality naturally.

“Constructed wetlands provide many advantages over conventional wastewater treatment systems,” she wrote. “They provide an economical, low technology, less expensive and high energy-saving treatment technology.”

In addition to providing clean water, constructed wetlands also support aquatic animals and can be a source of nutrients to replenish agricultural soils, she wrote.

            “The next step moving forward is to combine our respective expertise and implement the lessons embodied in ancient Maya reservoirs in conjunction with what is currently known about constructed wetlands,” she wrote.

Lidar map of Tikal highlighting some of its reservoirs.

CREDIT

(Image adapted Tankersley et al. 2020). Lidar-derived hillshade image created by Francisco Estrada-Belli of the PAQUNAM LiDAR Initiative. Used with permission. Graphic modified by Bryan Lin.

Maya vessel (c. 700-800 CE) from Guatemala depicting a king sitting on a throne wearing a water lily headdress. Water lilies (Nymphaea ampla) on reservoir surfaces indicated clean water and symbolized Classic Maya kingship (c. 250-900 CE).

CREDIT

Credit: Courtesy the Boston Museum of Fine Arts (www.mfa.org).

 

Editor’s note:  

To reach Lisa Lucero, email ljlucero@illinois.edu  

Once published, the paper “Ancient Maya Reservoirs, Constructed Wetlands, and Future Water Needs” is available online or from the U. of I. News Bureau.

DOI: 10.1073/pnas.2306870120

---

JOURNAL

Proceedings of the National Academy of Sciences

DOI

10.1073/pnas.2306870120 

METHOD OF RESEARCH

Commentary/editorial

SUBJECT OF RESEARCH

Not applicable

ARTICLE TITLE

Ancient Maya Reservoirs, Constructed Wetlands, and Future Water Needs

ARTICLE PUBLICATION DATE

9-Oct-2023

 

Evidence from the remains of 1918 flu pandemic victims contradicts long-held belief that healthy young adults were particularly vulnerable

Peer-Reviewed Publication

MCMASTER UNIVERSITY

 

IMAGE: 

AN EMERGENCY HOSPITAL AT CAMP FUNSTON, KANSAS DURING THE 1918 INFLUENZA PANDEMIC. NEW RESEARCH CONTRADICTS THE WIDESPREAD BELIEF THE FLU DISPROPORTIONATELY IMPACTED HEALTHY YOUNG ADULTS.

view more 

CREDIT: NATIONAL MUSEUM OF HEALTH AND MEDICINE, OTIS HISTORICAL ARCHIVES NEW CONTRIBUTED PHOTO COLLECTION / WIKIMEDIA COMMONS

Hamilton, ON, Oct. 9, 2023 – New analysis of the remains of victims of the 1918 influenza pandemic, which killed an estimated 50 million people worldwide, contradicts the widespread belief the flu disproportionately impacted healthy young adults. 

Because so many people fell ill so quickly, physicians at the time believed the healthy were as likely to die from the flu as those who had already been sick or frail. Despite numerous historical accounts, though, it turns out there is no concrete scientific evidence to support that belief.

Researchers at McMaster University and the University of Colorado Boulder who analyzed victims’ age of death and studied lesions on victims’ bones report that the most susceptible to dying of the flu had exhibited signs of previous environmental, social and nutritional stress.  

“Our circumstances – social, cultural and immunological – are all intertwined and have always shaped the life and death of people, even in the distant past,” explains Amanda Wissler, an assistant professor in the Department of Anthropology at McMaster and lead author on the study, published today in the journal PNAS. 

“We saw this during COVID-19, where our social backgrounds and our cultural backgrounds influenced who was more likely to die, and who was likely to survive,” she says.

Much of the research on the 1918 pandemic relies on historical documentation such as vital statistics, census data and life insurance records, none of which include information on pre-existing conditions, or general environmental, dietary or other chronic stressors which can impact one’s overall health over the course of a lifetime.

For the study, researchers examined the skeletal remains of 369 individuals from the Hamman-Todd Documented skeletal collection, which is housed at the Cleveland Museum of Natural History. All had died between 1910 and 1938. The sample was divided into two groups: a control group who had died before the pandemic, and those who died during the pandemic.

