The source population included beneficiaries of the Hospital Italiano de Buenos Aires Care Program (HIBACP), a prepaid health maintenance organization in the city of Buenos Aires, Argentina, that currently provides comprehensive care for 150,000 persons. All study patients received treatment at the Hospital Italiano de Buenos Aires between January 1, 2007, and December 1, 2012. The data sources for the study included information extracted from the institution's electronic health record (EHR) system (including data from a central laboratory), the institutional registry for sepsis (IRS, www.ClinicalTrials.gov Identifier: NCT01403935), and the institutional registry for thromboembolic disease (IRTD www.ClinicalTrials.gov Identifier: NCT01372514). All patients' health care information in the institution is stored in a Clinical Data Repository (CDR), which has been operating for more than 10 years. The CDR has mirrored databases with de-identified information to ensure the privacy and confidentiality of the data. To enable the secondary analysis, the CDR was used as an information source. This repository stores clinical documents from different services such as test results, images, clinical notes, drug prescriptions, pharmacy dispensations, outpatients visits, ER visits, in-hospital care, among other examples. Therefore, the information in this cohort was collected from high-quality secondary health care database systems of the Hospital network, integrated into the Electronic Medical Record (EMR) with a relational base model.

The study was approved by the Hospital Italiano de Buenos Aires' institutional review board and was carried out in compliance with the principles outlined in the Declaration of Helsinki. A waiver of informed consent was obtained due to the nature of the study.

A retrospective cohort study was conducted including all patients older than 18 years of age who underwent surgery for traumatic hip fracture and had ≥1 plasma sodium measured preadmission, at admission, and ≥1 measurement before surgery. All patients included in the study had preadmission plasma sodium determinations. Patients with a traumatic hip fracture were identified using relevant Systematized Nomenclature of Medicine-Clinical Terms (SNOMED-CT) codes from emergency department or hospital primary discharge diagnosis information found in the EHR (7). Each patient's primary treating inpatient physician reviewed relevant medical records, including radiographic imaging and operative reports, in order to confirm that there was a solitary traumatic hip fracture and to exclude pathologic hip fractures, such as bone metastases, or multiple traumatic fractures.

All plasma sodium concentration measurements were performed using the ion-selective electrode method (normal range 135–145 mmol/L). Plasma sodium levels were corrected for plasma glucose using the following formula: corrected plasma sodium (mEq/L) = measured plasma sodium (mEq/L) + 0.016 ^* (plasma glucose [mg/dL]−100). Using available information from a central laboratory as of the date of hospital admission, patients were classified as being normonatremic or hyponatremic, with further subclassification of hyponatremia. Using available preadmission and admission plasma sodium data, patients with plasma sodium fluctuations above or below 135 mmol/L were excluded from the analytic sample to avoid misclassification of natremia status, and those with persistent values >145 mmol/L consistent with hypernatremia were also excluded. Among the remaining patients, to be considered normonatremic, patients had to have all preadmission and admission plasma sodium values between 135 and 145 mmol/L. Patients with hyponatremia were categorized as having either chronic prolonged hyponatremia (defined as having ≥2 consecutive plasma sodium values <135 mmol/L over a period of more than 90 days before admission) or recent hyponatremia (defined as having one plasma sodium determination <135 mmol/L within 30 days before admission with all previous available values being between 135 and 145 mmol/L). Hyponatremia was not corrected, and patients remained hyponatremic during the entire follow-up period after surgery.

Preadmission patient characteristics that could be associated with increased in-hospital morality included age, gender, body mass index, ischemic heart disease, chronic heart failure, dementia, liver failure, gait disorders, and chronic kidney disease (CKD, defined as an estimated glomerular filtration rate <60 ml/min/1.73 m², using the Modification of Diet in Renal Disease Study equation). Clinical risk factors for fractures present in the FRAX algorithm were included. The fracture risk algorithm (FRAX) was developed by the World Health Organization to calculate the 10-year probability of a hip fracture and the 10-year probability of any major osteoporotic fracture (defined as clinical spine, hip, forearm, or humerus fracture) in a given patient. These calculations account for femoral neck bone mineral density (BMD) and other clinical risk factors, as follows: age, sex, body mass index, personal history of fracture, use of oral glucocorticoid therapy, secondary osteoporosis (e.g., coexistence of rheumatoid arthritis), parental history of hip fracture, current smoking status, and alcohol intake (three or more drinks per day). The World Health Organization risk assessment tool is available at http://www.shef.ac.uk/FRAX.

Preadmission medication use that may be associated with hyponatremia included chronic use of antidepressant medications (selective serotonin reuptake inhibitors, tricyclic antidepressants), anticonvulsants, thiazide diuretics, and proton pump inhibitors and corticosteroid use based on information from admission notes and medication lists.

Follow-up occurred through December 31, 2012. Outcomes of interest included both in-hospital and post-discharge outcomes. Post-operative complications (within 30 days after hip surgery) that were evaluated included the development of acute myocardial infarction, heart failure, significant atrial or ventricular arrhythmias, stroke, venous thromboembolism (deep venous thrombosis or pulmonary embolism), and sepsis. In-hospital mortality, length of hospital stay (in days), hospital readmission for any cause within 30 days after discharge, and death from any cause during follow-up were evaluated using comprehensive data from the EHR.

