INVESTIGACIÓN ORIGINAL ORIGINAL RESEARCH

Economic and health burden of rotavirus gastroenteritis for the 2003 birth cohort in eight Latin American and Caribbean countries

Carga económica y de morbilidad de la gastroenteritis por rotavirus en la cohorte de nacidos en 2003 en ocho países de América Latina y el Caribe

Richard D. RheingansI, *; Dagna ConstenlaI; Lynn AntilI; Bruce L. InnisII; Thomas BreuerIII

IDepartment of Global Health, Rollins School of Public Health, Emory University, Atlanta, Georgia, United States of America
IIGlaxoSmithKline, King of Prussia, Pennsylvania, United States of America
IIIGlaxoSmithKline Biologicals, Rixensart, Belgium

ABSTRACT

OBJECTIVE: To estimate the health and economic burden of rotavirus gastroenteritis in hospital and outpatient settings in eight Latin American and Caribbean countries (Argentina, Brazil, Chile, Dominican Republic, Honduras, Mexico, Panama, and Venezuela).
METHODS: An economic model was constructed using epidemiological data from published articles, national health administration studies, and country-specific cost estimates. For each of the eight countries, the model estimated the rotavirus outcomes for the 2003 birth cohort during the first five years of life. The main outcome measures included health care costs, transportation costs, lost wages, and disease burden expressed in disability-adjusted life years. Estimates were expressed in 2003 US dollars. All future costs and disability-adjusted life year estimates were discounted at a rate of 3%. Sensitivity analyses evaluated the impact of specific variables on the medical cost of treating rotavirus.
RESULTS: For every 1 000 children born during 2003 in the eight Latin American and Caribbean countries studied here, we estimated that rotavirus gastroenteritis would result in an average of 246 outpatient visits, 24 hospitalizations, 0.6 deaths, and US$7 971 in direct medical costs during their first five years of life. The incidence of rotavirus-associated outpatient visits and the cost of outpatient visits were predicted to have the largest impact on the total medical cost per child. CONCLUSIONS: Rotavirus gastroenteritis is likely to result in substantial disease and economic burden to health systems in Latin American and Caribbean countries, and the foreseeable burden should be an important consideration in evaluating the cost-effectiveness of vaccination. Key words: Rotavirus; gastroenteritis; costs and cost analysis; cost of illness; models, economic; Latin America; Caribbean Region. RESUMEN OBJETIVO: Estimar la carga económica y de morbilidad de la gastroenteritis por rotavirus en hospitales y servicios ambulatorios de ocho países de América Latina y el Caribe (Argentina, Brasil, Chile, Honduras, México, Panamá, República Dominicana y Venezuela). MÉTODOS: Se elaboró un modelo económico a partir de datos epidemiológicos de artículos publicados, estudios de autoridades sanitarias nacionales y los estimados de costos específicos de cada país. El modelo calculó las consecuencias de la infección por rotavirus en los primeros cinco años de vida de la cohorte de nacidos en 2003 en cada uno de los ocho países estudiados. Las principales medidas de valoración fueron los costos de la atención sanitaria, los costos de transportación, los salarios perdidos y la carga de morbilidad expresada en años de vida ajustados por discapacidad. Los estimados se expresaron en dólares estadounidenses del año 2003. Se empleó una tasa de descuento de 3% para los cálculos de costos y años de vida ajustados por discapacidad de los años siguientes. El impacto de las variables específicas sobre los costos clínicos del tratamiento de la infección por rotavirus se realizó mediante análisis de sensibilidad. RESULTADOS: Se estimó que durante los primeros cinco años de vida, la gastroenteritis por rotavirus provoca en promedio 246 consultas externas, 24 hospitalizaciones, 0,6 muertes y gastos médicos directos por US$ 7 971,00 por cada 1 000 niños nacidos en 2003 en los ocho países estudiados de América Latina y el Caribe. Se prevé que la frecuencia de visitas médicas asociadas con la infección por rotavirus y los costos por consultas externas tengan el mayor impacto en los costos clínicos totales por niño.
CONCLUSIONES: La gastroenteritis por rotavirus puede representar una considerable carga económica y de morbilidad para los sistemas sanitarios de los países de América Latina y el Caribe. Se debe prestar una especial atención a la carga previsible al evaluar la efectividad en función del costo de la vacunación contra rotavirus.

Palabras clave: Infecciones por rotavirus, gastroenteritis, costos y análisis de costo, costo de enfermedad, modelos económicos, América Latina, Región del Caribe.

Rotavirus is the most important viral cause of acute gastroenteritis in children under 5 years of age, causing severe dehydrating diarrhea and vomiting in infants and young children worldwide. In Latin America, it is estimated that rotavirus gastroenteritis is associated with more than 15 000 deaths in under children under 5 years of age (1). In addition, several studies have documented that rotavirus gastroenteritis is a common cause of hospitalization in children of this age group (17). A recent review of 28 studies of children hospitalized with severe diarrhea in Latin America found that 31% to 38% of those hospitalizations were attributed to rotavirus gastroenteritis (8).

Although the main impact of rotavirus gastroenteritis is the morbidity and mortality it causes in children, information on the economic burden of disease can aid decisionmakers in choosing interventions to improve health. With the development of rotavirus vaccines, information on the economic burden of rotavirus and the potential costs averted by vaccination can provide decisionmakers with an estimate of the net impact of vaccination on financial resources, either in the health sector or for society as a whole.

