Approaches to use the WHO respiratory syncytial virus surveillance platform to estimate disease burden

Abstract The World Health Organization (WHO) recently completed the first phase of a RSV surveillance pilot study in fourteen countries (two to three in each WHO region) building on the Global Influenza Surveillance and Response System (GISRS). This active surveillance strategy had several objectives including understanding RSV‐related health burden in a variety of settings. A range of approaches can be used to estimate disease burden; most approaches could not be applied by participating countries in the WHO surveillance pilot. This article provides the recommendations made by WHO for strengthening and expanding the scope of the RSV surveillance in the next phase to enable burden estimation.


| DE SCRIP TI ON OF WHO PILOT
The WHO RSV surveillance pilot (1) took place in fourteen countries ranging from high to low income and from tropical to temperate settings. The main aim was to assess whether global RSV surveillance could fit into the GISRS platform. A secondary aim was to assess the feasibility of collecting data points needed for burden estimates in participating countries. The focus of the pilot was to collect data to estimate the proportion of respiratory hospitalizations associated with RSV. In some settings, this was extended to outpatient or primary care settings. Many countries were successful in implementing the clinical and laboratory aspects of the surveillance including collection of the required sample of clinical specimens.

| BURDEN OF D IS E A S E DUE TO R S V
The burden of disease can be characterized in several different ways.
Burden on the health service is often quantified in terms of numbers of outpatient/primary healthcare consultations; emergency department attendances and hospital admissions. Burden on the population may also be described by the numbers of infections including short-term morbidity and any longer-term chronic consequences of infection plus deaths associated with infection.
Burden of disease estimates may also include the cost of diseases (either direct costs to the health service or including indirect cost to patients and carers due to work loss). For example, a study in Bangladesh demonstrated a median direct cost of USD 10 million of RSV-associated hospitalizations among children <5 years of age in 2010 and an indirect cost of USD 3 million. 2 Better description of the health burden due to RSV infection will assist policy-makers to make informed decisions on the allocation of scarce health resources when considering new vaccines or other interventions for RSV disease.
Obtaining an accurate estimate of the disease-specific healthcare burden can be challenging, as often the investigation, laboratory testing and reporting of infectious diseases such as RSV are incomplete or absent. RSV-associated infection is known to be very common as shown with ad hoc age-stratified serological studies, for example a study in Kenya indicated that 100% of the population are likely to be infected by 3 years of age. 3 Only a proportion of these infections, however, will be symptomatic and a smaller proportion of them will present to health services. In a prospective cohort study in Finland, the average annual symptomatic RSV infection rate was 275/1000 in children <3 years of age over two RSV seasons, of whom 58% developed acute otitis media, and only 3% of those symptomatic were hospitalized. 4 Outside of enhanced studies such as these, most of those with infections that contact the healthcare service including hospitalizations will remain untested for RSV. Use of routine surveillance data alone without enhanced RSV testing may thus significantly underestimate the burden of disease due to RSV-associated LRTI on the healthcare system and on the health of the general population.

