Article Text

other Versions

Prevalence of invasive bacterial infection in febrile infants ≤90 days with a COVID-19 positive test: a systematic review and meta-analysis
  1. Silvia Pérez-Porra1,
  2. Elena Granda1,2,
  3. Helvia Benito3,4,
  4. Damian Roland5,6,
  5. Borja Gomez7,
  6. Roberto Velasco8,9
  1. 1Pediatrics Department, Hospital Universitario Rio Hortega, Valladolid, Spain
  2. 2Pediatrics Department, Hospital Universitario de Burgos, Burgos, Spain
  3. 3Gerencia de Atención Primaria de Segovia, Segovia, Spain
  4. 4CAP Concòrdia. Consorci Corporació Sanitària Parc Tauli, Sabadell, Barcelona, Spain
  5. 5SAPPHIRE Group, Health Sciences, University of Leicester, Leicester, UK
  6. 6Paediatric Emergency Medicine Leicester Academic (PEMLA) Group, Children's Emergency Department, Leicester Royal Infirmary, Leicester, UK
  7. 7Pediatric Emergency Department, Biocruces Bizkaia Health Research Institute, Hospital Universitario de Cruces. University of the Basque Country, UPV/EHU, Barakaldo, Bilbao, Basque Country, Spain
  8. 8Pediatric Emergency Unit, Department of Pediatrics, Hospital Universitari Parc Tauli, Sabadell, Barcelona, Spain
  9. 9Department of Paediatrics & Child Health, University College Cork, Cork, Ireland
  1. Correspondence to Dr Elena Granda, Pediatrics Department, Hospital Universitario Rio Hortega, Calle Dulzaina 2, 47012, Valladolid, Spain; e_granda15{at}


Background Febrile infants with an infection by influenza or enterovirus are at low risk of invasive bacterial infection (IBI).

Objective To determine the prevalence of IBI among febrile infants ≤90 days old with a positive COVID-19 test.

Methods MEDLINE, Embase, Cochrane Central Register databases, Web of Science, and grey literature were searched for articles published from February 2020 to May 2023. Inclusion criteria: researches reporting on infants ≤90 days of age with fever and a positive test for SARS-CoV-2 (antigen test/PCR). Case reports with <3 patients, articles written in a language other than English, French or Spanish, editorials and other narrative studies were excluded. Preferred Reposting Items for Systematic Reviews and Meta-analysis guidelines were followed, and the National Institutes of Health Quality Assessment Tool was used to assess study quality. The main outcome was the prevalence of IBI (a pathogen bacterium identified in blood and/or cerebrospinal fluid (CSF)). Forest plots of prevalence estimates were constructed for each study. Heterogeneity was assessed and data were pooled by meta-analysis using a random effects model. A fixed continuity correction of 0.01 was added when a study had zero events.

Results From the 1023 studies and 3 databases provided by the literature search, 33 were included in the meta-analysis, reporting 3943 febrile infants with a COVID-19 positive test and blood or CSF culture obtained. The pooled prevalence of IBI was 0.14% (95% CI, 0.02% to 0.27%). By age, the prevalence of IBI was 0.56% (95% CI, 0.0% to 1.27%) in those 0–21 days old, 0.53% (95% CI, 0.0% to 1.22%) in those 22–28 days old and 0.11% (95% CI, 0.0% to 0.24%) in those 29–60 days old.

Conclusion COVID-19-positive febrile infants ≤90 days old are at low risk of IBI, especially infants >28 days old, suggesting this subgroup of patients can be managed without blood tests.

PROSPERO registration number CRD42022356507.

  • pediatric emergency medicine
  • bacterial
  • COVID-19
  • infections

Data availability statement

Data are available upon reasonable request.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Febrile infants with an infection caused by influenza or enterovirus are at low risk of having a concomitant invasive bacterial infection (IBI).


  • This systematic review and meta-analysis including 3943 COVID-19-positive febrile infants ≤90 days of age shows that they are at low risk of IBI, with a pooled prevalence of IBI of 0.14% (95% CI, 0.02% to 0.27%). By age groups, pooled prevalence of IBI was 0.56% (95% CI, 0.0% to 1.27%) in those 0–21 days of age, 0.53% (95% CI, 0.0% to 1.22%) in those 22–28 days of age and 0.11% (95% CI, 0.0% to 0.24%) in those 29–60 days of age.