Few studies have evaluated the economic burden imposed by rotavirus gastroenteritis in Latin America. Among publications retrieved by a MEDLINE search of health economic studies of rotavirus or gastroenteritis from 1995 to 2005, only one study was undertaken in Latin America (6). According to this study, the Peruvian Ministry of Health was estimated to spend over US$2.6 million annually to treat young children with rotavirus gastroenteritis (6). The purpose of this study was to estimate the impact of rotavirus gastroenteritis in children on the cost of direct medical treatment to the health care system in eight Latin American and Caribbean countries: Argentina, Brazil, Chile, the Dominican Republic, Honduras, Mexico, Panama, and Venezuela. In addition to health care costs, nonmedical direct (e.g., transportation) and indirect costs (e.g., productivity losses) were also considered. The results should provide insight into the reductions that could be achieved in the disease and economic burden by preventing rotavirus infection. METHODS The estimates provided in this analysis of the cost of rotavirus gastroenteritis to the health care system and society were obtained with a generalized model populated with epidemiological and cost data for the eight Latin American and Caribbean countries. The model estimated the expected rotavirus gastroenteritis-associated health outcomes and costs during the first five years of life for the 2003 annual birth cohort in each country (Figure 1). Country estimates were based on a combination of country-specific data and extrapolation from other countries for which data were lacking. Simulation techniques were used to calculate ranges for these estimates. The eight countries were selected to provide estimates for different representative settings within Latin America and the Caribbean, so geographic range and income diversity were considered in choosing the countries. Income levels ranged from a per capita gross domestic product (GDP) of 2003 US$ 1 001 in Honduras, a lower-middle-income country, to a per capita GDP of 2003 US$6 121 in Mexico, an upper-middle-income country (9). Larger countries were selected to cover a significant portion of the Latin American and Caribbean population. The availability of epidemiological data for a specific country was considered, although it was not a necessary condition for inclusion. The primary perspective for this analysis was that of society as a whole, including the direct medical costs, direct nonmedical costs, and indirect costs associated with medical treatment in formal inpatient and outpatient settings. Direct medical costs included the costs of tests, medication, supplies, facilities, and personnel needed for treatment. Nonmedical direct costs included the cost of transportation to and from the medical facility. Indirect costs were based on caregiver productivity losses that occur as a result of the child being sick. The main outcome measures were: (1) the health care costs of rotavirus gastroenteritis during the first five years of life for an annual birth cohort, (2) the nonmedical direct and the indirect costs for each cohort, and (3) the disease burden expressed in disability-adjusted life years (DALYs). All estimates were based on the expected events and costs for an annual birth cohort until 5 years of age, after which rotavirus infection rarely causes severe morbidity or mortality. Estimates were expressed in 2003 US$. All future costs and DALYs estimates were discounted at a rate of 3%.

Estimating disease burden of rotavirus gastroenteritis

For each country, the epidemiological burden of rotavirus gastroenteritis was estimated for the 2003 birth cohort until the age of 5 years. A quantitative model was used to predict the expected events during the first five years of life. Predicted estimates were modeled over time rather than by directly following the cohort prospectively in an epidemiological study for the five-year period. The key outcomes included hospitalizations, outpatient visits, and deaths. The risks of hospitalization, outpatient visit, and death were based on the cumulative risk of each event due to acute gastroenteritis, and the proportion of those events associated with rotavirus. Estimates of the number of events were calculated based on the 2003 birth cohort size and the estimated age distribution of each event.

Hospitalization. The 5-year cumulative risk of hospitalization for rotavirus gastroenteritis was estimated for each country based on epidemiological data from published articles and national health administration studies. Published estimates of hospitalization for diarrheal disease and rotavirus gastroenteritis were available for children under 3 years of age in Chile (4, 10), and for children under 5 years of age in Venezuela (4, 10). Because the Chilean study reported the rotavirus hospitalization rate for children under the age of 3 years, the incidence rate was converted to the under-five rate on the basis of published estimates (3, 7, 1113). The 5-year cumulative risk of hospitalization for rotavirus in Venezuela was adjusted since only one-third of the "hospitalized cases" reported in the study were actual admissions for overnight care or longer.1

The 5-year cumulative incidence of hospitalization for diarrhea in Argentina, Brazil, Honduras, and Mexico was based on national administrative data (1418). Because national administrative data do not include hospitalizations that occur in the private health care sector, the rates were adjusted to include this sector (5, 1820). Data for the incidence of hospitalization for diarrhea in the Dominican Republic and Panama were unavailable; therefore, cumulative incidence estimates for the two countries were based on the mean rate for Argentina, Brazil, Chile, Honduras, Mexico, and Venezuela (4, 10, 1418).

The total number of hospitalizations of children under 5 years of age for diarrhea recorded by the various sources was divided by the birth cohort (21) to obtain the five-year cumulative incidence of hospitalization for diarrhea in each of the eight countries. The proportion of hospitalizations for rotavirus infection was based on published estimates for Argentina (5), Brazil (2228), Chile (4), Mexico (29), and Venezuela (10). Because published estimates were not available for the Dominican Republic, Honduras, and Panama, the proportion of hospitalizations for diarrhea attributed to rotavirus for these countries was estimated based on other countries in Latin America and the Caribbean (46, 11, 29). The cumulative risk of hospitalization for rotavirus gastroenteritis was then calculated by multiplying the cumulative risk of hospitalization for diarrhea by the proportion attributed to rotavirus gastroenteritis.

Outpatient visits. The five-year cumulative incidence of outpatient visits for rotavirus gastroenteritis was estimated from published and administrative data. Published estimates of outpatient visits for rotavirus gastroenteritis were available in Chile for children under 3 years of age (4) and in Venezuela for children under 5 years of age (10). The Chilean data were adjusted to a five-year risk on the basis of published estimates (30, 31), and were adjusted to include data regarding nonemergency visits (32). The incidence of outpatient visits for rotavirus in Venezuela was adjusted to include outpatient visits in nonhospital settings, with data that were based on interviews with eight Venezuelan physicians.

The five-year cumulative incidence of outpatient visits for diarrhea in Argentina, Honduras, Mexico, and Panama was calculated from national administrative data (1618, 33, 34), and adjusted for inclusion of the private sector (5, 18, 20, 35). Because data were unavailable for Brazil and the Dominican Republic, the cumulative risk of an outpatient visit due to diarrhea in these two countries was based on the mean risk calculated for Argentina, Chile, Honduras, Mexico, Panama, and Venezuela.

For Brazil, Chile, and Venezuela the proportion of outpatient visits for diarrheal disease attributed to rotavirus was based on published data (4, 10, 3639). The proportion of outpatient visits attributed to rotavirus diarrhea for the Dominican Republic, Honduras, Mexico, and Panama was based on extrapolation from other countries in Latin America and the Caribbean (46, 10, 40). For Argentina, the rotavirus proportion was based on a published study (5) and a personal communication.2 The cumulative risk of an outpatient visit for rotavirus gastroenteritis was calculated by multiplying the cumulative risk of an outpatient visit for diarrhea by the proportion attributed to rotavirus gastroenteritis.