A living person’s skeletal structure may undergo lasting changes due to poor health, resulting in diminished height, irregular growth, developmental tooth defects and other indicators.

The team searched for lesions, or indicators of stress, on the shinbones of the pandemic victims.  New bone formation occurs in response to inflammation caused by physical trauma or infection, for example. Researchers can determine if a lesion had been active, in the midst of healing or had completely healed, all of which provide evidence of underlying conditions.

“By comparing who had lesions, and whether these lesions were active or healing at the time of death, we get a picture of what we call frailty, or who is more likely to die. Our study shows that people with these active lesions are the most frail,” says Sharon DeWitte, a biological anthropologist at the University Colorado Boulder and co-author on the study.

Preexisting medical conditions such as asthma or congestive heart failure are common risk factors which can contribute to poor outcomes from infectious diseases such as influenza. 

Racism and institutional discrimination can amplify these effects, as evidenced in the COVID-19 pandemic, say researchers.  During the Black Death in London, for example, individuals who had previously suffered environmental, nutritional and disease stressors were more likely to die from the plague than their healthier peers.

“The results of our work counter the narrative and the anecdotal accounts of the time,” says Wissler. “This paints a very complicated picture of life and death during the 1918 pandemic.”

The researchers plan to continue to explore the relationship between socioeconomic status and mortality in future work.

Key findings

• Contrary to popular belief, there is no evidence the 1918 influenza pandemic disproportionately killed healthy young people.
   
• Remains of victims indicate that, as in other pandemics, pre-existing medical conditions and socioeconomic factors increased likelihood of death from the flu pandemic.

-30-

Media contacts

Matt Innes-Leroux

Leroum2@mcmaster.ca

647-921-5461 (c)

 

Michelle Donovan

Associate Director, Media Relations

donovam@mcmaster.ca

905-512-8548

 

Lisa Marshall

Associate Director, Science Storytelling

Office of Strategic Relations and Communications

University of Colorado Boulder

303-492-3115

lisa.marshall@colorado.edu

 

 

---

JOURNAL

Proceedings of the National Academy of Sciences

DOI

10.1073/pnas.2304545120 

METHOD OF RESEARCH

Observational study

SUBJECT OF RESEARCH

People

ARTICLE TITLE

Frailty and survival in the 1918 influenza pandemic

ARTICLE PUBLICATION DATE

9-Oct-2023

 

History of parental infertility associated with small increased risk for birth defects among children conceived through fertility treatment

Peer-Reviewed Publication

AMERICAN COLLEGE OF PHYSICIANS

Annals of Internal Medicine Tip Sheet
@Annalsofim
Below please find summaries of new articles that will be published in the next issue of Annals of Internal Medicine. The summaries are not intended to substitute for the full articles as a source of information. This information is under strict embargo and by taking it into possession, media representatives are committing to the terms of the embargo not only on their own behalf, but also on behalf of the organization they represent.
----------------------------
1. History of parental infertility associated with small increased risk for birth defects among children conceived through fertility treatment

Experts urge informed decision-making when choosing a reproductive technology

Abstract: https://www.acpjournals.org/doi/10.7326/M23-0872 

URL goes live when the embargo lifts

A study of more than 850,000 children born in Australia found that parental infertility may be a factor for a small increased risk of birth defects in children conceived through fertility treatment. The authors also found that the use of intracytoplasmic sperm injection (ICSI) was associated with an increased risk for major genitourinary abnormalities. However, the authors caution that the overall increase in the relative risks is small. The study is published in Annals of Internal Medicine.

 

Many children are conceived through assisted reproductive technology (ART) cycles each year. Previous research has shown that children born from ART have a 25% to 50% increased risk for congenital abnormalities, particularly cardiac and genitourinary anomalies. However, it is unclear how much of this risk can be attributed to parental infertility problems compared to ART treatment.

 

Researchers from University of New South Wales, Sydney, Australia, conducted propensity score–weighted population-based cohort study of 851,984 infants born between 2009 and 2017 in New South Wales, Australia. The authors found that there were approximately 40 additional cases of any major congenital abnormality per 10,000 singleton ART births compared to singletons conceived naturally to parents without prior infertility problems. This risk became insignificant when ART-conceived children were compared to children conceived naturally to parents with a history of infertility. The authors say that these findings suggest that parental infertility may partly explain the increase in risk seen in ART-conceived children. The authors also found that ICSI treatment was a risk factor for genitourinary anomalies, even in couples without male infertility. They note that these findings strongly suggest that ICSI represents an independent risk factor for congenital abnormalities and should be reserved for patients with male infertility.

