Multiple electrolyte imbalances in hospitalized patients: a multimorbidity perspective from a large, retrospective cohort study

INTRODUCTION: Multiple electrolyte imbalances (MEIs) are underexplored in hospitalized patients. We aimed to determine: (1) the prevalence and prognostic impact of MEIs; (2) the associations between electrolyte imbalance (EI) combinations and adverse outcomes; and (3) the potential interactions among EI types. MATERIALS AND METHODS: Hospitalized patients at Peking Union Medical College Hospital were enrolled from 2015 to 2020. Adverse outcomes included in-hospital mortality or discharge against medical advice. Multivariable logistic regression models were used to evaluate the associations of number of EIs, EI types, and their combinations with adverse outcomes and to calculate the population-attributable fractions (PAFs). The additive and multiplicative interactions were examined for each EI combination. RESULTS: Among 324,056 hospitalizations, the prevalence of MEIs was 18.8%. Compared to patients without EIs, the odds ratios (ORs) for adverse outcomes were 2.18 (95% confidence interval [CI]: 1.83-2.59) for patients with 1 EI and 17.34 (95% CI: 15.32-19.62) for those with ≥2 EIs. The highest-risk EI combinations at the individual level were hypercalcemia-hypernatremia (OR = 14.96 [95% CI: 11.34-19.68]), hyponatremia-hypernatremia (13.00 [95% CI: 10.09-16.74]), and hypochloremia-hypernatremia (11.06 [95% CI: 8.51-14.34]), while hypokalemia-hypernatremia (PAF = 17.89% [95% CI: 17.37%-18.41%]), hyperchloremia-hypernatremia (15.29% [95% CI: 14.99%-15.59%]), and hypocalcemia-hypernatremia (12.40% [95% CI: 11.94%-12.86%]) contributed the most to adverse outcomes at the population level. Synergistic additive interactions were observed for hyperchloremia-hypernatremia (relative excess risk due to interaction = 4.80 [95% CI: 3.05-6.55]) and hypercalcemia-hypernatremia (6.98 [95% CI: 3.11-10.86]). CONCLUSIONS: MEIs are common and harmful in hospitalized patients. Prioritizing different intervention targets at individual and population levels may improve clinical outcomes.