Estimating the burden of influenza‐associated hospitalizations and deaths in Central America

Objectives Our objective was to estimate the incidence of influenza‐associated hospitalizations and in‐hospital deaths in Central American Region. Design and setting We used hospital discharge records, influenza surveillance virology data, and population projections collected from Costa Rica, El Salvador, Guatemala, Honduras, and Nicaragua to estimate influenza‐associated hospitalizations and in‐hospital deaths. We performed a meta‐analysis of influenza‐associated hospitalizations and in‐hospital deaths. Main outcome measures The highest annual incidence was observed among children aged <5 years (136 influenza‐associated hospitalizations per 100 000 persons). Results Annually, 7 625–11 289 influenza‐associated hospitalizations and 352–594 deaths occurred in the subregion. Conclusions Our results suggest that a substantive number of persons are annually hospitalized because of influenza. Health officials should estimate how many illnesses could be averted through increased influenza vaccination.


Introduction
Influenza is a vaccine preventable disease that annually causes approximately 1 million influenza-associated hospitalizations among children younger than 5 years and 500 000 deaths among all age-groups. 1,2 The burden of disease is concentrated predominantly in developing countries where an estimated 934 000 hospitalizations and 111 500 deaths attributable to influenza occur among children aged <5 years. 1 Few studies have estimated the burden of influenza illness in Central America, a tropical region comprised of lower to upper-middle-income countries. [3][4][5][6] Estimating the influenza disease burden is difficult because ill persons may not seek care or present for care with ill-defined symptoms where they are often treated empirically without respiratory sampling.
Information on disease burden can be useful for local public health practitioners exploring the potential value of prevention programs. 6,7 Annual influenza vaccination remains a key strategy for the prevention of influenza-related hospitalizations and deaths. In addition since 2004, countries in the Americas have expanded the use of influenza vaccine among subpopulations at increased risk of complications following influenza infections. 8 As annual influenza vaccination is a costly investment, influenza disease burden should be well documented, 9 as should be the assessment of influenza vaccine impact. 7 Our objective was to estimate the incidence of influenzaassociated hospitalizations and in-hospital deaths in five Central American countries.

Methods
We used hospital discharge records, influenza surveillance virology data and population projections collected during January 1, 2009, to December 31, 2012, from Costa Rica, El Salvador, Guatemala, Honduras, and Nicaragua to estimate influenza-associated hospitalizations and in-hospital deaths. 6 Calculating the number of pneumonia and influenza diagnoses In all countries except Guatemala, we obtained public hospital discharge records and in-patient survival data from national Health Information Systems available from the Ministries of Health. In Guatemala, we recruited a survey team to collect the information at all public hospitals. For each month during 2009-2012, we determined the number of patients hospitalized with pneumonia or influenza diagnoses corresponding to ICD-10 codes J09-J18, the WHO proxies for severe acute respiratory infections (SARI), 6 by age-group, that is, <5 years, 5-64 years, and >64 years, the age-groups typically targeted for vaccination. 8

Estimating the proportion of hospitalized patients with influenza virus infection using surveillance data
Influenza surveillance among SARI case-patients was conducted in the study countries by sentinel hospitals and National Influenza Centers according to PAHO-CDC 2006 generic protocol for influenza surveillance. 3,6 Laboratories tested respiratory samples using immunofluorescence and rtRT-PCR 3,6 after systematically identifying a sample of SARI cases through routine clinical practice. We determined the age-stratified percentage of samples from SARI case-patients testing positive for influenza virus by dividing the number of positive by the total sampled each same month.

Calculating the population at risk based upon population projections
We assumed that every inhabitant was at risk of developing severe influenza illness and requiring admission to a public hospital. Thus, we defined the population at risk per country as the most recent population estimate based on census projections, stratified by year and age-group. National Statistical Offices provided the data for each country. As a sensitivity analysis, we also assumed that in each country, only the proportion of the population typically seeking care at public hospitals network was at risk of developing severe influenza and being identified in national data sets. Public hospitals cover 87% of the population in Costa Rica, 80% in El Salvador, 70% in Guatemala, 60% in Honduras, and 61% in Nicaragua. 10

Statistical analysis
We estimated incidence of influenza-associated hospitalizations and in-hospital deaths for each year, age-group, and country by multiplying the age-stratified monthly number of patients with pneumonia or influenza clinical diagnoses by the age-stratified percentage of samples from SARI surveillance case-patients that tested positive for influenza. We then divided the estimated annual number of influenzaassociated hospitalizations and deaths by the population at risk. The formula is presented in the Data S1. We calculated 95% confidence intervals (95% CI) assuming a Poisson distribution. We performed a meta-analysis of individual country rates by pooling the information (2009-2012) from the five countries, using a random-effects model, and assuming significant heterogeneity in the data. We conducted a second sensitivity analysis stratifying influenza rates into pandemic (2009-2010) and seasonal influenza years (2011-2012).

