Hospitalizations for diarrhea in Quebec children from 1985 to 1998 : Estimates of rotavirus-associated diarrhea

1Public Health Research Unit, CHUL Research Centre, Laval University, Quebec City; 2Institut national de santé publique du Québec, Beauport; 3Ministère de la Santé et des Services Sociaux du Québec, Québec City; 4Centre Hospitalier universitaire de Québec, Québec City, Quebec Correspondence and reprints: Dr Bernard Duval, Institut national de santé publique, 2400 D’Estimauville, Beauport, Quebec G1E 7G9. Telephone 418-666-7000 extension 234, fax 418-666-2776, e-mail bernard.duval@ssss.gouv.qc.ca Received for publication July 17, 2001. Accepted November 3, 2001 RP Buigues, B Duval, L Rochette, et al. Hospitalizations for diarrhea in Quebec children from 1985 to 1998: Estimates of rotavirus-associated diarrhea. Can J Infect Dis 2002; 13(4):239-244.

D iarrhea is an important public health problem in chil- dren.Although no longer a frequent cause of death in developed countries, rotavirus remains the major cause of hospitalizations for diarrhea among children (1).Although the first commercial vaccine was withdrawn quickly from the market because of an association with intussusception (2), other vaccines are being developed and are expected soon.The benefits that are expected from these vaccines are a contentious issue in developed nations (3).Before considering the universal use of a new rotavirus vaccine, it is essential to estimate the benefit that may be expected from its introduction.Outpatient consultations and hospitalizations are the main contributors to the burden of diarrhea.The purpose of the present study was to estimate the number of hospitalizations that may be associated with rotavirus.These data are not directly available because routine microbiological testing is not performed for all children who are hospitalized for diarrhea.
In Quebec, complete computerized hospitalization data have been available since 1985.In the present study, three different methods were used to estimate the proportion of all hospitalizations with a diagnosis of diarrhea that could be attributable to rotavirus.

Source of data
Med-Echo is the administrative database that contains complete information on all hospitalizations in Quebec.It was created in 1982, but complete information has been available only since 1985.For each hospitalization, a principal diagnosis and up to 15 secondary diagnoses are recorded.

Study population
The study population consisted of all children from one to 59 months of age who were hospitalized for diarrhea in Quebec from April 1, 1985to March 31, 1998.Newborn babies and those up to 28 days of age were excluded because rotavirus is rare in that population and because the International Classification of Diseases-9th revision (ICD-9) provides a specific coding for health events that occur during the neonatal period.
The following variables were studied: primary and secondary diagnoses, age, sex, admission date, exit date, stay duration, death and area of residence.The encrypted health insurance number was used to eliminate double registration.
Two consecutive hospitalization episodes were considered as only one hospitalization when there were 15 or fewer days between them.This happened in 1.7% of the episodes.For calculation of the length of stay, only the first episode was considered.The global results were increased by an insignificant factor when the length of the repeat episode was added to the length of the first episode.

Rotavirus estimate
Three methods were used to estimate the number of hospitalizations for rotavirus-associated diseases.
The first method was described by Ho et al (4) and by Jin et al (5).The difference in the number of winter hospitalizations for diarrhea, minus the number of summer hospitalizations for diarrhea, was considered to be the number of hospitalizations caused by rotavirus.This approach was based on the seasonality of rotavirus that was observed in 88 health centres in Canada, Mexico and the United States, which revealed that there were few rotavirus cases during summer months (6).
The second method of estimation used the proportion of laboratory-confirmed rotavirus infection among diarrhearelated hospitalizations that was measured by Waters et al (7) in the Toronto, Ontario area in 1998.In that study, RAD cases represented approximately 37% of those patients tested by enzyme immunoassay and approximately 36% of those tested by electron microscopy.This etiological fraction was calculated for the period between November 1 and June 30, 1998.In the present study, this proportion was applied to all diarrhea cases registered in Med-Echo for these same months of each year.
The third method calculated RAD hospitalizations by using either monthly weighted estimates or age weighted estimates that were observed between 1974 and 1982 in a Washington, DC hospital (8).