Mortality. The five-year cumulative incidence of diarrheal mortality was based on country-specific estimates, using ICD-10 categories A00A09 (41). National administrative data on deaths from diarrhea were available for Argentina, Brazil, Mexico, Panama, and Venezuela (21, 4245). Estimates of the risk of death due to diarrheal disease in Chile and the Dominican Republic were based on the WHO Mortality Database (46). For Honduras, estimates were based on the Pan American Health Organization estimate of mortality in children under 5 years of age (21), and on a national survey estimate of the proportion attributable to diarrhea (47). The five-year cumulative risk of death due to diarrheal disease was then calculated by dividing the estimated number of deaths due to diarrheal disease in under-five children by the birth cohort of each country (21). The cumulative risk of death due to rotavirus gastroenteritis was calculated by multiplying the risk of death due to diarrheal disease by the proportion of deaths from diarrhea attributed to rotavirus. Because no studies estimated the proportion of deaths from diarrhea attributed to rotavirus, the proportion was assumed to be the same as that for hospitalized cases.

Disability-adjusted life years. In addition to estimating the number of events, the disease burden was also expressed in terms of DALYs, an index that makes comparisons with other diseases possible. The DALYs estimate includes two components: years of life lost (YLLs) due to premature mortality, and years lived with disability (YLDs). The former was calculated based on the average country-specific life expectancy at birth and 1 year of age (48). To calculate YLDs, only morbidity from a disease severe enough to require medical care was considered. This component was calculated from default disability weights from the Global Burden of Disease Study (49) and the World Health Organization's guidelines for cost-effectiveness studies (50), assuming an average duration of illness of six days (51). Because of the short duration of illness from gastroenteritis, YLDs contributed very little to the DALYs estimate, so the estimate was based primarily on the YLLs. To ensure comparability, a discount rate of 3% and age weighting were included (49).

Table 1 summarizes the values used to calculate the disease burden for rotavirus gastroenteritis.

Estimating medical costs associated with rotavirus gastroenteritis

The economic burden of rotavirus gastroenteritis to the health care system in the eight countries was estimated by combining the number of each type of event with information on the costs associated with the event. This information was combined in a spreadsheet-based decision-analytic model. For hospital and outpatient events, these were partitioned into the cost of the visit (including facilities and personnel) and the cost of the resources used for treatment (diagnostic tests and medication). For each country, baseline estimates were based on resource use information from a prospective rotavirus surveillance study and facility-specific cost data. To compare costs among countries, two alternative sources of cost estimates were used: (1) the WHO-CHOICE (Choosing Interventions that are Cost Effective) project (52) and (2) estimates obtained from physician interviews.

Prospective rotavirus surveillance study. Economic cost data were collected as part of an ongoing rotavirus gastroenteritis surveillance study done prior to a randomized vaccine trial (53). The prospective study was conducted in one or more sites in each of the eight countries. Participating sites generally consisted of secondary- or tertiary-level facilities that provided inpatient and outpatient care to children. Eligible cases were children less than 24 months of age who came to the center with severe gastroenteritis, defined as an episode of diarrhea (the passage of three or more loose or watery stools within a 24-hour period), with or without vomiting, that required overnight hospitalization or rehydration therapy equivalent to World Health Organization (WHO) plan B (oral rehydration therapy) or plan C (intravenous rehydration therapy), in a medical facility such as a hospital, clinic, or supervised rural health care center. At each site, economic data were collected for a sample of approximately 60 inpatients and 60 outpatients. Sample size was determined based on the recommendation of the WHO guide to costing diarrheal disease (54) in order to ensure an estimated mean direct medical cost within 15% of the actual mean. Stool samples were tested for rotavirus with an enzyme-linked immunoassay.

Health economic data were collected during 2003. The surveillance study contained an economic component to calculate the direct medical costs, direct nonmedical costs, and indirect costs associated with rotavirus gastroenteritis. A cost questionnaire was administered to the caregivers of patients with gastroenteritis. Questions asked about nonmedical direct costs (cost of transporting the child to the hospital and visiting the child in the hospital) and indirect costs (time lost from paid work to care for the child). Investigators also reported patient data regarding the length of stay and the utilization of diagnostics and medication to treat gastroenteritis.

Survey data were analyzed using SAS statistical software (SAS Institute Inc., Cary, North Carolina, United States of America). Separate estimates were generated for inpatients and outpatients. Inpatients were those who were classified as inpatients and spent a minimum of one day in the hospital, and outpatients were those who were not classified as inpatients and spent a maximum of one day at the hospital or clinic. For countries in which outpatients can stay at the hospital "under observation" but are not admitted, the outpatient definition was adapted to include patients who stayed for more than one day at the center's rehydration facility.

Health care costs were calculated as the sum of the visit cost at the hospital or outpatient setting, the cost of diagnostics, and the cost of medication. Estimates of the hospital per diem costs were based on information from hospital finance departments of public health care facilities, with the exception of Chile since per diem data were available from private facilities as well. The cost per stay as an inpatient was calculated by multiplying the per diem rate (which includes the cost of accommodation, administration, food, and personnel) by the duration of stay (between three and four days).

The cost per outpatient visit was calculated as the weighted mean of the cost of visiting a pediatrician, general practitioner, emergency room, or rehydration facility based on the proportion of patients seen in each of these four outpatient settings. This information was obtained from physician interviews done in 2003. Health care utilization patterns varied by country; however, the proportion seen by a pediatrician or in an emergency department was generally greater than the proportion seen by a general practitioner or in a rehydration facility. Costs for these types of outpatient visits were based on the average costs for public facilities.

Mean total diagnostic and medication costs for treating gastroenteritis were calculated for both inpatients and outpatients for each country. The unit cost for each diagnostic test was based on estimates provided by the finance departments of public hospitals, and costs associated with medications were based on the listing of medications and national formularies (55, 56).

