The historical reliance on inference rather than observation for hydrogen positions deserves some elaboration. X-ray crystallography detects electrons, not nuclei, and hydrogen atoms, with their single electron participating in the bond, scatter X-rays weakly relative to carbon, nitrogen, and oxygen. In small-molecule crystallography, where crystals are typically highly ordered and data can be collected to very high resolution, hydrogen positions can often be refined directly or placed with confidence using geometric constraints. In macromolecular crystallography, by contrast, protein crystals diffract to lower resolution, contain enormous amounts of solvent-disordered content, and rarely permit direct localization of hydrogens on ligands. Neutron crystallography can locate hydrogen nuclei explicitly because neutrons scatter from nuclei rather than electron clouds, but neutron sources are scarce, beamtime is limited, and growing large protein crystals suitable for neutron work is difficult. As a result, the tautomeric state of a ligand bound in a protein structure is usually assigned by software or by the crystallographer’s judgment, using rules of thumb about which heteroatoms are likely protonated at physiological pH. These assignments are then inherited by downstream users of the structure, often without re-examination.

The scale of the Cambridge Structural Database mining effort described in the study is worth appreciating in context. The CSD has accumulated small-molecule crystal structures for over a century, and its high-resolution entries preserve the geometric detail needed to distinguish, for example, a hydroxyl oxygen from a carbonyl oxygen, or an imidazole nitrogen bearing a proton from one that does not. By systematically extracting more than 1.1 million tautomeric states from these structures, the researchers effectively converted a repository built for crystal engineering and structural chemistry into a training corpus for machine learning. This kind of data repurposing has become an increasingly productive strategy in computational chemistry, where experimental archives originally collected for one purpose turn out to contain the labels needed to train models for another. The orders-of-magnitude increase over prior tautomer datasets, which typically contained a few hundred experimentally characterized examples in solution, changes what is statistically learnable: rare tautomeric motifs, substituent effects, and heteroaromatic patterns that would be absent or underrepresented in smaller collections can now inform predictions.

The choice of a graph neural network operating on two-dimensional molecular input is also consequential. Two-dimensional representations encode the connectivity of atoms and bonds but not the three-dimensional conformation, meaning the model must learn tautomer preferences from topology and chemical context alone. This is a deliberate trade-off. Requiring a 3D structure would mean generating conformations first, introducing an additional source of error and computational cost, and would tie the model’s accuracy to the quality of conformer generation. By working directly from the 2D form, the tool can be applied at the earliest stages of library triage, before any geometry has been computed, which is exactly where tautomer decisions have the greatest leverage since they propagate into every subsequent calculation.

The finding that roughly 2.5 percent of ligands in the PDBbind set may carry incorrect tautomer assignments warrants careful interpretation. PDBbind is a curated collection of protein-ligand complexes drawn from the Protein Data Bank, widely used as a benchmark for docking and binding affinity prediction. If a few percent of its ligands have misassigned protonation or tautomeric states, then benchmark results that appear to reflect method performance may in part reflect data noise. The researchers are careful to note that the experimental structures themselves are not necessarily wrong; rather, the chemical annotation layered onto them may need revision. This distinction matters because the electron density in a crystal structure constrains the heavy-atom framework, and a reassigned tautomer often fits the same density equally well while offering more chemically reasonable hydrogen bonding. The example highlighted in the release, where reassignment enables additional hydrogen bonds with nearby protein residues, illustrates the practical payoff: a corrected tautomer can transform an apparently suboptimal binding pose into a coherent one.

Hydrogen bonding is particularly sensitive to tautomeric state because it depends on which atoms carry hydrogens and which act as acceptors. A pyridone and a hydroxypyridine, for instance, present entirely different donor-acceptor patterns to a binding site despite sharing the same atomic composition. In structure-based drug design, where hydrogen bonds are often the dominant specificity-determining interactions, an incorrect tautomer can lead medicinal chemists to draw the wrong conclusions about why a ligand binds or how to improve it. Free energy perturbation calculations and other rigorous simulation methods are similarly affected, since they require a fully specified molecular topology including every hydrogen. An error introduced at the tautomer assignment stage cannot be corrected later by more sampling or better force fields; it is baked into the simulation from the start.

