How Commonly Are Children Hospitalized for Dehydration Eligible for Care in Alternative Settings?

Background  Avoiding unnecessary hospitalization has long been a goal of child health care providers. Managed care practice environments increasingly pressure the practicing pediatrician to avoid hospitalization.

Objectives  To estimate the proportion of childhood dehydration hospitalizations eligible for care in alternative settings (eg, short-stay treatment and triage units, home nursing) and to assess the type and duration of services that might be required for alternative setting care of children with these illness episodes.

Design  All dehydration hospitalizations for the 198,593 children (aged >1 month and <19 years) dwelling in Rochester, NY (Monroe County), between 1991 and 1995 were identified in county-wide hospital discharge computer files. Medical records were reviewed for a random sample of 380 of the hospitalizations. Children with major underlying conditions were excluded from analysis because of higher risk for deterioration, and greater complexity of medical care might render alternative settings inappropriate. Measures included a 4-item score estimating level of dehydration, serum bicarbonate level at presentation, and time to rehydration. Rehydration was defined as a drop in urine-specific gravity to 1.010 or less or reduction of fluid administration to the maintenance rate.

Results  Altogether, 1121 dehydration hospitalizations occurred during the study period. Based on medical record review for a random sample of 380 of these 1121, major underlying problems were present in 27.4% (104) of hospitalizations sampled. Simple, acute gastroenteritis accounted for 75.4% (208) of 276 hospitalizations remaining in the sample. Levels of dehydration for these children were estimated as at least 5% for 51.0% (106) and at least 10% for 16.3% (34) of hospital admissions, and serum bicarbonate levels were 12 mmol/L or less for 26.0% (54). Time from hospital admission to rehydration was no greater than 12 hours for 79.3% (165) and no greater than 24 hours for 94.7% (197). However, hospital stay was generally substantially longer. The time hospitalized following rehydration represented 85.8% of the average inpatient stay. Hospital discharge was heavily concentrated in daytime hours, although the children achieved rehydration at all hours of the day. No deterioration occurred during hospitalizations studied.

Conclusion  Nearly all children hospitalized for simple, acute gastroenteritis in Rochester might be eligible for care in alternative settings designed to provide hospital-level care for short periods.

GASTROENTERITIS and dehydration account for a large proportion of childhood hospitalizations. In US children younger than 5 years, 220,000 hospitalizations are attributed to diarrhea.¹ These represent 9% of all hospitalizations for this age group. In 1992, among California children aged 1 through 5 years, gastroenteritis and dehydration accounted for 10.6% of hospital discharges, ranking second to asthma, which accounted for 12.8%.² Estimates based on recent observations in a large children's hospital in Texas indicated that rotavirus, the most common etiology of gastroenteritis among children, produces 110,000 hospital admissions annually in the United States at a cost of $352 million.³ In the setting for the present study, Rochester, NY (Monroe County), gastroenteritis and dehydration was the third most common medical reason for hospital admission between 1994 and 1996 among children older than 1 month and younger than 19 years. This problem accounted for 3.9% of all childhood hospital admissions, following only asthma and other lower respiratory tract illness, which accounted for 7.3% and 5.8%, respectively (K.M.M. and C.W., unpublished data, 1997).

Current medical practice environments increasingly pressure the practicing pediatrician to avoid hospitalization. Moreover, hospitalization is disruptive to families and has been associated with notable psychosocial morbidity.⁴ Geographic variation studies suggest that a large proportion of hospitalizations for dehydration or gastroenteritis are discretionary^5-10 and may be potentially replaced through care in alternative settings or avoided altogether. Small-area variations as great as 14-fold in the province of Ontario⁵ and 18-fold among regions in Washington State⁷ have been observed. Alternative settings might include short-stay treatment and triage centers (or observation units), home nursing care, and sick-child day care.⁴ Extensive experience in the use of treatment and triage centers for avoidance of hospitalizations for acute asthma exacerbation has been reported.^11-17 Innovations in the treatment of dehydration, particularly oral rehydration,^18-22 may facilitate avoidance or replacement of dehydration hospital admissions.

