Diarrhea- and Rotavirus-Associated Hospitalizations Among Children Less Than 5 Years of Age: United States, 1997 and 2000
OBJECTIVE. A new rotavirus vaccine may be licensed in the United States in early 2006. Estimates of the burden of severe rotavirus disease, particularly hospitalizations, will help evaluate the potential benefits of a national rotavirus immunization program.
DESIGN. The Kids' Inpatient Database, a robust sample of 10% of the uncomplicated births and 80% of other pediatric discharges was used to estimate the number and rate of diarrhea- and rotavirus-associated hospitalizations among US children <5 years of age in 1997 and 2000.
RESULTS. In 1997 and 2000, diarrhea was coded in 13% of all childhood hospitalizations, for an estimated cumulative incidence of 1 diarrhea hospitalization per 23 to 27 children by age 5. Most diarrhea-associated hospitalizations (62%) were coded as unspecified etiology, and 35% as viral. Rotavirus was the most common pathogen recorded for 18% and 19% of diarrhea-associated hospitalizations in 1997 and 2000, respectively. Diarrhea-associated hospitalizations coded as unspecified or viral exhibited a marked winter peak similar to that of hospitalizations coded as rotavirus, suggesting that the rotavirus-specific code captures a fraction of all rotavirus hospitalizations. Using indirect methods, we estimated that rotavirus was associated with 51142–60155 and 46839–56820 hospitalizations in 1997 and 2000, respectively. By these estimates, rotavirus is associated with 4% to 5% of all childhood hospitalizations, and 1 in 67 to 1 in 85 children will be hospitalized with rotavirus by 5 years of age.
CONCLUSIONS. Diarrhea is an important cause of hospitalization in US children, and rotavirus is the most important etiology. Disease burden estimates have remained stable during the past decade. An effective rotavirus vaccine will likely reduce substantially the burden of severe rotavirus disease, estimated to account for 4% to 5% of all hospitalizations and ∼30% of hospitalizations for watery diarrhea among children <5 years of age.
Rotavirus is the leading cause of severe gastroenteritis among children worldwide. Globally, ∼500000 children die of rotavirus disease each year,1 but mortality from rotavirus is uncommon in developed countries because of better access to medical care and better nutrition. Nevertheless, rotavirus remains an important cause of morbidity in developed countries. For example, in the United States, previous studies have estimated that rotavirus causes ∼55000 hospitalizations and 500000 physician visits each year among children <5 years of age2–4 with substantial health care and societal costs. Because of this tremendous health burden, vaccines have been developed to prevent severe rotavirus disease and its economic consequences. A bovine-human reassortant rotavirus vaccine was shown recently to be safe and effective in preventing severe rotavirus disease in a large clinical trial of >70000 infants and may be licensed for use in US infants in 2006.5,6 Accurate data on the health and economic burden of rotavirus disease among US children will be essential to policymakers when making decisions regarding the use of the next rotavirus vaccine in the United States.
Previous estimates of rotavirus-associated hospitalizations among US children were based primarily on analysis of data from National Center for Health Statistics National Hospital Discharge Survey (NHDS).2,3,7 The NHDS is designed to provide estimates of national hospitalization trends and provide useful overall estimates of disease burden. However, because NHDS estimates are generated from a relatively small number of sampled discharges (0.5–1.0% of all hospital discharges), the NHDS may not provide robust estimates where hospitalization data are limited, which may occur when assessing patient characteristics or outcomes such as race or hospital mortality. In addition, the NHDS does not include information on hospital charges, which are needed to assess the economic impact of rotavirus disease and to evaluate the cost-effectiveness of vaccination.
In this study, we examined hospitalization data from the Kids' Inpatient Database (KID), a national sample of pediatric hospital discharges from community hospitals in the United States. The KID sampled 1.9 million pediatric discharges during 1997 and 2.5 million pediatric discharges during 2000 in contrast to the NHDS annual sample of ∼250000 discharges among all ages.8–10 We used KID data to confirm previous disease burden estimates derived from analyses of NHDS data and to assess the health and economic benefits of implementing a rotavirus immunization program in the United States.