The throughput figures reported for the released tool give a sense of its practical reach. Processing roughly 4.6 million compounds in 3.2 hours on a single GPU-enabled node corresponds to on the order of hundreds of molecules per second, fast enough to annotate entire corporate or public screening libraries overnight. This matters because tautomer enumeration is combinatorial: a molecule with several heteroatoms and exchangeable protons can generate many candidate tautomers, and exhaustively evaluating each with quantum mechanics is infeasible at library scale. A learned model that ranks candidate tautomers by predicted stability provides a practical filter, allowing expensive quantum or experimental follow-up to be reserved for the most plausible candidates.

It is also worth situating this work within the broader movement toward physics-aware machine learning in chemistry. Rather than attempting to predict binding affinities or biological activity end-to-end, the model targets a discrete, well-defined chemical inference problem that sits upstream of many downstream tasks. Problems of this kind, where the answer is a structured chemical choice rather than a continuous property, can be validated against experimental ground truth and integrated into existing workflows as drop-in improvements. The open-source release of the tool as Tautomer-Predictor follows a pattern increasingly common in computational chemistry, where reproducibility and community adoption are treated as integral to the research contribution rather than afterthoughts.

Certain limitations naturally accompany this approach. The training data derive from small-molecule crystals, which reflect tautomeric preferences in the solid state, where packing forces and crystal environments can shift equilibria relative to solution or to the protein binding pocket. The study’s authors and the field more generally recognize that tautomer populations are environment-dependent: a form favored in a crystal may not dominate in water, and the protein microenvironment, with its own hydrogen bond donors, acceptors, and dielectric properties, can stabilize forms that are minor in bulk solution. Models trained on crystallographic data therefore capture one informative slice of tautomer stability, and their predictions are best understood as strong priors to be refined in context rather than final answers. Nevertheless, the availability of a fast, data-driven prior represents a meaningful advance over rule-based tautomer perception, which has historically relied on hardcoded chemical heuristics that vary between software packages and can disagree with one another.

The funding context, an NIH grant supporting methods development for molecular modeling, reflects the recognized importance of infrastructure-level improvements in drug discovery research. Individual breakthroughs in target identification or clinical validation attract considerable attention, but cumulative progress often depends on quietly fixing foundational layers of the computational pipeline. Tautomer assignment is one such layer. As machine learning methods continue to permeate molecular design, the quality of the chemical representations fed into them becomes a limiting factor, and efforts to clean and standardize those representations, whether through better data, better models, or better experimental characterization, compound across the entire discovery enterprise. The work from NYU exemplifies this genre of contribution: not a new drug or a new target, but a repair to the substrate on which countless downstream calculations rest.

Subject of Research: Scientists use AI to find stable forms of molecules for drug discovery

Article Title: Scientists use AI to find stable forms of molecules for drug discovery

Article References: Scientists use AI to find stable forms of molecules for drug discovery. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: tautomers, deep learning, drug discovery, chemical science, hydrogen positions, graph neural network, Cambridge Structural Database, Protein Data Bank, molecular modeling, NYU, Scientists, find

The project, named Fragmented Legacy: Turning Broken 3D Materials into Archaeological Interpretations, is one of 19 selected for this year’s Humanities and AI Virtual Institute awards from Schmidt Sciences. It is led by Chen Feng, Institute Associate Professor at NYU’s Tandon School of Engineering and founding co-director of the NYU Center for Robotics and Embodied Intelligence, together with Jing Zhang, a faculty fellow at the same center. On the archaeological side, the team includes Radu Iovita and Scott Williams of NYU’s Department of Anthropology and Raiko Krauß, Professor of Central and Southeast European Prehistory at Tübingen, with additional collaborators at Leiden University, Arizona State University, and the Max Planck Institute for Evolutionary Anthropology.

The problem the group is attacking is deceptively simple to state and brutally hard to solve. Most archaeological collections are never systematically refitted at all, because the work is so slow that it simply does not get done. Reported success rates tell the story: roughly 1 percent for ceramics, and between 4 and 26 percent for bones. A human specialist may spend days or weeks trying to find the matching edge for a single flake, and in a museum setting the physical constraints make it worse. Drawers can only be opened one at a time, so an entire collection can never be laid out for comparison at once. The result is that vast stores of information sit locked inside unmatched fragments.

