Novel investigative tools (e.g. whole genome sequencing) help characterize host and viral genetic contributions to disease severity in pediatric viral respiratory infection. However, a validated scoring system for quantifying illness severity is needed to properly contextualize results. Existing scoring systems are outdated, unvalidated and underpowered. We thus developed a scoring system to address these concerns.
Children hospitalized with viral respiratory infections were prospectively enrolled over 2 years, with 51 clinical variables abstracted from the medical record. 7 variables felt to be most predictive of disease severity and significantly correlated with each other (Spearman correlation coefficient p<0.001) were included in the scoring system (duration of hospital and ICU stay, oxygen and high flow nasal cannula (HFNC) use and intubation; maximum nasal cannula and HFNC support), and combined into a disease severity index by converting each into an ordinal score and summing over the variables, with each variable sub-divided into 7 levels of exposure (based on equal interval length cutpoints).
For a validation comparison, sampling algorithms utilizing a linear model selected a subset of 96 patients whose disease severity would be randomly assessed by 8 pediatricians in blocks of 12, using D-optimality and space filling criteria to protect against non-linearity (severity scored 1 to 10; 80% power for detection of correlation >0.28, two-sided α= 0.05). Mixed model regression analyses compared clinician-scored disease severity with the scoring system. Akaike Information criteria (AIC) and coefficients of determination (R2) ranked severity indices.
445 subjects (56.2% male, median age 1.2 years) were enrolled. Clinician scores of disease severity averaged 6.2 (SD = 2.2, range 1-10). A scoring system using 7 variables with 7 levels of exposure per variable produced the lowest AIC (0.00, R2= 0.70 for predicting clinician-scored disease severity after adjustment for rater effects) (Fig. 1).
A 7-variable scoring system quantifying disease severity in pediatric viral respiratory infections correlates well with clinican assessment, and may advance the study of such infections.
Fig. 1 Fitted Model: Association of Clinician Score and Severity Scoring System
M. Woslager, None
A. Belmonte, None
A. Behnken, None
A. Subbaswamy, None
H. Agarwal, None
T. Vigil, None
F. Caffey, None
M. Muller, None
D. Dinwiddie, None
W. Dehority, None