Abstract
Background. Diarrhea remains an important cause of morbidity, but until the mid 1990s, hospital admissions for diarrhea in the US adult population were declining. We aimed to describe recent trends in gastroenteritis hospitalizations and to determine the contribution of norovirus.
Methods. We analyzed all gastroenteritis-associated hospital discharges during 1996–2007 from a nationally representative data set of hospital inpatient stays. Annual rates of discharges by age were calculated. Time-series regression models were fitted using cause-specified discharges as explanatory variables; model residuals were analyzed to estimate norovirus- and rotavirus-associated discharges. We then calculated the annual hospital charges for norovirus-associated discharges.
Results. Sixty-nine percent of all gastroenteritis discharges were cause-unspecified and rates increased by ≥50% in all adult and elderly age groups (≥18 years of age) from 1996 through 2007. We estimate an annual mean of 71,000 norovirus-associated hospitalizations, costing $493 million per year, with surges to nearly 110,000 hospitalizations per year in epidemic seasons. We also estimate 24,000 rotavirus hospitalizations annually among individuals aged ≥5 years.
Conclusions. Gastroenteritis hospitalizations are increasing, and we estimate that norovirus is the cause of 10% of cause-unspecified and 7% of all-cause gastroenteritis discharges. Norovirus should be routinely considered as a cause of gastroenteritis hospitalization.
Diarrheal disease mortality has been declining globally, but more than 1.3 million diarrheal deaths still occur each year [1], the great majority of which occur among young children in developing countries [2]. In the United States, the age pattern of diarrhea mortality is reversed; diarrhea-associated deaths are 5 times more common in elderly individuals than in children [3]. Although several studies have examined the burden of pediatric diarrhea hospitalizations in US children, the burden in adults has been less well evaluated. One study found that gastroenteritis was listed as a discharge diagnosis for 400,000–500,000 hospitalizations annually among US adults during 1979–1995, with a decline of 20% reported over the study period; almost four-fifths of the discharges were of undetermined etiology [4, 5]. Recognition of the organisms that cause disease in each age group is key to developing targeted care (eg, antibiotics or oral/intravenous rehydration therapy) or prevention activities (eg, vaccination or infection prevention/control). Noroviruses have been increasingly appreciated as a cause of disease, with periodic increases in norovirus activity every 3–4 years in association with the emergence of new viral strains [6]. The low-fertility, low-mortality populations in the United States and other developed countries continue to age, and increasing numbers live in long-term care facilities, where norovirus outbreaks are common [7]. Considering these demographic changes coupled with the potential emergence of new pathogens, including Clostridium difficile [8, 9], we aimed to assess whether gastroenteritis discharges have continued to decline since 1995 and specifically to assess the contribution of norovirus to hospitalization in all age groups.
Despite being the most common cause of gastroenteritis in the community, [10–12] norovirus is rarely laboratory-confirmed or recorded explicitly as a cause of hospitalization, because clinical laboratories in hospitals rarely test for norovirus. Molecular diagnostics and characterization techniques have revealed the high incidence, range of severity [13–16], and evolving epidemiology of norovirus infections, but diagnostics are generally restricted to outbreak investigations [6, 17]. For this reason, we extended a previously developed indirect method [16, 18–20] to quantify the contribution of norovirus to hospital discharges. Noroviruses have a wintertime seasonality that differs in terms of timing and intensity from year to year. Therefore, we deconstructed the seasonal patterns of cause-unspecified gastroenteritis hospital discharges to estimate the role of norovirus.
METHODS
Data Source
The Nationwide Inpatient Sample (NIS) is a nationally representative database of hospital inpatient stays occurring in the United States that is curated by the Healthcare Cost and Utilization Project (http://www.hcup-us.ahrq.gov/nisoverview.jsp). NIS data are collected from a national sample of >1000 hospitals located in 40 states, as of 2007. Approximately 20% of all US hospitals are captured in the sample. Data are collected irrespective of payer and include persons covered by Medicare, Medicaid, and private insurance, as well as uninsured persons. We extracted all records with an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code for cause-specified or nonspecified gastroenteritis-associated hospital discharges from 1996 through 2007 (see Table 1 for specific codes extracted).