Data Source and Definition of Cases
We examined the KID for 1997 and 2000 to identify hospital discharge records reporting diarrhea among children <5 years of age.8,9 KID is a product of the Healthcare Cost and Utilization Project and was produced by the Agency for Healthcare Research and Quality in collaboration with public and private statewide data organizations. The database is a robust sample of 10% of the uncomplicated births and 80% of other pediatric discharges in participating states (22 in 1997 and 27 in 2000) from short-term, nonfederal, general and specialty hospitals in the United States.8 National estimates of hospitalizations, hospital length of stay, hospital charges, and the number of procedures were calculated using the Healthcare Cost and Utilization Project weighting methodology.8 Discharges for all births were excluded from all of the analyses in this study.11 The KID does not contain individual identifiers, so hospital discharge records were the unit of analysis in this study.
Diarrhea-associated hospitalizations were defined as hospitalizations with discharge records that included any of the following International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnoses and codes: diarrhea of determined etiology (bacterial [001–005 and 008.0–008.5, excluding 003.2]; parasitic [006–007, excluding 006.3–006.6]; and viral [008.6 and 008.8], including rotavirus specifically [008.61]) and diarrhea of undetermined etiology, including cases presumed to be infectious (009.0–009.3) and noninfectious (558.9–787.91). Discharge records included ≤15 diagnostic fields, all of which were queried for a diarrhea code. A record that reported >1 diarrhea-associated etiologic agent was counted only once in the calculation to determine the total number of diarrhea-associated hospitalizations; however, such records were attributed to as many agents as applicable when data were aggregated by pathogen.
Age, Gender, Race/Ethnicity, Region, and Admission Month
Diarrhea-associated hospitalizations for the year 2000 were examined by gender, race, geographic region (standard census regions of Northeast, Midwest, South, and West), month and year of discharge, and age group (0–11, 12–23, 24–35, 36–47, and 48–59 months). SUDAAN software was used to determine the SEs for these estimates.8,12 Hospitalization estimates with a relative SE (SE/estimated number of discharges) >0.3 were not considered to be reliable and are not presented. Some variables had missing information; admission month was not reported for discharges from 2 participating states, and data on race were missing for 18% of all diarrhea discharges.8
Hospitalization rates were calculated as the estimated number of hospitalizations per 10000 children, based on estimates of the US resident population of children <5 years of age and estimates of live births for infants using the natality data for the corresponding year.13,14 Denominators obtained from vital records data were considered free from sampling error.13 SEs and 95% confidence intervals were calculated using SUDAAN software to account for the sampling design of the KID.8,12
We found previously that relying solely on the rotavirus ICD-9-CM code to identify these hospitalizations underestimates the true number of rotavirus hospitalizations.15–17 Consequently, in addition to recording the number of discharges coded with the specific ICD-9-CM code for rotavirus, the disease burden of rotavirus diarrhea was also estimated using 2 indirect methods. For the first estimate, we multiplied the monthly number of diarrhea-associated hospitalizations by the monthly proportion of rotavirus infections (among all of the diarrhea admissions) identified during a surveillance study among children hospitalized for diarrhea at Children's Hospital National Medical Center from 1974 through 1982.15 The residual estimate, as described previously by Ho et al,7 was calculated by subtracting the number of summer (May through October) diarrhea hospitalizations from the number of winter (November through April) diarrhea hospitalizations. Because of data use restrictions, 2 states did not report admission month (Florida and Connecticut), and we could not estimate the number of hospitalizations using the indirect methods. Therefore, for these 2 states, we estimated rotavirus hospitalizations by multiplying the total number of hospitalizations for diarrhea by the proportion of hospitalizations attributable to rotavirus calculated from the other states.