Ethical issues
The project was reviewed by Institutional Review Committees of the Universidad del Valle de Guatemala and the countries' Ministries of Health and judged to be a public health evaluation. Only records without personal identifiers were used for the analysis.  (Tables S1-S5).

SARI case-patients under surveillance
During the study period, surveillance identified and tested 43 951 SARI case-patients samples for influenza; 48% children aged <5 years, 41% persons aged 5-64 years, and 11% adults aged >64 years. The percentages of SARI casepatients that tested positive for influenza viruses varied significantly between age-groups (P-value<0Á001): 9% among children aged <5 years, 32% among persons aged 5-64 years, and 13% among adults aged >64 years. In 2009 and 2011 influenza A (H1N1) pdm09 was the most commonly identified subtype among SARI case-patients samples (Figure 1). In 2010, influenza A (H3) was detected in 6% and in 2012 influenza B appeared in 4% of all samples processed.
Influenza-associated hospitalization and in-hospital deaths rates were 1Á4-1Á5 higher if we assumed only the subpopulation seeking care at public hospitals was at risk of developing illness identifiable through our national data. (Table S6).

Discussion
Our results suggest that influenza, a vaccine preventable illness, is associated with substantial morbidity and mortality in Central America. In the five countries studied, 7563-11 397 influenza-associated hospitalizations and 292-473 inhospital deaths occurred annually among 40 million inhabitants.
Very young children were most commonly hospitalized as a result of influenza-associated hospitalizations and older adults most commonly died as a result of influenza. While persons aged 5-64 years were more commonly hospitalized during 2009 because of influenza A (H1N1) pdm09 influenza pandemic, 11 we did not identify a difference in the annual age-adjusted influenza-associated hospitalization rates by influenza type and subtype during sensitivity analyses (Table S6).
Influenza-associated hospitalizations rates in this study were similar to those published in a recent global metaanalysis 1 which reported a rate of 1 (95% CI: 1-2) per 1000 person-years among children aged <5 years. Costa Rica used similar methods 6 and obtained comparable results. El Salvador and Nicaragua 3,4 estimated slightly higher rates than our results, whereas another study conducted in Guatemala 5 estimated slightly lower rates. These differences could be explained by the use of active population-based surveillance in these studies which may have resulted in increased referrals of case-patients to the hospital.
Our estimated influenza-associated in-hospital mortality was similar to that reported previously in the region, except for the lower mortality rates we estimated among adults aged >64 years. 6,12,13 Our rates among children aged <5 years were also lower than those of a global meta-analysis using different methodology, 1 whereby 4Á9 children per 100 000 died from influenza-associated severe acute lower respiratory infection (ALRI) in developing countries. A plausible reason for these discrepancies could be the incompleteness and underreporting in mortality registries in study countries. 14 In addition, ALRI tends to be more sensitive than SARI in identifying influenza illness and tends to yield higher rate estimates.

Limitations
We extrapolated influenza laboratory results from SARI casepatients identified through surveillance to all untested hospitalized patients with SARI proxy discharge diagnoses. We were unable to assess differences between SARI casepatients identified through surveillance and the SARI proxies. In addition, surveillance systems and medical practices vary from country to country; we therefore used randomeffects models because we assumed that influenza-associated hospitalization rates incidence varied by country but followed same distribution. Our results might underestimate the burden of influenza-associated hospitalizations and mortality because underreporting of deaths is common in developing countries; a study estimated~50% underreporting in 2003 for some countries in the region. 14 Moreover, we assessed only in-hospital deaths and missed those that occurred in the community. Last approximately 20% of samples were analyzed using indirect immunofluorescence, an assay that has a lower sensitivity than rtRT-PCR. 15

Conclusion
Influenza is among the leading causes of hospitalizations and deaths. 2 Our results suggest that, each year, a substantial number of persons in Central America are hospitalized and died as a result of influenza. Our findings suggest that public health professionals should evaluate the potential value of giving young children and older adults influenza vaccines and empiric treatment with antivirals. Further studies should be conducted to explore the number of illnesses and cost averted through influenza vaccination and antiviral policies especially among subpopulations at high risk of developing complications as a result of influenza illness (e.g., pregnant women).

Supporting Information
Additional Supporting Information may be found online in the supporting information tab for this article: Data S1. Mathematical formula Table S1. Influenza-associated hospitalizations and deaths in Costa Rica during 2009-2012. Table S2. Influenza-associated hospitalizations and deaths in El Salvador during 2009-2012. Table S3. Influenza-associated hospitalizations and deaths in Guatemala during 2009-2012. Table S4. Influenza-associated hospitalizations and deaths in Honduras during 2009-2012. Table S5. Influenza-associated hospitalizations and deaths in Nicaragua during 2009-2012. Table S6. Sensitivity analysis of influenza-associated hospitalizations and deaths.