Objectives: Dehydration in older adults contributes to increased morbidity and mortality during hospitalization. As such, early diagnosis of dehydration may improve patient outcome and reduce the burden on healthcare. This prospective study investigated the diagnostic accuracy of routinely used physical signs, and noninvasive markers of hydration in urine and saliva.

Design: Prospective diagnostic accuracy study.

Setting: Hospital acute medical care unit and emergency department.

Participants: One hundred thirty older adults [59 males, 71 females, mean (standard deviation) age = 78 (9) years].

Measurements: Participants with any primary diagnosis underwent a hydration assessment within 30 minutes of admittance to hospital. Hydration assessment comprised 7 physical signs of dehydration [tachycardia (>100 bpm), low systolic blood pressure (<100 mm Hg), dry mucous membrane, dry axilla, poor skin turgor, sunken eyes, and long capillary refill time (>2 seconds)], urine color, urine specific gravity, saliva flow rate, and saliva osmolality. Plasma osmolality and the blood urea nitrogen to creatinine ratio were assessed as reference standards of hydration with 21% of participants classified with water-loss dehydration (plasma osmolality >295 mOsm/kg), 19% classified with water-and-solute-loss dehydration (blood urea nitrogen to creatinine ratio >20), and 60% classified as euhydrated.

Results: All physical signs showed poor sensitivity (0%-44%) for detecting either form of dehydration, with only low systolic blood pressure demonstrating potential utility for aiding the diagnosis of water-and-solute-loss dehydration [diagnostic odds ratio (OR) = 14.7]. Neither urine color, urine specific gravity, nor saliva flow rate could discriminate hydration status (area under the receiver operating characteristic curve = 0.49-0.57, P > .05). In contrast, saliva osmolality demonstrated moderate diagnostic accuracy (area under the receiver operating characteristic curve = 0.76, P < .001) to distinguish both dehydration types (70% sensitivity, 68% specificity, OR = 5.0 (95% confidence interval 1.7-15.1) for water-loss dehydration, and 78% sensitivity, 72% specificity, OR = 8.9 (95% confidence interval 2.5-30.7) for water-and-solute-loss dehydration).

Conclusions: With the exception of low systolic blood pressure, which could aid in the specific diagnosis of water-and-solute-loss dehydration, physical signs and urine markers show little utility to determine if an elderly patient is dehydrated. Saliva osmolality demonstrated superior diagnostic accuracy compared with physical signs and urine markers, and may have utility for the assessment of both water-loss and water-and-solute-loss dehydration in older individuals. It is particularly noteworthy that saliva osmolality was able to detect water-and-solute-loss dehydration, for which a measurement of plasma osmolality would have no diagnostic utility.