Hospital Stay and Charges
Average length of stay and charges for diarrhea-associated hospitalizations were also examined and compared; differentiation was performed using 2-sided t tests. Comparisons of proportions were made by using the χ2 test. All of the statistical tests incorporated the SEs of estimates before assessing statistical significance at P < .05.12
In 1997 and 2000, diarrhea was associated with 173220 and 150465 hospitalizations, respectively (corresponding with 70625 and 78772 unweighted discharge records, respectively) among US children <5 years of age (Table 1). For the 2 years combined, diarrhea was associated with ∼13% of all hospitalizations among children <5 years of age (13.1% in 1997 and 12.6% in 2000). Most (62%) diarrhea-associated hospitalizations were of unspecified etiology. Of diarrhea-associated hospitalizations, 35% were attributed to viral agents, followed by 5% to bacterial agents, and <1% to parasites. A specific code for rotavirus was reported on 18% (n = 31852) and 19% (n = 28536) of all of the diarrhea-associated hospitalizations in 1997 and 2000, respectively.
Age, Gender, Race/Ethnicity, and Region
In 2000, the overall rate of diarrhea-associated hospitalizations was 78 per 10000 children <5 years of age (Table 2). Examination of diarrhea-associated hospitalization rates by various demographic characteristics showed that the hospitalization rates were significantly greater for children <1 year of age (infants) compared with children 1 to 4 years of age (P < .001), males compared with females (P < .001), and Hispanic compared with black, white, and Asian/Pacific Islander children (P < .001 for all race/ethnicity comparisons). The rate was highest in the South, intermediate in the Northeast and West, and lowest in the Midwest (P = .01). In 2000, 161 deaths were reported among 150465 children hospitalized with diarrhea, for a hospital fatality rate of 1 death per 935 hospitalized children. Rotavirus-associated deaths were too infrequent to calculate reliable national estimates.
Seasonality and Age Distribution
In 2000, the monthly number of diarrhea hospitalizations by etiology demonstrated a clear winter peak from December through April (Fig 1). Hospitalizations presumed to be noninfectious and of viral etiology, which together accounted for 95.5% of all diarrhea-associated hospitalizations, also exhibited the same winter seasonal pattern. In contrast, the small fraction of hospitalizations of bacterial etiology (4.8%) showed a small peak during the summer (July through September). There were too few parasitic diarrhea hospitalizations (<1%) to graph.
Examination of the age distribution of diarrhea-associated hospitalizations in 2000 by race/ethnicity showed that, compared with whites and Asian/Pacific Islanders, black and Hispanic children were hospitalized at a relatively younger age (Fig 2). Of all childhood diarrhea hospitalizations, 55% among black children occurred in infants (<12 months), compared with 50% for Hispanic children, 41% for Asian/Pacific Islander children, and 39% for white children. The same pattern was observed to a slightly greater degree for hospitalizations with a specific discharge code for rotavirus (data not shown).
To account for the incomplete reporting of rotavirus-associated hospitalizations, we used 2 indirect methods to estimate these hospitalizations. The direct estimate (relying solely on the rotavirus ICD-9-CM code) was substantially lower for both study years (n = 31852 and 28536 in 1997 and 2000, respectively) than the estimates obtained using the monthly detection rate derived from the largest and most complete active surveillance study performed to date (n = 51142 and 46839 in 1997 and 2000, respectively) and the residual method (n = 60155 and 56820 in 1997 and 2000, respectively).
Hospital Stay and Charges
Diarrhea- and rotavirus-associated hospitalizations (adjusted estimate based on methods above) accounted for an estimated 454653 to 523411 and 146569 to 188237 inpatient days per year, respectively, with the median length of stay being 2 days for both groups. The median charges for a diarrhea-associated hospitalization were $2951 (interquartile range: $1844–5009) and $2999 (interquartile range: $1947–4875) for rotavirus. When we applied the median charge of a rotavirus hospitalization ($2999) to the adjusted number of rotavirus-coded hospitalizations (46839–60155), we found annual costs of $140 to $180 million for rotavirus hospitalizations.
This analysis of the robust hospital discharge data from KID, a sample of 80% of pediatric hospital discharges (exclusive of births) from approximately half of all US states, confirms and extends the findings of previous studies examining the burden of diarrhea- and rotavirus-associated hospitalizations.2,3,7,15–19 Our data demonstrate that diarrhea is associated with 150000 to 170000 hospitalizations annually and accounts for ∼13% of all hospitalizations among US children. Therefore, we estimate that 1 of every 23 to 27 children will be hospitalized with diarrhea by age 5. The observed winter peak and age distribution of diarrhea hospitalizations suggest that many of these events may be caused by rotavirus, and this pathogen was coded in one fifth (n = 31857 in 1997 and 28536 in 2000) of diarrhea hospitalizations.