Rotavirus infection indicators
Several indicators were suggested by Lebaron et al in 1990 (6) to measure trends in rotavirus epidemiology: the peak season (two consecutive months with the highest number of detections), the onset of rotavirus season (the month in which detections first exceeded the monthly mean), the duration of rotavirus season (the number of months during which detection exceeded the mean) and the epidemic intensity (defined as the ratio of the number of detections for the peak month to the number of detections for the mean observed during the season).
These indicators of rotavirus infection were extended to all diarrhea hospitalizations.For crude incidence rates, the total number of Quebec children from one month to five years of age in this same 13-year period was used as the denominator.

Statistical analysis
For data analysis, SAS Institute Inc (USA) software was used and one hospitalization stay was the statistical unit.Frequency distributions were studied and crude rates were calculated with a 95% CI to compare variables of interest.
For comparison, Pearson χ 2 tests were performed for qualitative variables.

RESULTS Diarrhea-related hospitalizations
There were 63,827 diarrhea-related hospitalizations among one-month to five-year-old children in Quebec from April 1, 1985 to March 31, 1998, for a cumulative denominator of 5.8 million person-years.The annual number of cases varied between 4113 and 6017 hospitalizations (Table 1).Quebec had an average birth cohort of 90,000 children during those years.
Although the onset of rotavirus season was generally in   years.In low incidence years, the peak numbers reached 400 to 600 monthly hospitalizations compared with 600 to 1000 monthly hospitalizations in high incidence years.Epidemic intensity varied from 1.287 in 1991 to 2.061 in 1990.For example, in 1991, the month with the highest number of hospitalizations had 28.7% more hospitalizations than the monthly mean.In 1990, the peak month had more than twice as many hospitalizations as the yearly average.The crude incidence rates decreased with increasing age (Table 3), from 2113.2/100,000 one-to 11-month-old children (95% CI 2085.9 to 2140.6) to 411.4/100,000 48-to 59-month-old children (95% CI 399.9 to 422.9).Boys were hospitalized more frequently for a diarrheal episode than were girls (54.8% and 45.1%, P<0.001).
For diarrheal episodes, mean hospital stay duration varied from 6.0 days in one-to 11-month-old children to 2.8 days in 48-to 59-month-old children (Table 4).The overall annual mean duration decreased from 5.2 days (in 1985) to 3.3 days (from 1996 to 1998).It also varied from 2.9 days to 5.8 days within various regions throughout the province.
A viral etiology was confirmed in 24.9% of all diarrhearelated hospitalizations; bacterial and parasitic infections were confirmed in 5.4% and 0.2% of all diarrhea-related hospitalizations, respectively.Presumed infectious etiologies and presumed noninfectious etiologies represented 8.2% and 62% of total hospitalizations for diarrhea, respectively.

RAD
The three methods gave relatively close estimates of the incidence of RAD: The estimated annual mean was 1505.9 hospitalizations for RAD with the Ho et al ( 4) and Jin et al ( 5) method, 1424.0 hospitalizations for RAD with coefficient 0.37, provided by Ford-Jones et al (7), and 1352.8 and 1849.0RAD with Brandt's (8) monthly and age-specific coefficients, respectively (Table 5).