Why does a single match matter so much? Because when two fragments fit, they carry information that neither piece holds alone. If a stone flake found in one part of a site fits back onto a core recovered from another, researchers can reconstruct how people moved through that space, how a tool was manufactured step by step, or whether the site’s layers were disturbed after deposition. Refitted animal bones can show how a carcass was butchered and how meat was shared across a group, a direct window into social behavior tens of thousands of years old. Refitted human fossils can reveal anatomical features that no single fragment preserves. Without refits, many of those questions simply go unanswered.

Feng’s lab specializes in teaching machines to understand three-dimensional shapes in the real world, where conditions are messy and incomplete. Broken archaeological material, he notes, is among the hardest versions of that problem: the pieces are worn, many are missing entirely, and there is no picture on the box showing what the finished object should look like. The AI systems his team is building must reason through that uncertainty, generating plausible reconstructions while flagging where the evidence is thin, with the archaeologists on the team judging whether the answers actually make sense in light of the archaeological record.

The technical plan unfolds in stages. First, the team will assemble a dataset of at least 100,000 real and digitally simulated broken objects, pairing 3D models with photographs, descriptions, and catalog data. Crucially, the simulated breaks will not be random: they will follow how stone, clay, and bone actually fracture, so that the synthetic fragments behave like the real ones archaeologists pull from the ground. Next, the researchers will develop AI models that work out where each fragment belongs while simultaneously generating a plausible version of the complete object. The models will reason across shape, images, and text, so that archaeological knowledge — typologies, raw material properties, contextual records — can rule out reconstructions that are geometrically possible but archaeologically implausible.

Archaeologists will remain in the loop throughout, comparing competing reconstructions and teaching the system which ones hold up. This human-in-the-loop design reflects a core philosophy of the project: the AI is not meant to replace expert judgment but to multiply its reach. As Zhang puts it, archaeologists are like Sherlock Holmes, piecing together clues to understand the past, and the goal is for Fragmented Legacy to be their Watson — matching fragments and suggesting possible reconstructions so that researchers have more time to interpret the evidence and ask new questions.

The project did not appear out of nowhere. It grew out of a collaboration between Feng’s AI4CE Lab and Iovita’s lab that began about four years ago with LUWA, an earlier system that used AI to classify microscopic wear on stone tools. Zhang, then a postdoctoral researcher working in both labs, connected the two groups. In 2024, she asked Iovita what archaeology’s most pressing technical problem was, and he pointed straight at refitting. That conversation produced GARF, a system that reassembles broken objects, and then CRAG, which goes further by also filling in missing pieces — work that laid the direct groundwork for the HAVI award now funding Fragmented Legacy.

Iovita’s ambitions for the technology extend well beyond reassembling objects within a single site. Refitting objects between sites, to demonstrate direct relationships between different locations, has only ever been done twice in the history of the discipline, he notes. His stated dream is to use AI to do this at scale, which he argues will revolutionize the field. If fragments from one site can be matched against fragments from another, researchers could trace the movement of raw materials, tools, and even the people who carried them across landscapes in ways that have previously been impractical or impossible.

Accessibility is built into the project from the start. All datasets, model weights, and benchmarks will be released under permissive open-source licenses, so that university researchers and cultural resource management firms — the companies that carry out much of the archaeological work required ahead of construction in the United States and abroad — can use the tools directly in their own workflows. The team also expects the methods to travel beyond archaeology, finding applications in other scientific fields and in practical repair and restoration work, wherever broken three-dimensional objects need to be reassembled and their missing parts intelligently reconstructed. For a discipline built on fragments, the promise is that the pieces sitting unmatched in museum drawers may finally begin to speak.

Subject of Research: AI-based refitting of fragmented archaeological artifacts such as stone tools, pottery, and bones

Article Title: NYU researchers are building AI to piece together broken archaeological finds

Article References: NYU researchers are building AI to piece together broken archaeological finds. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: artificial intelligence, archaeology, refitting, 3D modeling, stone tools, pottery, bones, NYU, Schmidt Sciences, machine learning, cultural heritage, open source

The research, led by Deborah Olarte, assistant professor of nutrition at New York University Steinhardt, interviewed 25 elementary school teachers in the Anchorage School District in Alaska. The timing of the study design was deliberate. During the 2021-2022 school year, Anchorage schools participated in the universal free school meal program, a federal initiative that provided breakfast and lunch to every student regardless of family income. In the 2022-2023 school year, that program ended, and schools reverted to the traditional model in which meals are subsidized only for students who qualify based on economic need. Because the same teachers worked across both years, they could serve as natural comparison observers, describing how the same children, classrooms, and school routines changed when the meal guarantee vanished.

