Our aim was to explore biological variation of serum sodium levels as a method of quantifying health risk in older adults. We investigated whether dynamic changes in serum sodium levels could provide additional prognostic information to standard predictors of mortality in older people. Analysis of routinely collected clinical datasets containing information on demographics, hospitalisation, biochemistry, haematology and physical function for Dundee in-patient rehabilitation services, between 1999 and 2011. Older people admitted to inpatient rehabilitation following an acute medical or surgical hospitalisation. Five dynamic measures of sodium levels homeostasis - minimum, maximum, standard deviation, and minimum and maximum deviation from mean - were derived for each individual, using biochemistry data from the year preceding their rehabilitation discharge. Cox regression models tested for associations with time to death. Covariates included age, sex, discharge Barthel score, co-morbid diagnoses, haemoglobin, albumin and eGFR. 3021 patients were included (mean age 84 years, 1776 (58.8%) females). 1651 (54.7%) patients experienced hyponatraemia and 446 (14.8%) became hypernatraemic. Mean sodium was correlated with all mean, minimum and SD of sodium. Kaplan-Meier survival curves showed that those without sodium perturbations had the best mortality outcomes, whilst those with both hyponatremia and hypernatremia had the worst. Multivariate Cox regression showed that standard deviation and hypernatraemia were significant predictors of death in non-adjusted models, but not fully adjusted models. All dynamic measures of dysnatraemia were associated with increased mortality risk, but failed to add predictive value to established static measures after adjusting for covariates.
Keywords: Frail elderly; hypernatremia; hyponatremia; mortality; sodium.