The primary objectives for this study were (1) to estimate the proportion of childhood dehydration hospitalizations eligible for care in alternative settings and (2) to assess the type and duration of services required for the care of these illness episodes, irrespective of setting. An important methodological challenge in meeting these objectives was to develop a measure of clinical stability in children with dehydration that could be based on medical record review. Such a measure should be useful in developing evidence-based approaches to more efficient treatment of various problems associated with dehydration.

To provide a broad view of the potential for avoiding hospitalization, dehydration attributed to both gastroenteritis and other causes was considered. Viral gastroenteritis was expected to account for most dehydration hospitalizations, but treatment requirements for dehydration owing to problems such as gingivostomatitis may be similar to those for dehydration attributed to viral gastroenteritis. Children with major underlying chronic problems contributing to acute dehydration were excluded from analysis because treatment complexity or risk for deterioration might often make hospitalization necessary despite the availability of alternatives.

We conducted a descriptive study based on analysis of medical record reviews of a random sample of dehydration hospitalizations as well as hospital discharge files. The sampling frame for the medical record reviews included all of the hospitalizations between 1991 and 1995 (N = 28,047) for children (aged >1 month and <19 years) who were residents of Monroe County (Rochester), NY. No programs designed to replace pediatric hospitalizations were active or developing during the study period. Hospital discharge data were obtained from the Rochester hospitals' database, a data set developed and maintained cooperatively by all Rochester-area hospitals. Among these hospital discharges, there were 1121 with a primary diagnosis suggesting dehydration. Dehydration hospitalizations were identified by a primary hospital discharge diagnosis coded according to the International Classification of Diseases, Ninth Revision²³ indicating gastroenteritis due to a specific agent, nonspecific viral gastroenteritis, noninfectious gastroenteritis or diarrhea, hypovolemia, or diarrhea. The following International Classification of Diseases codes falling in these clinical groups were encountered: 0030., 004.0, 004.3, 004.8, 004.9, 008.0, 008.00, 008.43, 008.45, 008.49, 008.5, 008.6, 008.61, 008.67, 008.69, 008.8, 009.0, 009.1, 276.5, and 558.9. Admissions with a primary hospital discharge diagnosis indicating vomiting alone (codes 787.0-787.03) were excluded for several reasons. The number of these vomiting hospitalizations (n = 16) was small compared with the number that met the definition of dehydration hospitalizations (1121). In our clinical experience, some children with simple, acute gastroenteritis manifest their illness primarily with vomiting at the time they are hospitalized, but before they are discharged from the hospital, they usually develop diarrhea as well; therefore, they receive a hospital discharge diagnosis of acute gastroenteritis. We believe that children who at hospital discharge have yet to develop illness manifestations that would support a diagnosis more specific than vomiting probably did not have acute gastroenteritis. They probably had an illness associated with a high level of diagnostic uncertainty, and thus might not make good candidates for care in alternative settings.

Information for characterizing diagnostic considerations on presentation, illness severity over time, and diagnostic and therapeutic services provided was obtained by detailed review of medical records from the time of presentation in the emergency department to the time of hospital discharge. In the interest of efficiency, computer-generated random samples for detailed medical record reviews were drawn from 2 of 3 hospitals that care for pediatric inpatients. These 2 hospitals accounted for more than 90% of dehydration hospital admissions during the study period. The goal of sampling was to obtain 300 illness episodes for medical record review in which children were potential candidates for care in alternative settings. Dehydration episodes in children with a major underlying chronic problem were excluded from primary analysis. Major underlying chronic problems included those in which the underlying condition might contribute directly to the dehydration (eg, short-gut syndrome, renal failure, or ventriculoperitoneal shunt failure) or might impair physiologic resilience (eg, swallowing disorder, bronchopulmonary dysplasia). The sampling goal was chosen so that major subgroups would each have approximately 100 cases and thus yield point estimates with reasonable precision. Based on a preliminary review of the first 130 medical records randomly selected, it was estimated that 380 records would require review to reach the goal of 300 illness episodes without a major underlying chronic problem. Altogether, 380 medical records were reviewed from the 1121 dehydration hospital admissions identified among hospital discharges, for an overall sampling fraction of 33.9%.

