Exploring respiratory tract infections in acute Irish hospitals (2016–2021)

Abstract

Introduction

Respiratory tract infections (RTIs) are a leading cause of morbidity and mortality. This study aimed to explore the underlying characteristics of inpatients with an RTI pre-and-during the COVID-19 pandemic and apply supervised machine learning (ML) to predict the RTI diagnoses during the pandemic.

Methods

Data on inpatients with an RTI from 55 acute Irish hospitals (2016–2021) was extracted from the Irish hospital inpatient enquiry (HIPE) dataset. Multivariable logistic regression, random forests and extreme gradient boosting models were applied.

Results

Out of 1,133,385 inpatients with an infection, 43.3 % had an RTI. Of which 65.2 % were before the pandemic and 34.8 % during the pandemic. In comparison to pre-pandemic, the median hospital length of stay (LOS) increased from 4 days to 5 (other RTIs) and 6 days (COVID-19). Deaths associated with an RTI increased from 3.3 % (pre-pandemic) to 7.6 % (during pandemic) (COVID-19: 11.3 %, and other RTIs: 3.5 %) and 5.2 % of COVID-19 infections were hospital acquired. Furthermore, inpatients with COVID-19 were generally younger than inpatients with other types of RTIs (75 % over the age of 45). Applying ML showed that a COVID-19 diagnosis was significantly associated with shortness of breath, cough, fever, chest pain, nausea and vomiting, malaise and fatigue, obesity, systemic inflammatory response syndrome, bradycardia, kidney diseases, and tachycardia. In addition, wheezing, smoking and heart diseases distinguished an RTI infection from COVID-19.

Conclusions

During the pandemic, inpatients with an RTI infection had longer LOS and higher mortality compared to before the pandemic. Supervised ML is helpful in predicting an RTI diagnosis, with wheezing, smoking and heart diseases being the main discriminating factors between other types of RTIs from COVID-19 diagnoses.