Pathogen Categories and Diagnostic Codes Used to Identify Gastroenteritis Discharges from the Nationwide Inpatient Sample
| First position | Any position | ||||
| Cause | ICD-9-CM code or codes | Annual mean (%) | No. of discharges per 10,000 persons (95% CI) | Annual mean (%) | No. of discharges per 10,000 persons (95% CI) |
| Cause unspecifieda | 009.0–009.3,558.9, 787.91, 008.8 | 187,274 (67.5%) | 6.5 (6.3–6.8) | 709,357 (68.6%) | 24.8 (24.0–25.6) |
| Cause specified | |||||
| Viralb | 008.61–008.69 | 20,754 (7.5%) | .73 (.67–.78) | 93,021 (9.0%) | 3.3 (3.1–3.4) |
| Bacterialc | 001.0–00.9, 002.0–002.9, 003.0–003.9, 004.0–004.9, 005.0–005.9, 008.0–008.5 | 17,927 (6.5%) | .63 (.61–.64) | 30,339 (2.9%) | 1.1 (1.0–1.2) |
| Clostridium difficile | 008.45 | 49,990 (18.0%) | 1.75 (1.68–1.82) | 196,777 (19.0%) | 6.9 (6.6–7.2) |
| Parasitic | 006.0–006.2, 006.9, 007.0–007.9 | 1,294 (0.5%) | .05 (.04–.06) | 4,054 (0.4%) | .14 (.13–.15) |
| All-cause gastroenteritis | – | 277,238 | 11.7 (11.3–12.1) | 1,033,547 | 37.5 (36.3–38.7) |
| First position | Any position | ||||
| Cause | ICD-9-CM code or codes | Annual mean (%) | No. of discharges per 10,000 persons (95% CI) | Annual mean (%) | No. of discharges per 10,000 persons (95% CI) |
| Cause unspecifieda | 009.0–009.3,558.9, 787.91, 008.8 | 187,274 (67.5%) | 6.5 (6.3–6.8) | 709,357 (68.6%) | 24.8 (24.0–25.6) |
| Cause specified | |||||
| Viralb | 008.61–008.69 | 20,754 (7.5%) | .73 (.67–.78) | 93,021 (9.0%) | 3.3 (3.1–3.4) |
| Bacterialc | 001.0–00.9, 002.0–002.9, 003.0–003.9, 004.0–004.9, 005.0–005.9, 008.0–008.5 | 17,927 (6.5%) | .63 (.61–.64) | 30,339 (2.9%) | 1.1 (1.0–1.2) |
| Clostridium difficile | 008.45 | 49,990 (18.0%) | 1.75 (1.68–1.82) | 196,777 (19.0%) | 6.9 (6.6–7.2) |
| Parasitic | 006.0–006.2, 006.9, 007.0–007.9 | 1,294 (0.5%) | .05 (.04–.06) | 4,054 (0.4%) | .14 (.13–.15) |
| All-cause gastroenteritis | – | 277,238 | 11.7 (11.3–12.1) | 1,033,547 | 37.5 (36.3–38.7) |
NOTE. Annual mean number of discharges and population rates. CI, confidence interval; ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification.
Excluding records in which another cause is specified in a subsequent diagnostic position.
Excluding viral, other not elsewhere classified (008.8).
Excluding C. difficile (008.45).
Our primary objective was to estimate norovirus-associated hospital discharges, and preliminary analyses demonstrated that the norovirus specific code (008.63, Norwalk virus or Norwalk-like agent) was very rarely used (∼200 admissions nationally per year). Specific codes, which underestimate the true burden of disease, are typically used when there is diagnostic confirmation of a pathogen. Since norovirus diagnostics are not routinely available in hospital settings, we sought to estimate the proportion of cause-unspecified gastroenteritis codes that were due to norovirus by means of indirect approaches. Nonspecific gastroenteritis potentially caused by norovirus was defined on the basis of the following ICD-9-CM codes: presumed infectious (009.0–009.3), presumed noninfectious (558.9), unclassified viral infection (008.8), and diarrhea (787.91) in any of the 15 diagnostic code positions. Presumed noninfectious discharges were included because it is likely that many episodes without a known etiology are incorrectly classified as such. In order to compare our findings with those in previous studies [5], we also analyzed all-cause gastroenteritis discharge rates restricted to the first 3 diagnostic code positions.