The researchers asked open-ended questions focused on the benefits and challenges of offering free meals to students, and the answers painted a consistent picture. During the universal meal period, teachers perceived reduced stress for both students and families, along with fewer visible signs of food insecurity in the classroom. One teacher explained the mechanism plainly: children who would have felt shy about admitting they were hungry or that they had no lunch no longer had to worry, because food was simply provided to everyone. The stigma that had long surrounded free or reduced-price meals—an enduring barrier documented in school nutrition research—dissolved when every tray was free.

That period was not without drawbacks, and the teachers were candid about them. The most commonly cited downside of universal meals was increased food waste, as students who were not hungry, or who took meals out of convenience, discarded uneaten portions. Food waste has been a persistent concern in school meal policy debates, and opponents of universal programs frequently point to it as evidence of inefficiency. Yet in the overall accounting offered by the teachers interviewed, the waste was outweighed by what the program prevented: hunger, shame, and the classroom disruptions that follow both.

After the program’s expiration, the observations shifted sharply. Teachers reported increased student hunger, more frequent behavioral issues, greater difficulty concentrating, and rising anxiety among students about paying for meals. They also described changes in their own behavior: teachers became more likely to reach into their own pockets and provide food for students, a quiet subsidy that rarely appears in any policy ledger. Perhaps most striking from a nutrition standpoint, teachers observed increased consumption of sugary snacks and beverages and ultra-processed foods, alongside a decrease in school meal participation—particularly among students who were eligible for free meals under the traditional income-based rules.

That last finding may seem paradoxical, but it aligns with a well-documented phenomenon in school nutrition research. Eligibility for subsidized meals does not guarantee participation. Application paperwork, enrollment lapses, household instability, language barriers, and the social stigma of being identified as a free-meal student all suppress take-up even among children who qualify. When universal meals removed those friction points, participation rose; when they returned, participation fell, and the children who lost the most were often those the safety net was nominally designed to protect. Meanwhile, the calories that filled the gap came disproportionately from cheap, shelf-stable, heavily marketed products rather than balanced meals.

The classroom consequences described by teachers were vivid. According to one teacher interviewed in the study, some children were simply hungry all afternoon, could not focus, and disruptive behaviors followed. Hunger’s effect on cognition and behavior is well established in the scientific literature: fluctuations in blood glucose, the physiological stress of food insecurity, and the mental preoccupation of worrying about a next meal all compete with attention, working memory, and self-regulation—the very capacities elementary classrooms depend upon. Teachers in the study effectively described these mechanisms playing out in real time, as afternoon lessons became harder to deliver and behavior management consumed more instructional time.

Olarte emphasized that teachers occupy a unique vantage point in the school meal policy landscape. Participation data can show that fewer students ate lunch, she noted, but teachers can explain why a student did or did not eat and how the consequences presented in the classroom. In her view, universal free school meals removed a worry from the school day, and when the meals went away, the worry came back—and teachers watched it play out in front of them. Her comments also highlighted a less visible dimension of the policy shift: the burden that school meal programs place on educators themselves. Teachers in the study described buying snacks, cleaning up, and absorbing the logistical and emotional fallout of student hunger, and Olarte argued that this burden may shape families’ decisions about whether to participate in school meals at all, an underexamined feedback loop in nutrition policy research.

The study’s authors are careful about the limits of their design. As a qualitative survey of 25 teachers in a single district, the research captures perceptions and experiences rather than measured dietary intake or standardized achievement outcomes. Teachers gave the researchers one window into the policy shift, Olarte noted, and she called for future work that includes the voices of students, parents, administrators, and food service staff across different districts and states. Anchorage itself is an instructive setting: Alaska’s high cost of living and long winters amplify food insecurity, and a single district’s experience may not generalize uniformly across the country. Still, the consistency of the themes—hunger, stress, stigma, junk food substitution, and lost focus—echoes concerns raised by school nutrition advocates in the many states where universal meal programs ended in 2022.

