Similar to the approach used in our previous review, we performed a computer search of the scientific literature (in English and other languages) published from January 2000 through June 2004 using the words rotavirus and the truncated stem rota-. We restricted the analysis to studies that met the following criteria: 1) were initiated after 1993; 2) were conducted for at least 1 full calendar year; and 3) examined rotavirus among at least 100 children <5 years of age hospitalized with diarrhea.

For each study, we determined the proportion of cases positive for rotavirus among children hospitalized with diarrhea. We plotted this proportion against the per capita gross national product (GNP) for the country in which the study was conducted. We then classified countries by per capita GNP into World Bank income groups (low, <US $756; low-middle, US $756–$2,995; high-middle, US $2,996–$9,265; and high, >US $9,265) (10), and calculated the median (interquartile range [IQR]) proportion of diarrhea hospitalizations attributable to rotavirus for each income group.

We next calculated an overall median detection rate by taking a weighted average of the median detection rates for each of the income groups. The weights assigned to each income group corresponded to the proportion of deaths from childhood diarrhea among countries in that income group as determined on the basis of our previous analysis (2): 85% in low-income countries, 13% in low-middle–income countries, 2% in high-middle–income countries, and <1% in high-income countries. To estimate deaths from rotavirus disease among children, we multiplied the overall median detection rate of rotavirus among children hospitalized with diarrhea by a recent World Health Organization estimate of deaths from diarrhea among children worldwide (5).

We abstracted information from 41 studies that met all the inclusion criteria (Table A1). Unlike the previous review of studies conducted for the period 1986–1999, in which the proportion of diarrhea-related hospitalizations attributable to rotavirus showed a distinct increasing trend with increasing income level, we found that the median detection rates increased only slightly with increasing income level (Figure 1). The median detection rate for rotavirus among children hospitalized with diarrhea was 39% in studies conducted in low-income countries, 40% for low-middle–income countries, 38% for high-middle–income countries, and 44% for high-income countries, for an overall weighted median estimate of 39% (Table).

If we multiply the greater median rotavirus detection rate of 39% (IQR 29%–45%) from this analysis by 1,566,000 recently estimated childhood diarrhea deaths (5), we find that rotavirus causes ≈611,000 childhood deaths (IQR 454,000–705,000). More than 80% of all rotavirus-related deaths were estimated to occur in low-income countries of south Asia and sub-Saharan Africa (Figure 2).

Compared with results from studies published from 1986 to 1999, the proportion of diarrhea hospitalizations attributable to rotavirus appears to have increased between 2000 and 2004. This phenomenon likely reflects a relatively slower rate of decrease in hospitalizations for rotavirus compared with other causes of severe childhood diarrhea. This finding could be accounted for by several factors. First, interventions to improve hygiene and sanitation are likely to have a greater impact on diarrhea caused by bacterial and parasitic agents, which are transmitted primarily through contaminated food or water, unlike rotavirus, which is often spread from person-to-person. This hypothesis is supported by data from the United States (11) and Mexico (12), which showed that as diarrhea-related childhood deaths decreased dramatically in both countries; the decrease was greatest during the summer months when diarrheal diseases caused by bacteria are more prevalent. In both countries, diarrhea-related deaths in recent years have exhibited peaks only in the winter when rotavirus infections are common. Second, oral hydration therapy to replace loss of body fluids, which many regard as a major factor responsible for the decrease in diarrhea deaths (13), is often more difficult to successfully administer in children with severe vomiting (14), a common manifestation of rotavirus disease. Third, unlike antimicrobial therapies that are effective against some bacterial and parasitic agents, no specific treatment for rotavirus infection is available.

We have derived preliminary updated estimates of rotavirus-related childhood deaths on the basis of the findings of our review. Because we wanted to assess the most recent trends in rotavirus incidence, we examined a relatively limited number of studies published in the last 5 years, particularly from upper-middle– and high-income countries. However, these 2 income groups account for only a small fraction (<5%) of all deaths from rotavirus disease, and the 28 studies available from low- and low-middle–income countries allowed for a reasonably robust analysis. Nevertheless, our findings should be updated as new data on rotavirus hospitalizations and updated estimates of childhood diarrhea deaths become available. In 2002, the World Health Organization published a generic protocol for hospital-based surveillance of rotavirus (15), and studies using this protocol are currently being conducted or planned in >30 countries in Asia, Africa, the Middle East, and Latin America. Data from these and other studies, particularly from countries such as India and China, which account for a large fraction of global rotavirus deaths, should be used to update our estimate of rotavirus-related deaths and further refine it to develop country-specific figures. These data, together with information on effects and costs of rotavirus disease, will allow policymakers to assess the magnitude of the problem of rotavirus and the value of new vaccines that may soon be available.