Population Data and Rate Calculations
Denominator data were obtained by using the National Center for Health Statistics’ Bridged Race population estimates for 1996 through 2007 [21, 22].
The NIS is a stratified sample of hospitals drawn from a subset of community hospitals (defined as non-Federal, short-term, general, and other specialty hospitals, excluding hospital units of institutions), stratified by region, location and teaching status, bed size category, and ownership (public or private). The NIS includes all discharges from the sampled hospitals. We used discharge sample weights to estimate national totals, and standard errors for the estimates were calculated by using the svy suite of commands in Stata (version 11.0; StataCorp).
Statistical Methods
Norovirus-associated discharges were estimated by extending a previously developed indirect method [16, 18, 20, 23] in order to estimate the main outcome of interest (norovirus-associated discharges) without a predictor time series of laboratory reports or confirmed cases. To determine the total number of cases of infection with a specific pathogen or pathogen group (discharges associated with rotavirus, bacterial gastroenteritis [excluding C. difficile], parasitic gastroenteritis, and C. difficile), the number of cause-unspecified discharges in a month was modeled as a function of the monthly number of discharges associated with that specific pathogen or pathogen group in the same month (Table 1; Figure S1). We assumed that the total number of cases of infection with a specific pathogen or pathogen group in a given month was proportional to the number of discharges due to that pathogen in the same month. Primary diagnoses (reported in the first diagnosis field) were used for these predictor pathogen groups, because they were believed to be the most reliable diagnoses of community-acquired infections. Separate models were fitted for each age group (0–4, 5–17, 18–64, 65–74, 75–84, and ≥85 years). Rotavirus infection counts in the 0–4 year age group were used as the predictor for all age groups because laboratory confirmation of rotavirus infection is rare outside pediatric populations. We included rotavirus in the model because preliminary visual inspection highlighted a late-winter–spring peak in nonpediatric age groups that was consistent with the pediatric rotavirus season, and since rotavirus is known to cause disease in adults, although the incidence is not known [24].
Poisson regression models were fitted with an identity link (such that model coefficients were on the natural scale) in Stata as follows:
Model residuals were calculated. Norovirus-associated discharges were estimated by taking the difference between the residual for each month and the seasonal minimum (from July through June) of the residuals. This approach assumes that any remaining seasonality not otherwise captured in the model is due to norovirus and that there is a month in which there are 0 norovirus-associated discharges (the month of annual minimum residual).
Estimated discharges from these nonspecific discharges were added to the coded rotavirus-associated discharges for an estimate of total rotavirus-associated discharges.
Economic Analysis
The health care costs of hospitalizations were calculated by multiplying the number of estimated hospital discharges by the charges of the hospital stay. We used the median of the total charges submitted by the hospital (reported in NIS) for a coded viral, pathogen-unspecified discharge (ICD-9-CM code 008.8) to estimate the norovirus-associated charges by age group. Where rotavirus was specified as the cause, reported charges was used. All charges were expressed in 2007 US dollars to adjust for inflation.
RESULTS
From seasonal years 1996/97 through 2006/07, an annual mean of 709,109 discharges per year (24.8 discharges per 10,000 persons [95% confidence interval, 24.0–25.6 discharges per 10,000 persons]) with a cause-unspecified gastroenteritis code in any position occurred, which make up the majority (69%) of all-cause gastroenteritis discharges (Table 1). The rate of cause-unspecified discharges increased by 41% from 1996/97 to 2006/07 (from 21.5 to 30.4 discharges per 10,000 persons) (Figure 1). Rates decreased in children (decrease in 0–4–year-olds, 21%; decrease in 5–17-year-olds, 8%) whereas rates increased by >50% in all adult and elderly age groups. In all adult and elderly age groups (≥18 years old), cause-unspecific gastroenteritis rates were higher in women than in men (rate ratio, 1.49).
A, Cause-specified and cause-unspecified gastroenteritis discharges. B, Age-specific cause-unspecified gastroenteritis discharges from seasonal years 1996/97 through 2006/07. The number of discharges per 10,000 persons is calculated by seasonal year (July–June).
