Hospitalization rates for pneumococcal disease in Brazil , 2004-2006

OBJECTIVE: To estimate hospitalization rates for pneumococcal disease based on the Brazilian Hospital Information System (SIH). METHODS: Descriptive study based on the Hospital Information System of Brazilian National Health System data from January 2004 to December 2006: number of hospitalizations and deaths for pneumococcal meningitis, pneumococcal sepsis, pneumococcal pneumonia and Streptococcus pneumoniae as the cause of diseases reported in Brazil. Data from the 2003 Brazilian National Household Survey were used to estimate events in the private sector. Pneumococcal meningitis cases and deaths reported to the Notifi able Diseases Information System during the study period were also analyzed. RESULTS: Pneumococcal disease accounted for 34,217 hospitalizations in the Brazilian National Health System (0.1% of all hospitalizations in the public sector). Pneumococcal pneumonia accounted for 64.8% of these hospitalizations. The age distribution of the estimated hospitalization rates for pneumococcal disease showed a “U”-shape curve with the highest rates seen in children under one (110 to 136.9 per 100,000 children annually). The highest hospital case-fatality rates were seen among the elderly, and for sepsis and meningitis. CONCLUSIONS: PD is a major public health problem in Brazil. The analysis based on the SIH can provide an important input to pneumococcal disease surveillance and the impact assessment of immunization programs. DESCRIPTORS: Pneumococcal Infections, epidemiology. Pneumonia, Pneumococcal. Meningitis, Pneumococcal. Sepsis. Hospitalization. Epidemiologic Surveillance. Artigos Originais 540 Pneumococcal disease in Brazil Novaes HMD et al Streptococcus pneumoniae is a major cause of illness and death worldwide. Clinical manifestations of pneumococcal infection include serious life-threatening invasive disease – meningitis, sepsis, pneumonia with bacteremia, and other conditions such as bone and joint infection and peritonitis. Pneumococcus also causes non-invasive diseases such as pneumonia without bacteremia, bronchitis, sinusitis, and otitis media, which accounts for most pneumococcal-related disease. S. pneumonia is the most common cause of invasive bacterial disease in children in countries with universal childhood immunization programs against Haemophilus infl uenza type b.1 There are available safe and effective vaccines to prevent pneumococcal disease (PD) in children under two. The heptavalent pneumococcal conjugate vaccine (PCV7) has shown signifi cant impact in reducing PD rates in countries where universal childhood immunization programs have been implemented.8,21 Both the 10-valent (PCV10) and the 13-valent conjugate pneumococcal vaccines have recently been made available.5,26 In Brazil, the PCV7 has been available in the Brazilian National Health System (Sistema Único de Saúde, SUS) RESUMO OBJETIVO: Estimar as hospitalizações por doença pneumocócica com base em dados do Sistema de Informações Hospitalares (SIH). MÉTODOS: Estudo descritivo com base em dados do SIH de janeiro de 2004 a dezembro de 2006: números de hospitalizações e mortes por meningite pneumocócica, sepse pneumocócica, pneumonia pneumocócicca e Streptococcus pneumoniae como causa de doenças ocorridas no Brasil. Dados da Pesquisa Nacional por Amostras de Domicílios 2003 foram utilizados para o setor privado. Casos e mortes por meningite pneumocócica notifi cados no Sistema Nacional de Agravos de Notifi cação no mesmo período também foram analisados. RESULTADOS: A doença pneumocócica foi responsável por 34.217 hospitalizações no Sistema Único de Saúde (0,1% de todas as hospitalizações). Pneumonia pneumocócica foi responsável por 64,8% dessas hospitalizações. A distribuição das estimativas de hospitalizações segundo faixa etária mostrou curva em “U”, com maior freqüência entre crianças < 1 ano (110-136,9/100.000 crianças/ano). A letalidade hospitalar foi mais alta entre idosos, e entre casos de meningite e sepse. CONCLUSÕES: Doença pneumocócica é importante problema de saúde pública no Brasil. Análise baseada no SIH pode contribuir para a vigilância epidemiológica da doença pneumocócica e para a avaliação do impacto do programa de vacinação. DESCRITORES: Infecções Pneumocócicas, epidemiologia. Pneumonia Pneumocócica. Meningite Pneumocócica. Sepse. Hospitalização. Vigilância Epidemiológica.