 

Media contacts: For an embargoed PDF, please contact Angela Collom at acollom@acponline.org. To speak with the corresponding author Georgina M. Chambers, Ph.D., please contact g.chambers@unsw.edu.au.

----------------------------

2. Researchers describe horrific effects of new drug threat, xylazine, or "tranq" 

An animal sedative approved by the FDA, xylazine has now made its way to the illicitly manufacture fentanyl supply, creating new challenges for physicians caring for its victims 

Abstract: https://www.acpjournals.org/doi/10.7326/M23-2001   
Editorial: https://www.acpjournals.org/doi/10.7326/M23-2299  

URL goes live when the embargo lifts 
Xylazine, an animal sedative that is approved by the U.S. Food and Drug Administration (FDA) for veterinary use only, has made its way into the illicitly manufactured fentanyl (IMF) supply and has significantly increased in prevalence in recent years, likely due to its low cost, easy availability, and presumed enhanced "high." Researchers reviewed pertinent xylazine research and pulled from their own clinical experience to offer new guidance on the care of patients exposed to this dangerous drug. Their review is published in Annals of Internal Medicine. 

 

The authors provide several critical recommendations for health care institutions and agencies to help combat the threat of xylazine and improve awareness among providers. They suggest educating providers in many settings (emergency departments, primary care, low-barrier clinics, opioid treatment 

programs, office-based opioid treatment programs) on the alterations in recognition, acknowledgment, prophylaxis, and treatment in the presence of chronic xylazine use; conducting animal and human research to further investigate xylazine–fentanyl pharmacology, toxicology, adverse effects, withdrawal syndrome, and treatment strategies; expanding screening to include xylazine on standard urine drug testing; further define the test characteristics to understand the parameters and timing of testing; harm-reduction efforts including increased surveillance of the drug supply and xylazine test strip distribution; expanding access to low-barrier treatment settings with collocated substance use disorder and wound care treatment; and expanding access to inpatient and residential settings where both wound care and substance use disorder treatment are offered. 

 

An accompanying editorial by authors from New York University and University of Florida College of Medicine emphasizes the importance of comprehensive surveillance of xylazine use and poisonings. The authors note that most data on xylazine use comes from decedents and an increasing number of studies monitoring hospitalization, very little is known about the more common outcome—use that does not lead to hospitalization or death. The authors suggest several courses of action that can be taken to increase understanding of xylazine use and risk, including monitoring drug-using persons who are not currently at risk; including xylazine in drug testing efforts; and educating users, clinicians, and researchers about exposure.    

 

Media contacts: For an embargoed PDF, please contact Angela Collom at acollom@acponline.org. To speak with the corresponding author Joseph D’Orazio, M.D., please contact Joseph D’Orazio at dorazio-joseph@cooperhealth.edu

----------------------------

3. AI predictive models shown to be unreliable over time in clinical settings 

Abstract: https://www.acpjournals.org/doi/10.7326/M23-0949  
Editorial https://www.acpjournals.org/doi/10.7326/M23-2345 

URL goes live when the embargo lifts 

A simulation study of artificial intelligence (AI) predictive models using electronic health record (EHR) data in the ICU setting found that the successful use of such models may impair the function of other models – present and future, as well as themselves. This impaired function will manifest as unnecessary investigations and procedures. The authors also report that retraining these models did not significantly reduce these errors. The study is published in Annals of Internal Medicine.  

Researchers from Icahn School of Medicine at Mount Sinai Health System and the University of Michigan School of Medicine simulated 3 common scenarios of model implementation and associated changes in model performance using data from 130,000 critical care admissions. The scenarios each consider deployment of models to predict the risk for death or acute kidney injury (AKI) in the first 5 days after admission to the ICU. Scenario 1 considers the result of implementing and retraining a mortality prediction model, scenario 2 considers the implementation of an AKI model followed by the creation of a new mortality prediction model, and scenario 3 considers the simultaneous implementation of both an AKI and mortality prediction model. The authors found that the model in scenario 1 lost 9 to 39% specificity after retraining once. The mortality model in scenario 2 lost 8% to 15% specificity when created after the AKI model had been in use. In scenario 3, models for AKI and mortality prediction implemented simultaneously, each led to reduced effective accuracy of the other by 1% to 28%. The authors report that in each scenario, performance for models trained on data from populations that benefit from interventions afforded by model prediction is inferior to performance of the original model.