The policy stakes remain live. Since the federal waivers expired, a number of states have enacted their own universal free school meal programs, while others have not, creating a natural patchwork for future comparative research. The Anchorage study suggests that the consequences of that patchwork are not abstract. They arrive in classrooms as afternoon hunger, as children anxious about lunch balances, as teachers quietly stocking granola bars in desk drawers, and as students trading balanced meals for sugary drinks and ultra-processed snacks. Funding for the study was provided by the National Institute of Child Health and Human Development. As districts and lawmakers weigh the costs of feeding every child, the teachers’ testimony offers a ground-level answer to a question that participation statistics alone cannot: what the end of universal free meals actually felt like, for the children living it and the educators watching it happen.

Subject of Research: The impact of ending the pandemic-era universal free school meal program on student hunger, diet, and classroom behavior as perceived by teachers

Article Title: End of free school meal program led to increased stress and hunger, poorer diets, and reduced focus, say teachers

Article References: End of free school meal program led to increased stress and hunger, poorer diets, and reduced focus, say teachers. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: school meals, child nutrition, food insecurity, teachers, classroom behavior, universal free meals, student hunger, ultra-processed foods, education policy, Anchorage, NYU, Journal of Nutrition Education and Behavior

The central finding is deceptively simple but consequential for how educators and policymakers think about diversity in higher education. In male-majority fields such as physics, engineering, computer science, business, and economics, straight cisgender men with lower-than-average high-school grades and test scores were substantially overrepresented compared with straight cisgender women who had similar levels of academic performance. Among students at the top of the achievement distribution, that gap narrowed dramatically. In other words, the familiar underrepresentation of women in quantitative and technical fields is driven disproportionately by what happens to students with weaker academic credentials, not by a uniform filtering process that operates equally at every level of preparation.

Lead author Joseph Cimpian, professor of economics and education policy at NYU Steinhardt and NYU Wagner, emphasized that this heterogeneity is easy to miss if researchers only examine overall representation rates. Across both datasets, he and his colleagues found that when a representation gap varied across the achievement distribution, that gap tended to favor the major’s historically dominant demographic group most strongly among students with lower pre-college grades and test scores, and least so at other points in the distribution. Those gaps narrowed considerably among students with the strongest pre-college performance. Cimpian argued that looking closely at how gaps differ by achievement level, rather than relying on aggregate statistics, is essential for understanding who is actually entering these different fields—and who is being screened out or discouraged before they ever arrive.

The methodological foundation of the study is one of its distinguishing strengths. The researchers analyzed the High School Longitudinal Study of 2009, which followed 6,340 students from their first year of high school in 2009-2010 through their junior year of college, capturing demographic information including sex and LGBTQ+ identity alongside measures of academic preparation, interests, and outcomes. They complemented this with the Baccalaureate and Beyond Longitudinal Study of 2016, which provided academic and demographic data for 12,950 students who earned a bachelor’s degree in 2016. Using two independent, nationally representative datasets allowed the team to test whether the same patterns emerged in different cohorts and at different stages of the college pipeline, from initial major selection through persistence to degree completion.

Crucially, the study is among the first to use nationally representative data to examine how LGBTQ+ students sort across the full range of college majors, not just the STEM disciplines that have dominated conversations about demographic disparities. The results showed that LGBTQ+ students, regardless of sex, were underrepresented in male-majority fields such as engineering, computer science, and business. LGBTQ+ men were also underrepresented in nursing and other health fields—a finding that mirrors, in reverse, the pattern of straight cisgender women being underrepresented in male-dominated technical majors. The consistency of these across-group patterns suggests that the dynamics shaping who enters and stays in a major operate along multiple demographic dimensions simultaneously, not solely along the lines of gender.

In female-majority fields, the evidence for a parallel pattern of overrepresentation among straight cisgender women was mixed, and the authors are careful not to overstate symmetry between the two directions of the analysis. However, some statistical models revealed a striking result: LGBTQ+ women tended to be overrepresented in female-majority fields among students with the highest pre-college achievement. The researchers interpret this cautiously, but it raises the possibility that LGBTQ+ women may need stronger academic credentials to gain a foothold even in fields that might superficially appear welcoming to them. If confirmed by future work, this would suggest that the barriers facing sexual and gender minority students are not confined to traditionally hostile environments but extend into disciplines where their demographic group holds numerical dominance.