| P OTENTIAL ME THODS TO ME A SURE BURDEN OF D IS E A S E US ING A SURVEILL AN CE PL ATFORM
A range of approaches, many deployed for influenza, can be used to estimate local-, country-and international-level RSV-related burden in terms of numbers of cases and disease incidence either using established disease surveillance systems and vital statistics or through special studies. Their potential application in the context of the WHO surveillance pilot (1) is explored: • Method 1: Regression modelling using routine data sources. This ecological approach uses national or large-scale routine laboratory surveillance and administrative health service data that have been collected consistently over several years. It has been utilized in some countries to measure the burden of a wide range of infections including influenza and RSV. 5-7 Based on weekly data on primary care consultations, hospitalizations and deaths due to ICD-coded respiratory disease, the regression model attributes a proportion of the respiratory infection such as pneumonias and bronchiolitis to RSV (and influenza) using weekly laboratory data adjusting for key confounders such as meteorological conditions. This statistical approach requires seasonality in the infection of interest and health data availability over several years. The method has also been applied to derive international burden estimates for the 2009 influenza pandemic where parameter values were used to derive estimates for those countries where data were lacking. 8 Those WHO pilot countries with access to such routine health data could potentially apply such methodology.
• Method 2: Multiplicative modelling. This approach has been applied to estimate influenza-related mortality for the 2009 pandemic by multiplying age-specific symptomatic infection attack rate and symptomatic case fatality rate from a range of income settings.
With this method, a respiratory mortality multiplier was defined to consider differences in outcome according to setting. Simulation models with these parameters were then applied across all countries to derive global estimates of influenza-related mortality. 9 A similar approach has been used in a recently published modelling study to estimate the global hospital admission rate and mortality rate due to RSV-related acute lower respiratory tract infection (ALRI) using information on the proportion of hospital ALRI admissions that were due to RSV and the case fatality ratio (CFR) for RSV-ALRI. 1 However, gaps in RSV knowledge remain. This article noted that very limited RSV hospital and mortality data were available from low-income settings, particularly in sub-Saharan Africa and the Eastern Mediterranean region. The WHO pilot surveillance aims to gather relevant hospital surveillance data that will strengthen burden estimates, particularly from low-income settings.
• Method 3: Data linkage. Linkage of routine hospital administrative data to laboratory records has been used in developed settings such as Australia to better describe RSV epidemiology. 10 This approach can provide more detailed information on patients with laboratory-confirmed RSV, such as prevalence and type of underlying risk factors, length of stay and likelihood of different outcomes. Estimates of RSV-related hospital admissions can be derived, considering differences in laboratory testing practice by age and other key risk factors. These individual-level data allow more accurate cost information to be generated. 10,11 To our knowledge, few such studies have been undertaken in LMIC settings and the approach is unlikely to be applicable in most countries participating in the WHO pilot surveillance due to lack of electronic administrative databases in hospitals or primary care settings.
• Method 4: Enhanced surveillance/prospective cohorts. The establishment of prospective population cohorts, with the comprehensive systematic collection of epidemiological information and biological samples from participants, allows RSV epidemiology and burden to be studied more directly-both in the community and in primary and secondary care settings. The Influenza and RSV in Infants Study (IRIS) is a multi-country prospective study in four low-income settings and was set up to assess the frequency of influenza and RSV in children <1 year of age who have been hospitalized and to ascertain predictors of more severe disease. 12 Prospective population-based studies have also been created, including household studies to describe the burden and transmission of RSV in both low 13 -and high-income settings. 4 Such studies can also provide information on key parameters needed for realistic transmission models and provide the opportunity to gather information on quality of life in young RSV-positive children, information which at present is largely lacking. At present, there are only a very limited number of such studies as they are resource intensive to establish and maintain and thus tend to cover very limited periods.
• Method 5: Simple, rapid assessment approach. This methodology has been used to estimate the national burden of influenza using surveillance data from sentinel sites and by applying several adjustments to obtain a national estimate. [14][15][16] This approach is simple and involves using the severe acute respiratory illness (SARI)/ LRTI rate from an established sentinel hospital network, adjusting for key risk factors for pneumonia (such as malnutrition, indoor air pollution and HIV infection) at the provincial or district level and multiplying by the RSV proportion positive from hospital surveillance to derive the RSV-associated SARI rate. This was intended to be the approach applied in the WHO pilot.

| E XPERIEN CE S IN COLLEC TING HOS PITAL AND COMMUNIT Y MORB ID IT Y DATA IN THE R S V SURVEILL AN CE PILOT AND THE WAY FORWARD IN THE NE X T PHA S E
The results of the pilot were reviewed at a face-to-face meeting    (Table 2) including the simple rapid assessment approach (method 5) to estimate the number of RSV-related respiratory admissions. Regression modelling methods (method 1) will be possible once an adequate time series of these data has been collected, which will allow cross-verification of the rapid assessment method. In addition, some study sites may be able to provide independent burden estimates at a later stage using one or more of the other methods (methods [2][3][4] outlined in this paper, which will allow further cross-validation.
The next phase of the WHO Global RSV surveillance programme will assist countries to better describe RSV health burden by collecting the necessary data needed. The critical data still to be collected

| CON CLUS IONS
Accurate in-country disease burden estimates are important for local policy-makers to make decisions on the introduction of prevention strategies and to employ cost-effectiveness models when new RSV vaccines or monoclonal antibodies become available.
Even simple estimates of proportion of hospitalizations associated with RSV are important to raise awareness of RSV disease burden among policy-makers and providers. The WHO Global RSV surveillance programme aims to provide countries with a platform based on established approaches to collect these data. The next phase of the programme will assist countries with tools for collection of burden data, closely monitor sampling strategies for patient selection and testing, and strengthen the surveillance to provide simple, robust burden estimates for informed immunization policy decisions.

ACK N OWLED G EM ENTS
The WHO global RSV surveillance pilot project was supported by an