Abnormal Sodium Before Surgery Flags Risk in Older Patients but Adds Little Predictive Power

A routine blood test that costs pennies and is performed on virtually every patient entering an operating room may carry far more information than clinicians currently extract from it. Serum sodium, the most common electrolyte measured in preoperative screening, has long been suspected of signaling trouble after surgery, particularly in older adults. Now one of the largest prospective studies of its kind has put that suspicion to a rigorous statistical test, and the results are a masterclass in the difference between a risk factor and a useful predictor. The finding, drawn from the European BioCog cohort, is likely to reshape how perioperative physicians think about the humble sodium value sitting at the top of every preoperative lab report.

The study, published in European Geriatric Medicine by a team led by Marinus Fislage of Charité – University Medical Center Berlin, examined 618 patients aged 65 and older who were scheduled for elective surgery expected to last at least 60 minutes. None of the participants had dementia, and all had routinely obtained preoperative sodium measurements available. Using standard clinical thresholds, the researchers classified 31 patients, or 5.0 percent of the cohort, as hyponatremic, with sodium values ranging from 125 to 134 mmol/L, and 16 patients, or 2.6 percent, as hypernatremic, with values between 146 and 150 mmol/L. In total, only 7.5 percent of these older surgical patients carried any sodium abnormality before their operation, a figure that itself is informative about how uncommon clinically relevant dysnatremia is in an elective surgical population.

The raw numbers, however, were striking. Patients with abnormal sodium levels fared dramatically worse across every outcome the investigators tracked. Postoperative delirium, an acute and often devastating disturbance of attention and cognition that affects roughly one in five older surgical patients, occurred in 38.3 percent of the dysnatremia group compared with 20.8 percent of those with normal sodium. In-hospital mortality was more than tenfold higher, at 17.0 percent versus 1.6 percent. By 90 days after surgery, 27.7 percent of patients with abnormal sodium had died, compared with just 2.8 percent of the normonatremic group. Severe complications, graded as Clavien-Dindo class II or higher, affected 65.2 percent of the dysnatremia group against 38.4 percent of the rest. On the surface, these are the kinds of associations that make headlines and prompt calls for routine sodium-targeted interventions.

But the BioCog team asked a more demanding question than whether abnormal sodium and bad outcomes travel together. They wanted to know whether adding sodium to prediction models that already contain established clinical risk factors actually improves the models’ ability to distinguish which individual patients will suffer delirium, complications, or death. This distinction matters enormously in clinical practice. A variable can be strongly associated with an outcome in aggregate yet add almost nothing to individual-level prediction if its information is already captured by other variables, or if the abnormality is simply too rare to move the needle for the population as a whole. Confusing these two properties, association and incremental predictive value, is one of the most common pitfalls in biomarker research, and the study was explicitly designed to avoid it.

The methodological machinery the researchers deployed reflects that ambition. For each outcome, they built multivariable prediction models incorporating established clinical risk factors and then measured what happened when abnormal sodium was added. Discrimination, the model’s ability to separate patients who experience an outcome from those who do not, was quantified using the area under the receiver-operating characteristic curve, or AUROC, with bootstrap correction to guard against over-optimism. Calibration, whether predicted probabilities match observed event rates, was assessed with Brier scores, which penalize both discrimination failures and miscalibration. Postoperative delirium, the primary outcome, was assessed not by impression but with a battery of four validated diagnostic tests, a rigor that distinguishes BioCog from many chart-based delirium studies.

The verdict on delirium and complications was sobering. Adding dysnatremia to the delirium model nudged the AUROC from 0.71 to 0.72, but the bootstrap-corrected difference was a mere 0.0045, with a 95 percent confidence interval stretching from −0.0027 to 0.0199, a range that includes zero. For postoperative complications, the picture was nearly identical, with an optimism-corrected gain of 0.004 and a confidence interval of −0.0011 to 0.0124. Brier scores showed only marginal differences for these outcomes as well. In plain terms, knowing a patient’s sodium status before surgery tells a clinician almost nothing about that individual’s risk of delirium or complications beyond what age, comorbidity, type of surgery, and other established factors already reveal. The 38 percent delirium rate among dysnatremic patients is alarming, but with only 47 such patients in the cohort, the signal is too sparse and too entangled with other risk markers to sharpen anyone’s individual forecast.

