The research, led by Furlow and colleagues, tackles one of the most consequential prediction problems in critical care. Up to 80 percent of patients who regain spontaneous circulation after cardiac arrest are comatose on admission, and their recovery is frequently uncertain. Because brain injury is the most common cause of death and disability in this population, accurate neuroprognostication is essential, yet reliable predictors remain limited and many apply only to narrow subsets of patients. Current guidelines emphasize a multimodal approach, combining examination findings, EEG patterns, imaging, and biomarkers rather than relying on any single signal. Even so, the process remains as much art as science, and the reasoning behind a withdrawal-of-care decision is often scattered across hundreds of narrative notes.

To build their model, the researchers extracted unstructured text from notes spanning the entire hospitalization of comatose cardiac arrest survivors treated at three hospitals. Natural language processing algorithms processed this text and identified potentially informative keywords, which were then used to train a logistic regression model to predict outcomes on a stratified Cerebral Performance Category scale, the standard framework for grading neurologic recovery after cardiac arrest. The baseline model performed well, achieving an area under the receiver operating characteristic curve of 0.9 and an area under the precision-recall curve of 0.89. Those figures place the note-based model in the range of respectable clinical prediction tools, though the authors caution that performance fell when the model was restricted to notes from only the first 24 hours of admission, with the decline especially pronounced for predicting good outcomes.

That last detail matters enormously, and it is where the accompanying commentary by Sarah Nutman and Christopher Horvat of the University of Pittsburgh sharpens the discussion. Post-cardiac arrest guidelines from the American Heart Association, the European Resuscitation Council and European Society of Intensive Care Medicine, and the Neurocritical Care Society recommend waiting at least 72 hours before attempting to prognosticate, because sedation, targeted temperature management, and the natural evolution of injury can cloud early assessment. A model trained on notes from the first day of admission is therefore being asked to make a judgment that guidelines explicitly say should not yet be made. Nutman and Horvat suggest that the drop in performance on early notes is unsurprising, and that a 72-hour window might have been more instructive for evaluating the model’s true clinical utility.

The commentary also raises a subtler statistical concern: temporal leakage. When a model is trained on notes drawn from the entire hospitalization, the patient’s eventual outcome can become progressively clearer as time passes, so the text may simply echo a conclusion that clinicians have already reached rather than genuinely predicting it. The authors of the original study worried that their models’ weaker performance on good outcomes creates a risk of self-fulfilling prophecies, in which a pessimistic algorithmic prediction influences the decision to withdraw care, thereby making the prediction come true. Nutman and Horvat propose an alternative interpretation: the asymmetry may reflect clinicians incorporating information that never makes it into the written record, leaving the model blind to the very signals that would allow it to recognize patients on a path to recovery.

Perhaps the most provocative contribution of the work is not the AUROC at all, but the window it opens into the hidden architecture of prognostic decision-making. By identifying language patterns associated with withdrawal of life-sustaining treatment, the researchers demonstrated that natural language processing can do more than forecast outcomes; it can interrogate the clinical reasoning, and the possible biases, embedded in routine documentation. Two findings stand out. Keywords related to intimate partner violence were significantly more common in the notes of patients on whom life-sustaining treatment was withdrawn. And although several guidelines recommend against using myoclonus as a prognostic marker, myoclonus appeared significantly more often in the records of patients whose care was withdrawn, raising the possibility that the finding is still quietly influencing bedside decisions despite the guideline warnings.

These observations point toward a use case that extends well beyond prediction. Even without a predictive model, the approach suggests that natural language processing could be deployed to systematically audit differences in clinical outcomes using variables that structured data never captures. In an era when health systems are increasingly attentive to equity, the ability to mine the narrative record for evidence of inconsistent decision-making represents a genuinely novel form of quality control. The same technique that flags a guideline-discordant reliance on myoclonus could, in principle, reveal whether prognostic pessimism is distributed unevenly across patient populations, institutions, or clinical teams.

The study also highlights an underappreciated dimension of prognostication: whose observations count. Neuroprognostication has historically been driven by physician perspectives, yet nurses, therapists, and other health professionals spend far more continuous time at the bedside and often notice changes, in spontaneous movements, responsiveness to family, or sleep-wake cycling, that physicians document less frequently. Incorporating these complementary observations through interdisciplinary evaluation may improve predictions, and the note-mining approach provides a structured mechanism to capture and integrate perspectives from across the clinical team. In that sense, the model is not merely reading the chart; it is aggregating the collective perception of everyone who wrote in it.