Previous studies indicate that ≥94% of discharges coded as rotavirus are indeed confirmed by laboratory testing, indicating that the rotavirus code seems to have a high positive predictive value.16,18 However, several observations indicate that rotavirus-coded discharges likely capture only a fraction of all rotavirus hospitalizations. First, an analysis of hospital-specific discharge data from New York state demonstrated that 54% of all hospitals in the state never used the rotavirus code, whereas 12% of hospitals coded rotavirus on >30% of diarrhea hospitalization discharge records, suggesting substantial variation in testing and/or coding practices in different hospitals.18 Second, a multicenter, laboratory-based, active surveillance study demonstrated rotavirus detection rates of 56% in children hospitalized with vomiting, diarrhea, and fever, a proportion substantially greater than the proportion of discharges coded with rotavirus in this analysis.17 Third, our finding that the seasonal patterns of diarrhea hospitalizations coded as viral or presumed noninfectious were similar to those coded as rotavirus further suggests incomplete identification of rotavirus cases.3,7 Finally, a recent study that examined hospital discharge records of children admitted with laboratory-confirmed rotavirus disease at a hospital in Cincinnati found that discharge data captured fewer than half of all rotavirus hospitalizations.16 By use of indirect methods, we estimated that rotavirus accounted for 47000 to 60000 hospitalizations among US children <5 years of age, which suggests that 1 in 67 to 1 in 85 children will be hospitalized with rotavirus by age 5.
The rates of diarrhea- and rotavirus-associated hospitalizations estimated from the KID are comparable to those reported in other analyses of national and state hospitalization data.2,3,18 However, we noted that the estimates of total diarrhea hospitalizations for both study years (n = 173220 and n = 150465 in 1997 and 2000, respectively) were lower than the NHDS annual average estimate of 185742 hospitalizations per year for the period 1979–1992.3 Because the total number of diarrhea hospitalizations exhibit annual variation (eg, NHDS analyses for 1993, 1994, and 1995 identified 150222, 171466, and 165745 annual hospitalizations, respectively2), it is possible that the lower figures from the KID data might reflect study years with fewer total hospitalizations. In addition, it is possible that, during 2000, a fraction of diarrhea hospitalizations might have been prevented by the use of >0.5 million doses of a rotavirus vaccine that was introduced in October 1998 but withdrawn from the market in June 1999 because of its association with intussusception.20 Updated NHDS data for the same study years as the KID (1997 and 2000) are needed to carefully compare national diarrhea hospitalization estimates derived from these 2 data sources. It must also be noted that the indirect methodology used to estimate the number of rotavirus-associated hospitalizations is based on data from an active surveillance conducted nearly 2 decades ago, and the proportion of diarrhea-associated hospitalizations among children that are attributable to rotavirus may have changed since that time.
In 1983, Brandt et al15 reported that black children and those on Medicaid were hospitalized with viral diarrhea at an earlier age relative to other children, and the authors postulated that these differences might be attributable to greater exposure to enteric infections in the former groups. Our data support these observations in that black and Hispanic children had higher overall rates of diarrhea-associated hospitalization and were more likely to be hospitalized during infancy compared with whites and Asian/Pacific Islanders. We also noted that the diarrhea-associated hospitalization rate was higher in the South compared with other geographic regions, although the proportion attributed to each pathogen was no different, indicating that other factors, such as greater diarrheal disease burden or different hospitalization criteria, may account for this difference. Although we cannot fully explain the reasons for these phenomena, the relatively younger age of hospitalization among black and Hispanics highlights the need to avoid delays in immunization of these infants with rotavirus vaccines when they become available.
