Partnering for enhanced digital surveillance of influenza‐like disease and the effect of antivirals and vaccines (PEDSIDEA)

Abstract Background Standardised clinical outcome measures are urgently needed for the surveillance of influenza and influenza‐like illness (ILI) based on individual patient data (IPD). Objectives We report a multicentre prospective cohort using a predefined disease severity score in routine care. Patients/Methods The Vienna Vaccine Safety initiative (ViVI) Disease Severity Score (“ViVI Score”) was made available as an android‐based mobile application to three paediatric hospitals in Berlin and Athens between 2013 and 2016. Healthcare professionals assessed ILI patients at the point of care including severity, risk factors and use of antibiotics/antivirals/vaccines. RT‐PCR for influenza A/B viruses was performed at the Hellenic Pasteur Institute and the Robert Koch Institute. PCR testing was blinded to severity scoring and vice versa. Results A total of 1615 children aged 0‐5 years (54.4% males) were assessed at the three sites. The mean age was 1.7 years (SD 1.5; range 0‐5.9). The success rate (completion of the scoring without disruption to the ER workflow) was 100%. ViVI Disease Severity Scores ranged from 0 to 35 (mean 13.72). Disease severity in the Berlin Cohort was slightly higher (mean 15.26) compared to the Athens Cohorts (mean 10.86 and 11.13). The administration of antibiotics was most prevalent in the Berlin Cohort, with 41.2% on antibiotics (predominantly cefuroxime) as opposed to only 0.5% on neuraminidase inhibitors. Overall, Risk‐adjusted ViVI Scores were significantly linked to the prescription of both, antibiotics and antivirals. Conclusions The Risk‐adjusted ViVI Score enables a precision medicine approach to managing ILI in multicentre settings. Using mobile applications, severity data will be obtained in real time with important implications for the evaluation of antiviral/vaccine use.


| INTRODUC TI ON
Acute respiratory viral infections and influenza-like illness (ILI) are among the most common reasons for primary care visits and hospitalizations in children. Traditionally, hospitalization and admission to intensive care units have been considered criteria for "severe disease," but clinical management decisions may differ from site to site. The European Respiratory Society emphasized that clinical outcomes, in particular mortality and hospitalization rates due to respiratory illness, vary significantly across Europe. 1 For example, mortality appears to be higher in Eastern Europe, for reasons yet unknown. 1 Improved understanding of regional differences will require validated, standardized disease severity measures. 2 Standardized severity measures will allow cross-cohort comparison and a precision medicine approach to managing individual influenza infections in different risk groups. 3 Quality improvement programmes focused on optimizing treatment and prevention efforts depending on a patient's individual status will benefit from timely diagnostics and consistent use of standardized measures and operating procedures. 4 Severity measures thus must be sufficiently granular to capture disease progression in patients with very mild to very severe disease, including within the intensive care unit. 5 We present the first multicentre quality improvement programme implementing a standardized clinical severity measure for ILI in routine care. QI efforts are designed to induce system-level change. The participating departments agreed to introduce an institution-wide standard operating procedure, which is implemented in specific case scenarios (in this case, ILI) with regular analysis and evaluation. The PEDSIDEA operating procedure was introduced into routine care as a "standing order" for predefined standardized disease severity assessments and diagnostic testing in all patients with ILI, regardless of the rea-

| Severity assessments
As reported previously, the ViVI Disease Severity Score is a standardized clinical outcome measure that can be used independent of clinical treatments or interventions. The ViVI Score mobile application provides a uniform approach to defining ad hoc disease severity at any given time point, based on extensive literature review as well as WHO Criteria for uncomplicated and complicated influenza. 6 The ViVI Score consists of nine unweighted symptoms/items reflecting uncomplicated disease (DSU1-9) plus 13 weighted items reflecting complicated disease (DSC 1-13) resulting in overall scores ranging from 0-48. 2 only 0.5% on neuraminidase inhibitors. Overall, Risk-adjusted ViVI Scores were significantly linked to the prescription of both, antibiotics and antivirals.

Conclusions:
The Risk-adjusted ViVI Score enables a precision medicine approach to managing ILI in multicentre settings. Using mobile applications, severity data will be obtained in real time with important implications for the evaluation of antiviral/ vaccine use.

