AbstractBackground: Hyponatremia is the most common electrolyte disturbance in heart failure patients. It significantly affects morbidity and mortality, also increasing expenditure in heart failure patients. It is rarely recognised or treated sufficiently due to inadequate diagnostic measurements and doubts concerning the effectiveness of treatment. An important step to proper management is to recognise patients at high risk.
Aims: To develop a reliable prediction model for the risk of hyponatremia in patients hospitalised with heart failure.
Design and methods: Information was retrieved manually from medical records of patients, and hyponatremia was defined as sodium level <135 mmol/L. A nested case-control design was applied and logistic regression analysis was performed to derive the prediction model. Purposeful selection was used to select predictors and statistical analysis was performed using SPSS™ IBM(R) and the R software.
Results: From 464 patients included, 102 patients (22%) were hyponatremic during hospitalisation. Hyponatremia was significantly associated with longer stay (OR = 2.1, 95% confidence interval [1.3–3.3]) and higher mortality rate (3.4, [1.8–6.4]). Other than fluid restriction and diuretics, sodium chloride based therapies were used in the research site, but >50% of patients with severe hyponatremia did not receive active treatment. Six variables were significantly associated with increased risk: serum sodium at admission (0.77, [0.72–0.83]), fatigue (3.71, [1.99–6.9]), ascites (3.73, [1.55–8.99]), inotropes (2.95, [1.38–6.34]), heparin (2.98, [1.33–6.66]) and antibiotics (2.87, [1.56–5.29]). These were included in the prediction model with good predictive ability both overall (Brier-score = 0.107, NR2 = 0.531).The prediction model was then presented in a regression formula format as Hyponatremia = 128.1 – Sodium + 5.2 Fatigue + 5.2 Ascites + 4.3 Positive inotropes + 4.3 Heparin + 4.2 Antibiotics.
Conclusion: A risk-prediction model to stratify the risk for developing hyponatremia during hospitalisation was derived by including predictors selected from patient- and medication-related factors identified. The prediction model exhibits good predictive performance indicating that it can be practically used.
Secondary Supervisor : Akhmad Fauzy
|Date of Award||May 2016|
|Supervisor||Patrick Ball (Supervisor), Hana Morrissey (Supervisor) & Akhmad Fauzy (Supervisor)|