Streptococcus pneumoniae is a major cause of illness and death worldwide.Clinical manifestations of pneumococcal infection include serious life-threatening invasive disease -meningitis, sepsis, pneumonia with bacteremia, and other conditions such as bone and joint infection and peritonitis.Pneumococcus also causes non-invasive diseases such as pneumonia without bacteremia, bronchitis, sinusitis, and otitis media, which accounts for most pneumococcal-related disease.S. pneumonia is the most common cause of invasive bacterial disease in children in countries with universal childhood immunization programs against Haemophilus infl uenza type b. 1 There are available safe and effective vaccines to prevent pneumococcal disease (PD) in children under two.The heptavalent pneumococcal conjugate vaccine (PCV7) has shown signifi cant impact in reducing PD rates in countries where universal childhood immunization programs have been implemented. 8,21Both the 10-valent (PCV10) and the 13-valent conjugate pneumococcal vaccines have recently been made available. 5,26 Brazil, the PCV7 has been available in the Brazilian National Health System (Sistema Único de Saúde, SUS) RESUMO OBJETIVO: Estimar as hospitalizações por doença pneumocócica com base em dados do Sistema de Informações Hospitalares (SIH).

INTRODUCTION
since 2002, but only for children under fi ve with high risk of PD.This strategy has resulted in low vaccine coverage as we estimated based on administrative data that less than 5% of children under 12 months were vaccinated in 2009.The 23-valent pneumococcal polysaccharide vaccine is also available for the elderly (≥ 60 years), and adults and children two years of age and older with chronic conditions.In March 2010, the PCV10 was introduced for routine immunization of infants (four doses at two, four, six and 12 months of age) with catch-up vaccination for children under two in the fi rst year of the program.Effective surveillance of PD is essential to monitor this program's impact.
Pneumococcal surveillance systems focusing on invasive pneumococcal disease (IPD) are very heterogeneous and have different notifi cation rules, case definitions, case reporting coverage, data collection, health information system utilization, quantity and quality of laboratories so data comparability is a problem. 18laboratory-based surveillance system of IPD has been active in Brazil since 1993 (Network Surveillance System for the Bacterial Agents Responsible for Pneumonia and Meningitis, Sistema Regional de Vacinas -SIREVA II).This network provides data on pneumococcus serotype distribution and antimicrobial resistance but with limited population coverage.The linking of laboratory data with epidemiological and clinical information needs to be strengthened.3 The use of data based on clinical diagnosis, hospital admissions and outpatient consultations in pneumococcal burden of disease studies and surveillance systems has been limited as recent reviews of vaccine impact studies in the US and Europe have shown.8,21 The clinical diagnosis has high sensitivity but low specifi city for PD, and laboratory-based systems have been prioritized, particularly for IPD.Hospital admissions and consultations have more often been used to estimate pneumococcal pneumonia and otitis due to the limited contribution of diagnostic laboratory tests in these conditions.
A population-based time series analysis based on hospital admissions data have shown that this is a feasible, valid and useful approach for measuring the impact of pneumococcal conjugate immunization programs, especially when there are good quality national hospital databases with clear inclusion criteria and appropriate methodological strategies.Studies in England and Wales, US and Australia have shown similar tendencies for PD decline and seasonal variation when comparing active surveillance and hospital information data. 8,15,19e Brazilian National Health System provides a freeaccess, good quality national hospital database (Sistema de Informações Hospitalares -SIH) and estimates of hospitalization and case-fatality rates for PD in Brazil may effectively and effi ciently contribute to the assessment of the potential impact of an universal childhood pneumococcal immunization program. 13e present study aimed to estimate hospitalization rates for PD in Brazil based on SIH-SUS data.