A, Cause-specified and cause-unspecified gastroenteritis discharges. B, Age-specific cause-unspecified gastroenteritis discharges from seasonal years 1996/97 through 2006/07. The number of discharges per 10,000 persons is calculated by seasonal year (July–June).
If only the first 3 diagnostic code positions were considered, then the all-cause gastroenteritis discharge rate for adults and elderly persons was 29.6 discharges per 10,000 persons in seasonal year 1996/97, which is broadly similar to figures reported in another study from 1979 through 1995 (27.0 discharges per 10,000 persons) [5]. Rates increased substantially over the study period to 49.2 discharges per 10,000 persons in 2006/07.
In the 0–4 year age group, a regular peak in cause-unspecified discharges occurred in March every year (Figures 2 and S2), which is consistent with the rotavirus season. With each older age group, a January peak was increasingly apparent, although there remained a smaller but distinct March peak in all age groups. In the regression models, rotavirus and C. difficle were associated with cause-unspecified gastroenteritis in all age groups, whereas bacterial pathogens were only associated with cause-unspecified gastroenteritis in adults aged 18–64 years and parasites in adults aged 18–64 years and ≥85 years (Wald test; P < .001 for all where significant). Discharges coded as C. difficle infection increased markedly over the study period but did not account for all of the increase in cause-unspecified discharges; a variable indicating a secular time trend remained significant in all models after inclusion of the pathogen categories (P < .001 for all age groups). Thirty-one percent of cause-unspecified gastroenteritis was attributed to a pathogen in all age groups, with 50% being attributed in 0–4-year-olds.
Seasonal and long-term trends of cause-unspecified gastroenteritis, by age, in discharges per 100,000 persons, 1996–2007.
Seasonal and long-term trends of cause-unspecified gastroenteritis, by age, in discharges per 100,000 persons, 1996–2007.
Model residuals peaked most years in December or January in all age groups (Figures 3 and 4), which is highly consistent with the norovirus season. The peak was most pronounced in years when norovirus epidemics are known to have occurred in the United States (seasonal years 2002/03 and 2006/07) [7, 25, 26]. These epidemic seasons appeared to have different effects by age group: in 2002/03 there was an increased number of discharges in all age groups, whereas in 2006/07 heightened activity was restricted to older age groups (Table 2).
Estimated Norovirus-associated Hospital Discharges, by Age Group and Season (July–June), 1996/97–2006/07
| Age group, number of discharges (per 100,000 persons) | |||||||
| Seasonal year | 0–4 years | 5–17 years | 18–64 years | 65–74 years | 75–84 years | ≥85 years | All ages |
| 1996/97 | 10,335 (53) | 4,443 (9) | 7,636 (5) | 5,525 (29) | 9,878 (86) | 7,536 (199) | 45,354 (17) |
| 1997/98 | 12,749 (66) | 4,235 (8) | 15,702 (9) | 7,469 (40) | 8,740 (74) | 4,712 (121) | 53,608 (20) |
| 1998/99 | 22,148 (116) | 5,493 (11) | 19,756 (12) | 8,716 (47) | 6,846 (57) | 4,290 (106) | 67,250 (24) |
| 1999/00 | 13,792 (72) | 5,650 (11) | 16,233 (9) | 5,409 (29) | 7,390 (60) | 8,354 (201) | 56,827 (20) |
| 2000/01 | 13,326 (69) | 3,108 (6) | 13,298 (8) | 5,875 (32) | 8,006 (64) | 7,693 (179) | 51,306 (18) |
| 2001/02 | 21,003 (109) | 7,698 (14) | 19,487 (11) | 7,376 (40) | 8,267 (66) | 5,741 (130) | 69,571 (24) |
| 2002/03 | 26,967 (138) | 8,146 (15) | 19,553 (11) | 9,331 (51) | 13,245 (104) | 9,552 (210) | 86,794 (30) |
| 2003/04 | 14,725 (74) | 4,660 (9) | 21,468 (12) | 8,288 (45) | 6,417 (50) | 6,918 (147) | 62,477 (22) |