Figure 1 shows the relationships among study subgroups. Dehydration episodes in children with no major underlying chronic problem were labeled as dehydration and no chronic problem. Dehydration and no chronic problem included 3 subgroups, 2 of which were based on the causes of dehydration that were apparent at the time of hospital admission. These 2 subgroups included children whose dehydration was attributed to simple, acute gastroenteritis, generally viral, and children whose dehydration was attributed to some other acute problem, such as oral infection (eg, gingivostomatitis, pharyngitis) or recent tonsillectomy. The final subgroup, diagnostic uncertainty, was characterized at the time of presentation by uncertainty about diagnosis and consideration of a serious acute problem (eg, appendicitis) other than dehydration.

To allow generalizability to alternative care settings to be assessed, severity of fluid and electrolyte derangement was characterized by the level of dehydration and presence of acidosis and hypernatremia. Severe hypernatremia and severe acidosis were defined, respectively, by serum sodium values greater than 150 mmol/L and serum bicarbonate values of 12 mmol/L or less. Level of dehydration was based on a 4-item scale, recently described by Gorelick et al,²⁴ and on the presence of acidosis or hypernatremia. The 4 items on the Gorelick scale are ill appearance, capillary refill more than 2 seconds, absent tears, and dry mucous membranes. Based on this scale, moderate or greater dehydration (5% dehydrated) was defined as 2 or more of these findings, and severe dehydration (10% dehydrated) was defined as 3 or more findings. Using rehydration weight as the criterion standard, Gorelick et al found sensitivity and specificity of this scale for 5% dehydration to be 0.79 and 0.87, respectively. For 10% dehydration, sensitivity and specificity were 0.82 and 0.83, respectively. Any item on the severity scale that was not mentioned in the medical record was assumed to be normal. Clinical observations in emergency departments at which children presented were made by house staff and attending physicians. House staff were supervised by attending physicians at all times. Unlike observations performed by Gorelick et al while developing and validating their scale, there was no attempt to standardize observations through procedures, such as using a stopwatch while assessing capillary refill.

Several measures addressed the potential for care in alternative settings, including time to rehydration, hospital length of stay, and hospital stay beyond rehydration. Times were measured in 4-hour periods. Four-hour periods were chosen as the unit of time because the point at which key observations were made, while often somewhat ambiguous in medical records, could almost always be placed within the first or second half of a standard, 8-hour nursing shift. Nursing shifts, in general 8 hours long, measure a critical health care resource. Four-hour periods were converted to hours to simplify the presentation of results. Observations were abstracted from medical records for as many as 24 periods (4 days). Nearly all (96.0%) (n = 265) hospital admissions for dehydration and no chronic problem lasted less than 4 days.

The primary outcome measure, rehydration, was chosen because it represents a cardinal index of physiologic stability in children with dehydration. Once rehydration has been achieved, the treatment of children presenting with simple, acute dehydration generally is straightforward, with the primary focus usually becoming the administration of sufficient fluids to meet maintenance requirements and to replace ongoing losses. Achievement of rehydration was based on observations of urine-specific gravity and on fluid administration rate. A child was considered to be rehydrated in the first 4-hour period when (1) isotonic or more dilute urine was documented, or (2) the rate of intravenous (IV) fluid administration was reduced to the maintenance rate. During the study period, urine-specific gravity was measured only by refractometer. The maintenance fluid rate was estimated for each child by the Holiday-Segar method.²⁵ The maintenance IV fluid rate was reached when IV fluids were reduced to a rate that was no greater than 5 mL/h above the maintenance rate. Unintentional interruptions in IV fluid administration, due to fluid infiltration, for example, were not counted as reductions in fluid administration.

