Effect of antibiotics in preventing hospitalizations from respiratory tract infections in children with Down syndrome

Children with Down syndrome (DS) are at high risk of respiratory tract infections (RTIs) due to anatomical variations, comorbidities, and immune system immaturity. Evidence on interventions to reduce this risk is incomplete. This study aims to quantify the effect of antibiotics prescribed for RTIs in primary care on the subsequent risk of RTI‐related hospitalization for children with DS versus controls.


| INTRODUCTION
With an incidence of 1 in 1000 live births, and prevalence of 6.3 in 10,000 people, Down syndrome (DS) is one of the most common genetic conditions in the UK. 1,2 As of 2011, an estimated 37,090 people were living with DS in England and Wales, of whom approximately 10,438 were aged 0-18. 3 It has been projected that the number of children with DS is increasing in the UK, with one analysis projecting 11,592 children 0-15 to be living in England and Wales by 2020. 4 In addition, life expectancy of people with DS has doubled in the past six decades, increasing from 30 to 60 years, 5,6 alongside advances in medical and surgical treatment, improved social inclusion and support, and general quality of life. 5,6 Children with DS are thought to be at an increased risk of frequent and severe respiratory tract infections (RTIs) due to anatomical variations (such as a narrow upper airway), complications from comorbidities (including congenital heart disease and reflux), and immune system immaturity. 7,8 A study of 22 children with DS age-matched to 22 healthy siblings found children with DS had a significantly higher frequency of lower RTIs (LRTIs) compared to their siblings alongside observed immune parameter differences. 7 These respiratory infections account for a large proportion of healthcare utilization in children with DS, especially at younger ages. 9 In an Australian study of 3786 hospitalizations in 405 children with DS, 26.7% of all admissions were due to RTIs, with 52.6% of all children with DS experiencing any hospital admission due to an RTI. 8 When compared to published admission rates for the pediatric population in Western Australia, significant differences were noted-the rate ratio for hospitalizations with respiratory system-related diagnoses in children with DS was 17.9. In particular, LRTIs result in longer length of stay and a high proportion of intensive care unit admissions (43% in one study) in children with DS. 10 Despite this perceived risk, there is remarkably little evidence on interventions to reduce RTIs, with most trials for RTI treatments worldwide being less open to children and adults with DS than those without DS. There are many implicit barriers to research for individuals with learning disabilities, 11,12 and barriers may also be explicit in the form of exclusion criteria. A 2019 review of over 26,000 studies in the NIHR portfolio found that 60.3% of all studies excluded learning disability groups, all studies investigating pneumonia excluded learning disability groups, and only 1.4% of studies were specifically targeted towards individuals with learning disabilities. 13 Recent research on preventative and therapeutic interventions for RTIs in adults and children with DS is limited; a 2015 systematic review identified only five studies, from a search of 13,575 records, none of which focused on the effect of antibiotics. 14 The impact of antibiotics prescribed for RTIs in primary care has been previously quantified for the general UK population, 15 but given the differing prevalence and severity of RTIs in children with DS, the lack of evidence for the role of antibiotics in treating children with DS and RTIs therefore represents an important research gap.
Evidence in this area would be beneficial for patients, their families, and healthcare professionals (HCPs) to guide appropriate, timely and personalized treatment of RTIs in children with DS.
The present study utilizes routinely collected primary and secondary UK National Health Service health care data to address this study gap, estimating the effect of antibiotic prescription following RTI-related General Practitioner (GP) consultations in terms of reducing RTI-related hospitalizations in children with DS and controls.

| Participants
CALIBER was searched between January 1st, 1997 and March 25th, 2010 for all adults and children with DS, as identified by all Read and ICD-10 codes related to DS in CPRD and HES. Individuals with an exit date before their entry date were removed, due to suspected data quality issues. For each remaining individual with DS, five controls without DS were frequency matched by GP, gender, birth year (±5 years), and start of follow-up. Those who were over 18 years old at the entry date were excluded after this point.

