Objective. To determine whether length of stay (LOS) for asthma admissions at a local university-affiliated children's hospital (UACH) is similar to that of community hospitals within the same county.
Methods. A retrospective analysis was performed using computerized hospital abstract records from 1989 through 1994. The study population was children 1 to 18 years old whose first or only hospitalization for a primary diagnosis of asthma occurred during the study period at either the UACH or one of the 17 community hospitals in King County, WA, that admit pediatric patients (n = 2491). Transfers and patients with chronic obstructive asthma or secondary diagnoses such as cystic fibrosis were not included in the study. Asthma patients were compared by sociodemographic and health risk characteristics such as age, sex, insurance status, and a comorbidity severity score. Differences between the two hospital populations were tested by χ2 and t tests. The effect of hospitalization at the UACH or the community hospitals on LOS was determined using analysis of covariance after adjusting for the sociodemographic and health risk covariates.
Results. Sixty-two percent (62%) of the asthma patients in the study population were discharged from the UACH. Compared with patients discharged from the community hospitals, the UACH patients were significantly younger, more often male, used public insurance, and resided in areas with lower median household incomes. The severity of comorbidities was not different between the two hospital groups. Overall, adjusted mean LOS was not significantly longer at the UACH (2.1 days) than at the community hospitals (2.0 days); however, adjusted mean LOS for specific subgroups, most notably poor children and those with public insurance, was significantly longer at the UACH.
Conclusion. LOS for first or only asthma hospitalizations during 1989 through 1994 at the UACH was similar to local community hospitals within the same county. Specific subgroups of children were hospitalized for a longer period at the UACH, but children with private insurance and from areas with higher median household incomes had similar LOS, and presumably costs, at the UACH and the community hospitals.
The aim of this study was to determine whether length of stay (LOS), as a proxy for cost, was similar for a common pediatric diagnosis at a university-affiliated children's hospital (UACH) compared with community hospitals in King County, WA. In studies that focused mainly on adult patients, teaching hospitals delivered care at higher costs than other types of hospitals.1 Although the reasons for higher costs varied among the many studies, the two major contributors were more adverse case-mix/severity of illness2 and a larger proportion of poor patients at teaching hospitals compared with other hospitals.7Estimates of the importance of case-mix varied from 9% to 30% of costs,12 but even after adjusting for case-mix, teaching hospitals still cost more.1,3,6,10,13,14
Teaching hospitals perform most of the charity care in the United States.7,8,15 Within the same pediatric facility, academic physicians were found to care for a higher percentage of uninsured patients.11 Both adult and pediatric patients of lower socioeconomic status tend to be sicker, have longer LOS, greater hospital resource consumption, and thus, higher costs, after adjusting for other sociodemographic and illness-related characteristics.16 Studies on children and adults with asthma in New York City, Chicago, and Boston, and an analysis based on the 1988 National Hospital Discharge Survey all found significant associations between increased hospitalizations and Medicaid or other public sources of funding, residence in areas with lower median household incomes, and other measures of low socioeconomic status.20
In addition, teaching hospitals have had difficulty maintaining their volume of admissions and referrals as managed care becomes the norm.24 Health plans have been reluctant to contract with teaching hospitals for all but the most intensive care because they perceive teaching hospitals as too costly and inefficient. The problem is intensified for children's hospitals. In the past, children's hospitals provided much of the inpatient pediatric care in the communities where they were located,11 but as community hospitals expand their pediatric services, they are able to market themselves as less expensive alternatives to the specialist-centered focus of a children's hospital.25
The UACH that is the subject of this article is a 208-bed private, not-for-profit regional tertiary care center and teaching hospital for the University of Washington School of Medicine. It finds itself in a changing, competitive marketplace that did not exist 5 years ago. For example, a recent unpublished study found that community hospitals in King County are increasingly likely to admit pediatric patients.25 In addition, the same evaluation, based on 1991 discharges, showed that the UACH had higher mean costs for many, but not all, diagnosis-related groups, although the analysis did not take case mix into account.25
Using a larger data set than the 1991 analysis and a measure of case-mix, this study compares LOS between the UACH and the 17 community hospitals that discharged pediatric patients with the diagnosis of asthma between 1989 and 1994.