| 2004/05 | 13,968 (70) | 5,141 (10) | 14,713 (8) | 10,031 (54) | 14,258 (110) | 8,900 (184) | 67,010 (23) |
| 2005/06 | 37,601 (185) | 10,203 (19) | 25,611 (14) | 14,362 (77) | 15,883 (122) | 8,905 (176) | 112,566 (38) |
| 2006/07 | 16,856 (82) | 5,615 (11) | 25,640 (14) | 12,725 (67) | 28,042 (215) | 20,048 (379) | 108,927 (36) |
| Seasonal mean | 18,497 (94) | 5,854 (11) | 18,100 (10) | 8,646 (47) | 11,543 (92) | 8,423 (185) | 71,063 (24) |
| Age group, number of discharges (per 100,000 persons) | |||||||
| Seasonal year | 0–4 years | 5–17 years | 18–64 years | 65–74 years | 75–84 years | ≥85 years | All ages |
| 1996/97 | 10,335 (53) | 4,443 (9) | 7,636 (5) | 5,525 (29) | 9,878 (86) | 7,536 (199) | 45,354 (17) |
| 1997/98 | 12,749 (66) | 4,235 (8) | 15,702 (9) | 7,469 (40) | 8,740 (74) | 4,712 (121) | 53,608 (20) |
| 1998/99 | 22,148 (116) | 5,493 (11) | 19,756 (12) | 8,716 (47) | 6,846 (57) | 4,290 (106) | 67,250 (24) |
| 1999/00 | 13,792 (72) | 5,650 (11) | 16,233 (9) | 5,409 (29) | 7,390 (60) | 8,354 (201) | 56,827 (20) |
| 2000/01 | 13,326 (69) | 3,108 (6) | 13,298 (8) | 5,875 (32) | 8,006 (64) | 7,693 (179) | 51,306 (18) |
| 2001/02 | 21,003 (109) | 7,698 (14) | 19,487 (11) | 7,376 (40) | 8,267 (66) | 5,741 (130) | 69,571 (24) |
| 2002/03 | 26,967 (138) | 8,146 (15) | 19,553 (11) | 9,331 (51) | 13,245 (104) | 9,552 (210) | 86,794 (30) |
| 2003/04 | 14,725 (74) | 4,660 (9) | 21,468 (12) | 8,288 (45) | 6,417 (50) | 6,918 (147) | 62,477 (22) |
| 2004/05 | 13,968 (70) | 5,141 (10) | 14,713 (8) | 10,031 (54) | 14,258 (110) | 8,900 (184) | 67,010 (23) |
| 2005/06 | 37,601 (185) | 10,203 (19) | 25,611 (14) | 14,362 (77) | 15,883 (122) | 8,905 (176) | 112,566 (38) |
| 2006/07 | 16,856 (82) | 5,615 (11) | 25,640 (14) | 12,725 (67) | 28,042 (215) | 20,048 (379) | 108,927 (36) |
| Seasonal mean | 18,497 (94) | 5,854 (11) | 18,100 (10) | 8,646 (47) | 11,543 (92) | 8,423 (185) | 71,063 (24) |
Illustration of deconstruction of cause-unspecified discharges into pathogen groups, based on model fitting. This figure illustrates how cause-unspecified discharges (black line) were attributed to pathogen groups, using the ≥85 year age group (1996–2007), as an example. The background discharges and secular increase in discharges (α + γ × Time) (gray bars), predicted rotavirus-associated discharges (β1 × Rota0−4,y) (red bars), predicted parasite-associated discharges (β4 × Parax,y,) (orange bars), and predicted Clostridium difficile–associated discharges (β3 × CDiffx,y,) (blue bars) do not capture the predominantly wintertime increase in discharges. Bacterial pathogens were not significant predictors of cause-unspecified discharges in this age group, so they are not included in the graph. Based on model residuals, norovirus-associated discharges (green bars) were calculated on the assumption that the winter increase was due to this pathogen.
Illustration of deconstruction of cause-unspecified discharges into pathogen groups, based on model fitting. This figure illustrates how cause-unspecified discharges (black line) were attributed to pathogen groups, using the ≥85 year age group (1996–2007), as an example. The background discharges and secular increase in discharges (α + γ × Time) (gray bars), predicted rotavirus-associated discharges (β1 × Rota0−4,y) (red bars), predicted parasite-associated discharges (β4 × Parax,y,) (orange bars), and predicted Clostridium difficile–associated discharges (β3 × CDiffx,y,) (blue bars) do not capture the predominantly wintertime increase in discharges. Bacterial pathogens were not significant predictors of cause-unspecified discharges in this age group, so they are not included in the graph. Based on model residuals, norovirus-associated discharges (green bars) were calculated on the assumption that the winter increase was due to this pathogen.