Based on their findings, rather than adopting a universal strategy, model developers should simulate each model’s updating strategy at each site where a model is to be implemented. They also recommend measures to track how and when predictions influence clinical decision making, because most suggested mitigation strategies rely on this information being available, and EHR data may be rendered unsuitable for training models otherwise.

An accompanying editorial by authors from Johns Hopkins University provides important context to these findings. They note that the drift observed in the models included in this study appear in AI in other contexts, including popular large language models like Chat GPT. The authors highlight that these models collapse if recursively trained on their own output, and this “noise” introduced to other clinical models may further degrade clinical predictions in the future. The editorial authors also suggest that fixing model drift starting with inspection, rather than immediate correction, may help. They suggest that, as with other interventions, clinical trials may be required to evaluate the effect of AI models on relevant clinical outcomes. 

 

Media contacts: For an embargoed PDF, please contact Angela Collom at acollom@acponline.org. To speak with the corresponding authors Akhil Vaid, M.D., and Girish N. Nadkarni, M.D. please contact Karin Eskenazi at karin.eskenazi@mssm.edu.  

----------------------------

4. Physicians debate best screening tools and practices for patients with potential dementia and cognitive impairment 

‘Beyond the Guidelines’ features are based on the Department of Medicine Grand Rounds at Beth Israel Deaconess Medical Center 

Abstract: https://www.acpjournals.org/doi/10.7326/M23-1808  

URL goes live when the embargo lifts 

In a new Annals ‘Beyond the Guideline’s feature, two experts review the available evidence about cognitive impairment to determine effective screening tools, interventions to improve patient outcomes, and the circumstances under which they would recommend screening for cognitive impairment (CI). All ‘Beyond the Guidelines’ features are based on the Department of Medicine Grand Rounds at Beth Israel Deaconess Medical Center (BIDMC) in Boston and include print, video, and educational components published in Annals of Internal Medicine.   

 

Dementia, according to the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders, is defined by a significant decline in 1 or more cognitive domains that interferes with a person’s independence in daily activities. Mild cognitive impairment differs from dementia in that the impairment is not sufficient to interfere with independence.  A variety of screening tests are available for Cognitive Impairment. A positive screening test does not diagnose CI; rather it should lead to additional testing to confirm the diagnosis. Upon review of the evidence, the United States Preventive Services Task Force (USPSTF) concluded in 2020 that the evidence was insufficient to assess the balance of benefits and harms of screening for cognitive impairment in older adults (“I statement”). The USPSTF did clarify that although there is insufficient evidence, there may be important reasons to identify CI. 

 

BIDMC Grand Rounds discussants, Michael Barry, M.D., Chair of the U.S. Preventive Services Task Force and a Professor of Medicine at Harvard Medical School, and Deborah Blacker, M.D., ScD, member of the Department of Psychiatry and the Alzheimer's Disease Research Center at Massachusetts General Hospital and Professor of Psychiatry at Harvard Medical School recently debated the case of Ms. B., a 75-year-old woman who, along with her primary care physician, were wondering whether she should be screened for memory loss given multiple risk factors and her desire to be proactive about her medical care. 

 

Dr. Barry explained the rationale for the “I” statement and the lack of sufficient evidence from either the direct or indirect evidence pathway to recommend for or against screening. Dr. Blacker agrees that the available screening tests have issues with reliability, especially in certain patient populations. Both discussants also agree that the established agents have limited efficacy and have not been observed to change the course of the disease. Dr. Blacker reviewed that the newer antiamyloid agents may change the course of the disease, but any cognitive benefits are modest and carry the risk for significant adverse effects. Dr. Barry reviewed that although the USPSTF found that there is insufficient evidence to recommend for or against screening, there may be important reasons to identify CI, as early detection may allow for identification and treatment of reversible causes, help clinicians be aware of patients who might have difficulty understanding and adhering to medical treatment plans, and provide useful information for patients and families as they begin advance care planning. Dr. Blacker concurs and adds that these potential benefits add to the importance of developing a system that conducts and supports screening in primary care. In summary, Dr. Barry would leave the informed decision to screen our patient, Ms. B, to her and her clinician. Dr. Blacker would grant Ms. B's request for screening, as she thinks Ms. B's request may reflect greater difficultie than are otherwise apparent and that, given her family responsibilities and complex medical regimen, the prompt recognition of CI is of particular importance.