Several clinical situations were of particular interest because they might present special challenges to care in alternative settings. These included severe dehydration, severe hypernatremia, severe acidosis, and critical care unit (CCU) transfers. Many clinicians might be unwilling to care for a child in alternative settings until it was clear that such conditions were resolving. Severe dehydration, severe hypernatremia, and severe acidosis are defined above. A CCU transfer was defined by admission of a child initially to a regular inpatient unit, followed by transfer to the pediatric CCU. A CCU transfer, when it occurs, is often an appropriate response to deterioration despite usual inpatient care. Deterioration during care in alternative settings obviously is undesirable, even though careful monitoring in such settings should prevent adverse physical outcomes.

Denominators for calculation of county-wide hospitalization rates were based on the 1990 US census. Over the 5-year observation period, the 198,593 children younger than 19 years in the 1990 census accounted for 992,965 child-years at risk. The t test was used to assess statistical significance for differences between 2 means, and the F test was used to assess significance for differences among multiple means.

During the study period, 1121 hospitalizations in Monroe County were attributed to dehydration, yielding a hospitalization rate of 1.13 (95% confidence interval, 1.06-1.93) per 1000 child-years. Table 1 displays the frequencies of primary hospital discharge diagnoses among all dehydration hospital admissions and in the medical record sample. As shown in Figure 1, among the 380 records reviewed, 104 (27.4%) were excluded from the primary analysis because a major underlying condition contributed directly to the dehydration or may have impaired physiologic resilience. Many of these children had multiple chronic problems. The 10 most common chronic problems included failure to thrive and feeding disorders, 17; neuromuscular disorders (eg, cerebral palsy, congenital quadriplegia, Werdnig-Hoffmann disease, hydrocephalus), 16; seizure disorder, 13; asthma and bronchopulmonary dysplasia, 8; malignancy, 8; chronic renal disease, 7; chronic liver disease, 6; congenital heart disease, 5; postsurgical states (eg, gastrostomy, ileostomy), 5; and endocrine and/or metabolic disorders, 4. The remaining 276 illness episodes included the group with dehydration and no chronic problem. Assuming the number of children with underlying problems occurred among all dehydration hospitalizations in the same proportion that they occurred in the record review sample, the county-wide hospitalization rate for dehydration and no chronic problem was estimated as 0.82 (95% confidence interval, 0.76-0.87) per 1000 child-years.

The mean (SD) ages of all children hospitalized with dehydration and those in the record review sample were 4.4 (5.2) and 4.2 (4.8) years, respectively. Sex distributions among all children hospitalized and those in the record review sample were also similar, with males constituting 53.9% (604) overall and 55.3% (210) in the sample.

Among the 276 children most likely to be candidates for care in alternative settings—those with dehydration and no major chronic problem—the diagnosis was apparent at the time of hospital admission for most, including the 208 (75.4%) with simple, acute gastroenteritis and the 25 (9.5%) with other acute problems. Children with other acute problems included 3 who were dehydrated following tonsillectomy, 14 with oral (eg, gingivostomatitis) or pharyngeal (eg, streptococcal pharyngitis) conditions that interfered with oral intake, and 8 with viral syndromes associated with vomiting. For the remaining 43 (15.6%), diagnostic uncertainty at the time of hospital admission appeared to be an important factor in justifying the hospitalization. Diagnostic considerations for these 43 included appendicitis, 21; bacterial gastroenteritis, 3; other serious bacterial infection (eg, sepsis, urinary tract infection), 11; and a diverse group of other problems, 8. Males predominated (56.9%) (n = 118) among 208 children with simple, acute gastroenteritis. Differences in sex distribution among the 3 subgroups were not statistically significant. As shown in Table 2, the mean age for all children with dehydration and no chronic problem was 4.1 years. Age distributions varied distinctly among subgroups (F₂ = 20.1, P<.001); children with simple, acute gastroenteritis were younger (3.2 years) than those with diagnostic uncertainty (7.6 years) and other acute problems (5.2 years). As expected for gastroenteritis predominately due to rotavirus, the largest proportion (28.4%) (n = 59) of simple, acute gastroenteritis hospital admissions occurred in March, and the months of January through April accounted for 71.2% (n = 84) of hospital admissions in this subgroup. Among children in the remaining 2 dehydration subgroups, little seasonal predominance was observed; 55.9% (n = 38) of hospital admissions occurred during the first 6 calendar months.