Statistical analyses were performed using SPSS 19.0 statistical software, and a two-sided P <0.05 was considered significant in all analyses. Continuous variables were reported as either mean and standard deviation (SD) or median and interquartile range (IQR) if the variable was non-normally distributed. Categorical variables were reported as frequencies and proportions. Pairwise comparisons were made between groups for continuous variables using a t-test or the Kruskal–Wallis test, as appropriate, and for categorical covariables using a chi-square test or Fisher's exact test. The proportion of patients experiencing in-hospital death and readmission within 30 days post-discharge across groups was compared using chi-square tests. Index hospitalization length of stay was compared across groups using a t-test. Incidence of all-cause death was compared across groups using Kaplan–Meier survival curves and a log-rank test.

To evaluate the independent association between natremia status and clinical outcomes, a propensity score was generated to predict the probability of presenting with any hyponatremia by considering as candidate variables the baseline patient characteristics and medications shown in Table 1. The final propensity score logistic regression model had a c statistic of 0.72.

A multivariable logistic regression was performed for the dichotomous outcomes of in-hospital death and readmission for any cause within 30 days post-discharge. A multivariable linear regression was performed for index hospitalization length of stay. A multivariable Cox proportional hazard model was performed for the outcome of time to death from any cause. All multivariable models were adjusted for the propensity score for presenting with hyponatremia as a continuous variable in addition to directly adjusting for age, gender, body mass index, preadmission diagnosed dementia, prior heart failure, preexisting CKD, and risk factors for hip fracture, given that these variables may have differential prognostic information on the outcomes of interest. For the outcome of death from any cause, an additional model was conducted that included post-operative complications which may mediate, at least in part, any observed association between hyponatremia and mortality. The follow-up period was up to 5 years.

Between 2007 and 2012, 1,571 eligible adults were admitted for hip fracture repair, with 76.7% presenting with normonatremia (median plasma sodium 138 mmol/L), 14.2% presenting with chronic prolonged hyponatremia (median plasma sodium 131 mmol/L), and 9.1% presenting with recent hyponatremia (median plasma sodium 133 mmol/L) (Figure 1). Of note, among the 222 patients presenting with chronic prolonged hyponatremia, the estimated mean (SD) time with hyponatremia was 1,027 (649) days and the median (interquartile) time of hyponatremia was 932 days (445.25–1,484) before admission.

Compared with patients who were normonatremic, those presenting with chronic prolonged hyponatremia were older and were more likely to have prior heart failure, a history of alcoholism, preadmission anticonvulsant drug use, and a lower body mass index (Table 1). Compared with those who were normonatremic on admission, those presenting with recent hyponatremia were less likely to have preadmission anticonvulsant drug use and also had a lower body mass index (Table 1). Of note, there was no significant difference in the median time between admission and surgery in those who presented with normonatremia (1 day), chronic prolonged hyponatremia (1 day, P = 0.86), or recent hyponatremia (1 day, P = 0.78).

Compared to patients presenting with normonatremia (median length of stay = 7 days), the unadjusted median length of hospital stay was 8 days (P = 0.15) in those presenting with chronic prolonged hyponatremia and 8 days (P = 0.78) in those with recent hyponatremia. After adjustment for the propensity to present with any hyponatremia and other potentially confounding variables, no differences were found in length of stay with chronic prolonged hyponatremia (P = 0.97) nor in recent hyponatremia (P = 0.93) in the length of stay.

Crude risk of in-hospital death was 14.7% in patients presenting with normonatremia compared with 24.3% (P < 0.01) in those with chronic prolonged hyponatremia and 13.2% (P = 0.63) in patients with recent hyponatremia. In multivariable analyses, compared to those presenting with normonatremia, having chronic prolonged hyponatremia or recent hyponatremia was not significantly associated with in-hospital death (Table 2).

The frequencies of post-operative complications within the first 30 days among groups are shown in Table 2. In multivariable analyses adjusting for the likelihood to present with any hyponatremia in addition to age, gender, and Charlson comorbidity score, chronic prolonged hyponatremia was not significantly associated with any of the post-operative complications, while recent hyponatremia was only associated with higher odds of post-operative sepsis (adjusted OR 1.84, 95% CI: 1.01–3.35) (Table 2).

The crude risk of readmission for any cause at 30 days post-discharge was higher in patients presenting with chronic prolonged hyponatremia (18%, P = 0.04) but not recent hyponatremia (12.5%, P = 0.90) compared to patients with normonatremia (12.9%). After adjustment for potential confounders, there was no significant association with 30-day readmission for chronic prolonged hyponatremia (adjusted OR 1.29, 95% CI: 0.86–1.60) or recent hyponatremia (adjusted OR 0.93, 95% CI: 0.54–1.60).

Unadjusted survival was lower in patients presenting with chronic prolonged hyponatremia (P < 0.01) or recent hyponatremia (P = 0.046) (Figure 2). In Cox proportional hazards regression that adjusted for the likelihood of presenting with hyponatremia, as well as age, gender, Charlson comorbidity score, baseline medication use, and occurrence of post-operative complications, patients presenting chronic prolonged hyponatremia had a 53% significantly higher adjusted rate of death from any cause (adjusted hazard ratio 1.53, 95% CI: 1.12–2.09), while there was no significant association for recent hyponatremia (adjusted hazard ratio 1.29, 95% CI: 0.86–1.93). The mean Charlson comorbidity score in patients with 30-day in-hospital mortality or post-operative complications was 4 ± 2 SD.