K E Y W O R D S
children, disease severity, ILI, influenza, mobile health, standardization

RT-PCR for influenza A/B viruses was performed at the Hellenic
Pasteur Institute, Attikon Hospital and the Robert Koch Institute: At the National Reference Centre for Influenza in Berlin, nasopharyngeal swabs were received and eluted in 3.0 mL cell culture medium. After RNA extraction and cDNA synthesis, real-time PCR was performed using the Light Cycler 480 real-time PCR system.
Primer and probes for amplification as well as typing and subtyping were used as described recently. 8 At the Attikon Hospital laboratory, RNA was extracted with QIAamp Viral RNAmini (Qiagen) using the QIAcube technology for automated extraction. All specimens were analysed to assess the quality of the specimen and extraction procedure, as well as for the presence of influenza virus by real-time RT-PCR with primers and probes as described in WHO molecular diagnostic protocols. 9 Aghia Sophia samples were analysed at the Hellenic Pasteur Institute 10 using NucliSENS® easyMAG® platforms (bioMérieux Hellas) and an in-house multiplex real-time RT-PCR. The PCR protocol is validated according to ISO 15189 requirements and deposited with the European Influenza Surveillance Network. 11 Virological laboratories were blinded to ViVI Scores, and influenza PCR results were made available after patient discharge, that is after severity scoring was completed and uploaded.

| Data analysis
Descriptive statistics (percentages, summary measures and histograms) were used to map the distribution of the ViVI Disease Severity Score (ViVI Score) and risk factors (ViVI Risk Factor Score, RF-Score) across the three PEDSIDEA sites. The correlation between RF-Scores, the VIVI Scores and treatment decisions was assessed using mean differences and t tests to assess significance. Pearson's correlation coefficient was used to assess the correlation between the RF-scores and age. Finally, the ViVI Score/RF-Score Index was developed to take into account both disease severity and pre-existing risk factors so as to better predict patient outcomes (see Supporting Information). All analyses were conducted using stata version 14.  one in Berlin) were included in the QI programme and analysis.

| Population and demographics
The success rate (completion of the scoring without disruption to the ED workflow) was 100%. The mean age was 1.7 years (SD 1.5; range 0-5.9), and the median age (IQR) was 1.3 (0.5-2.7) years for the overall PEDSIDEA cohort, and there were 54.4% males.
The mean RF-Score was 0.86 (SD 0.74, range 0-4) given a maximum possible RF-Score of 16, while the median RF-Score (IQR) was 1 (0-1). The demographic characteristics and distribution of risk factors for the overall cohort and by study site are summarised in Table 1, while the distribution of RF-Scores is plotted in Figure 1A,B.

| Prescribing practices across PEDSIDEA sites
Oseltamivir was the preferred antiviral across all three sites ( Table 2).
The most commonly used antibiotic class across the three sites was cephalosporins (cefotaxime, cefuroxime, ceftriaxone and cefprozil) followed by penicillins (amoxicillin, ampicillin, penicillin and amoxicillin + clavulanate). In a few cases, antibiotic combinations were prescribed which included vancomycin, erythromycin, azithromycin, ciprofloxacin, metronidazole or gentamicin in addition to a cephalosporin or penicillin. There appeared to be a slight preference for using cephalosporins in the Berlin site as compared to the two Athens sites.

| Correlation of ViVI Disease Severity Score with the ViVI Risk Factor Score
There was a significant but weakly positive correlation between the RF-Score and the ViVI Score (Pearson's correlation coefficient 0.2404; P < 0.001).

| Risk-adjusted ViVI Score: a new score based on disease severity and patient risk factors to predict patient outcomes and need for treatment
The mean Risk-adjusted ViVI Score was 8.29 (SD 4.56; range 0-32), while the median (IQR) was 7.5 (5)(6)(7)(8)(9)(10). Figure 3A,B shows the distribution of the Risk-adjusted ViVI Score for the overall PEDSIDEA cohort and by study site.
Patients who received antiviral treatment had a significantly higher Risk-adjusted ViVI Score than those who did not receive an-

| Distribution of Risk-adjusted ViVI Score by age and by viral aetiology
Pearson's correlation coefficient r = 0.3323; P < 0.001 shows a significant but weak positive correlation between age and Risk-adjusted ViVI Score (Supporting Information).

| D ISCUSS I ON
This is the first report of the use of the ViVI Score in a paediatric multicentre setting in Europe. The PEDSIDEA Network proof-of-concept  The ViVI Score App (https ://score.vi-vi.org) provides a useful instrument to harmonize severity assessments in multicentre clinical trials and observational studies. Future studies will explore use of the ViVI Score App in adult patients and for patient/parent-reported outcomes. Interesting differences have been observed between sites: patients with the same level of severity did not necessarily receive the same treatment. Standardization will provide a useful path forward with important implications for best practice and policy.

| CON CLUS IONS
The Risk-adjusted ViVI Score allows the consistent measurement of disease severity in urgent care and multicentre settings. The significance of the Risk-adjusted ViVI Score indicates that physicians may be more likely to resort to antibiotics or antivirals if they perceive a patient as "too ill" in relation to the number of risk factors. Standardized risk factor data and severity data have important implications for influenza surveillance and the critical evaluation of antibiotic and antiviral use, as well as vaccine effectiveness. 29 Surveillance programmes are strengthened enabling public health authorities to detect highly pathogenic viruses early on, even if they are prevalent at low rates. 30 Future studies will include clinical trials, adult ILI surveillance studies, and the alignment of patient-and physician-reported outcome measures.

ACK N OWLED G EM ENTS
The authors kindly thank the Vienna Vaccine Safety Initative Think