A substantial portion of children admitted to the hospital with dehydration and no chronic problem were moderately or severely ill, as indicated by observations that 44.2% (122) were at least 5% dehydrated, 13.8% (38) were at least 10% dehydrated, 19.9% (55) had an initial serum bicarbonate value of 12 mmol/L or less, and 6.5% (18) were hypernatremic (Table 2). Also apparent in Table 2, the worst fluid and electrolyte disturbances were found among children with simple, acute gastroenteritis.

Isotonic urine production and maintenance fluid administration identified the time that rehydration was achieved in 55.1% (152) and 44.9% (124) of illness episodes, respectively. Rehydration was achieved long before children were discharged from the hospital. As indicated in Table 2, the mean time for all children with dehydration and no chronic problem from admission to the inpatient unit to rehydration was 6.4 hours, whereas the mean time from hospital admission to hospital discharge was 44.0 hours. The severity of dehydration and time to achieve rehydration varied substantially among the dehydration subgroups, but the hospital length of stay did not. Measured from the time of admission to the inpatient unit, time to rehydration varied significantly (F₂ = 3.29, P = .04) among subgroups, but time to hospital discharge did not (F₂ = 1.66, P = .19).

Differences in time to achieve rehydration and length of stay were also examined by the level of dehydration and by the presence of severe acidosis and hypernatremia at presentation. As expected, the level of dehydration among all children with dehydration and no chronic problem predicted the time required for rehydration (F₂ = 3.68, P = .03). Mean (SD) times from hospital admission to rehydration for children estimated to be less than 5% dehydrated, 5% to 9% dehydrated, and 10% dehydrated or greater were 5.4 (8.1) hours, 7.3 (8.0) hours, and 8.9 (7.5) hours, respectively. Mean (SD) times from hospital admission to rehydration of children with (n = 55) and without (n = 221) severe acidosis were 7.5 (8.0) hours and 6.2 (8.1) hours, respectively (unpaired, 2-tailed t₂₇₄ = 1.09, P = .28).

The subgroup with simple, acute gastroenteritis represents a large, clinically homogeneous category for which challenges of treatment are well defined. Accordingly, estimates for the potential impact of alternative settings on care for this group seemed most reliable and most useful. For children with simple, acute gastroenteritis, Figure 2 compares proportions of those remaining dehydrated and those remaining hospitalized over time.

Times from presentation in the emergency department and from arrival at the inpatient unit are both used in Figure 2. These times provide upper and lower bounds for the length of hospital stay potentially avoided. The time of arrival at an inpatient unit may vary considerably with factors, such as bed availability, that are unrelated to the treatment of a specific illness episode. If care in alternative settings began at the same time that the children arrived at the inpatient unit, the area between the dashed lines estimates the total avoidable hospital stay. If care in alternative settings began immediately following arrival in the emergency department, the corresponding area between the solid lines estimates the total avoidable hospital stay. In this latter example, the term hospital stay includes time spent in the emergency department.