Alternative health care cost estimates. To compare the findings across countries, we also estimated inpatient and outpatient visit costs with a standardized approach developed by WHO for its WHO-CHOICE project (52) and data from cost studies of diarrhea (57). The WHO-CHOICE project provided estimates of the per-diem and per-visit cost of public health care facilities for 14 epidemiological categories based on geographical region and mortality stratum. The per-diem and per-visit cost estimates included the cost of the facility, personnel, equipment, and food, and excluded the cost of medications and diagnostic tests (52). Based on information from cost studies of diarrheal disease conducted in Mexico and the Philippines, we estimated that these costs would add 33% to the cost of hospitalization, and 65% to the cost of outpatient visits (57).

Because the WHO-CHOICE model was developed using year-2000 international dollars, all costs were converted to 2003 US$using the consumer price index, purchasing power parity conversion factors, and official exchange rates (9, 58). "International dollars" are used to express the value of a quantity in a local currency based on its purchasing power compared to a US$ rather than the actual exchange rate. A hospital per-diem rate was estimated for each country as a weighted average of estimates for public primary-, secondary-, and tertiary-level facilities, using the proportions of each from the multicountry study (59). The cost of hospitalization was calculated as the product of the per-diem cost and a duration of 3.5 days (6, 60, 61). Country-specific outpatient visit costs were calculated by adjusting the WHO- CHOICE regional estimate proportionally to the country-specific hospital per-diem cost.

The second alternative approach used physician interviews to estimate resource utilization for the treatment of rotavirus gastroenteritis, along with facility- and country-specific unit cost estimates. Interviews of four to eight physicians in each country, mostly from public health settings, were conducted in 2003 to characterize the typical management of rotavirus gastroenteritis. Interviews included questions regarding care for patients with moderate rotavirus gastroenteritis, and for patients with severe rotavirus gastroenteritis.

Estimating societal costs of rotavirus gastroenteritis

In addition to health care costs, nonmedical direct (e.g., transportation costs) and indirect costs (e.g., productivity losses) were also calculated based on the prospective rotavirus surveillance study. Caregivers of inpatients and outpatients were asked how much money they spent to transport their child or themselves to the health facility. Caregivers were also asked to quantify time lost from paid work due to their child's illness. The average indirect cost was estimated by multiplying the mean hours lost by the mean female hourly wage for each of the countries (6265).

Uncertainty and sensitivity analyses

Although country-specific data on the epidemiology of the disease and the associated health care costs were available for most countries, the quality and relevance of the data were limited for others, a limitation that created uncertainties regarding the final economic burden estimates. Two approaches were used to address this uncertainty: sensitivity and uncertainty analyses. A one-way sensitivity analysis was conducted by varying individual input parameters by ± 20% to assess their impact on the economic burden in each country. An uncertainty analysis was conducted to evaluate the overall impact of uncertainty on the estimates, to develop upper and lower bounds for the estimates, and to assess the contribution of individual parameters to overall uncertainty.

For each country, burden estimates were calculated in a spreadsheet-based decision-analytic model that estimated the frequency of different events and the costs associated with each. A Monte Carlo model was developed based on this model. In Monte Carlo analysis, individual point estimates of parameters are replaced with distributions of potential values (66). In a series of iterations, individual values are randomly selected from each of the distributions, and results are calculated and stored. The process is repeated for a large number of iterations (10 000 in this case). The final product is a distribution of potential outcomes that describe the likely range of actual expected results.

For national disease burden variables, distributions were used to characterize the cumulative incidence of illness outcomes (hospitalization, outpatient visits, and death) in each country, and the proportion due to rotavirus. The distributions specified a range around the value chosen for analysis, and described the likelihood that the true population value would be the value chosen. Wider distributions were used for countries for which the estimates were extrapolated from other countries. For the cost variables, input distributions were based on the mean costs and standard errors from the baseline analysis described above. Table 1 shows the best estimates and distributions used in the analysis.

Upper- and lower-bound uncertainty limits (5% and 95%) were estimated for the key output parameters: direct medical costs and total societal costs for each country. In addition, a contribution-to-variance analysis was done to determine the contribution of the individual input parameters to the overall uncertainty regarding total societal costs.

RESULTS

Disease burden

In the eight countries together, which represent 68% of the total annual births in Latin America, rotavirus gastroenteritis was predicted to be responsible for approximately 2 000 000 outpatient visits, 190 000 hospitalizations, and 5 000 deaths for the annual birth cohort. In each of the countries, per 1 000 children, rotavirus gastroenteritis was predicted to result in 205 to 359 outpatient visits, 7 to 35 hospitalizations, and 0.05 to 3.2 deaths (Table 2). The DALYs resulting from these events ranged from 2 to 107 per 1 000 children, with lower rates observed in higher-income countries due to lower mortality.

Economic burden

The mean predicted costs (direct medical, direct nonmedical, and indirect) for hospital and outpatient events associated with gastroenteritis are shown in Table 3. In each of the countries, direct medical costs were estimated to account for the majority of the total costs of hospitalizations and outpatient visits. The cost of medications and diagnostics was expected to account for 5% to 25% of the medical costs of hospitalization, and 30% to 75% of the costs of outpatient visits. The total estimated medical costs per hospitalization ranged from US$72.49 in the Dominican Republic to US$ 190.12 in Panama, and the total estimated medical costs per outpatient visit ranged from US$6.54 in the Dominican Republic to US$ 33.99 in Chile. The calculated treatment costs for hospital and outpatient events were consistent with estimates provided by the two alternative methods (WHO-CHOICE and physician interviews).

Table 4 shows the estimated health care system and societal costs of these rotavirus-associated events for the 2003 birth cohort in each country, and upper and lower uncertainty limits for these estimates. From the health care system perspective, the average estimated direct medical costs ranged from US$3 400 to over US$ 16 000 per 1 000 births, with higher costs tending to occur in the higher-income countries. In all countries except Argentina and Brazil, where rotavirus hospitalization rates are higher, the majority of the total costs were predicted to result from outpatient visits. From the societal perspective, total estimated costs (including nonmedical and indirect costs) of rotavirus gastroenteritis ranged from US$5 000 to almost US$ 19 000 per 1 000 births. Overall, direct medical costs accounted for 73% of the total estimated cost in the eight countries. Hospitalization events accounted for 17% to 70% of direct estimated medial costs, and 12% to 60% of total estimated costs in each country.

