Inflammatory bowel disease, which encompasses Crohn’s disease and ulcerative colitis, is a chronic immune-mediated condition of the digestive tract that produces abdominal pain, diarrhea, bleeding, and profound impairment of quality of life. Its relationship with psychiatric illness runs deep: anxiety is markedly overrepresented among people with IBD, and the two conditions appear to feed each other through inflammatory, neural, and behavioral pathways. Yet in patients whose primary clinical picture is dominated by anxiety, the gastrointestinal symptoms of IBD can be dismissed as somatic expressions of psychological distress, delaying diagnosis and treatment. The researchers behind the new study set out to determine whether a simple laboratory value already collected in virtually every hospitalized patient could help close that diagnostic gap.

Serum osmolality reflects the balance of sodium, glucose, urea nitrogen, and other osmotically active solutes in the blood, and it is a fundamental indicator of hydration status and metabolic homeostasis. It has previously been linked to a range of chronic inflammatory and immune-mediated disorders, but evidence connecting it specifically to inflammatory bowel disease in anxious populations was scarce. To address this, the team turned to the Medical Information Mart for Intensive Care IV database, known as MIMIC-IV, a large, de-identified repository of clinical data from patients admitted to Boston-area hospitals. They identified 9,511 inpatients with anxiety who had been admitted to general wards, of whom 211 carried a diagnosis of inflammatory bowel disease and 9,300 did not.

The analytical strategy was unusually thorough for a cross-sectional study. The researchers began with univariate screening and variance inflation factor tests to eliminate redundant variables and assess multicollinearity. To correct for baseline differences between patients with and without IBD, they applied stabilized inverse probability of treatment weighting, a technique that reweights the comparison groups so that confounders such as age, comorbidities, and medication use are statistically balanced. They then modeled the association between serum osmolality and IBD using both conventional multivariable logistic regression and Firth penalized logistic regression, the latter being a method designed to produce stable estimates when outcomes are rare—an important consideration given that only about 2 percent of the cohort had IBD.

The headline finding was striking. After full adjustment, patients with high serum osmolality, defined as a value at or above 296.71 milliosmoles per kilogram, had roughly fourteen times the odds of having inflammatory bowel disease compared with those below that threshold, with an odds ratio of 14.36 and a 95 percent confidence interval of 9.77 to 21.12, a result that remained highly significant. Restricted cubic spline analysis, which allows the data to reveal the shape of a relationship rather than forcing it into a straight line, showed a significant nonlinear dose-response pattern, with the association between osmolality and IBD varying across the physiological range rather than rising uniformly.

To test how well serum osmolality alone could discriminate between patients with and without IBD, the team employed nested 10-fold cross-validation, a rigorous machine learning practice in which the data are repeatedly split into training and testing sets to guard against over-optimistic performance estimates. The resulting area under the receiver operating characteristic curve was 0.695, indicating moderate discriminative ability. In parallel, the researchers trained an XGBoost gradient-boosting model and interrogated it with SHapley Additive exPlanations, a game-theoretic framework that quantifies each feature’s contribution to individual predictions. In that analysis, serum osmolality ranked as the second most important predictor of IBD, surpassed only by platelet count—a biologically plausible companion finding, since platelets participate directly in the inflammatory cascade driving bowel disease.

Robustness checks reinforced the result. The association held when the researchers reran the models without weighting, when they applied inverse probability of selection weighting, and when they used a narrower definition of anxiety to classify the cohort. An E-value estimation, which quantifies how strong an unmeasured confounder would need to be to explain away the observed association, suggested considerable resilience. Subgroup analyses across demographic and clinical strata showed consistent associations, indicating that the relationship between osmolality and IBD was not confined to a particular age group, sex, or clinical profile within the anxious inpatient population.

The biological story behind the statistics remains speculative but coherent. Inflammatory bowel disease involves barrier dysfunction, fluid shifts, diarrhea-driven dehydration, and systemic inflammation, all of which can perturb serum osmolality. Elevated osmolality may reflect subclinical dehydration from chronic intestinal fluid loss, or it may track with the inflammatory and neuroendocrine stress responses that characterize both IBD and anxiety. Conversely, some investigators have proposed that hyperosmolar states can themselves promote inflammatory signaling, suggesting a possible bidirectional loop. The authors are careful to note that the cross-sectional design of the study makes causal inference impossible: high osmolality could be a consequence of IBD, a contributor to it, or a shared downstream marker of another process entirely.

Those caveats matter, and the study’s limitations deserve emphasis. The cohort consisted exclusively of hospitalized patients with anxiety drawn from a single academic medical center database, so the findings may not generalize to outpatients, community populations, or people without psychiatric comorbidity. Serum osmolality is influenced by hydration, renal function, medications, and acute illness, factors that weighting and adjustment can only partially neutralize. The authors explicitly state that serum osmolality should be viewed as a potential adjunctive quantitative marker rather than a stand-alone diagnostic test, pending validation in prospective studies that follow patients over time and can establish whether osmolality changes precede the onset or flares of bowel disease.

Even with those limitations, the implications are appealing in their simplicity. Serum osmolality is already measured or easily calculated in nearly every hospitalized patient, requires no specialized equipment, and adds essentially no cost to routine care. If prospective research confirms the association, a value above the threshold identified in this study could serve as a flag prompting clinicians caring for anxious patients to look more closely for gastrointestinal inflammation—ordering fecal inflammatory markers, referral to gastroenterology, or endoscopic evaluation—rather than attributing symptoms solely to anxiety. In a field where diagnostic delay measurably worsens outcomes, an inexpensive number already sitting in the electronic health record could prove a surprisingly powerful clue, and the study’s blend of classical epidemiology with modern machine learning offers a template for how overlooked routine biomarkers might be systematically rediscovered.

Subject of Research: Association between serum osmolality and inflammatory bowel disease in patients with anxiety using the MIMIC-IV database

Article Title: Serum osmolality as a potential marker of inflammatory bowel disease in patients with anxiety: a cross-sectional analysis based on the MIMIC-IV database

Article References: Wu, S., Chen, L., Shen, Y., & Qi, Y. (2026). Serum osmolality as a potential marker of inflammatory bowel disease in patients with anxiety: a cross-sectional analysis based on the MIMIC-IV database. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08611-y

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08611-y

Keywords: serum osmolality, inflammatory bowel disease, anxiety, MIMIC-IV, biomarker, cross-sectional study, machine learning, XGBoost, SHAP, logistic regression, BMC Psychiatry, Serum