of prescribed medication, diuretic use or antidepressant medication was similar in those considered by nurses to be dehydrated or hydrated. Patients who were dehydrated had a significantly higher admission median blood sodium (142 vs. 138 mmol/L), U:C ratio (105 vs. 62 mmol:mmol), and median plasma osmolality (308 vs. 298 mOsm/kg) than those who were not dehydrated (all P < 0Æ05, Mann–Whitney U-tests) (Table 1). Dehydrated patients also had a higher median test strip USG (1Æ020 vs. 1Æ010 SG units) and refractometer USG (1Æ022 vs. 1Æ014 SG units) and Ucol (6 vs. 5) than those hydrated, but these differences were not statistically significant (all P > 0Æ05, Mann–Whitney U-test) (Table 1). Overall the within-subject agreement between the refractometer USG being >1Æ020 SG units (indicator of dehydration) and nurses opinion that patients were dehydrated on admission was 84% (95% confidence intervals, CI, 60%, 97%).