Limitations

This study has several limitations. Firstly, the retrospective design of this study may have induced bias. Also, the multivariate logistic regression model for LOFR almost reached statistical significance for all the factors that reached statistical significance in the univariate analyses. This suggests that our results could have been different with a larger sample size.
Secondly, some variables of interest may be absent or poorly reported in the medical records, particularly some features related to the infant’s environment (parental smoking, familial history of atopy). Moreover, other variables of interest could be influenced by the pediatric services’ usual practices, such as the prescriptions of bronchodilators. Also, socioeconomic factors were not retrieved in our analysis, because these variables were not reported in the medical records. Still, common socioeconomic criteria, namely the level of education, the occupation, the age, and the average parental income, have been shown to be associated with hospital referral and LOS, and could explain why 9% of the children in our cohort were admitted with only a mild bronchiolitis27,28.
Thirdly, the clinical severity scale proposed by the HAS has not been validated yet, and this study was the first to compare its results with LOS. Furthermore, the clinical severity was retrospectively rated, which is not the initial intended use of this scale and could have led to some inaccuracies. Nonetheless, some criteria from this scale (namely the assessment of respiratory distress) could be highly subjective from one caregiver to another. Since this sole criterion is sufficient to rate the clinical severity as ”severe”, the complete reliability of this tool should be investigated in the future. In our study, 52% of infants had a ”severe” clinical severity on admission, mainly because of a RR ”≥ 70/min or < 30/min, or presence of apnea” or an altered general condition.
Lastly, the cut-off value used in the present study to describe adequate infant feeding was decided in the absence of a validated standard but could have influenced our results.