What might explain why representation gaps are sharpest among students with weaker academic credentials? The researchers point toward social norms and institutional environments that subtly favor a field’s historically dominant demographic group. Students with lower grades and test scores may have less margin for error when navigating environments where they are demographically outnumbered: they may receive less encouragement, face harsher judgment when they struggle, or perceive fewer role models and fewer signals that people like them succeed in the field. High-achieving students, by contrast, may possess a kind of buffer—strong credentials, confidence, and external validation—that allows them to persist even in environments that are less than fully welcoming. The study is correlational, and the authors explicitly note that it cannot establish that these norms directly cause the observed gaps; it demonstrates only that individual academic and demographic factors on their own do not fully account for the patterns in the data.

That caveat matters, but it does not diminish the study’s practical implications. If prior performance alone cannot explain who stays in a major and who leaves, then interventions focused exclusively on academic preparation—tutoring, prerequisite pipelines, admissions criteria—are unlikely to close representation gaps on their own. The findings point instead toward the climate of departments and classrooms, particularly for students who arrive with less academic capital and therefore less resilience against discouragement. They also suggest that aggregate diversity statistics may mask the true geography of exclusion. A department that appears moderately diverse overall may, in fact, be nearly homogeneous among its less-prepared students, with demographic diversity concentrated only among those strong enough to overcome environmental headwinds.

Cimpian framed the results as a story of missed opportunity rather than simple loss. If talented students are leaving—or never entering—majors where they would do well and contribute, partly based on factors unrelated to their preparation or even their interest in those fields, that is a loss both for the students themselves and for the disciplines they might have enriched. The phrase ‘unrelated to their preparation or even interest’ is the study’s sharpest edge: it implies that the conventional explanations for demographic sorting, which center on ability and preference, are incomplete. Something in the environment surrounding a major appears to interact with a student’s demographic identity and achievement level to shape whether that student completes a degree in the field.

The research, co-authored by Zachary T. McDermott at Abt Global, Jo R. King at Boston University, Nathaniel Woznicki at the NYU Department of Psychology, and Taek H. Kim at the Korean Educational Development Institute, was supported by grants from the Institute of Education Sciences at the US Department of Education. Its publication in Nature Human Behaviour reflects growing scientific interest in moving beyond binary questions of whether demographic gaps exist toward more granular questions of where, for whom, and under what conditions those gaps emerge. For universities seeking to diversify engineering, computer science, nursing, education, and every field in between, the message is that the students most at risk of exclusion are not the strongest performers but those with the least academic cushion—and that the environments those students encounter, not just the students themselves, may be the most promising target for change.

Subject of Research: Demographic representation and persistence across college majors by achievement level

Article Title: Students with lower grades and test scores earn degrees more often when they belong to an academic major’s majority group

Article References: Students with lower grades and test scores earn degrees more often when they belong to an academic major’s majority group. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: higher education, college majors, gender representation, LGBTQ+ students, STEM diversity, academic achievement, degree completion, Nature Human Behaviour, longitudinal study, education policy, student persistence, NYU

The 988 Suicide and Crisis Lifeline was designed to solve a fundamental access problem. Before its launch, people experiencing a mental health crisis had to navigate a ten-digit hotline number, and capacity varied widely. The three-digit dialing code allows anyone in the country to call, text, or chat with trained counselors through a single, easy-to-remember number. When people contact 988, they are routed to local crisis call centers, and those centers depend heavily on funding from individual states. How states raise that money has become one of the central policy questions surrounding the lifeline’s long-term sustainability, because the federal government established the number but left most of the financing responsibility to the states themselves.

Most states finance 988 and related crisis services through their general budget appropriations, the ordinary pool of tax revenue that legislators allocate each year. But twelve states have taken a different approach, creating dedicated telecom fees modeled on the mechanism that has long paid for 911 emergency services in many states. These fees are deliberately small and flat, for example thirty cents per cell phone line per month, spread across millions of subscribers. The study’s lead author, Jonathan Purtle, professor of public health policy and management at NYU School of Global Public Health, noted that mental health advocates have recommended telecom fees to fund 988, but until now it has not been fully understood whether they are making a difference. The answer, he said, is empirically yes, because states that have these telecom fees put more money toward funding suicide prevention.

To reach that conclusion, the research team examined annual state budgets, which may include both state and federal funds, for all fifty states and Washington, DC, across six fiscal years from 2021 through 2026. The researchers looked at what each state allocated specifically for suicide prevention and mental health crisis services, and then analyzed whether those allocations varied according to a range of state-level factors, including geography, budget structure, and the presence of a dedicated fee. This budget-based approach is methodologically demanding because states organize their finances in very different ways, and the authors are careful to acknowledge the limits of comparing line items across fifty-one distinct budgeting systems.