Sensitivity and uncertainty

The results of the one-way sensitivity analysis are shown in Table 5. The table reports the estimated cost per 1 000 children in 2003 US$for each country, using three scenarios: baseline, a 20% increase in the selected value, and a 20% decrease. Compared to the baseline figures, a 20% increase or a 20% decrease in rotavirus hospitalization or outpatient visit incidence resulted in a 2% to 12% and an 8% to 18% change in total costs, respectively. Similarly, a 20% increase or decrease in average direct medical costs for hospitalization or outpatient visits resulted in a 2% to 11% and a 4% to 13% change in total costs, respectively. The results of the uncertainty analyses are shown in Table 4 and Table 6. Upper and lower uncertainty bounds are provided in Table 4 for the estimates of direct medical and societal costs of rotavirus gastroenteritis. The uncertainty limits were smallest when country-specific data were available for estimating key epidemiological parameters (Argentina, Chile, and Venezuela). The contribution-to-variance analysis summarized in Table 6 examined the effect of selected input parameters on the total estimated cost of rotavirus gastroenteritis in each country. Uncertainty in the incidence of outpatient visits for diarrhea and the proportion of these visits ascribed to rotavirus diarrhea contributed at least 7% to the overall uncertainty in the economic burden for each country, although the contribution was much higher in some countries such as the Dominican Republic, Honduras, Mexico, and Panama. The incidence of hospitalization for diarrhea and the proportion of hospitalizations ascribed to rotavirus diarrhea were highest in Brazil, where these factors accounted for 16% to 17% of the overall uncertainty. Direct medical costs were a relatively larger source of variability in countries for which there was more certainty regarding country-specific epidemiological inputs (e.g., hospitalizations in Argentina or Chile, and outpatient visits in Honduras). Indirect costs for hospitalizations contributed a small fraction to the overall uncertainty; however, indirect costs for outpatient visits made larger contributions to overall uncertainty, especially in Brazil, Mexico, Panama, and Venezuela. DISCUSSION The study demonstrates that rotavirus gastroenteritis can be expected to result in a large health and economic burden in the eight Latin American and Caribbean countries studied. For every 1 000 children born, rotavirus gastroenteritis was estimated to result in an average of 246 outpatient visits, 24 hospitalizations, 0.6 deaths, and 2003 US$ 7 971 in direct medical costs during their first five years of life. These rates translated into estimates of approximately 2 000 000 outpatient visits, 190 000 hospitalizations, 5 000 deaths, and US$62.4 million in direct medical costs for the 2003 annual birth cohort by the age of 5 years. Direct medical expenses from hospitalizations and outpatient visits were predicted to account for costs of up to US$ 16 per child (US$3.4 to US$ 16.2, depending on the country). When productivity losses for caregivers and additional out-of-pocket expenses were also considered, the estimated societal cost rose to almost US$19 per child (US$ 5.1 to US$18.9, depending on the country). For the eight countries together, rotavirus gastroenteritis was estimated to lead to costs of US$ 85 million (including societal costs) for the 2003 birth cohort. This estimate did not include additional costs associated with cases that are treated at home or by traditional healers and pharmacists.

The estimated number of deaths due to rotavirus in the current study is lower than the estimates by Parashar and colleagues, which were based on regression analysis rather than country-specific data (1). The observed differences were greatest for Chile, Argentina, and Mexico, all of which are upper-middle-income countries. Because most of the studies used in Parashar's regression model were from low-income and lower-middle-income countries (1), their model may not perform as accurately for higher-income countries.

In terms of public health, deaths and hospitalizations for severe cases of gastroenteritis are of greatest concern. However, from an economic perspective, outpatient visits are also of concern since they can generate a large portion of the costs of rotavirus gastroenteritis. Outpatient visits accounted for an estimated 30% to 83% of the direct medical costs in each country, and accounted for an even greater fraction of the total estimated societal costs (40% to 88%). The proportion of total costs associated with outpatient visits was higher in countries with lower rates of rotavirus hospitalizations, such as Mexico and Honduras.

Rotavirus gastroenteritis was predicted to result in substantial disease and economic burdens in each of the eight countries studied; however, the pattern differed slightly among countries. As expected, the greatest disease burden (in terms of DALYs ) was predicted for lower-middle-income countries, especially the Dominican Republic and Honduras. In contrast, the greatest economic burden was foreseen in upper-middle-income countries due to their higher treatment costs per child and higher productivity losses. Variability in health care costs may be due to differences in treatment patterns (medications, diagnostics, and relative importance of inpatient and outpatient treatment) and input costs (primarily labor).

Figure 2 shows the predicted health and economic burdens of rotavirus (DALYs per 1 000 births and total cost per child) by per-capita income in each of the countries. Although higher-income countries tended to have higher economic burdens, and lower-income countries tended to have higher health burdens, many middle-income countries had large health and economic burdens associated with rotavirus gastroenteritis. This may reflect real differences between countries. For children who have access to timely medical attention, rotavirus gastroenteritis is likely to generate economic costs. For children with little access to medical care, gastroenteritis may be more likely to result in mortality (an increase in DALYs ).

The sensitivity and uncertainty analyses provide an important tool for assessing the validity of the burden estimates presented here. Although there is uncertainty about the country-specific epidemiological and economic inputs used in the model, the overall uncertainty bounds presented in Table 4 suggest that the economic burden estimates are quite robust. In countries where uncertainty bounds are wider, estimates could be improved with additional data for input variables that contribute the most to overall uncertainty. This is particularly true for countries where epidemiological estimates were extrapolated from other countries.

The primary limitations of this study relate to the availability of country-specific incidence data for severe rotavirus gastroenteritis events. The potential impact of these data limitations was systematically assessed in the sensitivity and uncertainty analyses to determine their importance. Another limitation is that the study did not include the costs borne by families for treatment of rotavirus gastroenteritis in less formal settings, including treatment at home or by traditional healers. As a result, the estimates presented here are likely to underestimate the total economic burden of rotavirus gastroenteritis.

