Presentations

Abstract

Introduction:

Hypernatremia is a poorly understood clinical condition recognized in hospitalized patients. The largest observational retrospective cohort study observed that 6% of patients developed hypernatremia and they were 14 times more likely to die in the hospital. Unfortunately, this study included patients in whom the development of hypernatremia could be therapeutic or in patients that may have been administered hypertonic saline. A retrospective review of a large electronic health record database can be undertaken to evaluate an association of mortality in patients admitted to the hospital with an observed hypernatremia (serum sodium (SNa) above 145 mmol/L) during the first 14 day of admission. Patients in whom hypernatremia could be a therapeutic goal and those who received hypertonic saline (3% saline or higher) should be excluded from evaluation.

Methods:

For this retrospective cohort study deidentified electronic health record information from approximately 115 million individuals in 85 healthcare organizations of the TriNetX Research Network was accessed on 27 June 2024. To assess mortality, patient records for this study were linked to the United States Social Security Death Index. Patients 18 years and older who had an inpatient encounter with two serum sodium assays 24 hours apart during the first 14 days of hospitalization were included. Patients with an ICD-10 diagnosis related to traumatic brain injury; ischemic and hemorrhagic cerebral infarcts; and renal failure were excluded from the review. Patients with SNa below 135 mmol/L at admission or within the first 14 days of hospitalization were excluded. Patients were divided into two groups for comparison. Patients with SNa above 145 at admission or with in the first 14 days of hospitalization were considered the hypernatremia groups. Patients that maintained SNa between 135 and 145 at admission and the subsequent 14 days were the comparison group. Propensity score match was applied considering age and sex, with a 1:1 ratio within the TriNetX platform. The primary end point for the study was all cause 30-day mortality. All analyses were completed within TriNetX Advanced Analytics using the Compare Outcomes functionality. Univariate time-to-event analysis using the Kaplan-Meier analysis option was implemented for each outcome. Hazard ratios (HR) and associated 95% confidence intervals were estimated using the Cox proportional hazards model.

Results:

714,407 patients were identified; 42,717 had SNa above 145 mmol/L. The hypernatremia group were 9.0 years older (p=0.01) and had 5.2 % more males (p<0.01). After 1:1 PSM. 41,580 patients were in the hypernatremia group and 41,958 were in the comparator. All-cause mortality at 30 days was observed in 12.2% of patients in the hypernatremia group but only 1.3% of the comparison group. The largest group was the SNa 146-150 mmol/L patients accounting for 74.1% of the entire population. The risk of death at 30-days increased as the sodium level increased but due to the shrinking population in each group the 95% confidence interval widened. The Kaplan-Meier survival curve suggests that the risk of death in the entire hypernatremia cohort continues to increase throughout the 30-day observation period. The hypernatremia group was divided into 5 quartiles of 5 mmol SNa. The risk of death increased as the SNa increased with HR 9.755 (95% CI 8.934, 10.652) for the entire group.

Conclusion:

This is the first retrospective database review of hypernatremia that excluded patients in whom hypernatremia may have been a therapeutic goal or occurred after hypertonic saline injection. Hypernatremia in inpatients significantly increased the risk of death at 30-days. There is no level of hypernatremia that does not increase the risk of death. The risk of death increases as the SNa increases. In summary, hypernatremia should be avoided. Inpatients with hypernatremia should be recognized and judiciously treated to avoid additional increases in SNa. This study is not without limitations. The use of the TriNetX database provided a large patient population for this study; however, the database has profound limitations that limited the analysis of the results. First, the database only allowed for propensity score matching on three variables, therefore age and sex were the only two selected. Race was not used in propensity scoring matching because the authors did not feel that it was a reliable variable. Additionally, the database only allowed for a 1:1 propensity score matching between the two groups although the comparison group was large enough to have had a 1:2 or 1:3 propensity score comparison. Finally, all analysis in this study were univariate, as the database did not allow for multivariant analysis which limited the precision of the analysis.