Still, significant questions remain before such an algorithm could approach the bedside. Post-arrest guidelines insist on a multimodal approach rather than reliance on any single predictor or algorithm, so the critical unknown is the additive benefit of this tool when used alongside established predictors, including the clinical examination, EEG and neuroimaging findings, and serum biomarkers. The original study compared its model’s performance to other natural language processing models, but not to known neuroprognostic predictors, which limits how much can be said about its incremental value. It also remains unknown whether the algorithm captures anything genuinely new, or whether it simply rediscovers, in narrative form, the same information clinicians already extract from structured testing. A model trained on clinical notes may be learning patient biology, clinician interpretation, family preferences, institutional practice patterns, or an inseparable mixture of all four, and in post-arrest care that distinction is not academic.

The stakes of getting this right could hardly be higher. A falsely pessimistic prediction does not merely misclassify a patient; it can trigger decisions that make the prediction come true, converting an algorithmic error into an irreversible outcome. As Nutman and Horvat conclude, the next step is not to build better note-based models but to determine what these models are actually learning, when their predictions become reliable, and how they should be used alongside established multimodal prognostic tools. Used carefully, natural language processing may help clinicians extract real signals from the clinical record and strengthen one of the most difficult judgments in medicine. Used carelessly, it risks transforming documentation artifacts into clinical authority. The enduring promise of this work lies in making the invisible parts of prognostication visible, and then deciding, with appropriate humility, which of those signals deserve to guide care.

Subject of Research: Natural language processing of clinical notes for neurologic outcome prediction after cardiac arrest

Article Title: Reading Between the Lines After Cardiac Arrest

Article References: Nutman, S. K., & Horvat, C. M. (2026). Reading Between the Lines After Cardiac Arrest. Neurocritical Care. https://doi.org/10.1007/s12028-026-02649-2

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02649-2

Keywords: cardiac arrest, neuroprognostication, natural language processing, machine learning, electronic health records, clinical notes, coma, withdrawal of life-sustaining treatment, EEG, critical care, predictive models, healthcare bias

The story began with an elective abdominal operation in a patient whose medical history was already extraordinarily complex. She carried a diagnosis of VACTERL association, a congenital syndrome involving vertebral defects, anal atresia, cardiac malformations, tracheoesophageal fistula, renal anomalies, and limb abnormalities, and she had undergone multiple reconstructive surgeries as well as treatment for advanced gynecologic cancer. The planned exploratory laparotomy included removal of both ovaries and fallopian tubes, lysis of adhesions, repair of a small bowel leak, and placement of a temporary ureteral stent. For eight days her recovery proceeded uneventfully. Then, during a failed attempt to place a nasogastric tube for a suspected ileus, she suffered a pulseless electrical activity arrest. After eight minutes of cardiopulmonary resuscitation, her heartbeat returned, she was intubated and sedated, and surgeons re-entering her abdomen found frank pus. Septic shock was diagnosed as the likely cause of the arrest, and broad-spectrum antibiotics were started.

The neurological trouble surfaced on postoperative day 9, when abnormal movements concerning for seizure appeared during an attempt to wean her sedation. Neurologists ordered continuous electroencephalography, which recorded unreactive theta slowing consistent with severe encephalopathy, along with myoclonic movements that occurred without any evolving electrographic seizure activity. The official interpretation described findings consistent with postanoxic myoclonus, a phenomenon long regarded as an ominous sign after cardiac arrest. On examination the patient was comatose, her pupillary reflexes were the only preserved brainstem findings, and she exhibited severe stimulus-induced generalized myoclonus. Laboratory markers told a mixed story: elevated liver enzymes and troponin suggested ongoing hypoperfusion, but neuron-specific enolase, a blood biomarker of neuronal injury, remained reassuringly low at 16, 14, and 15 nanograms per milliliter over the first three days after the arrest. Head computed tomography showed nothing abnormal, and magnetic resonance imaging was scheduled three to five days after the arrest in keeping with neuroprognostication guidelines.

What followed was a cascade of findings that failed to converge on the grim prognosis the early exam had implied. Bacteremia from Enterobacter was confirmed on day 10, prompting broader antibiotics. Brain MRI on day 11, three days after the arrest, revealed no cerebral ischemia at all, an outcome that directly contradicted the expected signature of severe hypoxic-ischemic injury. By day 13, nurses were reporting violent stimulus-induced myoclonus accompanied by autonomic changes that demanded repeated midazolam boluses every four hours, and her mental status did not improve even after propofol sedation was withdrawn. When neurology was reconsulted on day 15, the team recognized that none of the highly reliable predictors of poor outcome had actually been met, raising the possibility that a reversible metabolic or iatrogenic process was obscuring the true picture. Laboratory review then uncovered worsening acute kidney injury, a pivotal clue.