The evidence presented in this study lends support to the conclusion that rotavirus gastroenteritis can be expected to produce substantial morbidity, mortality, and health care system costs in Latin America and the Caribbean. For every 1 000 children born in 2003, we estimate that rotavirus will result in 246 outpatient visits, 24 hospitalizations, 0.6 deaths, and US$7 971 in direct medical costs during their first five years of life. The eight countries studied here may spend an estimated total US$ 62.4 million annually in direct medical costs arising from the treatment of rotavirus gastroenteritis. When the costs of transportation and time lost from work are included, the total estimated societal burden of rotavirus gastroenteritis is approximately US\$ 86 million for the eight countries. This estimate does not include the additional nonmedical costs borne by families, or the costs of treatment in less formal settings, which may comprise an important proportion of the total economic burden imposed by rotavirus. Although rotavirus is often described as a democratic virus because it infects almost all children by the age of 5 years, the nature of the resulting burden can differ greatly. In higher-income countries the economic burden is likely to be greater, whereas in lower-income countries it is the health burden that can be expected to be greater.

As vaccines become available to control rotavirus gastroenteritis, information on the economic and health burden of disease in Latin America and the Caribbean, as well as in other regions, becomes more important. Health care resources used to treat rotavirus gastroenteritis could be used to address other health challenges or to offset the costs of vaccination. These economic burden estimates provide the foundation for further evaluations of the cost-effectiveness of rotavirus vaccination.

Acknowledgments. Financial support for this project was provided by GlaxoSmithKline Biologicals. The authors would like to acknowledge the help of GSK Medical Advisors (salaried employees) for their valuable input in generating epidemiological data. The GSK Medical Advisors were: Yolanda Cervantes, GSK-Mexico; Marisol Navarrete, GSK-Chile; Eduardo Ortega, GSK-Caribbean and Central American Region; Pilar Rubio, GSK-Brazil; Ricardo Ruttiman, GSK-Argentina; José Tavares, GSK-Brazil; and Juan Pablo Yarzabal, GSK-Venezuela. We are indebted to Miguel O'Ryan, Maribel Rivera, Jorge Gómez, Irene Pérez-Schael, and Alexandre Linhares for their assistance with identifying epidemiological data and for their general support of the study. We also thank Ralf Clemens for reviewing the manuscript.

Note on conflict of interest. The Pan American Health Organization (PAHO) has purchased vaccines from GlaxoSmithKline, and PAHO has also received contributions from GlaxoSmithKline. While the Revista Panamericana de Salud Pública/Pan American Journal of Public Health is affiliated with PAHO, the Revista/Journal is an independent scientific publication whose articles do not necessarily reflect the opinions or official positions of PAHO on specific issues. The mention of particular companies or of certain manufacturers' products in the Revista/Journal does not imply that they are endorsed or recommended by PAHO in preference to other ones of a similar nature. As with all other research articles published in the Revista/Journal, this article went through the regular process of peer review by outside experts.

REFERENCES

1. Parashar UD, Hummelman EG, Bresee JS, Miller MA, Glass RI. Global illness and deaths caused by rotavirus disease in children. Emerg Infect Dis. 2003;9(5):56572.

2. Linhares AC, Bresee JS. Rotavirus vaccines and vaccination in Latin America. Rev Panam Salud Publica. 2000;8(5):30531.

3. Pérez-Schael I, González R, Fernández R, Alfonzo E, Inaty D, Boher Y, et al. Epidemiological features of rotavirus infection in Caracas, Venezuela: implications for rotavirus immunization programs. J Med Virol. 1999;59(4): 5206.

4. O'Ryan M, Perez-Schael I, Mamani N, Pena A, Salinas B, Gonzalez G, et al. Rotavirus-associated medical visits and hospitalizations in South America: a prospective study at three large sentinel hospitals. Pediatr Infect Dis J. 2001;20(7):68593.

5. Gomez JA, Sordo ME, Gentile A. Epidemiologic patterns of diarrheal disease in Argentina: estimation of rotavirus disease burden. Pediatr Infect Dis J. 2002;21(9):84350.

6. Ehrenkranz P, Lanata CF, Penny ME, Salazar-Lindo E, Glass RI. Rotavirus diarrhea disease burden in Peru: the need for a rotavirus vaccine and its potential cost savings. Rev Panam Salud Publica. 2001;10(4):2408.

7. Bok K, Castagnaro N, Borsa A, Nates S, Espul C, Fay O, et al. Surveillance for rotavirus in Argentina. J Med Virol. 2001;65(1):1908.

8. Kane EM, Turcios RM, Arvay ML, Garcia S, Bresee JS, Glass RI. The epidemiology of rotavirus diarrhea in Latin America. Anticipating rotavirus vaccines. Rev Panam Salud Publica. 2004;16(6):3717.

9. World Bank Group. World development indicators online 2004. Available from: http:// www.worldbank.org/data/wdi2004/. Accessed 16 August 2004.

10. Salinas B, Gonzalez G, Gonzalez R, Escalona M, Materan M, Schael IP. Epidemiologic and clinical characteristics of rotavirus disease during five years of surveillance in Venezuela. Pediatr Infect Dis J. 2004;23(10 Suppl): S1617.

11. Cardoso DD, Soares CM, Dias e Souza MB, de Azevedo MDS, Martins RM, Queiroz DA, et al. Epidemiological features of rotavirus infection in Goiânia, Goiás, Brazil, from 1986 to 2000. Mem Inst Oswaldo Cruz. 2003;98(1): 259.

12. Barraza P, Avendaño LF, Spencer E, Calderón A, Prenzel I, Duarte E. Infección intrahospitilaria por rotavirus en lactantes, Santiago, Chile. Bol Oficina Sanit Panam. 1986;101(4): 32838.

13. Bok K, Castagnaro NC, Diaz NE, Borsa A, Cagnoli MR, Nates S, et al. Red de laboratorios de rotavirus: resultados del primer año de vigilancia. Rev Argent Microbiol. 1999; 31(1):112.