 

A complete list of ‘Beyond the Guidelines’ topics is available at www.annals.org/grandrounds. 

 

Media contacts: For an embargoed PDF, please contact Angela Collom at acollom@acponline.org. For an interview with the discussants, please contact Kendra McKinnon at kmckinn1@bidmc.harvard.edu.  

 

 

 

---

JOURNAL

Annals of Internal Medicine

DOI

10.7326/M23-0872 

METHOD OF RESEARCH

News article

SUBJECT OF RESEARCH

People

ARTICLE TITLE

Risk for Congenital Anomalies in Children Conceived With Medically Assisted Fertility Treatment

ARTICLE PUBLICATION DATE

10-Oct-2023

What is the impact of predictive AI in the health care setting?

Findings underscore the need to track individuals affected by machine learning predictions

Peer-Reviewed Publication

THE MOUNT SINAI HOSPITAL / MOUNT SINAI SCHOOL OF MEDICINE

 

IMAGE: 

MODEL USE LEADS TO MIXED ASSOCIATIONS BECAUSE AT-RISK PATIENTS AVOID ADVERSE OUTCOMES, AND THE EHR CAPTURES THIS. FUTURE MODELS TRAINED ON DATA CONTAINING THESE MIXED ASSOCIATIONS PERFORM WORSE.

view more 

CREDIT: VAID ET AL., ANNALS OF INTERNAL MEDICINE ©2023 AMERICAN COLLEGE OF PHYSICIANS. USED WITH PERMISSION.

Models built on machine learning in health care can be victims of their own success, according to researchers at the Icahn School of Medicine and the University of Michigan. Their study assessed the impact of implementing predictive models on the subsequent performance of those and other models. Their findings—that using the models to adjust how care is delivered can alter the baseline assumptions that the models were “trained” on, often for worse—were detailed in the October 9 online issue of Annals of Internal Medicine: https://www.acpjournals.org/doi/10.7326/M23-0949.

“We wanted to explore what happens when a machine learning model is deployed in a hospital and allowed to influence physician decisions for the overall benefit of patients,” says first and corresponding author Akhil Vaid, M.D., Clinical Instructor of Data-Driven and Digital Medicine (D3M), part of the Department of Medicine at Icahn Mount Sinai. “For example, we sought to understand the broader consequences when a patient is spared from adverse outcomes like kidney damage or mortality. AI models possess the capacity to learn and establish correlations between incoming patient data and corresponding outcomes, but use of these models, by definition, can alter these relationships. Problems arise when these altered relationships are captured back into medical records.”

The study simulated critical care scenarios at two major health care institutions, the Mount Sinai Health System in New York and Beth Israel Deaconess Medical Center in Boston, analyzing 130,000 critical care admissions. The researchers investigated three key scenarios:

1. Model retraining after initial use

Current practice suggests retraining models to address performance degradation over time. Retraining can improve performance initially by adapting to changing conditions, but the Mount Sinai study shows it can paradoxically lead to further degradation by disrupting the learned relationships between presentation and outcome.

       2. Creating a new model after one has already been in use

Following a model’s predictions can save patients from adverse outcomes such as sepsis. However, death may follow sepsis, and the model effectively works to prevent both. Any new models developed in the future for prediction of death will now also be subject to upset relationships as before. Since we do not know the exact relationships between all possible outcomes, any data from patients with machine-learning influenced care may be inappropriate to use in training further models.

      3. Concurrent use of two predictive models

If two models make simultaneous predictions, using one set of predictions renders the other obsolete. Therefore, predictions should be based on freshly gathered data, which can be costly or impractical.