A repeat EEG changed the diagnostic landscape. This time it captured centrally predominant spikes time-locked to the clinical myoclonic jerks, a pattern the consulting team interpreted as cortical myoclonus with potential epileptogenicity, in contrast to the earlier recording that had suggested a subcortical origin without cortical spikes. Repeat MRI remained unchanged, deepening the puzzle. The case report uses this evolution to highlight a fundamental problem in post-arrest neurology: myoclonus can arise from the cortex or the subcortex, and the two are notoriously difficult to disentangle. Cortical myoclonus is classically associated with single midline or parasagittal EEG spikes against a continuous background, thought to reflect bilateral motor strip involvement transmitted through the corpus callosum. Subcortical posthypoxic myoclonus, by contrast, has been conceptualized as reticular reflex myoclonus, a stimulus-sensitive phenomenon arising from the medulla oblongata. In practice, EEG findings in subcortical myoclonus may be absent or nonspecific, movement artifacts contaminate recordings, electromyography leads that would help localize the jerk are often not used, and no clinical examination can reliably separate the two categories. Recent studies have even reported similar rates of favorable outcomes between the groups, eroding confidence in the dichotomy itself.

The report also addresses the question hanging over every such case: does postanoxic myoclonus doom a patient? Historically the answer was assumed to be yes, but contemporary evidence cautions strongly against relying on myoclonus, especially within the first 48 to 72 hours after return of spontaneous circulation, to guide irreversible decisions. Early motor findings can be distorted by sedatives, metabolic derangements, and drug-induced encephalopathy, and current guidelines explicitly warn against premature prognostication, urge clinicians to exclude modifiable factors, and recommend waiting at least 72 hours before using EEG findings prognostically. Importantly, the authors note that the vertex spike-wave discharges synchronized with the patient’s jerks have been described in the literature as a characteristic EEG phenotype of Lance-Adams syndrome, a chronic and typically survivable form of posthypoxic myoclonus. Recognizing that pattern earlier, they suggest, might have reframed the entire prognostic conversation.

That reframing opened the door to the case’s most consequential insight. Among the differential diagnoses was toxic myoclonus, and the team noted that the patient’s antibiotics had recently been broadened to cefepime, a fourth-generation cephalosporin with a well-documented neurological footprint. Beta-lactam antibiotics can induce encephalopathy by antagonizing GABA-A receptors, the brain’s principal inhibitory signaling system, producing cortical hyperexcitability that manifests as myoclonus and seizures; the full mechanism of the accompanying encephalopathy remains incompletely understood. The randomized ACORN trial further strengthened the suspicion, having demonstrated greater neurological dysfunction among hospitalized adults treated with cefepime than with piperacillin-tazobactam. Critically, the patient had possessed none of the recognized risk factors for cefepime neurotoxicity when the drug was first chosen, including advanced age, renal impairment, or prolonged high-dose therapy. Only as her kidney function deteriorated did the drug’s accumulation become a credible threat, a vivid reminder that critically ill patients have rapidly shifting metabolic states that demand continual reassessment of neuroactive prescriptions.

On day 16, cefepime was discontinued and replaced with metronidazole and piperacillin-tazobactam, while valproate was initiated for symptomatic myoclonus control based on literature supporting its use in posthypoxic cortical reflex myoclonus. The response was striking. By day 17, brainstem reflexes had returned. By day 20, she opened her eyes to voice, and over the following days she progressed to full alertness and consistent command-following, although stimulus-induced myoclonus stubbornly persisted. That persistence was diagnostically decisive: it supported a final diagnosis of Lance-Adams syndrome, likely layered with a toxic-metabolic overlay of cefepime-induced neurotoxicity in the setting of acute kidney injury, rather than cefepime toxicity alone. The chronic myoclonus continued to obstruct rehabilitation, and levetiracetam restarted alongside up-titrated valproate, a commonly reported first-line strategy, produced no benefit. Clonazepam was eventually added. Her mentation, however, kept improving, and she was discharged to a rehabilitation facility awake, communicative, and bearing a diagnosis that carries a typically favorable neurological outlook rather than the devastating prognosis her first days in the ICU had suggested.