14. Brasil, Ministério da Saúde, Departamento de Informática do SUS (Sistema Único de Saúde) (DATASUS). Morbidade hospitalar do SUS: diarréia e gastroenterite origem infecc presumív; 2002. Available from: http://tabnet. datasus.gov.br/cgi/sih/mimap.htm. Accessed 20 September 2004.

15. Argentina, Ministerio de Salud y Ambiente de la Nación, Departamento Nacional de Estadísticas de Salud, Dirección de Estadísticas e Información de Salud. Egresos de establecimientos oficiales según variables seleccionadas. Buenos Aires: Ministerio de Salud y Ambiente; 2003.

16. Honduras. Secretaría de Salud, Departamento de Estadísticas. Informe de enfermedades transmisibles (TRANS). Tegucigalpa: Secretaría de Salud; 2002.

17. México, Instituto Mexicano de Servicios Sociales. Daños. In: Dirección de prestaciones médicas: información estadística en salud; 2002. Available from: http://www.imss.gob. mx/IMSS/IMSS_SITIOS/DPM/. Accessed 25 January 2004.

18. México, Servicio Social de Salud, Dirección de Estadísticas e Información de Salud. Anuario Estadístico 20002001. In: Sistema Nacional de Información en Salud; 20012002. Available from: http://www.salud.gob.mx/. Accessed 12 February 2004.

19. Carvalho DM. Grandes sistemas nacionais de informação em saúde: revisão e discussão da situação atual. Inf Epidemiol SUS. 1997;5: 746.

20. Pan American Health Organization. Volume II: Health in the Americas, Honduras. 1998 ed. Washington, D.C.: PAHO; 1998. Pp. 331342. Available from: http://www.paho.org/ english/HIA1998/Honduras.pdf. Accessed 8 October 2004.

21. Pan American Health Organization. Regional core health data systemtable generator 2004. Washington D.C.: PAHO; 2004. Available from: http://www.paho.org/Project. asp? SEL=HD&LNG=ENG&ID=379. Accessed 6 August 2004.

22. Carmona RC, Timenetsky Mdo C, da Silva FF, Granato CF. Characterization of rotavirus strains from hospitalized and outpatient children with acute diarrhoea in Sao Paulo, Brazil. J Med Virol. 2004;74(1):16672.

23. da Silva Domingues AL, da Silva Vaz MG, Moreno M, Camara FP. Molecular epidemiology of group A rotavirus causing acute diarrhea in infants and young children hospitalized in Rio de Janeiro, Brazil, 19951996. Braz J Infect Dis. 2000;4(3):11925.

24. Linhares AC, Moncao HC, Gabbay YB, de Araujo VL, Serruya AC, Loureiro EC. Acute diarrhoea associated with rotavirus among children living in Belem, Brazil. Trans R Soc Trop Med Hyg. 1983;77(3):38490.

25. Gusmão RH, Mascarenhas JD, Gabbay YB, Lins-Lainson Z, Ramos FL, Monteiro TA, et al. Rotavirus subgroups, G serotypes, and electrophoretypes in cases of nosocomial infantile diarrhoea in Belem, Brazil. J Trop Pediatr. 1999;45(2):816.

26. da Rosa e Silva ML, Naveca FG, Pires de Carvalho I. Epidemiological aspects of rotavirus infections in Minas Gerais, Brazil. Braz J Infect Dis. 2001;5(4):21522.

27. Coiro JR, Bendati MM, de Almeida Neto AJ, Heuser CF, Vasconcellos VL. Rotavirus infection in Brazilian children with acute enteritis: a seasonal variation study. Am J Trop Med Hyg. 1983;32(5):11868.

28. Cardoso DdD, Martins RM, Kitajima EW, Barbosa AJ, Camarota SC, Azevedo MS. Rotavirus e adenovirus em crianças de 05 anos hospitalizadas com ou sem gastrenterite em Goiânia-GO., Brasil. Rev Inst Med Trop Sao Paulo. 1992;34(5):4339.

29. Weuthrich B, ed. Proceedings of the Sixth International Rotavirus Symposium. Washington, D.C.: Albert B. Sabin Vaccine Institute; 2005.

30. González FS, Sordo ME, Rowensztein G, Sabbag L, Roussos A, De Petre E, et al. Diarrea por rotavirus: impacto en un hospital de niños de Buenos Aires. Medicina 1999;59(4):3216.

31. Urrestarazu MI, Liprandi F, Pérez de Suárez E, González R, Pérez-Schael I. Características etiológicas, clínicas y sociodemográficas de la diarrea aguda en Venezuela. Rev Panam Salud Publica. 1999;6(3):14956.

32. Chile, Ministerio de Salud, Dirección de Estadísticas e Información de Salud. Notificaciones SNSS recepcionadas en DEIS/MINSAL. Santiago de Chile: DEIS; 2004.

33. Panamá, Ministerio de Salud, Departamento de Vigilancia de Factores Protectores y de Riesgos a la Salud y Enfermedad, Dirección de Epidemiología. Estadísticas de salud 2002. Panamá: MINSA; 2002.

34. Argentina, Ministerio de Salud. Boletín epidemiológico nacional. Buenos Aires: Ministerio de Salud; 2002.

35. Panamá, Contraloría General de la República, Dirección de Estadística y Censo. Situación demográfica. Estimaciones y proyecciones de la población total del país, por sexo y edad: años 19502050. Ciudad de Panamá: Dirección de Estadística y Censo; 2002. (Boletín n°7, noviembre de 2002).

36. Orlandi PP, Silva T, Magalhaes GF, Alves F, de Almeida Cunha RP, Durlacher R, et al. Enteropathogens associated with diarrheal disease in infants of poor urban areas of Porto Velho, Rondônia: a preliminary study. Mem Inst Oswaldo Cruz. 2001;96(5):6215.

37. Bittencourt JA, Arbo E, Malysz AS, Oravec R, Dias C. Seasonal and age distribution of rotavirus infection in Porto Alegre, Brazil. Braz J Infect Dis. 2000;4(6):27983.

38. Stewien KE, da Cunha LC, Alvim Ade C, dos Reis Filho SA, Alvim MA, Brandão AA, et al. Rotavirus associated diarrhoea during infancy in the city of S. Luís (MA), Brazil: a two-year longitudinal study. Rev Inst Med Trop Sao Paulo. 1991;33(6):45964.

