Peripheral Perfusion Index in Ugandan Children With Plasmodium falciparum Severe Malaria: Secondary Analysis of Outcomes in a 2014-2017 Cohort Study

Abstract

Objectives: Continuous, noninvasive tools to monitor peripheral perfusion, such as perfusion index (PI), can detect hemodynamic abnormalities and assist in the management of critically ill children hospitalized with severe malaria. In this study of hospitalized children with severe malaria, we aimed to assess whether PI correlates with clinical markers of perfusion and to determine whether combining PI with these clinical measures improves identification of children with greater odds of mortality.

Design: Post hoc analysis of a prospective, multicenter, cohort study conducted between 2014 and 2017.

Setting: Two referral hospitals in Central and Eastern Uganda.

Patients: Six hundred children younger than 5 years old with severe malaria and 120 asymptomatic community children.

Interventions: None.

Measurements and main results: PI was measured at 6-hour intervals for the first 24 hours of hospitalization. We compared PI to standard clinical perfusion measures such as capillary refill time, presence of cold peripheral limbs, or temperature gradient. Admission PI was highly correlated with clinical measures of perfusion. Admission PI was lower in children with severe malaria compared with asymptomatic community children; and, among the children with severe malaria, PI was lower in those with clinical features of poor perfusion or complications of severe malaria, such as shock and hyperlactatemia (all p < 0.02). Among children with severe malaria, lower admission PI was associated with greater odds of mortality after adjustment for age, sex, and severe malaria criteria (adjusted odds ratio, 2.4 for each log decrease in PI [95% CI, 1.0-5.9]; p = 0.045). Diagnostically, the presence of two consecutive low PI measures (< 1%) predicted mortality, with a sensitivity of 50% and a specificity of 76%.

Conclusions: In severe malaria, PI correlates with clinical complications (including shock and elevated serum lactate) and may be useful as an objective, continuous explanatory variable associated with greater odds of later in-hospital mortality.

Figure 1.. Study flow diagram.

Low PI indicates a perfusion index of less than 1%. Time points (hours) indicate the intervals at which perfusion index follow-up measurements were taken and included in the analysis.

Figure 2.. Percent perfusion index (PI%) at A) admission in children with severe malaria (SM) compared to community controls (CC) and at B) follow-up time points during hospitalization in children with SM compared to baseline CC.

A) Bars indicate median and interquartile range. Circles represent PI% in individual children. Dark gray shaded area is indicating PI <1% (5th percentile of CC group), and percentages are the percent of children under 1% for each group. Wilcoxon rank-sum test used to compare PI in SM vs. CC. B) Boxes indicate median and interquartile range. Black bars represent upper and lower fence (3rd quartile + [1.5*IQR] and 1st quartile – [1.5*IQR], respectively). Outliers are shown as individual points above the upper fence. The dotted line is the median PI% in CC at baseline, which differed from SM at each time point (P<0.001). Wilcoxon rank-sum test used to compare PI in SM vs. CC. Wilcoxon signed-rank test used to compare PI in SM at different time points; PI in the SM group at baseline was significantly lower than each time point, P<0.001.

Figure 3.. Admission perfusion index (%) in children with severe malaria who died in hospital compared to those who survived (A) and perfusion index at 6-hour intervals post-admission in children who died vs. survived (B).

Circles represent perfusion index in individual children and bars show median and interquartile range. Logistic regression models adjusted for age, sex, and disease group; †P<0.10, *P<0.05, **P<0.01, ***P<0.001; log₁₀-transformed perfusion index used in logistic regression models. Percentages in A are % with perfusion index <1.

Figure 4.. Association between clinical or laboratory characteristics and perfusion index in children with severe malaria.

Gray background circles and error bars represent the β and 95% confidence intervals from univariable unadjusted linear regression models, where the dependent variable is log₁₀-transformed perfusion index %. The black circles and error bars are β and 95% confidence from a multivariable model that includes the measures that were significant in univariable models (P<0.05, bolded on y-axis). An Asterix indicates significant association with perfusion index in the multivariable model (P<0.05). Abbreviations: HRP-2, histidine-rich protein 2; AKI, acute kidney injury; LDH, lactate dehydrogenase