The lessons the authors distill reach well beyond this single bedside. Clinical myoclonus alone, they argue, should never be treated as a definitive marker of severe anoxic brain injury, even though myoclonic status epilepticus with truly malignant EEG patterns remains a poor prognostic sign; neuroprognostication must instead weave together evolving EEG findings, examination, neuroimaging, and evoked potentials while actively hunting for modifiable contributors such as antibiotic-associated encephalopathy. Cefepime-induced neurotoxicity, they emphasize, can worsen mental status and provoke myoclonus and seizures, and it is exceptionally easy to miss in patients who already have neurological impairment, because every new abnormality gets attributed to the original injury. And Lance-Adams syndrome, though rare, should stand in every clinician’s mind as the favorable exception among postanoxic movement disorders. For a field in which prognostic statements can drive decisions about withdrawal of life-sustaining therapy, the case is a powerful demonstration that diagnostic clarity, patience, and a willingness to question early assumptions can quite literally change the predicted future of a patient who wakes up against the odds.

Subject of Research: Neuroprognostication after cardiac arrest complicated by postanoxic myoclonus and antibiotic-induced neurotoxicity

Article Title: Neurotoxicity in Neuroprognostication: The Importance of Differential Diagnostic Clarity

Article References: Tocco, E., Gruccio, P., Davidow, J., & Graziano, A. (2026). Neurotoxicity in Neuroprognostication: The Importance of Differential Diagnostic Clarity. Neurocritical Care. https://doi.org/10.1007/s12028-026-02654-5

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02654-5

Keywords: neuroprognostication, cardiac arrest, postanoxic myoclonus, Lance-Adams syndrome, cefepime neurotoxicity, electroencephalography, encephalopathy, septic shock, myoclonus, neurocritical care, beta-lactam antibiotics, GABA-A receptors

The concept of alpha coma dates to 1975, when investigators at the Mayo Clinic described comatose patients whose EEGs displayed diffuse, unreactive alpha-frequency activity, typically 8 to 12 hertz, that looked deceptively like the background rhythm of an awake brain. Theta coma, a slower variant operating in the 4 to 7 hertz range, was recognized soon afterward as a related pattern. In the classic descriptions, particularly following hypoxic-ischemic injury after cardiac arrest, the rhythm was fixed and unresponsive to external stimulation, and mortality approached nearly one hundred percent. Generations of clinicians learned to regard the tracing as evidence that the thalamocortical circuits sustaining consciousness had been irreversibly dismantled, even as the cortical surface continued to generate an organized rhythm.

Yet scattered case reports over the decades kept complicating the picture. Patients occasionally woke up. Some recovered fully, particularly after drug intoxications, metabolic derangements, or brainstem strokes rather than cardiac arrest. These exceptions raised an uncomfortable question: was alpha and theta coma truly a single malignant entity, or was it a heterogeneous collection of states lumped together by an electrographic appearance that concealed radically different underlying pathologies and prognoses? The Thai-led team, led by Somkiat Phutinart, Penpisut Kedgan, and colleagues, set out to answer that question quantitatively, searching PubMed, EMBASE, and the Cochrane Library through December 2025 for original human studies reporting neurological outcomes in comatose patients with alpha, theta, or alpha-theta coma patterns.

The pooled numbers confirm that the overall prognosis remains sobering. Across the six included studies, 83.7 percent of patients experienced severe disability or death, with a confidence interval of 76.8 to 88.8 percent and remarkably low between-study heterogeneity. Mortality itself was estimated at 74.0 percent, though here the studies diverged more widely, with the confidence interval stretching from 43.2 to 91.4 percent. In other words, roughly one in four patients survived, and a meaningful minority avoided the most devastating functional outcomes. For a pattern historically described as nearly uniformly fatal, that margin of uncertainty is clinically consequential, because prognostication in comatose patients directly shapes decisions about the intensity and duration of intensive care.

The most striking finding concerns EEG reactivity, the capacity of the brain’s rhythmic activity to change in response to external stimuli such as sound, pain, or touch. Reactivity was present in a surprising 45.4 percent of patients, a figure far higher than the classical descriptions would suggest. When the researchers stratified outcomes by this feature, the separation was dramatic. A nonreactive EEG was significantly associated with mortality, carrying an odds ratio of 5.67 with a tight confidence interval of 1.22 to 26.31 and no measurable heterogeneity across studies. Nonreactivity also sharply reduced the odds of a good outcome, with an odds ratio of 0.12. Conversely, a reactive EEG was associated with roughly a ninefold increase in the odds of a favorable outcome, at an odds ratio of 8.68.

