Pediatric Critical Care Fellow University of Utah University of Utah Health Salt Lake City, Utah, United States
Background: Severe acute malnutrition (SAM) remains a high mortality disease of resource limited settings. Pediatric Early Warning Scores (PEWS) have been shown to accurately predict clinical decompensation in resource rich settings. Low resource settings with high prevalence of mortality and SAM may benefit from PEWS, particularly in settings with high patient to provider ratios. The PEWS for Resource Limited settings (PEWS-RL) was provisionally validated in Rwanda, with a score of ≥3 as a predictor of clinical decompensation. Data reporting the performance of PEWS-RL in resource limited settings is lacking.
Objective: To evaluate the validity of PEWS-RL in a large cohort of patients and compare validity between SAM and non-malnourished patients.
Design/Methods: The PEWS-RL was retrospectively applied to the admission vital signs of all patients age 6 to 59 months admitted to Kamuzu Central Hospital (KCH) in Lilongwe, Malawi in 2019. PEWS-RL includes six equally weighted variables including heart rate, respiratory rate, respiratory distress, need for supplemental oxygen, mental status, and temperature, resulting in a cumulative score of 0 to 6. Mortality was used as a surrogate outcome measure for clinical decompensation. Performance of the PEWS-RL score ≥3 as a predictor of mortality was evaluated in the full cohort and nutrition subgroups using area under the curve (AUC) and its 95% confidence interval (CI) from a univariable logistic regression model. Multiple imputation was used to account for missingness in our analysis.
Results: 8,280 patients met inclusion criteria. Of these 5,838 had non-missing nutrition status information and 4,034 had non-missing values for both nutrition status and PEWS-RL. 881/5,838 (15.1%) children had SAM. Mortality for non-SAM was 2.2% (n=109). Mortality for SAM was 10.0% (n=88). A PEWS-RL score of ≥3 demonstrated an AUC of 0.56 (95% CI: 0.53, 0.59). Subgroup analysis demonstrated AUC for patients with SAM, moderate acute malnutrition (MAM) and without malnutrition of 0.57 (0.52, 0.62), 0.54 (0.45, 0.62), 0.58 (0.53, 0.62) respectively. No statistically significant difference was determined when comparing PEWS-RL performance in patients with SAM to non-malnourished peers. Conclusion(s): We found that PEWS-RL score performs poorly in predicting mortality for all patients, including those with SAM, admitted to a tertiary hospital in Malawi. Future research should focus on identifying variables that predict decompensation in this population.
Table 1. Patient characteristics by malnutrition status
Authors/Institutions: Laura A. Carr, University of Utah Health, Salt Lake City, Utah, United States; Elizabeth M. Keating, University of Utah Health, Salt Lake City, Utah, United States; Guo Wei, University of Utah Health, Salt Lake City, Utah, United States; Angela P. Presson, University of Utah Health, Salt Lake City, Utah, United States; Msandeni E. Chiume, Kamuzu central hospital-Malawi, Lilongwe, , Malawi; Wilfred Gaven, Kamuzu central hospital-Malawi, Lilongwe, , Malawi; Alexander Kondwani, Kamuzu central hospital-Malawi, Lilongwe, , Malawi; Elizabeth C. Fitzgerald, University of North Carolina System, Chapel Hill, North Carolina, United States; Jeff Robison, University of Utah Health, Salt Lake City, Utah, United States; Andrew G. Smith, University of Utah Health, Salt Lake City, Utah, United States