METHODS
A descriptive study based on SIH-SUS data was conducted.Data on hospital admissions, which occurred from January 1 st , 2004 to December 31 st , 2006, due to the clinical syndromes of interest -pneumococcal meningitis, pneumococcal sepsis, pneumococcal pneumonia and S. pneumoniae as the cause of diseasesas the main diagnosis were obtained from the SIH-SUS.This database has information on individual cases of illness with data on patient demographic characteristics, length of hospital stay, diagnoses at discharge, disease outcome, and hospital reimbursement rates.
Since the SIH-SUS covers only the public sector, estimates of hospitalizations in the private sector were calculated to obtain national population estimates.Data from the 2003 Brazilian National Household Survey showed 72.9% of all clinical hospitalizations were in the SUS and were used to estimate hospitalization rates in the private sector. 4 Meningitis is a mandatory notifi able disease in Brazil, and data from the Notifi able Diseases Database (Sistema de Informação de Agravos de Notifi cação -SINAN) for the same period were also analyzed.SINAN database includes aggregate data on population frequency of notifi able diseases and their outcomes according to age group, clinical syndromes and etiological agents, and good quality information on meningitis. 13nual age-specifi c hospitalization rates for PD were calculated using the total number of hospitalizations for each clinical syndrome based on SIH-SUS data and population estimates in Brazil for each year studied obtained from the Instituto Brasileiro de Geografi a e Estatística (Brazilian Institute of Geography and Statistics).
For population estimates of annual age-specifi c hospitalization rates for PD, in addition to estimating the rates in the private sector, we also included the estimated number of cases of pneumococcal meningitis reported under the diagnosis "unspecifi ed bacterial meningitis".
Case-fatality was defined as in-hospital death.Age-specifi c case-fatality rates (CFR) were estimated based on the number of deaths/confi rmed cases for each clinical syndrome registered in the SIH-SUS and SINAN.

RESULTS
In the three-year study period there were 34,217 hospitalizations for PD in the SIH-SUS database, accounting for 0.1% of all hospitalizations (34,260,055) in the public sector in Brazil.Of these, 47.1% were in children under fi ve.Variations in hospitalization rates for PD in the public sector (SUS) by year of diagnosis were small, except in children under one and 1 to 4, with possibly a general downward tendency (Table 1).The highest hospitalization rates were seen in infants (Figure 1).The most common clinical syndrome was pneumonia (64.8% of all hospitalizations for PD) (Figure 2).S. pneumoniae as the cause of diseases (ICD-10, B95.3), sepsis and meningitis accounted for 13.7%, 12.3%, and 9.3% of PD hospitalizations in the SUS, respectively.
During the study period, there were 3,165 hospitalizations for pneumococcal meningitis registered in the SIH-SUS and 4,336 hospitalizations when estimates for the private sector were included.During this same period, 4,032 cases of pneumococcal meningitis were registered in the SINAN, which is 7% lower than that seen in the SIH-SUS including estimates for the private sector.
A causal pathogen was identifi ed in 41.4% of hospitalizations for bacterial meningitis reported in the SIH-SUS.Pneumococcus accounted for 31.3% of all cases of bacterial meningitis with an identifi ed pathogen, with slight variation by year of diagnosis and age group.Assuming the same proportion of pneumococcal meningitis in the hospitalizations for "unspecified bacterial meningitis", Pneumococcus accounted for an additional 4,384 hospitalizations due to bacterial meningitis, totalizing 7,549 hospital admissions for pneumococcal meningitis in the SUS.
The annual estimated hospitalization rates for PD by clinical syndrome and age group was based on SIH-SUS data and included estimates of hospitalization in the private sector for all clinical syndromes and estimates of additional pneumococcal meningitis cases reported under "unspecifi ed bacterial meningitis" (Table 2).This approach led to a more signifi cant increase in hospitalization rates among children under one, but the general trend was unchanged.There were 4,204 hospitalizations for pneumococcal sepsis in the public sector and other 1,555 were estimated for the private sector, totaling 5,759 hospital admissions over the three-year study period.
Pneumococcus accounted for 2.6% of overall sepsis hospitalizations in the SIH-SUS.A causal pathogen was identifi ed in most sepsis cases reported in the SIH-SUS."Unspecifi ed sepsis" accounted for only 6.6% of all hospitalizations for sepsis.The proportion of cases in which the etiologic agent was not identifi ed increased by age group, reaching 17.1% in adults aged 60 to 69 years.The rate of Pneumococcus among all causes of sepsis varied according to age, and was higher in children aged 1 to 4 (5.2%).During the three-year period studied, 2,746 hospitalizations for PD in the SIH-SUS resulted in death: 1,674 from sepsis, 672 from pneumonia, 377 from meningitis and 26 due to "S. pneumonia as the cause of diseases" (Table 3).The highest CFRs were seen for pneumococcal sepsis (39.8%), followed by pneumococcal meningitis (11.9%), pneumococcal pneumonia (3.0%) and "S.pneumonia as the cause of diseases" (2.1%) and in the older age groups (Figure 1).Pneumococcal CFRs did not change signifi cantly by year of diagnosis (Table 3).
During the same period, 1,276 deaths caused by pneumococcal meningitis were reported in the SINAN, resulting in an overall case-fatality rate of 31.7% in all age groups (Table 3).