Mortality told a subtler story. For in-hospital death, incorporating dysnatremia raised the optimism-corrected AUROC from 0.864 to 0.889, a difference of 0.023 with a confidence interval of 0.0046 to 0.0435 that excludes zero. For 90-day mortality, the bootstrap-validated C-index climbed from 0.91 to 0.93, a median increase of 0.021 with a confidence interval of 0.001 to 0.057. These are small but statistically reliable gains, and the authors are careful to frame them as exploratory findings that warrant further study rather than as a mandate for clinical change. Even here, the Brier scores barely moved, suggesting that while sodium helps rank patients by mortality risk, it does little to improve the accuracy of the predicted probabilities themselves. A gain of roughly two hundredths of a C-index, in the framework of decision-analytic literature on new markers, rarely justifies changes in care pathways on its own.

Why does a marker so visibly linked to death and delirium fail as a predictor? The answer lies in biology and in arithmetic. Dysnatremia in older adults is rarely an isolated electrolyte quirk; it is typically the downstream consequence of heart failure, kidney disease, malignancy, diuretic use, or general frailty, all of which are already captured by the covariates in a well-constructed prediction model. Sodium, in this sense, behaves as a messenger summarizing information the model already possesses. Meanwhile, its rarity, affecting fewer than one in twelve patients in this elective cohort, mathematically caps how much any population-level discrimination metric can improve. Prior large studies, including an analysis of major surgical populations by Cecconi and colleagues and a Danish cohort of nearly 280,000 acutely hospitalized patients, have consistently shown that even mild hyponatremia carries elevated mortality risk, yet the BioCog results suggest that in elective surgery the marker’s predictive ceiling is low precisely because the sickest patients are filtered out before they reach the operating room.

The clinical implications cut in two directions. On one hand, the study argues against investing heavily in sodium-based risk stratification tools for delirium prevention or complication prediction in older elective surgical patients; existing algorithms, including one recently developed and internally validated from the same BioCog cohort and published in the British Journal of Anaesthesia, already perform well without it. On the other hand, the strong associations with mortality serve as a reminder that a preoperative sodium value outside the 135 to 145 mmol/L range should never be dismissed as an incidental finding. It may not refine a risk calculator, but it flags a patient whose physiology is already compromised, prompting clinicians to consider whether the underlying cause, whether medication-related, endocrine, or organ-specific, can be identified and corrected before surgery. Trials of perioperative fluid management and prehabilitation programs, both active areas of research, may eventually determine whether intervening on modifiable contributors to dysnatremia can convert this marker of risk into a target of therapy.

For the growing field of perioperative medicine, the BioCog analysis offers a methodological template as much as a clinical message. By formally quantifying incremental predictive value with bootstrap-corrected discrimination metrics and Brier scores, rather than resting on relative risks alone, the researchers demonstrate how the surgical community should evaluate any candidate biomarker, from inflammatory panels to genetic signatures, before it earns a place in preoperative workflows. The study was conducted under the EU-funded BioCog project, registered as NCT02265263, and approved by the local ethics committee with written informed consent from all participants. As populations age and surgical volumes rise, the pressure to predict, and prevent, postoperative delirium and death will only intensify. This study tempers the enthusiasm for one tempting candidate while modeling the discipline required to find predictors that genuinely work.

Subject of Research: Preoperative serum sodium abnormalities and their association with and predictive value for adverse postoperative outcomes in older surgical patients

Article Title: Preoperative sodium abnormalities and adverse postoperative outcomes in older surgical patients: association and incremental predictive value in the BioCog cohort

Article References: Fislage, M., Feinkohl, I., Lammers-Lietz, F., Pischon, T., Winterer, G., Spies, C. D., Borchers, F., Zacharias, N., the BioCog Consortium, Wolf, A., Müller, A., Hadzidiakos, D., Yürek, F., Lachmann, G., Ofosu, K., Heinrich, M., Mörgeli, R., Gallinat, J., Kühn, S., … Fillmer, A. (2026). Preoperative sodium abnormalities and adverse postoperative outcomes in older surgical patients: association and incremental predictive value in the BioCog cohort. European Geriatric Medicine. https://doi.org/10.1007/s41999-026-01610-y

Image Credits: AI Generated

DOI: 10.1007/s41999-026-01610-y

Keywords: preoperative sodium, hyponatremia, hypernatremia, postoperative delirium, postoperative complications, mortality, older surgical patients, BioCog cohort, risk prediction, perioperative medicine, geriatric surgery, electrolyte disorders