Estimated norovirus-associated discharges per 10,000 persons from seasonal years 1996/97 through 2006/07. Gray regions represent the estimated monthly (July–June) minimum and maximum discharges per 10,000 persons for all non-epidemic years. The epidemic seasons of 2002/03 and 2006/07 are represented by a separate green and blue line, respectively.
Estimated norovirus-associated discharges per 10,000 persons from seasonal years 1996/97 through 2006/07. Gray regions represent the estimated monthly (July–June) minimum and maximum discharges per 10,000 persons for all non-epidemic years. The epidemic seasons of 2002/03 and 2006/07 are represented by a separate green and blue line, respectively.
We estimated an annual mean of 71,063 norovirus-associated discharges, with 26% occurring among 0–4-year-olds (mean, 18,497 discharges per year) and 40% among those aged ≥65 years (mean, 28,612 discharges per year). Norovirus was estimated to cause 20% of cause-unspecified and 13% of all gastroenteritis admissions in children aged 0–4 years, and 13% of cause-unspecified and 8% of all gastroenteritis admissions among adults aged ≥65 years (Table 3). The number of norovirus-associated discharges during epidemic years was substantially larger than the annual mean (eg, n = 108,927 discharges in seasonal year 2006/07; increase above the seasonal mean, 53%).
Estimated Annual Norovirus- and Rotavirus-associated Hospital Admissions by Hospital Charges by Age Group, Seasonal Years 1996/97–2006/07
| Gastroenteritis discharges | Norovirus-associated discharges | Rotavirus-associated discharges | |||||||||||||
| Cause unspecified | Modeled | Coded | Modeled | Total | |||||||||||
| Age group | Cause specified | N | Median duration of stay, d | Median charge, $a | Total | N | Cause unspecified, % | Total, % | Charge, million $b | N | N | Cause unspecified, % | N | Total, % | Charge, million $b |
| 0–4 years | 50,473 | 93,097 | 2 | 3,922 | 143,570 | 18,497 | 20 | 13 | 73 | 30,523 | 32,509 | 35 | 63,032 | 47 | 247 |
| 5–17 years | 13,482 | 36,665 | 2 | 4,526 | 50,147 | 5,854 | 16 | 12 | 27 | 1,951 | 3,925 | 11 | 5,876 | 13 | 27 |
| 18–64 years | 98,429 | 292,460 | 3 | 8,137 | 390,889 | 18,100 | 6 | 5 | 147 | 317 | 8,211 | 3 | 8,528 | 2 | 69 |
| 65–74 years | 50,063 | 102,906 | 4 | 8,111 | 152,969 | 8,646 | 8 | 6 | 70 | 92 | 2,824 | 3 | 2,916 | 2 | 24 |
| 75–84 years | 70,114 | 120,746 | 4 | 8,777 | 190,860 | 11,543 | 10 | 6 | 101 | 88 | 3,699 | 3 | 3,787 | 2 | 33 |
| ≥85 years | 40,236 | 63,235 | 4 | 8,972 | 103,471 | 8,423 | 13 | 8 | 76 | 73 | 2,870 | 5 | 2,943 | 2 | 26 |
| All ages | 322,797 | 709,109 | … | … | 1,031,906 | 71,063 | 10 | 7 | 493 | 33,044 | 54,038 | 8 | 87,082 | 8 | 427 |
| Gastroenteritis discharges | Norovirus-associated discharges | Rotavirus-associated discharges | |||||||||||||
| Cause unspecified | Modeled | Coded | Modeled | Total | |||||||||||
| Age group | Cause specified | N | Median duration of stay, d | Median charge, $a | Total | N | Cause unspecified, % | Total, % | Charge, million $b | N | N | Cause unspecified, % | N | Total, % | Charge, million $b |
| 0–4 years | 50,473 | 93,097 | 2 | 3,922 | 143,570 | 18,497 | 20 | 13 | 73 | 30,523 | 32,509 | 35 | 63,032 | 47 | 247 |
| 5–17 years | 13,482 | 36,665 | 2 | 4,526 | 50,147 | 5,854 | 16 | 12 | 27 | 1,951 | 3,925 | 11 | 5,876 | 13 | 27 |
| 18–64 years | 98,429 | 292,460 | 3 | 8,137 | 390,889 | 18,100 | 6 | 5 | 147 | 317 | 8,211 | 3 | 8,528 | 2 | 69 |
| 65–74 years | 50,063 | 102,906 | 4 | 8,111 | 152,969 | 8,646 | 8 | 6 | 70 | 92 | 2,824 | 3 | 2,916 | 2 | 24 |
| 75–84 years | 70,114 | 120,746 | 4 | 8,777 | 190,860 | 11,543 | 10 | 6 | 101 | 88 | 3,699 | 3 | 3,787 | 2 | 33 |
| ≥85 years | 40,236 | 63,235 | 4 | 8,972 | 103,471 | 8,423 | 13 | 8 | 76 | 73 | 2,870 | 5 | 2,943 | 2 | 26 |