DISCUSSION
The annual hospitalization rates by age in Brazil are similar to those reported in other studies in developed countries before the implementation of pneumococcal immunization program for children.PD age distribution has a "U"-shape curve with the highest rates seen in children under two and in those aged 60 and more. 11,15,20he burden of PD is considerable with a large annual number of hospital admissions.
Comparisons of national IPD hospitalization rates are diffi cult since studies used different case defi nitions (main diagnosis or all diagnosis), hospital admission databases (administrative or specifi c) and rate calculation methods.However, three comparable studies found similar average national annual PD hospitalization rates in the US, England and Wales and Singapore before the implementation of immunization programs. 14,15,22,24gional studies or those including only specific health services also call for careful comparisons.
A study in the metropolitan area of Santiago, Chile based on confi rmed cases found lower average IPD annual hospitalization rates in children under 14, but PD surveillance in Chilean health care centers found higher rates than in our study. 1,12mparisons for specific clinical conditions can help identify population and health care differences not evident with all-inclusive PD data.The annual hospitalization rates for pneumococcal meningitis in children and hospital CFR rates in the present study are considerably higher than those reported in the US and Spain. 6,25Antibiotic use is high in Brazil, even before hospital care, preventing laboratory confi rmation. 23,27his justifi es the inclusion of the proportion of pneumococcal meningitis among bacterial meningitis cases with unidentifi ed pathogen.We estimated two CFRs for pneumococcal meningitis, one based on hospital data (SIH-SUS), which was higher than that in developed Including estimates of "unspecifi ed bacterial meningitis" attributed to Pneumococcus and estimates of hospital admissions in both the public and private sector, assuming that the public sector accounts for 72.9% of all clinical hospitalizations.
countries, and the other one based on mandatory notifi cation data (SINAN), which was even higher especially in infants and children under fi ve.SINAN data includes confi rmed cases with positive laboratory data, and possibly more serious cases, cases treated in tertiary care hospitals with higher co-morbidities rates, and deaths in the emergency department, not included in hospitalization data.
Comparisons of national annual pneumonia hospitalization rates are complex due to different clinical diagnostic practices and medical nomenclatures and require careful analysis. Grijalva's study in the US found a higher national annual pneumococcal pneumonia hospitalization rates among adults (18 years and more) than observed in this study, even when correction to include private sector was considered. 9Differences in the access to secondary care in hospital settings, in care practices, health information systems, and true variations in population PD incidence and lethality rates may account for the differences in hospitalization rates between the two countries.
Our study has some limitations.The validity of administrative databases depends on data quality of SIH and SINAN databases. 13Our search for ICD-10 codes in the main diagnosis at discharge may have underestimated PD hospitalization rates as PD may be listed as a secondary diagnosis or even not listed at all.Errors and inconsistencies in diagnostic codes may occur.Non-confi rmation of Pneumococcus as the etiological agent is also a limitation.PD rates may have been affected by different clinical and laboratory practices, such as the frequency of collecting clinical specimens for culture in suspected cases, antibiotic use and culture techniques, which may vary among different health facilities and Brazilian regions.Information on specifi c pneumococcal serotypes causing disease is not available in the SIH and SINAN databases.
Since information on hospitalization rates in the private sector is not available, we used data from a national household survey to estimate them in private care settings.This approach may have overestimated PD rates since PD is more frequent among poor populations who are predominantly SUS users.The proportion of hospitalizations for acute serious infectious diseases is probably higher in the SUS.Nevertheless, meningitis is a notifi able disease in Brazil and we were able to use SINAN database to check our estimates, assuming that bacterial meningitis cases always require hospitalization.There was a small difference between the estimates based on SIH-SUS data including hospitalizations in the private sector (4,336) and confi rmed cases registered in the SINAN database