Much of the rehydration for children with simple, acute gastroenteritis (n = 208) was accomplished in the emergency department. Eighty percent of children (166) were rehydrated within 12 hours of admission to the hospital. Mean (SD) time spent in the emergency department was 7.2 (3.8) hours. As demonstrated in comparing dashed lines in Figure 2, by the time children arrive at the inpatient unit, already 22.7% of children (47) are no longer dehydrated. Twelve hours later, only 20.7% (43) are still dehydrated, while 95.7% (199) are still hospitalized. Twenty-four hours later, only 5.3% (11) are still dehydrated, while 70.2% (146) remain hospitalized. Considering time spent and rehydration achieved in the emergency department (Figure 2, solid lines), almost half (46.2%) (n = 96) are rehydrated by 12 hours after arrival at the emergency department, whereas virtually all remain in the hospital at that time. By 24 hours from presentation, only 9.1% (19) remain dehydrated, whereas 86.1% (179) remain hospitalized. For times from presentation, the mean (SD) difference between rehydration and hospital discharge was 36.3 (29.0) hours.

Among the 276 children with dehydration and no chronic problem, none were transferred to a CCU. Given the sample size of 276, the upper limit of the 95% confidence interval for the rate of CCU transfers is 1.1%.^26,27

This study addresses the potential to avoid or replace dehydration hospitalizations in children with no major underlying conditions. Although simple, acute gastroenteritis accounted for most (75.4%) (n = 208) dehydration hospitalizations with no chronic problem, substantial proportions were attributable to other acute problems identified on hospital admission (9.1%) (n = 25) or to circumstances characterized by diagnostic uncertainty (15.6%) (n = 43).

Some treatment challenges for the latter 2 groups differ from those for the simple, acute gastroenteritis group. Based on the frequencies of severe dehydration, severe acidosis, and severe hypernatremia, physiologic derangement was generally less severe in children with other acute problems and in children with diagnostic uncertainty. Primary treatment needs for the group with other acute problems, eg, rehydration and maintenance fluids owing to the inability or refusal to drink, seem straightforward and readily met in alternative settings that are capable of administering IV fluids. Difficulty with oral intake limits the usefulness of oral rehydration in this group.

Treatment needs for the diagnostic uncertainty group seem more complex. When diagnostic uncertainty includes concern for a serious acute problem that requires frequent physician examination or surgical consultation (eg, acute appendicitis, which accounted for almost half this group), the most practical alternative to hospitalization probably would involve close proximity to such services.

Available data provided the most useful information about the potential to avoid or replace hospitalization for the large and relatively homogeneous group with dehydration due to simple, acute gastroenteritis. As demonstrated in Table 2 and Figure 2, rehydration was achieved rapidly despite the fact that large proportions in this group had dehydration that was at least moderately severe or had severe acidosis. Although the times to rehydration varied significantly by level of dehydration, the mean difference in time to rehydration between children with less than 5% dehydration and children with 10% or more dehydration was only 3.5 hours. Most notably, children with simple, acute gastroenteritis remained hospitalized for long periods after rehydration was achieved. The mean time to hospital discharge following rehydration was 36.3 (29.0) hours. This period represented 85.8% of the mean hospital stay as measured from the time of arrival in the inpatient unit to the time of hospital discharge, and 73.2% of the mean hospital stay as measured from the time of arrival in the emergency department to the time of hospital discharge (Table 2). If one assumes that entry to care in an alternative setting would probably occur at some time earlier in the course of the illness than current hospital admissions, then 73.2% and 85.8% represent the lower and upper bounds for the proportion of the hospital experience potentially eliminated for children with simple, acute gastroenteritis.