| All ages | 322,797 | 709,109 | … | … | 1,031,906 | 71,063 | 10 | 7 | 493 | 33,044 | 54,038 | 8 | 87,082 | 8 | 427 |
Per discharge, in 2007 US dollars, based on the median hospital charges of pathogen-unspecified discharges coded for viral enteritis (008.8).
In 2007 US dollars.
We estimated that rotavirus caused 35% (annual mean, 32,509 discharges) of cause-unspecified gastroenteritis in children aged 0–4 years. By combining this figure with explicitly coded rotavirus-associated discharges, we estimate ∼63,000 annual rotavirus-associated hospitalizations in children aged 0–4 years. We also estimated that rotavirus causes 11% of cause-unspecified gastroenteritis in older children (aged 5–17 years) and 3%–5% in adult and elderly age groups (aged ≥18 years), totaling ∼24,000 discharges among individuals ≥5 years old.
Norovirus is a slightly less common cause of hospital discharges than rotavirus, but because it disproportionally affects the elderly, for whom hospitalization is relatively expensive, norovirus is more costly at $493 million per year, compared with $427 million per year for rotavirus (Table 3).
DISCUSSION
In contrast to the trend of decrease that was observed in the 1980s and early 1990s [5], all-cause gastroenteritis hospitalizations have increased markedly across adult and elderly age groups from 1996 through 2007. Our modeling indicates that neither C. difficile nor norovirus—which have both been increasing causes of hospitalization—fully account for this general increase. We estimate that norovirus causes >70,000 hospitalizations in the United States in an average year, but this number is highly variable, and in epidemic seasons hospitalizations may surge by 50% to ∼110,000 discharges. Norovirus affects individuals of all ages, but the highest rates of hospitalization for norovirus infection are among the elderly, for whom hospitalization is costly. Hospital charges attributable to norovirus were estimated at nearly $500 million per year. Norovirus-associated hospitalizations appear to have increased over the study period from 1996 through 2007. We also note a clear seasonal pattern of rotavirus-associated activity in nonpediatric age groups, which amounts to ∼24,000 hospital discharges per annum in the population >5 years of age.
The NIS is nationally representative, and due to the coding consistency and longevity of the data set, we were able to assess long-term trends in gastroenteritis and presumed norovirus-associated discharges. The principle limitation is the lack of etiologic data for the cause of gastroenteritis in the majority of patients, particularly among the elderly. In particular, norovirus is rarely tested for outside of public health laboratories investigating outbreaks, so there are neither etiologic fraction studies that we can apply to cause-unspecified discharges nor a time series of norovirus hospitalizations as there are for other pathogens that are routinely coded. Although the methods we used are indirect, the pattern of presumed norovirus-associated discharges is highly consistent with what is known about norovirus epidemiology over this study period. The model residuals attributable to norovirus peaked during the early part of winter—a seasonal pattern that has been well documented in emergency department visits and in long-term care settings [7, 27]. Epidemic seasons (associated with the emergence of new variants of genogroup 2, type 4) occurred in the United States and internationally in seasonal years 2002/03 and 2006/07 [7, 26, 28–30]; there was heightened activity in our estimated norovirus-associated discharges during those seasons. In the latter epidemic, the impact appeared to be focused in long-term care facilities [7], and here we note a disproportionate effect on elderly persons, compared with the effect on children during the previous year.
