New Bedside Cognitive Test Tracks Brain Fog After Electroconvulsive Therapy

Electroconvulsive therapy has long carried a paradoxical reputation. It is one of the most effective treatments in psychiatry, capable of lifting patients out of severe, life-threatening depression when medications have failed, yet it is also associated with a side effect that patients and families fear most: cognitive impairment. Memory gaps, slowed thinking, and difficulties with concentration can follow a course of treatment, and clinicians have historically struggled to measure these changes with the blunt instruments available at the bedside. A new prospective study from Peking University Sixth Hospital, published in BMC Psychiatry, argues that the field has been using the wrong ruler, and it offers detailed evidence that a purpose-built tool, the Electroconvulsive Therapy–Cognition Inventory, or ECT-CI, can do the job far better than the general cognitive screens in widespread use today.

The research team, led by Qinghua Guo, Yong Wang, Chuan Shi, and Shaomei Shang, followed 136 adults undergoing an acute course of ECT in a real-world clinical setting. Each participant completed the ECT-CI and the Montreal Cognitive Assessment, better known as the MoCA, at three time points: before treatment began, within 48 hours after the acute course ended, and again at a three-month follow-up. Crucially, the study did not rely on test scores alone to define impairment. Blinded psychiatrists adjudicated whether each patient met criteria for cognitive impairment using a framework based on the DSM-5-TR, the current edition of the American Psychiatric Association’s diagnostic manual, and those clinician judgments served as the reference standard against which both instruments were judged.

The clinical picture that emerged was sobering. Clinician-adjudicated cognitive impairment rose from 12.5 percent of patients at baseline to 33.1 percent immediately after the acute treatment course, and then climbed further to 50.7 percent at the three-month follow-up. That trajectory, in which the proportion of patients judged cognitively impaired continues to rise well after the last seizure has been induced, underscores how incomplete our understanding of post-ECT cognition remains. Many patients do not simply recover in a straight line after treatment ends, and some appear to experience worsening difficulties that only become apparent weeks later, a pattern that brief in-hospital testing at discharge would entirely miss.

Against this backdrop, the performance of the two cognitive instruments diverged in ways that matter for everyday practice. ECT-CI scores fell from baseline to the post-course assessment and then partially rebounded at follow-up, and the magnitude of those observed score changes was larger than what the MoCA captured over the same period. The ECT-CI also took less time to administer, a nontrivial advantage at the bedside of acutely ill psychiatric inpatients. At the domain level, the inventory registered changes in memory, executive function, and caregiver-rated cognition, offering a multidimensional view of how patients were actually functioning rather than a single composite number. In statistical comparisons of discrimination, essentially the ability of a test score to separate patients who have impairment from those who do not, the ECT-CI outperformed the MoCA against the clinician-adjudicated reference standard at every assessment point.

Perhaps the most clinically consequential finding concerned prediction. Baseline ECT-CI scores showed greater prognostic discrimination than baseline MoCA scores for identifying which patients would meet the clinician-adjudicated criteria for cognitive impairment three months later. In practical terms, this means a brief test administered before the first seizure could help clinicians flag patients at elevated risk of lasting cognitive difficulty, opening a window for interventions that do not currently exist in routine care: closer monitoring, adjusted treatment parameters, or earlier supportive measures. The MoCA, designed as a general screen for dementia and delirium across clinical populations, simply was not built to detect the specific, often subtle cognitive signature that ECT produces.

To move beyond averages and capture the diversity of individual recovery patterns, the researchers turned to growth mixture modelling, a statistical technique that identifies hidden subgroups within a population based on how their scores change over time. The analysis revealed three distinct ECT-CI trajectories. One group maintained high, stable scores throughout, suggesting resilience to the cognitive effects of treatment. A second group experienced an acute decline in scores immediately after the treatment course followed by a partial rebound at follow-up, a pattern consistent with transient, treatment-related effects that fade with time. The third and most concerning group started from a low baseline and suffered a marked acute decline, entering treatment already cognitively vulnerable and deteriorating further.