| Definitions
Hospitalization rates were acquired from HES and prescription and consultation rates from CPRD, within CALIBER. Hospitalizations were defined according to HES coding of an event as a "hospital admission," which were then coded by diagnosis in line with the ICD-10 coding system. Length of stay was calculated as "discharge date − admission date + 1 day," and all admissions including those lasting only 1 day were included. When calculating baseline risk of hospitalization from a consultation, all GP consultations for RTIs were the exposure and all RTI-related hospitalizations were the outcome. Each RTI-related GP consultation was followed up for 28 days or until the first RTI-related hospitalization within that time period. Rates of hospitalization were calculated by dividing the number of episodes of hospitalization in 28 days by the total number of person-years in this time.
A ranking system was used based on RTI-type (LRTI > URTI > unclassified), setting (secondary > primary care) and whether it was probable or possible (probable > possible) in the event of multiple RTI events being noted on the same day for the same patient (e.g., a probable URTI consultation and a possible LRTI hospitalization).

| Sample size
The Fitzgerald et al. 8 Australian study of hospitalizations for children with DS was utilized to inform our sample size calculation. They found an average of 0.8 and 0.1 RTI-attributable hospital admissions in children with and without DS, respectively (with numbers representing the proportion of each population with an outcome of interest (e.g., a record of hospitalization). 8 To calculate our sample size, with the assumption UK hospitalization rates were similar, we estimated that at least 20 individuals per group were required to identify this difference in hospitalization rates between children with DS and controls at 80% power using a significance level of 0.05. The number of individuals required increases by 10% for each variable considered for confounding.

| Statistical models
Univariate logistic regression was undertaken as the initial model to assess the effect of antibiotic prescriptions on the risk of subsequent RTI-related hospitalization in patients consulting for RTIs.
Multivariate logistic regression was undertaken in the final model.
The covariates included in the final model (entered in a single step) were antibiotic prescription, age group, gender, presence of congenital heart disease, presence of asthma, and number of prior RTI-related hospitalizations and RTI-related GP consultations in the preceding 6 months. Comorbidities were identified if a relevant Read (i.e., GP consultation) or ICD-10 (i.e., hospitalization) code was recorded for a child at any point between their CALIBER entry and exit dates. Two further covariates were considered, but ultimately were excluded from the final model because there was insufficient power to detect difference due to the ratio of covariates to outcomes. These were the 28-day RTI-related Consultation Average, and the 28-day RTI-related Hospitalization Average.
Subgroup analysis was performed to assess the effect of antibiotics across age groups and RTI types. Where there was a reduction of sample size in subgroup analysis, post hoc power calculations were conducted to assess the risk of type II error. 20 Where a significant protective effect of antibiotic prescription was seen, the number needed to treat (NNT) was estimated. All data management and analyses were performed using STATA statistical software version 13 and R version 3.2.3 via the UCL Data Safe Haven.

| Study registration and ethics
The protocol for this study was approved by the CPRD independent scientific advisory committee, reference number 15_041R. Over the study period, there were 2874 RTI-related consultations in which children with DS were prescribed antibiotics, and 1811 where they were not (see Table 2).

| Prescription of antibiotics following an RTI-related GP consultation and risk of subsequent hospitalization
In the overall population of children with DS, without adjustment for covariates, the prescription of antibiotics following an RTI-related GP consultation did not show any significant evidence for reduction in the risk of RTI-related hospitalization in the subsequent 28 days Among subgroups, without adjustment for prespecified covariates, antibiotics did not provide a significant protective effect against hospitalization across any type of RTI and any age group for children with DS or controls (see Table 3). After adjustment for all seven covariates (see Section 2.6), antibiotics did not protect against RTI-related hospitalization for the overall group of children with DS or for controls. However, subgroup analysis by age after adjustment for covariates revealed a protective effect for infants (0-1 years; see Table 4). RTIrelated hospitalizations for infants with DS were reduced when antibiotics were prescribed (adjusted odds ratio, 0.260 (95% CI, 0.077-0.876); NNT, 11.9 (95% CI 6.0-1708.7)). There was no protective effect at other ages, or by type of RTI, or in controls (see Table 4).
Post hoc power calculations were conducted across RTI subgroups for both children with DS and controls. Study power was higher for analyses in children with DS (range, 11.4%-37.3%) than in controls (range, 7%-21.2%), due to the relative infrequency of hospitalization in controls (see power calculations in Table S2).

| DISCUSSION
In