“Our findings reinforce the complexities and challenges of maintaining predictive model performance in active clinical use,” says co-senior author Karandeep Singh, MD, Associate Professor of Learning Health Sciences, Internal Medicine, Urology, and Information at the University of Michigan. “Model performance can fall dramatically if patient populations change in their makeup. However, agreed-upon corrective measures may fall apart completely if we do not pay attention to what the models are doing—or more properly, what they are learning from.” 

“We should not view predictive models as unreliable,” says co-senior author Girish Nadkarni, M.D., MPH, Irene and Dr. Arthur M. Fishberg Professor of Medicine at Icahn Mount Sinai, Director of The Charles Bronfman Institute of Personalized Medicine, and System Chief of Data-Driven and Digital Medicine. “Instead, it's about recognizing that these tools require regular maintenance, understanding, and contextualization. Neglecting their performance and impact monitoring can undermine their effectiveness. We must use predictive models thoughtfully, just like any other medical tool. Learning health systems must pay heed to the fact that indiscriminate use of, and updates to, such models will cause false alarms, unnecessary testing, and increased costs.”

“We recommend that health systems promptly implement a system to track individuals impacted by machine learning predictions, and that the relevant governmental agencies issue guidelines,” says Dr. Vaid. “These findings are equally applicable outside of health care settings and extend to predictive models in general. As such, we live in a model-eat-model world where any naively deployed model can disrupt the function of current and future models, and eventually render itself useless.”

The paper is titled “Implications of the Use of Artificial Intelligence Predictive Models in Health Care Settings: A Simulation Study.”

The remaining authors are Ashwin Sawant, M.D.; Mayte Suarez-Farinas, Ph.D.; Juhee Lee, M.D.; Sanjeev Kaul, M.D.; Patricia Kovatch, BS; Robert Freeman, RN; Joy Jiang, BS; Pushkala Jayaraman, MS; Zahi Fayad, Ph.D.; Edgar Argulian, M.D.; Stamatios Lerakis, M.D.; Alexander W Charney, M.D., Ph.D.; Fei Wang, Ph.D.; Matthew Levin, M.D., Ph.D.; Benjamin Glicksberg, Ph.D.; Jagat Narula, M.D., Ph.D.; and Ira Hofer, M.D.

The work was supported by Clinical and translational award for infrastructure UL1TR004419.

-####-

About the Icahn School of Medicine at Mount Sinai

The Icahn School of Medicine at Mount Sinai is internationally renowned for its outstanding research, educational, and clinical care programs. It is the sole academic partner for the eight- member hospitals* of the Mount Sinai Health System, one of the largest academic health systems in the United States, providing care to a large and diverse patient population.  

Ranked 14th nationwide in National Institutes of Health (NIH) funding and among the 99th percentile in research dollars per investigator according to the Association of American Medical Colleges, Icahn Mount Sinai has a talented, productive, and successful faculty. More than 3,000 full-time scientists, educators, and clinicians work within and across 44 academic departments and 36 multidisciplinary institutes, a structure that facilitates tremendous collaboration and synergy. Our emphasis on translational research and therapeutics is evident in such diverse areas as genomics/big data, virology, neuroscience, cardiology, geriatrics, as well as gastrointestinal and liver diseases. 

Icahn Mount Sinai offers highly competitive MD, PhD, and Master’s degree programs, with current enrollment of approximately 1,300 students. It has the largest graduate medical education program in the country, with more than 2,000 clinical residents and fellows training throughout the Health System. In addition, more than 550 postdoctoral research fellows are in training within the Health System. 

A culture of innovation and discovery permeates every Icahn Mount Sinai program. Mount Sinai’s technology transfer office, one of the largest in the country, partners with faculty and trainees to pursue optimal commercialization of intellectual property to ensure that Mount Sinai discoveries and innovations translate into healthcare products and services that benefit the public.

Icahn Mount Sinai’s commitment to breakthrough science and clinical care is enhanced by academic affiliations that supplement and complement the School’s programs.

Through the Mount Sinai Innovation Partners (MSIP), the Health System facilitates the real-world application and commercialization of medical breakthroughs made at Mount Sinai. Additionally, MSIP develops research partnerships with industry leaders such as Merck & Co., AstraZeneca, Novo Nordisk, and others.