Who ended up in that worst trajectory was not random. Lower baseline MoCA scores and a diagnosis of schizophrenia were associated with membership in the low-baseline, marked-decline group. This finding carries real weight for treatment planning. Patients with schizophrenia who undergo ECT, often for refractory symptoms, may represent a population in whom cognitive reserve is already diminished and in whom the additional burden of a treatment course pushes function below a critical threshold. The results support a model of risk-stratified cognitive monitoring, in which patients are assessed before treatment and assigned to different intensities of follow-up based on their baseline profile and diagnostic background, rather than receiving a uniform, one-size-fits-all check.

The study also ventured into territory that has attracted growing interest across neuropsychiatry: the role of inflammation. In exploratory analyses, the researchers examined systemic immune-inflammatory markers, including the systemic immune-inflammation index and related indices derived from routine blood counts, along with C-reactive protein, a well-known marker of systemic inflammation. Their analyses suggested that systemic immune-inflammatory burden characterized the unfavourable cognitive trajectories, and, intriguingly, that baseline CRP levels interacted with the cumulative ECT charge, the total amount of electrical stimulation a patient received over the course. In other words, the relationship between how much stimulation a patient’s brain absorbed and whether that patient developed clinician-adjudicated cognitive impairment appeared to depend on the inflammatory state of the body before treatment began.

These inflammatory findings must be read with appropriate caution, as the authors themselves frame them as exploratory, and a single cohort study cannot establish mechanism. But the underlying hypothesis is biologically plausible and resonates with a broader literature linking neuroinflammation to vulnerability of the hippocampus and other memory-critical structures. If confirmed in independent samples, the idea that inflammatory status modifies the cognitive cost of cumulative electrical stimulation could reshape how clinicians think about dosing strategies. It might eventually support the use of inexpensive blood tests, already drawn routinely in pre-ECT workups, as part of a personalized risk assessment, and it could motivate trials of anti-inflammatory or lifestyle interventions aimed at lowering inflammatory burden before an elective treatment course.

What the study delivers most concretely is a validation case for disease-specific measurement. The ECT-CI tracked score changes that the MoCA missed, discriminated better against a rigorous clinical reference standard, predicted three-month outcomes from baseline, and lent itself to trajectory-based risk stratification, all while requiring less administration time. For a procedure administered to hundreds of thousands of patients worldwide each year, the absence of a dedicated cognitive monitoring standard has been a conspicuous gap. This prospective real-world cohort, registered with the Chinese Clinical Trial Registry and conducted under ethics approval at Peking University Sixth Hospital, provides the strongest longitudinal evidence to date that filling that gap is both feasible and clinically worthwhile. The next steps, replication in larger and more diverse populations, external validation of the three-trajectory model, and prospective testing of whether risk-stratified monitoring actually improves patient outcomes, will determine whether the ECT-CI moves from promising research instrument to standard of care. For now, the message for clinicians is clear: when it comes to the cognitive price of electroconvulsive therapy, the right questions, asked with the right tool, reveal risks that general screening has been quietly missing.

Subject of Research: Longitudinal validation of the ECT-CI for monitoring cognitive impairment after electroconvulsive therapy

Article Title: Longitudinal evaluation of the ECT-CI after electroconvulsive therapy: score changes, clinical discrimination, and latent score trajectories in a prospective real-world cohort

Article References: Guo, Q., Wang, Y., Guo, L., He, X., Pu, C., Shi, C., & Shang, S. (2026). Longitudinal evaluation of the ECT-CI after electroconvulsive therapy: score changes, clinical discrimination, and latent score trajectories in a prospective real-world cohort. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08718-2

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08718-2

Keywords: electroconvulsive therapy, ECT-CI, cognitive impairment, MoCA, growth mixture modelling, prognostic discrimination, inflammation, CRP, psychiatry, neuropsychological assessment, schizophrenia, longitudinal cohort