  • Febrile infants older than 28 days of life with a COVID-19 positive test can possibly be managed without blood testing.


Approximately 10%–20% of febrile infants under 3 months of age who arrive at the ED have a bacterial infection,1 2 with urinary tract infection (UTI) being the most common. The reported prevalence of invasive bacterial infection (IBI), considered as bacteraemia or meningitis, is 2%–4% and is even higher in the first weeks of life.3 The approach taken with these patients generally consists of extensive evaluations, hospitalisation and antimicrobial treatment. However, the management of febrile infants under the age of 3 months has changed considerably over the last two decades and it is possible to identify those at very low risk for UTI, bacteraemia or meningitis (ie, not requiring antibiotic treatment) who are thus suitable for outpatient management.4–6

Nevertheless, prior studies have shown that half of febrile infants will have a positive respiratory viral test and that infants with positive viral tests are less likely to have a bacterial infection.7–9 Thus, some authors recommend the use of these tests in the management of the febrile infant, given their high negative predictive value for IBI.10 11 In contrast, the most recent guidelines published by the American Academy of Pediatrics (AAP) state that viral testing should not affect entrance into their recommended pathway consisting of blood, urine and sometimes lumbar puncture (LP), although the guidelines allow for individualisation of assessment and management of virus-positive infants older than 28 days of age.4

The finding of a positive COVID-19 test in assessing febrile infants is unclear. Previous studies suggest that febrile infants with a positive COVID-19 test are less likely to have an IBI.12–15 However, to our knowledge, the published studies lack sample sizes large enough to accept this hypothesis.

Our primary aim was to determine the prevalence of IBI among febrile infants up to 90 days of age with a COVID-19 positive active infection test (PCR or antigen test). As secondary objectives, we aimed to determine the prevalence of UTI among infants less than 90 days of life and to determine the prevalence of IBI stratifying by the age groups determined by the recent AAP Febrile Infant Guidelines.4


This systematic review and meta-analysis was registered in the International Prospective Register of Systematic Reviews (CRD42022356507) and followed Preferred Reposting Items for Systematic Reviews and Meta-analysis reporting guidelines.16

Data sources

We performed a quadruple search strategy. First, with the assistance of a medical librarian, we performed a systematic search of PubMed, EMBASE, Cochrane Library, Scopus, Web of Science and We included all articles without restriction from 1 February 2020 to 1 May 2023. We decided to begin with articles published in February 2020 since the pandemic was declared in January. Online supplemental file 1 includes more details about the search strategy. Second, the references of the included papers were reviewed in order to identify potential studies, which could have been missed in the first search. Third, we reviewed records of the main paediatric meetings of 2021 and 2022. Finally, well-known researchers with expertise in the management of febrile infants with active databases were personally contacted to access unpublished data.

Study selection

Three researchers (SP-P, EG and RV), who were not blinded to journal, institution or author, independently screened the title and abstract of resulting papers. Disagreement about study selection was solved by consensus. Potentially eligible articles were evaluated for inclusion by full-text review (SP-P, EG and RV). Studies were included if they reported on infants up to 90 days old with fever (registered temperature at home or in the study setting ≥38°C) and a positive test for SARS-CoV-2 (antigen test or PCR). We decided to set 90 days old as the cut-off age because although some recent guidelines are more restrictive (ie, AAP guidelines consider patients younger than 60 days old4), most studies published on this topic and many European guidelines still establish 90 days old as the cut-off.17 Case reports with total sample size of less than three patients, articles written in other languages apart from English, French or Spanish, editorials and other narrative studies were excluded. If papers had an overall sample size bigger than three patients, they were included even though they had less than three patients younger than 90 days of age. In the case of more than one paper reporting patients from the same database, as verified with the corresponding author, only one of the publications was selected, always the one with the larger sample size. Studies with adult and paediatric population were included if data of infants could be extracted separately.