39. Teixeira JM, de Figueiredo RB, dos Santos HM, Ferreira MN, Camara GN. Aspectos epidemiológicos das infecções por rotavirus no Distrito Federal, Brasil. Rev Soc Bras Med Trop. 1991;24(4):22330.

40. Guardado JA, Clara WA, Turcios RM, Fuentes RA, Valencia D, Sandoval R, et al. Rotavirus in El Salvador: an outbreak, surveillance and estimates of disease burden, 20002002. Pediatr Infect Dis J. 2004;23(10 Suppl):S15660.

41. World Health Organization. International statistical classification of diseases and health related problems: ICD-10. Second edition. Geneva: WHO; 2005.

42. Venezuela, Ministerio de Salud y Desarrollo Social (MSDS), Dirección de Información Social y Estadística. Anuario de mortalidad 2001. Available from: http://www.msds.gov. ve/msds/direcciones_msds/Epidemiologia/ Estadistica/Anuarios/Anuario01.pdf. Accessed 6 September 2004.

43. México, Secretaría de Salud, Sistema Nacional de Información en Salud. Principales causes de mortalidad infantil y edad preescolar 2002. Available from: http://www.salud.gob.mx/ apps/htdocs/estadisticas/mortalidad/ mortalidad.htm. Accessed 13 September 2004.

44. Brasil, Ministério da Saúde. Anuário estatístico de saúde do Brasil; 2001. Available from: www. datasus.gov.br. Accessed 6 September 2004.

45. Argentina, Ministerio de Salud. Agrupamiento de causas de mortalidad por división, político territorial de residencia, edad y sexo. Buenos Aires: Ministerio de Salud; 2002.

46. World Health Organization, WHO Statistical Information System. WHO mortality database. Available from: http://www3.who.int/ whosis/menu.cfm?path=whosis,mort& language=english. Accessed 1 June 2004.

47. Honduras, Ministerio de Salud. Encuesta Nacional de Epidemiología y Salud Familiar (ENESF-2001) y Encuesta Nacional de Salud Masculina (ENSM-2001). Tegucigalpa: Ministerio de Salud; 2002.

48. World Health Organization, WHO Statistical Information System. Life tables for 191 countries. Geneva: WHO; 2001. Available from: http://www3.who.int/whosis/menu.cfm? path=whosis,bod,burden_statistics,life& language=english. Accessed 21 October 2004.

49. Murray CJL, Lopez AD. The global burden of disease: a comprehensive assessment of mortality and disability from diseases, injuries and risk factors in 1990 and projected to 2020. Cambridge: Harvard University Press; 1996.

50. Baltussen R, Adam T, Tan Torres T, Hutubessy R, Acharya A, Evans D, et al. Generalized cost effectiveness analysis: a guide. Geneva: World Health Organization; 2002.

51. Liddle JL, Burgess MA, Gilbert GL, Hanson RM, McIntyre PB, Bishop RF, et al. Rotavirus gastroenteritis: impact on young children, their families and the health care system. Med J Aust. 1997;167(6):3047.

52. World Health Organization, WHO-CHOICE. Prices for hospitals and health centres. Available from: http://www3.who.int/whosis/ cea/prices/unit.cfm?path=evidence,cea,cea_prices,cea_prices_unit&language=english. Accessed 20 January 2004.

53. Ruiz-Palacios GM, Pérez-Schael I, Velázquez FR, Abate H, Breuer T, Clemens SC, et al. Safety and efficacy of an attenuated vaccine against severe rotavirus gastroenteritis. N Engl J Med. 2006;354(1):1122.

54. World Health Organization. Guidelines for estimating the economic burden of diarrhoeal disease with focus on assessing the costs of rotavirus diarrhoea. Geneva: WHO, Department of Vaccines and Biologicals; 2005.

55. World Health Organization. Anatomical therapeutic chemical (ATC) classification index with defined daily doses (DDDs), 2000. Oslo: WHO Collaborating Centre for Drug Statistics Methodology; 2000.

57. World Health Organization, Department of Vaccines and Biologicals. Review of treatment cost protocol studies. Report of a meeting, Geneva, 1112 December 2000. Geneva: WHO, Department of Vaccines and Biologicals; 2001.

58. United States, Department of Labor, Bureau of Labor Statistics. Inflation and consumer spending, 2003. Available from: http://www. bls.gov/bls/inflation.htm. Accessed 19 August 2004.

59. Adam T, Evans DB, Murray JL. Econometric estimation of country-specific hospital costs. Cost Eff Resour Alloc. 2003;1(3):110.

60. Bittencourt SA, Leal Mdo C, Santos MO. Hospitalizações por diarréia infecciosa no Estado do Rio de Janeiro. Cad Saude Publica. 2002; 18(3):74754.

61. Gómez JA, Nates S, De Castagnaro NR, Espul C, Borsa A, Glass RI. Anticipating rotavirus vaccines: review of epidemiologic studies of rotavirus diarrhea in Argentina. Rev Panam Salud Publica. 1998;3(2):6978.

62. Organización Internacional del Trabajo. Honduras. Tasas de salarios mínimos vigentes, por tamaño de empresa, según rama de actividad desde el 1 de julio de 1999. Available from: http://www.oit.or.cr/oit/papers/sal_ hn99.htm. Accessed 5 October 2004.

63. International Labor Organization. Labour statistics. Volume 2: employment, wages, hours of work and labour cost (establishment surveys). Available from: http://laborsta.ilo. org/. Accessed 14 September 2004.

64. Argentina, Instituto Nacional de Estadística y Censos. Encuesta Permanente de Hogares. Buenos Aires: INDEC; 2003.

65. Venezuela, Instituto Nacional de Estadística, Ministerio de Sanidad y Departamento Social, Dirección de Información Social y Estadística. Encuesta de hogares. Caracas: INE; 2003.

66. Baltussen R, Hutubessy R, Evans D, Murray C. Uncertainty in cost-effectiveness analysis: probabilistic uncertainty analysis and stochastic league tables. Int J Technol Assess Health Care. 2002;18(1):1129.