DISCUSSION
In routine inpatient hospital practice, laboratory tests are not used consistently to microbiologically confirm rotavirus infection.Therefore, direct data are lacking to determine the proportion of diarrheal episodes due to rotavirus, and other methods must be used to estimate the burden of hospitalizations that may be prevented by a new vaccine.
Because 92% of rotavirus infections are observed among children under five years of age (9), the present study was limited to this age group.The present case definition was similar to what has been used by other investigators (10).
One remarkable feature of the epidemic curve is the annual alternation of low and high monthly peaks.The reason for this periodicity is not known.One hypothesis is that it takes two years to accumulate a sufficient number of susceptible children to produce a full-blown epidemic.Brandt et al (8) reported this, specifically for rotavirus-proven infection, which implied that studies of RAD incidence should be conducted over at least two years.
Because the average annual incidence rate of diarrhearelated hospitalizations was 11/1000 children younger than five years of age (Table 3), the cumulative incidence rate over five years would then be 55/1000 children.In other words, a child's risk for hospitalization for diarrhea in the first five years of life would be 1/18.This is approximately twice the risk that was observed in Toronto in 1998 (11).The Toronto data were collected over one year only.The lower estimate in that study may have been due to a different pattern of hospitalization in the Toronto area, a 'small year' in terms of diarrhea or the differences in the methodology used by the two studies.The authors of the Toronto study mentioned that they likely underestimated the hospi-  talization rate.The results of the present study seem to confirm this statement.
It is interesting to note that the three methods that were used in the present study provided relatively similar estimates of the number of RAD-related hospitalizations.Each method has its weaknesses.The first one, used by Ho et al (4) and Jin et al (5), was based on the hypothesis that all excess diarrhea-related hospitalizations observed in winter were due to rotavirus infections, which is an oversimplification of reality.The second method used a coefficient that was measured prospectively in Toronto hospitals in 1998.Unfortunately, the present study showed that the incidence of rotavirus diarrhea varies from year to year, while this observation was made over a one-year period only.The third method may be the most satisfactory because Brandt et al (8) collected prospectively the proportion of RAD over all hospitalizations for diarrhea over a seven-year period and provided age-and month-specific proportions.The age-specific estimate was the only one that diverged slightly from the others.Put together, the three methods provide a robust estimate of the mean annual number of hospitalizations due to RAD (Table 5) -approximately 320 hospitalized RAD cases/100,000 children or 1500 RAD-related hospitalizations/year for the whole province of Quebec.
The mean duration of hospitalizations for all diarrheas decreased over the years of the study despite significant differences between regions.In the past few years, it stabilized at 3.3 days.This figure may be slightly inflated by the presence of a few noninfectious chronic diarrhea cases and by the secondary diarrhea cases, which were likely to have been primarily nosocomial.The 15,155 diarrhea cases that were registered as secondary diagnoses represent 23.7% (95% CI 23.4 to 24.0) of all diarrhea-related hospitalizations.The large number of noninfectious diarrheas is an artefact of the coding practice that puts all diarrheas without specific laboratory evidence of infection in this category.

CONCLUSIONS
The present retrospective study showed the importance of diarrhea-related hospitalizations among children up to five years of age, and highlighted the potential impact that might result from the introduction of a rotavirus vaccine.

Figure 1 )
Figure 1) The number of monthly diarrhea-related hospitalizations (all diagnoses) among children younger than five years of age, Quebec, April 1, 1985 to March 31, 1998

TABLE 1 Diarrhea-related hospitalizations (all diagnoses) among children younger than five years of age in Quebec, April 1, 1985 to March 31, 1998
*Partial Year-April to May 1985; † Partial year-December 1997 to March 1998; Dec December; Nov November

TABLE 2 Diarrhea-related hospitalizations (all diagnoses) among children younger than five years of age in Quebec, April 1, 1985 to March 31, 1998 by civil year
Onset of rotavirus season -month in which detections first exceeded the monthly mean; Duration of rotavirus season -number of months during which detection exceeded the mean.Peak season -two consecutive months with the highest number of detections; Epidemic intensity -defined ratio of the number of detections for the peak month to the number of detections for the mean.Dec December; Feb February; Nov November

TABLE 3 Numbers and crude incidence rates, by age group, of diarrhea-related hospitalizations (all diagnoses) among children younger than five years of age in Quebec, April 1, 1985 to March 31, 1998
Crude incidence rate by age group and 95% CI per 100,000 children of the same age groups *

TABLE 5 Methods of comparing rotavirus-associated diarrhea hospitalizations among children younger than five years of age, Quebec, April 1, 1985 to March 31, 1998 Ho et al and Jin et al Ford-Jones Brandt et al (reference 8)
Coefficient of 0.37 applied for all columns for eight months; Parentheses indicate partial year; NA Not applicable