Data extraction

Corresponding authors of included studies were contacted by email in order to obtain cumulative data from studies that reported prevalence rather than SARS-CoV-2 test results and/or blood/urine/cerebrospinal fluid (CSF) culture data specifically and those enrolling patients with broader inclusion criteria. We did not ask for individual patient data, but only for the aggregate data in our study population. If data were not clear enough, the article was excluded. Data extraction was performed by one study researcher (SP-P) using standardised data extraction criteria and another researcher (EG) reviewed data extraction to validate accuracy. Extracted data included the number of COVID-19-positive febrile infants up to 90 days old, the number of blood cultures obtained and the number of them which were positive, the number of LPs done and how many grew pathogenic bacteria, and the number of urine cultures obtained and how many were positive. We considered as positive the isolation of 10 000 or more colony-forming unit (CFU) per milliliter of a single bacterium in a urine culture, regardless of urine dipstick and method of obtaining the urine sample, as we expected that in very few studies it would be possible to obtain these data. For positive cultures, information regarding isolated bacteria was also obtained, and only bacteria considered as true pathogens (online supplemental file 2) were considered as positive cultures. Patients considered COVID-19 positive by seroconversion or by epidemiological contact, patients without registered fever and patients without an obtained blood cultured were excluded.

Appraisal of methodological quality

Studies were assessed for methodological quality using the National Institutes of Health Quality Assessment Tools for Observational Cohort and Cross-Sectional Studies and for Case Series Studies (online supplemental file 3). The items numbered 6, 8, 10 and 14 of the Observational Cohort Checklist were omitted since they are not applicable to cross-sectional studies. Two researchers (RV and HB) qualified the studies as good (≤2 ‘no’ answers), fair (3–4 ‘no’ answers) and poor (>4 ‘no answers’). Disagreement about study quality was resolved by consensus. Methodological quality assessment was considered not applicable to unpublished databases.

Main outcomes and measures

The primary outcome was IBI prevalence, defined as a pathogen bacterium identified in blood and/or CSF by culture or PCR. The secondary outcomes were UTI prevalence, defined as a positive urine culture, and IBI prevalence stratified by the age groups defined by the AAP Febrile Infant Guidelines.4

Statistical analysis

Publication bias was evaluated by funnel plot. Meta-analysis was performed to assess the pooled prevalence of IBI; a random effects model was used because substantial heterogeneity was anticipated. The random effects meta-analysis model considers the contribution and weight of each study to the pooled estimate rather than the simple proportion. The heterogeneity across studies was measured with the I2 statistic. Higher I2 values mean a greater degree of heterogeneity between studies. I2 heterogeneity is interpreted as potentially unimportant (0%–19%), moderate (20%–49%), substantial (50%–79%) and considerable (>80%).18 The CI was set at 95%. For secondary outcomes, a meta-analysis of random effects model prevalence of UTI was conducted, with the same method as described above. In addition, number needed to screen (NNS) was calculated for IBI, bacteraemia, meningitis and UTI.19

Sensitivity analysis was performed since one of the studies had a sample size much larger than the others, and patients from its sample were removed to confirm that it did not suppose a bias to the results.

All analyses were performed using Stata (V.17.0; StataCorp, College Station, Texas, USA).

Patient and public involvement

No patient involved.


Search results

A flow diagram of search results is shown in figure 1. A total of 1023 studies and 4 databases were identified through initial search methods and 33 articles were included in the analysis.

Figure 1

Flowchart of the study selection process.

Included study characteristics

In total, 33 studies met inclusion criteria,12–15 20–44 the main characteristics of which are detailed in table 1. The range in the number of febrile infants included ranged from 1 to 3349, the latter study by Aronson being removed in the sensitivity analysis because its sample size was much larger than the rest. Online supplemental file 4 shows more details regarding the setting, country and number of patients from each study. The included studies were observational and case reports. Three were prospective, one was a prospective registry of febrile infants from one hospital with several publications45–47 and the rest were retrospective. Two other reports were unpublished data presented in Pediatric Emergency Medicine meetings. Seven were multicentre. As for the period of recruitment, 20 studies included patients only from 2020, 1 of the studies from 2021 to 2022, 1 from a non-specified period and the rest from a period which always included 2020. The complete quality assessment of each included study is detailed in online supplemental file 5. Although two studies were deemed low-quality because of the proposed conclusions, the data collection and results were sound and were therefore included in the analysis.