The headline finding is a clear upward trend. Average state spending on suicide prevention and crisis services rose from $1.05 per resident in 2021 to $6.55 per resident in 2025, more than a sixfold increase in per-capita investment in just four years. The proportion of states’ annual budgets devoted to suicide prevention likewise increased, from 0.022 percent of overall spending to 0.079 percent. Those percentages may look tiny, but they represent a meaningful shift in how state governments prioritize crisis response, and they coincide with the period during which 988 launched and public and policymaker attention to mental health crises intensified nationwide.

The regional patterns in the data are striking. Spending was highest in Western states, which invested an average of $7.19 per person per year, followed by the Northeast at $3.44, the South at $2.12, and the Midwest at $2.07. That means residents of the average Western state enjoyed roughly three and a half times the crisis-service investment of residents of the average Midwestern state. Even more consequential for policy debates, the researchers found that spending was twice as large in states that had enacted telecom fees compared with states that relied on other mechanisms. Purtle observed that this finding suggests telecom fees are a promising strategy to enhance crisis service infrastructure, giving legislators who are weighing such fees concrete evidence that the mechanism is associated with substantially higher investment.

One of the study’s more sobering findings concerns the relationship between need and spending. The researchers examined whether states with higher suicide rates are spending more money on suicide prevention and found that this was not necessarily the case. In other words, the states where the problem is most acute are not automatically the states investing the most in response. That mismatch matters because suicide risk is not evenly distributed across the population. Veterans, for example, die by suicide at more than double the rate of nonveterans, and the Department of Veterans Affairs spends $177 per veteran per year on suicide prevention, a figure that underscores just how modest state-level investments remain even after five years of growth.

The authors are careful about the caveats that surround their estimates. Because each state structures its budget differently, and not every state has specific line items or fees for suicide prevention, the classification process is conservative. If a state budgeted for broader mental health programs but did not specify how much of that funding would support suicide prevention and crisis services, the researchers did not count that state as funding suicide prevention that year. Purtle acknowledged that as a result, the study’s findings are likely an underestimate of what states actually spend. For readers, that means the true investment picture may be somewhat brighter than the headline numbers suggest, but it also means the study measures only explicit, designated commitments, which is precisely what advocates say is needed to guarantee stable crisis-line capacity.

The research arrives at a moment when early evidence suggests the 988 system may be having an effect. Emerging research is beginning to show a decline in suicide rates since the launch of 988, although that work has not yet explored whether the decline varies among states, which would be the crucial test of whether state-level funding decisions translate into lives saved. Purtle’s ongoing research program aims to answer exactly that question, examining whether state-level funding for 988 actually reduces suicide deaths and, in particular, whether telecom fees move the needle as a policy solution. If subsequent studies confirm that fee-financed states see larger declines, the twelve states with fees could become models for the rest of the country.

The study was conducted by a team spanning several leading institutions. In addition to Purtle, the authors include Michal Weiss and Jennifer Pomeranz of NYU School of Global Public Health, Ryan McBain and Jonathan Cantor of RAND, and Sachini Bandara, Fernanda Montoya, and Elizabeth Stuart of the Johns Hopkins Bloomberg School of Public Health. The research was funded by the National Institute of Mental Health under grant R01 MH131649, and the authors note that the content is solely their responsibility and does not necessarily represent the official views of the National Institutes of Health. For the growing community of crisis-center operators, state legislators, and advocates watching 988’s development, the study’s message is twofold: state investment in suicide prevention is rising faster than at any point in recent memory, but dedicated, fee-based financing mechanisms appear to be the most reliable way to turn that momentum into durable, well-funded crisis infrastructure that reaches everyone who dials three digits in their darkest moment.

Subject of Research: State-level budget allocations for suicide prevention and 988 crisis services in the United States, including the role of dedicated telecommunications fees

Article Title: State funding for suicide prevention has increased in recent years

Article References: State funding for suicide prevention has increased in recent years. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: suicide prevention, 988 lifeline, state budgets, mental health crisis services, telecom fees, health policy, JAMA Network Open, NYU, public health funding, crisis intervention, state government spending, veterans suicide risk