Diagnostic performance of calculated serum osmolarity versus clinical impression for identifying dehydration in older hospitalized patients

Objectives This study aimed to evaluate the diagnostic performance and agreement of calculated serum osmolarity for identifying dehydration in older hospitalized patients, using clinical impression of dehydration as the reference standard, and to examine glucose levels across concordant and discordant classification groups. Design and setting This prospective, observational study was conducted in a geriatric acute care unit. Participants The study included 455 older hospitalized patients with a mean age of 82.2 ± 6.8 years and 70% were female. Measurements Clinical impression of dehydration (dehydration vs. no dehydration [normal hydration + hyperhydration]) was judged and recorded by the attending physician within 24 h of admission. Calculated serum osmolarity was derived from routine laboratory values and classified as high (≥ 295 mOsmol/l) or normal (< 295 mOsmol/l). Diagnostic accuracy (sensitivity, specificity, predictive values, accuracy) and agreement (Cohen’s κ) were calculated. Glucose was compared between concordant/discordant groups. Results Clinically, 23% of patients were classified as dehydrated. High calculated osmolarity occurred in 38%. Using the ≥ 295 mOsmol/l cut-off, sensitivity was 48.6% and specificity 65.1%; positive and negative predictive values were 29.7% and 80.7%, respectively, and accuracy was 61.3%. Agreement beyond chance was poor (κ = 0.112). Mean glucose was higher in osmolarity-positive groups, including false positives (all p < 0.001 for planned pairwise comparisons). Conclusion In older hospitalized patients, calculated serum osmolarity (≥ 295 mOsmol/l) showed limited agreement with clinical dehydration assessment with frequent osmolarity-positive classifications in clinically non-dehydrated patients. These findings support cautious clinical interpretation of calculated osmolarity in acute care, particularly when glucose is elevated.