Table 1

Main characteristics of the studies finally included in the meta-analysis


Among the 33 studies, there were a total of 4496 febrile infants with COVID-19. A blood culture was obtained in 3943 and therefore was included in the analysis. The weighted prevalence of IBI among COVID-19-positive febrile infants was 0.14% (95% CI, 0.00% to 0.27%), and the NNS was 714. The meta-analysis detected no heterogeneity (χ2=14.96, df=32; p=1.00), with an I2=0.0% and an estimated between-study variance of 0.0. There were 14 bacteraemia and 5 meningitis. The weighted prevalence of bacteraemia was 0.10% (95% CI, 0.00% to 0.20%) with an NNS of 1000. The weighted prevalence of meningitis was 0.06% (95% CI, 0.00% to 0.15%) with an NNS of 1667. Figure 2 shows the forest plot for IBI with individual study event rate, and online supplemental file 6 shows the forest plot for bacteraemia and meningitis separately. Since not all authors were able to provide the number of patients with a performed LP, we performed a sensitivity analysis including only those articles with this data, with a weighted prevalence of meningitis of 0.01% (95% CI, 0.00% to 0.32%).

Figure 2

Forest plots for invasive bacterial infections with individual study event rates. ES, estimated prevalence; I2, heterogeneity; IBI, invasive bacterial infection.

When analysed by age groups, the prevalence of IBI was as follows (figure 3): 0.56% (95% CI, 0.0% to 1.27%) in those 0–21 days of life, 0.53% (95% CI, 0.0% to 1.22%) in those 22–28 days old and 0.11% (95% CI, 0.0% to 0.24%) in those 29–60 days old.

Figure 3

Forest plot for IBI with individual study event rates, stratified by age groups. ES, estimated prevalence; I2: heterogeneity; IBI, invasive bacterial infection.

A total of 3975 urine cultures were included in the analysis, with a weighted prevalence of UTI of 0.76% (95% CI, 0.50% to 1.03%), with an NNS of 131. Online supplemental file 7 shows the forest plot for UTI with individual study event rates.

Finally, given that the study by Aronson et al had a significantly larger sample size than the rest, a sensitivity analysis was performed, eliminating their patients. The estimated prevalence of IBI was 0.02% (95% CI, 0.00% to 0.18%) and 0.45% (95% CI, 0% to 0.97%) of UTI.

Funnel plots for all outcome measures demonstrated symmetrical distributions around all pooled estimates, indicating no evidence of publication bias for IBI and a minimal bias for UTI (online supplemental file 8).


In this systematic review and meta-analysis, we combined patients from 33 data sets with nearly 4000 febrile infants in a very diverse population from more than 20 countries in North and South America, Europe and the Middle East. Our results suggest that febrile infants with a positive COVID-19 test have a low prevalence of IBI as well as UTI. This is especially noticeable in the case of infants older than 28 days of age, so this subgroup of patients could be managed in a less invasive manner, without blood testing.

Young febrile infants account for a great number of attendances in the ED. Given their risk for bacterial infection, clinicians usually perform blood testing, urine culture and LP, increasing costs as well as potential adverse events related to these procedures.48 Additionally, a high proportion of the children are admitted to hospital and given broad spectrum antibiotics.

Several clinical decision rules have been developed to identify low risk infants who could be managed without performing LP and hospitalisation, or prescribing antibiotics.4–6 Additionally, a high proportion of febrile infants are positive for respiratory viral testing,7 8 and positive respiratory viral tests have been associated with a lower prevalence of IBI.9 49 For this reason, some of those rules include viral testing.10 However, the most recent guidelines published by the AAP did not consider rapid viral testing in their pathway, although they suggest the possibility of individualising the management of virus-positive febrile infants older than 28 days of age and highlight the importance of research on this topic.4

Different rates of bacterial infection have been found with associated viral infections. In the study by Mahajan et al,8 1200 (40.7%) febrile infants had a viral test performed and among those with a positive result, rates of bacterial infection ranged from 0.0% (95% CI, 0.0% to 4.2%) in infants positive for parainfluenza to 9.8% (95% CI, 3.3% to 21.4%) in the patients positive for ‘other’ viruses. Blaschke et al, in their cohort of 4037 febrile infants with a viral test positivity of 54.8%, found IBI in 1.6% of those positive only to human rhinovirus and 0.9% among infants who tested positive to viruses other than rhinovirus.7 In another study of 844 febrile infants tested for influenza, 14% were positive and there were no cases of bacteraemia nor meningitis among these children.49