The Icahn School of Medicine at Mount Sinai is located in New York City on the border between the Upper East Side and East Harlem, and classroom teaching takes place on a campus facing Central Park. Icahn Mount Sinai’s location offers many opportunities to interact with and care for diverse communities. Learning extends well beyond the borders of our physical campus, to the eight hospitals of the Mount Sinai Health System, our academic affiliates, and globally.

------------------------------------------------------- 

* Mount Sinai Health System member hospitals: The Mount Sinai Hospital; Mount Sinai Beth Israel; Mount Sinai Brooklyn; Mount Sinai Morningside; Mount Sinai Queens; Mount Sinai South Nassau; Mount Sinai West; and New York Eye and Ear Infirmary of Mount Sinai.

JOURNAL

Annals of Internal Medicine

DOI

10.7326/M23-0949. 

METHOD OF RESEARCH

Computational simulation/modeling

SUBJECT OF RESEARCH

Not applicable

ARTICLE TITLE

Implications of the Use of Artificial Intelligence Predictive Models in Health Care Settings: A Simulation Study

ARTICLE PUBLICATION DATE

9-Oct-2023

AI language models could help diagnose schizophrenia

Peer reviewed | Experimental study | People

Peer-Reviewed Publication

UNIVERSITY COLLEGE LONDON

Scientists at the UCL Institute for Neurology have developed new tools, based on AI language models, that can characterise subtle signatures in the speech of patients diagnosed with schizophrenia.

The research, published in PNAS, aims to understand how the automated analysis of language could help doctors and scientists diagnose and assess psychiatric conditions.

Currently, psychiatric diagnosis is based almost entirely on talking with patients and those close to them, with only a minimal role for tests such as blood tests and brain scans.

However, this lack of precision prevents a richer understanding of the causes of mental illness, and the monitoring of treatment.

The researchers asked 26 participants with schizophrenia and 26 control participants to complete two verbal fluency tasks, where they were asked to name as many words as they could either belonging to the category “animals” or starting with the letter “p”, in five minutes.

To analyse the answers given by participants, the team used an AI language model that had been trained on vast amounts of internet text to represent the meaning of words in a similar way to humans. They tested whether the words people spontaneously recalled could be predicted by the AI model, and whether this predictability was reduced in patients with schizophrenia.  

They found that the answers given by control participants were indeed more predictable by the AI model than those generated by people with schizophrenia, and that this difference was largest in patients with more severe symptoms.

The researchers think that this difference might have to do with the way the brain learns relationships between memories and ideas, and stores this information in so called ‘cognitive maps’. They find support for this theory in a second part of the same study where the authors used brain scanning to measure brain activity in parts of the brain involved in learning and storing these ‘cognitive maps’.

Lead author, Dr Matthew Nour (UCL Queen Square Institute of Neurology and University of Oxford), said: “Until very recently, the automatic analysis of language has been out of reach of doctors and scientists. However, with the advent of artificial intelligence (AI) language models such as ChatGPT, this situation is changing.

“This work shows the potential of applying AI language models to psychiatry – a medical field intimately related to language and meaning.”

Schizophrenia is a debilitating and common psychiatric disorder that affects around 24 million people worldwide and over 685,000 people in the UK.

According to the NHS, symptoms of the condition may include hallucinations, delusions, confused thoughts and changes in behaviour.

The team from UCL and Oxford now plan to use this technology in a larger sample of patients, across more diverse speech setting, to test whether it might prove useful in the clinic.

Dr Nour said: “We are entering a very exciting time in neuroscience and mental health research. By combining state-of-the-art AI language models and brain scanning technology, we are beginning to uncover how meaning is constructed in the brain, and how this might go awry in psychiatric disorders. There is enormous interest in using AI language models in medicine. If these tools prove safe and robust, I expect they will begin to be deployed in the clinic within the next decade.”

The study was funded by Wellcome.

---

JOURNAL

Proceedings of the National Academy of Sciences

DOI

10.1073/pnas.2305290120 

METHOD OF RESEARCH

Experimental study

SUBJECT OF RESEARCH

People

ARTICLE TITLE

Trajectories through semantic spaces in schizophrenia and the relationship to ripple bursts

ARTICLE PUBLICATION DATE

9-Oct-2023