Equally important, we estimate that the proportion of hospital admissions for simple, acute gastroenteritis that might be avoided through care in alternative settings would approach 100%. The mean time from presentation to rehydration was 13.3 hours. By 24 hours after presentation, more than 90% of the patients had been rehydrated. There was no indication that entry to care in an alternative setting should be delayed until rehydration was achieved. No children admitted with simple, acute gastroenteritis received care other than oral or IV fluids and monitoring, care that could be provided readily through a home nursing service or in a short-stay unit. The mean times to achieve rehydration for children whose dehydration was 10% or more or who had severe acidosis were only marginally (<2 hours) longer than for children whose physiologic derangement was milder. Deterioration to the point that a child was transferred to the CCU did not occur. No child with dehydration or a chronic problem was admitted to the CCU for any portion of his or her hospital stay.

Previous reports include observations that are consistent with our findings and support our interpretations. Several studies indicate that rehydration may be achieved in less than 8 hours through either IV^28,29 or oral rehydration techniques.^18,30-32 Failure of oral rehydration, defined as replacement of this approach with IV therapy, has been estimated as only 3.6% (95% confidence interval, 1.4%-5.8%).²⁰ Difficulty in predicting persistent vomiting and ongoing fluid losses from diarrhea presents a challenge in the treatment of acute gastroenteritis,^28,33 but these issues could be addressed effectively in an alternative approach if careful follow-up is built in.

Observations support the validity of our measure of rehydration. The level of physiologic derangement, as indicated by the level of dehydration and severity of acidosis, was greater in the group of children with simple, acute gastroenteritis than in the other 2 subgroups analyzed. Longer time to rehydration for the simple, acute gastroenteritis group would be expected on this basis. Time to rehydration was close to 3 hours longer for the children with simple, acute gastroenteritis than for children in the other 2 subgroups (Table 2), consistent with this expectation. That our measure for time to rehydration performs as expected suggests that it provides a useful index of the phenomenon it is intended to measure (return to normal fluid balance).

Time to rehydration as measured by medical record review overestimates the time required for rehydration to be achieved. Output of isotonic or more dilute urine was considered the best available index of rehydration. In persons with normal renal function, urine concentration provides a direct measure of hydration, although this index may provide a somewhat conservative estimate for the time that rehydration is achieved because of the lag between urine production by the kidneys and bladder emptying. Unfortunately for purposes of this study, urine-specific gravity was not always recorded after rehydration had apparently been achieved. Many children had no recorded urine-specific gravity for many periods before or after the time when clinical observations and treatment (eg, discontinuation of IV fluids) clearly indicated that rehydration had occurred. The second criterion for achievement of rehydration, reduction of IV fluid administration to the maintenance rate, was adopted because of this practical limitation in the use of urine concentration as the indicator for this outcome. Assuming valid clinical assessments, maintenance IV fluid administration clearly provides a conservative estimate for the time that rehydration was achieved. Considerable lag time is expected from when a child achieves rehydration to when a physician's order to reduce IV fluid administration is implemented. Despite this lag, the time at which rehydration is achieved should be understood as a time when the level of monitoring can be reduced, not necessarily as a time for discharge from inpatient or alternative services. Diminished fluid intake and excess fluid loss commonly continues well beyond the time that rehydration is achieved.

The most important limitation is that studies with observational designs cannot provide precise estimates for the proportion of hospital admissions that could be avoided or replaced. Precise estimates can only be provided through randomized clinical trials of well-defined alternative services. End points available through medical record review provide imperfect measures of the duration of care needed in alternative settings. For example, children with simple, acute gastroenteritis are generally not yet well at the time they become rehydrated, and some of these children will not be ready for discharge from an alternative setting at the time rehydration is achieved. This difficulty in interpreting our findings is compounded by that the fact that we have not specified a particular alternative setting. The capacity to provide particular services and the ability to respond to unforeseen events may vary among alternative settings, such as short-stay units located within the medical center and home nursing care. We also recognize that external validity is limited by our one community sample. We have focused on physiologic end points as measures of potential rather than services delivered (eg, hospital length of stay) because the biologic characteristics of children with gastroenteritis are much less likely to vary than patterns of care.