The findings of this analysis indicate that rotavirus remains the major etiologic cause of diarrhea among children hospitalized in the United States and underscore the potential benefits of a rotavirus vaccine that may soon be licensed. Over the past 2 decades, a series of successive assessments of disease have failed to demonstrate a substantial decline in pediatric diarrhea hospitalizations3,7 despite extensive efforts to improve treatment with such strategies as oral rehydration therapy. The promise of effective, safe rotavirus vaccines may provide the best opportunity to reduce the morbidity and associated direct and indirect economic burden attributable to severe diarrhea among US children. Hospital discharge data might also allow for assessment of the impact of rotavirus vaccines after their introduction, but the year-to-year variability in total hospitalizations, incomplete coding of rotavirus diagnosis on hospitalization records, and the 1- to 2-year delay in availability of data might limit their use for this purpose. Therefore, a more timely system using active surveillance is essential to determine accurate rotavirus disease burden estimates. A network of sentinel hospitals performing surveillance of laboratory-confirmed cases of rotavirus diarrhea would be well-suited for this task, providing not only valuable information to refine estimates of rotavirus hospitalizations among US children but also an opportunity to assess vaccine effectiveness through case-control studies.
- Accepted November 16, 2005.
- Address correspondence to Mark A. Malek, MD, Respiratory and Enteric Viruses Branch, National Center for Infectious Diseases, Centers for Disease Control and Prevention, 1600 E Clifton Rd, NE Building 3, Room 108, MS A-34 Atlanta, GA 30333. E-mail:
The views in this article are those of the authors and do not necessarily represent the views of the funding agency.
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Parashar UD, Holman RC, Clarke MJ, Bresee JS, Glass RI. Hospitalizations associated with rotavirus diarrhea in the United States, 1993 through 1995: surveillance based on the new ICD-9-CM rotavirus-specific diagnostic code. J Infect Dis.1998;177 :13– 17
- ↵Merck Co Inc. Merck Updates pipeline and focuses on future growth [press release]. Whitehouse Station, NJ: Merck Co Inc; December 14, 2004
- ↵Vesikari T, Matson D, Dennehy P, et al. Protection against rotavirus gastroenteritis of multiple serotypes by a pentavalent (human-bovine) reassortant vaccine (PRV). Paper presented at: Annual Meeting of the European Society for Pediatric Infectious Diseases; May 18–20, 2005; Valencia, Spain
- ↵Ho MS, Glass RI, Pinsky PF, Anderson LJ. Rotavirus as a cause of diarrheal morbidity and mortality in the United States. J Infect Dis.1988;158 :1112– 1116
- ↵Healthcare and Cost Utilization Project. Kids' Inpatient Database, 1997 and 2000 Data [CD-ROMs]. Rockville, MD: Agency for Healthcare Research and Quality; 2002
- ↵Dennison C, Pokras R. Design and operation of the National Hospital Discharge Survey: 1988 Redesign. Vital Health Stat 1.2000;(39):1– 42
- ↵Public Health Service and Health Care Financing Administration. International Classification of Diseases. 1998 [CD-ROM]. 6th ed. 9th Revision, Clinical Modification. Washington, DC: Public Health Service; 1998
- ↵Shah BV, Barnwell BG, Bieler GS. SUDAAN User's Manual: Software for Analysis of Correlated Data, Release 6.40. Research Triangle Park, NC: Research Triangle Institute; 1995
- ↵Bureau of Census. Intercensal Estimates of the Population of States by Age, Sex, and Race: 1970–2000. Washington, DC: Bureau of Census; 2002
- ↵US Department of Health and Human Services. Detailed Data 1979–2000: Public Use Data Tape Documentation: Natality. Hyattsville, MD: Centers for Disease Control and Prevention, National Center for Health Statistics; 2000
- ↵Brandt CD, Kim HW, Rodriguez WJ, et al. Pediatric viral gastroenteritis during eight years of study. J Clin Microbiol.1983;18 :71– 84
- ↵Hsu VP, Staat MA, Roberts N, et al. Use of active surveillance to validate international classification of diseases code estimates of rotavirus hospitalizations in children. Pediatrics.2005;115 :78– 82
- ↵Chang HG, Glass RI, Smith PF, et al. Disease burden and risk factors for hospitalizations associated with rotavirus infection among children in New York State, 1989 through 2000. Pediatr Infect Dis J.2003;9 :808– 814
- Copyright © 2006 by the American Academy of Pediatrics