The residual method employed here potentially has both over- and underestimating biases. We assume that all the residual seasonality is due to norovirus; this may lead to an overestimate, but as noted above, the residual seasonality is highly consistent with norovirus seasonality. It is also possible that in using this approach, some norovirus-associated discharges were attributed to rotavirus, given some overlap in their seasonal patterns. This approach also assumes there is 1 month in each year in which there are no norovirus-associated discharges. Norovirus is known to circulate and cause disease year-round, so this will lead to an underestimate. More than half of all gastroenteritis discharges were not assigned to any pathogen, and some of these could well be due to norovirus. Finally, it is not possible with these data to determine whether infections are community- or hospital-acquired. Norovirus outbreaks commonly occur in health care facilities [31–33]; if a case of norovirus infection is hospital-acquired, then the entire cost of hospitalization would not be due to the infection, which may lead to an overestimate of the economic impact.
The high burden of diarrhea-associated hospitalization and mortality among elderly patients in the United States has been described elsewhere [4]. Mounts et al [5] observed a trend of decreasing hospitalizations from 1979 to 1995. Using similar data sources and methods, we observed a reversal of that trend in all adult and elderly age groups. Our estimate of norovirus-associated discharges is higher than previous estimates (70,000 discharges compared with 56,000 discharges [11]), and we highlight for the first time (to our knowledge) how the burden on hospitals varies from year to year. Population-based studies have consistently documented that norovirus is the most common pathogen causing acute gastroenteritis in the community [12, 14, 34, 35]. Although severe gastroenteritis is a relatively uncommon manifestation [36, 37], there is a substantial burden on medical resources due to high population incidence. General practitioner consultations are as numerous for norovirus infections as for rotavirus infections among children <5 years old in England [10]. We estimate that norovirus is the cause of 7% of all hospital discharges in which gastroenteritis is recorded. This figure may be lower than estimates based on hospital admission studies for 2 reasons. First, there are high background levels of gastroenteritis symptoms among hospitalized adult and elderly populations. Patients—particularly among the elderly—may be admitted for a non-gastroenteritis-related condition but may experience gastroenteritis symptoms during their stay. Second, studies of hospital admissions, which have detected norovirus in as many as 23% of adult gastroenteritis admissions [38], have not tested healthy controls in order to determine whether norovirus is actually the cause of disease [39].
The time-series models predict associations between cause-specified discharge categories and cause-unspecified discharges; from this we also gain an estimate for other pathogens. The estimated rate of rotavirus hospitalizations in 0–4-year-olds (63,000 discharges per year) is consistent with previous estimates using other methods [40, 41] and therefore provides confidence of the validity of the method. We estimated 24,000 discharges due to rotavirus in persons ≥5 years old. This disease burden from rotavirus in older age groups, particularly among older children 5–17 years of age in whom rotavirus accounted for 13% of gastroenteritis discharges, is poorly recognized. Given early evidence of indirect benefits to nonvaccinated persons after implementation of rotavirus vaccination in the United States, it will be of interest to evaluate whether vaccination of young children has resulted in a reduction of rotavirus-associated discharges among older children and elderly adults.
All-cause gastroenteritis admissions are increasing in the aging US population, and our data indicate that the well-recognized increase in discharges from C. difficile infection do not fully explain this trend. Discharges attributable to norovirus also appear to be increasing, and with more of the population living in long-term care institutions, these individuals may be exposed to a higher risk of being involved in an outbreak and transmission from within their living environment. These results point to the need for increased diagnostic testing for norovirus in routine emergency and inpatient settings, including long-term care facilities. Norovirus infection diagnosis upon admission may lead to increased infection prevention, including isolation. Clinical management of a patient with norovirus infection is different than that of patients with C. difficile–associated diarrhea, for which testing is common. Rapid and reliable norovirus diagnostics need to be developed and made available. Focused studies with standard clinical definitions using state-of-the-art diagnostic testing are needed in order to establish the burden of norovirus-associated hospitalizations and health care–acquired infections and to monitor the changing epidemiology of the virus.
The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
Potential conflicts of interest. All authors: no conflicts.
Comments