The COVID-19 pandemic led to the need for immediate identification of patients with SARS-CoV-2 and rapid viral testing was considered a standard when attending to patients with infectious symptoms for public health reasons.50 This resulted in obtaining a body of strong evidence developed in record time regarding the association of this virus with IBI prevalence among febrile infants. Some papers reported a globally higher prevalence of IBI during the pandemic in febrile infants,15 36 probably related to a decrease in virus circulation during lockdown.51 In the epidemiology, severity and outcomes of children presenting to emergency department across Europe during the SARS-CoV-2 pandemic (EPISODES) study, there was a 78% reduction in bronchiolitis cases during the pandemic.52 Similarly, Burstein compared two prepandemic cohorts with a pandemic cohort from March 2020 to March 2021 and reported an increase in IBI (1.5% and 0.8% pre pandemic vs 3.4% during the pandemic) and a simultaneous decrease in positivity in multiplex viral testing (64.6% and 62.4% pre pandemic vs 36% in the pandemic cohort).15

A recent article by Aronson robustly analysed trends in the prevalence of bacterial infections in febrile infants during the COVID-19 pandemic.53 This study included febrile infants visiting the ED or admitted to hospital from 97 hospitals in the USA and Canada and described the prevalence of IBI by month from November 2020 to March 2022. The highest prevalence of IBI was 6.1% in February 2021 and it decreased over time, with overall prevalences of bacteraemia and meningitis of 1.8% (95% CI, 1.5% to 2.1%) and 0.5% (95% CI, 0.3% to 0.6%), respectively, which are similar levels to those described in prepandemic studies.

Regardless of overall IBI prevalence, the previously mentioned studies also found IBI prevalence to be low among COVID-19-positive febrile infants, with a high NNS. Although we consider our results to be sufficiently robust, it is important, when implementing them in clinical practice, to take into account the pre-test and post-test probability. The data collected from the studies did not include the epidemic situation in each country at the time each study was conducted. Aronson et al found that for every 5% increase in the prevalence of COVID-19 among infants in the month that a given infant presented, the age-adjusted odds of the infant having UTI and IBI decreased (OR 0.97; 95% CI, 0.96 to 0.98 and OR 0.94; 95% CI, 0.89 to 0.98, respectively).53 Also, Mintegi et al recommended using influenza tests only during influenza epidemic months, since a positive influenza test in a non-epidemic month is very unlikely to be a true positive given the relationship between prevalence and false positives.10 On the other hand, considering that COVID-19 pandemic reached incidence levels much higher than those reached by influenza,54 thus a positive test during extremely high incidence of COVID-19 could represent an incidental positive. Therefore, a future line of research could be to determine the influence of the population incidence of COVID-19 on the predictive value of the viral tests for IBI.

Apart from that, it is also important to consider the differences in IBI prevalence with respect to age subgroups, since prevalence is much higher among patients of 28 days of life and younger. After the neonatal period, IBI prevalence declines sharply to less than 0.5%. Hence, we think that clinicians should be cautious when attending to neonates. On the other hand, we consider results from this meta-analysis robust enough to safely state that well-appearing infants older than 28 days old could be managed less invasively without performing a blood test.

The prevalence of UTI associated with viral infections has been controversial. Previous data of febrile infants showed a non-negligible prevalence also in older patients, even though they had viral symptoms.55 On the other hand, a meta-analysis of UTI and bronchiolitis revealed a lower prevalence of UTI when a strict criterion for diagnosis was established.56 In our study, the prevalence of UTI was below 1%, although the variation in UTI prevalence between studies was much greater than that found between IBIs. Also, higher heterogeneity was found in meta-analysis for UTI and publication bias could not be ruled out. Not forgetting the fact that urine collection can be done by non-invasive methods and the NNS obtained is much lower than that for IBI. Thus, we believe that a conservative approach could include UTI screening in these patients without adding high iatrogenicity.