REFERENCES
(4,032).Underreporting in the SINAN is concerning.There were differences in both databases in the number of cases registered according to age with more cases reported in the SIH-SUS than SINAN for some age groups, particularly children (data not shown).In studies conducted in other countries, the proportion of underreported meningitis cases varied by etiological agent, and was lower for meningococcus and higher for other bacteria, including Pneumococcus and virus. 10th invasive and non-invasive diseases may be registered under the pneumococcal pneumonia code.The diagnosis at discharge of pneumococcal pneumonia probably accounts for identifi ed cases of S. pneumoniae.The low proportion of pneumococcal pneumonia in all hospitalizations for pneumonia in the SUS (1.0% for all ages; 1.4% among infants) corroborates this hypothesis.In a prospective study conducted in Salvador, northeastern Brazil, Pneumococcus was identifi ed in 1.8% of blood cultures from children hospitalized for pneumonia. 16 hospitalization rates are probably underestimated.Less severe syndromes (pneumonia and other diseases), for which blood cultures are less frequently taken, are probably underdiagnosed in routine care and underrepresented in this study.On the other hand, more serious cases with death but not hospitalized were not included either.Despite these limitations, this study provides valid nationwide estimates of PD hospitalization rates.
Mild IPD, particularly pneumonia and bacteremia without focus may be treated at outpatient settings.These cases were not included in the present study.
Non-invasive PD rates, although more frequent, are much more diffi cult to be estimated.The etiological agent is not identified in the majority of cases of pneumonia and otitis media.The Brazilian National Outpatient Database does not include diagnoses and standardized diagnostic data for outpatient consultations and notifi cation systems should be developed for these cases to be included in the estimates.
Surveillance of IPD may be improved by integrating information from different sources and increasing its sensitivity and specifi city, as seen in other countries. 7he SIH-SUS database has a nationwide coverage, easy access and a great volume of data, thus it is a potential source to complement information from laboratorybased and sentinel surveillance, especially in the less developed areas with weak surveillance systems. 2 PD is a major public health problem in Brazil.This study points out that the use of hospitalization annual rates should be an important part of a PD surveillance that can monitor the implementation and impact of PD vaccination program.

ACKNOWLEDGMENT
To Miriam Regina de Souza of Faculdade de Medicina da USP for her help with data collection.

Table 1 .
Annual hospitalization rates (per 100,000 population) for pneumococcal disease in the public sector according to year of diagnosis, clinical syndrome and age group.Brazil, 2004-2006 Figure 1.Annual hospitalization rates and in-hospital case-fatality rates for pneumococcal disease in the Brazilian National Health System, by age group and year of diagnosis.Brazil, 2004-2006.Figure 2. Distribution of hospitalizations for pneumococcal disease in the public sector, by clinical syndromes and age groups.Brazil, 2004-2006.Source: The Brazilian National Health System Hospital Database (SIH-SUS)

Table 2 .
Estimated overall hospitalization rates a (per 100,000) for pneumococcal disease according to year of diagnosis, clinical syndrome and age group.Brazil, 2004-2006.

Table 3 .
Case fatality rates (%) for pneumococcal disease by year of diagnosis, clinical syndrome and age group.Brazil, 2004-2006.In-hospital case fatality rates for sepsis, pneumonia, other syndromes and all syndromes are based on data retrieved from the Brazilian Hospital Database (SIH/SUS).For meningitis, case-fatality rates (CFRs) are based on both SIH and data from the Notifi able Diseases Database (SINAN).
a bCase-fatality rates for the age group 40-59 years as a whole.