The physiological logic behind these numbers is instructive. Reactivity implies that at least some of the cortical and subcortical machinery responsible for processing sensory input remains functionally intact, even if the patient appears behaviorally unresponsive. A brain that can modulate its own activity in response to the outside world retains, in a measurable sense, a connection to that world. A fixed, unreactive rhythm, by contrast, suggests that the apparent wakefulness of the tracing is a hollow imitation, generated by residual circuitry that can no longer be recruited by external events. The meta-analysis thus reframes alpha and theta coma not as a single electrographic diagnosis but as a spectrum whose position on that spectrum is revealed by a simple bedside test performed during routine EEG recording.

Etiology emerged as the second powerful prognostic modifier. Hypoxic-ischemic injury following cardiac arrest predominated among the pooled patients, and it carried the worst outlook, consistent with the diffuse cortical and thalamic damage that global oxygen deprivation inflicts. By contrast, toxic or metabolic causes, including drug overdoses and severe metabolic disturbances, were disproportionately represented among patients who achieved better outcomes. This makes biological sense: sedative drugs and metabolic derangements can suppress consciousness and distort EEG rhythms reversibly, without destroying neurons, whereas anoxia kills cells outright. The finding echoes decades of case reports describing full recovery from alpha coma after lorazepam, imipramine, or bupropion overdose, and it underscores that the same electrographic pattern can reflect a transient pharmacological state in one patient and irreversible necrosis in another.

The authors are careful to temper their conclusions with an important methodological caveat. The included studies span more than four decades, and the diagnostic criteria for alpha and theta coma have evolved considerably since the 1970s. Early definitions relied on visual inspection of paper EEG traces and did not always distinguish true alpha coma from so-called alpha pseudocoma, in which a patient is actually awake but locked in, unable to move while retaining full awareness. Contemporary critical care EEG terminology, standardized by the American Clinical Neurophysiology Society, classifies these patterns within a much richer framework that incorporates reactivity, continuity, and frequency in ways the original descriptions never attempted. Whether the historical cohort of 152 patients would be classified identically under modern criteria remains genuinely uncertain, and the authors explicitly call for contemporary studies that redefine alpha and theta coma using current concepts of disorders of consciousness.

The practical implications for neurocritical care are nonetheless immediate. Modern post-cardiac-arrest guidelines emphasize multimodal prognostication, combining EEG with somatosensory evoked potentials, neuron-specific enolase, brain imaging, and clinical examination, precisely because no single test is infallible. This meta-analysis adds quantitative weight to the principle that a reactive alpha or theta pattern should prompt caution before a prognosis of futility is declared, while a nonreactive pattern in the setting of confirmed hypoxic-ischemic injury strengthens the case that the outlook is grim. The finding that nearly half of these patients exhibit reactivity also serves as a reminder that EEG should be recorded and interpreted with stimulation protocols, since reactivity that goes untested is prognostic information left on the table.

Beyond the intensive care unit, the study speaks to a broader theme in neuroscience: the danger of treating an electrographic appearance as a disease. Alpha and theta coma arose as descriptive labels in an era when the EEG was one of the few windows into the injured brain, and the labels hardened into prognostic dogma. Five decades of accumulated exceptions, now synthesized with modern meta-analytic methods, reveal the heterogeneity that was always there. For the families of comatose patients, and for the clinicians advising them, the message is one of calibrated hope. The rhythm on the screen may look like wakefulness without a person behind it, but the brain’s response to a spoken word or a painful pinch can reveal whether that person still has a fighting chance to return.

Subject of Research: Prognostic value of EEG reactivity and etiology in alpha and theta coma following severe brain injury

Article Title: EEG Reactivity and Etiology as Prognostic Modifiers in Alpha and Theta Coma: A Systematic Review and Meta-analysis

Article References: Phutinart, S., Kedgan, P., Kosonboon, P., Sinaroj, K., Siranart, N., Anukoolwittaya, P., & Surawattanawong, T. (2026). EEG Reactivity and Etiology as Prognostic Modifiers in Alpha and Theta Coma: A Systematic Review and Meta-analysis. Neurocritical Care. https://doi.org/10.1007/s12028-026-02646-5

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02646-5

Keywords: alpha coma, theta coma, EEG reactivity, neuroprognostication, cardiac arrest, hypoxic-ischemic injury, coma prognosis, meta-analysis, neurocritical care, disorders of consciousness, electroencephalography, systematic review