Limitations and strengths

This study has several limitations. First, only COVID-19-positive febrile infants are included in this meta-analysis, so we could not calculate OR and estimate risks. However, most studies about IBI and COVID-19 in febrile infants only included COVID-19-positive patients, and therefore very few patients could have been included in the meta-analysis if we had limited our search strategy to articles including both COVID-19-positive and COVID-19-negative febrile infants. Given that there is sufficient literature on the prevalence in overall febrile infants to compare with, we think that this is a minor limitation and that our results are robust enough to support the hypothesis of low prevalence of IBI in this COVID-19-positive febrile infants. Second, the data regarding patient’s general condition, biomarker values or personal background were not available in most studies, so it was not possible to perform a stratified analysis by these variables. In any case, given the low prevalence obtained, we believe it is unnecessary to perform complementary tests on these patients, except in the case of not-well appearing patients, in whom it seems reasonable to perform a full-sepsis work-up and the administration of empirical antibiotics, regardless of the results of the COVID-19 tests. Third, we did not have information related to prevalence of COVID-19 in the general population at the moment each study took place and we should be cautious when implementing results from this meta-analysis in clinical practice. Fourth, the data obtained from the studies did not include the presence or absence of the source of the fever. The variability in the consideration of mild upper respiratory tract symptoms as a sufficient source of fever means that the prevalence of UTI and IBI varies between studies.2 5 6 This means our results should be interpreted with caution, especially in febrile infants with no focus of any kind and in periods of very high incidence of COVID-19, in which coinfection may be a less relevant finding. Fifth, we did not consider results from urine dipstick for the diagnosis of UTI because these data were not available in the majority of studies, so this result could change as it has been seen in bronchiolitis when urinalysis result was added as a diagnostic criterion, decreasing the estimated prevalence of concomitant UTI previously reported56; despite this, we see this as a minor limitation since a true pathogen with 10 000 CFU/mL or more was isolated in all urine cultures considered as positive. Regarding the method of urine culture collection, this data was missing in most studies, but current guidelines recommend collecting it by a sterile method57; thus, it is presumed that it was performed in this way in most instances, minimising this limitation. Sixth, we limited data search to articles in English, Spanish and French, so some relevant data could have been omitted. Another limitation might be that, although the vast majority of patients included had a complete sepsis work-up, and therefore those under 28 days of age with a blood culture also had an LP, some patients above this age were managed without LP, and it is possible that some of those who received antibiotics could be misdiagnosed meningitis. However, we believe that this probability is low, and furthermore, it represents current clinical practice, so we believe that it does not affect the external validity of our results. Finally, one of the studies provided almost 80% of the patients included in the meta-analysis. However, in a sensitivity analysis without this study the prevalence of IBI was even lower, so this bias does not jeopardise the direction of the results. We included patients from different settings (ED, hospitalisation and ICU), and although this could be seen as a limitation, we consider it a strength, as it gives even more weight to our hypothesis of managing COVID-19-positive febrile infants less invasively especially as the prevalence of IBI in patients admitted to the ICU or on the hospital ward is probably higher than that of patients seen in the ED. Another strength is the inclusion of geographically diverse patients from different periods of the pandemic. Additionally, we used a strict definition of IBI that required identification of bacteria in blood and/or CSF cultures. The clinical applicability of our results and their impact on the management of febrile infants is strong, as COVID-19 positivity among febrile infants has been found to be higher than 25%.44


COVID-19-positive febrile infants up to 90 days of age are at low risk of IBI, mainly infants older than 28 days old, suggesting this subgroup of age can be managed without blood tests. However, UTI should be ruled out in all patients.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication


We thank Maria Luz de Andrés Loste (Cuca), the Hospital librarian, for her help with this project.


Supplementary materials


  • BG and RV are joint senior authors.

  • SP-P and EG are joint first authors.

  • Handling editor Gene Yong-Kwang Ong

  • Twitter @sperezp95, @egranda15, @damian_roland, @BorjaGomez79, @RoberVelasco80

  • Presented at This study was presented in the PAS meeting 2023 in Washington DC (USA), in the SEUP annual meeting 2023 in Las Palmas de Gran Canarias (Spain) and in EuSEM annual meeting 2023 in Barcelona (Spain).

  • Contributors SP-P made substantial contribution to conception and design, collected the data and reviewed multiple manuscript drafts for important intellectual content. EG made substantial contribution to conception and design, collaborated in data collection, wrote the initial draft of the manuscript and reviewed multiple manuscript drafts for important intellectual content. HB made substantial contribution to conception and design, helped with methodological analysis and reviewed multiple manuscript drafts for important intellectual content. DR and BG made substantial contribution to conception and design, collaborated in data collection and reviewed multiple manuscript drafts for important intellectual content. RV conceptualised and designed the study, collaborated in data collection, analysed the data, reviewed multiple manuscript drafts for important intellectual content and acts as the guarantor.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Linked Articles