The generalizability of our findings may be assessed by comparing distributions for levels of dehydration, serum bicarbonate, and serum sodium with those for children in other settings. It should be recognized that Gorelick's scale,²⁴ used for characterizing the level of dehydration, has been validated among children between 1 month and 5 years of age. Children in this age range constituted 81.7% (170) of the group with simple, acute gastroenteritis. Given the components of this scale (appearance, capillary refill, mucous membranes, tears), it may be valid for most of the remaining 18.3% (38) as well, but this has not been established. Children younger than 9 years constituted 91.3% (190) of those with simple, acute gastroenteritis.

Despite clear differences in severity among the 3 dehydration subgroups, there was no significant difference in hospital length of stay. Even more striking is that the time following rehydration represented 85.8% of the hospital inpatient stay. These observations suggest that factors other than physiologic derangement are key determinants of length of hospital stay. Bed availability, an issue that frequently delays hospital admissions, would generally not be a problem when admitting a child to home nursing. Although Figure 2 demonstrates that the achievement of rehydration is a continuous process, hospital discharges concentrate at certain times of the day. Figure 3 demonstrates a pattern for hospital discharge times that suggests hospital discharge is heavily influenced by physicians' rounds schedules and office routines of attending physicians, continuity clinic schedules of residents, and work and family responsibilities of parents. Most hospital admissions (74%) occur late in the evening or in the middle of the night. Although these children are rapidly and steadily rehydrated throughout the night they are admitted and in the daylight hours that immediately follow, hospital discharges occur most commonly in the afternoon the day after hospital admission. Given current inpatient care routines, this pattern of hospital discharges is not surprising. Our clinical experience suggests that many providers are reluctant to discharge from the hospital a child with remaining illness symptoms, even when remaining symptoms would be insufficient to justify hospital admission. We speculate that if children were admitted directly from the emergency department to a home nursing service that reduction and eventual withdrawal of services for children already in their own homes would occur sooner than current hospital discharge.

Findings indicate that rehydration is rapidly achieved for virtually all children hospitalized for dehydration due to simple, acute gastroenteritis. Virtually all hospital stays for dehydration could safely be shortened or eliminated, given modest health system innovations. Given that childhood hospitalization rates in Rochester have historically been substantially lower than those of other urban communities, potentially large cost reductions in Rochester might be still larger in many other communities. In 1982, for example, the hospitalization rate for childhood gastroenteritis in this community was less than half that in Boston, Mass, and almost half that in New Haven, Conn.⁹ Allocation of resources for a randomized trial to estimate carefully the efficacy and quality of care alternatives may be a wise investment.

Accepted for publication April 19, 1999.

This study was supported in part by grants MCJ-360831 from the Maternal and Child Health Program (Title V, Social Security Act), Health Resources and Services Administration, and 1 RO1NRO45O1-O1 from the National Institute of Nursing Research, National Institutes of Health, Department of Health and Human Services, Bethesda, Md; and by the Ronald McDonald Children's Charities Grants Program, Oak Brook, Ill.

Presented in part at the 38th Annual Meeting of the Ambulatory Pediatric Association, New Orleans, La, May 2, 1998.

The authors gratefully acknowledge the assistance of Ruth Stewart, RN, and Diane Kober, RN, for completion of medical record reviews; Nancy Wood, MPA, and Krista Lyman, BA, for data management; and the Rochester Healthcare Information Group for provision of hospitalization data.

Editor's Note: The hospital is no place for a child (or anyone else) who is not severely ill. The development of 24-hour outpatient units would eliminate a lot of unnecessary hospitalization time for dehydration and probably numerous other problems.—Catherine D. DeAngelis, MD

Corresponding author: Kenneth M. McConnochie, MD, MPH, Department of Pediatrics, Box 777, University of Rochester Medical Center, 601 Elmwood Ave, Rochester, NY 14642.