Data for this study were obtained from the 1989–1994 Washington State Department of Social and Human Services Comprehensive Hospital Abstract Reporting System (CHARS), a mandatory reporting system for all Washington State hospitals. Hospital and patient information were collected at discharge for every hospitalization. Race and ethnicity were not available from this data set. In a comparison of the CHARS computerized abstract and hospital records at 23 hospitals throughout the state of Washington, the hospital discharge abstracts that form the basis of CHARS were found to be quite accurate.26
The study design was a retrospective comparison of LOS of children discharged between 1989 and 1994 with the primary diagnosis of asthma from the children's hospital (the UACH) or one of the 17 community hospitals in King County (KCCH) without pediatric teaching affiliations. The KCCH did not include either of the two major teaching hospitals that are also affiliated with the University of Washington. Although both admit pediatric patients, one of the hospitals specializes in neurosurgery, burns, orthopedics, rehabilitation, and trauma, while the other admits pediatric patients mostly for pregnancy-related and newborn services. Only 8 pediatric patients with asthma were discharged from these two hospitals between 1989 and 1994. Attending physicians and practicing physicians at the UACH generally do not admit patients to the community hospitals.
There were 4583 discharges of pediatric patients from King County, who ranged in age from 28 days to 19 years, with a primary diagnosis of asthma (International Classification of Diseases, Ninth Revision [ICD-9] codes 493.0, 493.1, 493.2 and 493.9) from the UACH and the KCCH during the time frame of the study. In this data set, records for readmissions that occurred within 1 day of a discharge for the same primary diagnosis were merged with the preceding record and counted as a single admission. We excluded children with chronic obstructive asthma (ICD-9 code 493.2), because children with this more severe form of the asthma are almost exclusively hospitalized at the UACH and our interest was in the more common forms of the disease. Patients coded as transfers were removed because of the difficulty in assigning LOS to the hospitals involved. We limited the study to King County residents who were discharged from King County hospitals because the UACH receives referrals from the entire state while the community hospitals rarely receive out-of-area referrals for asthma. We assumed the out-of-area referrals would include children with more severe asthma. Cases with incorrect zip codes were deleted because we used a zip code-based income measure for some of the analyses. We excluded children under the age of 1 year, because of the difficulty in diagnosing asthma in infants. Patients with secondary diagnoses that could be a complicating factor in the diagnosis of asthma were also excluded. These secondary diagnoses included cystic fibrosis, bronchiolitis, respiratory syncytial virus infection, bronchopulmonary dysplasia, gastrostomies, congenital heart defects other than ventricular septal defect, and mental retardation. We then limited the study population to first or only asthma discharges during 1989 through 1994.
Several sociodemographic and health risk characteristics were used or constructed from the CHARS data set to compare the patient population at the UACH and the KCCH: age, gender, insurance status, median household income based on zip code of residence, the severity of listed comorbidities, whether or not a procedure was coded on the abstract, and readmissions for asthma within 1 week and within the period of the study.
Age groups of children 1 to 2 years old, 3 to 9 years old, and 10 to 18 years old were used for several of the analyses. These age groups were chosen because many pediatricians are reluctant to diagnose asthma in children <3 years old, and teenagers hospitalized for asthma demonstrate different patterns of hospitalization and LOS than younger children.27
Insurance status was dichotomized into a public and a private group. Public insurance status included Medicaid, Medicare, self-pay, and charity care; Medicaid accounted for 98% of this category. The private category included health maintenance organizations, other commercial insurance, health care service contractors, and other sponsored patients. Median household income based on zip code was assigned to each patient, and quartile ranges were determined based on the distribution of those values. These were also used as a measure of socioeconomic status.28
A comorbidity severity score was assigned to each patient based on an assessment of the effect of listed comorbidities on the length of an asthma hospitalization. A score was generated for each patient by using a panel of 11 pediatricians (including two of the authors, B.N.S. and A.H.N.). The panel were given a list of all comorbidities that occurred at least one time among all patients in the order of their ICD-9 code. The instructions asked them to rate all listed comorbidities as 0 (no effect on the LOS), 1 (mild or modest effect on LOS), 2 (moderate effect on LOS) to 3 (strong/strongest effect on the LOS), and examples were provided (eg, respiratory arrest might be assigned a 3 and ringworm, a 0). Interrater reliability was 95% (Cronbach's α). Every patient could have 0 to 8 listed comorbidities. A comorbidity severity score was assigned to each patient using the score of the highest rated comorbidity among all listed secondary diagnoses for that patient.
Procedures were dichotomized into two groups based on coding: no procedures coded or at least one procedure coded. Other health risk variables constructed or used included the number of times an individual was discharged from the hospital for asthma during 1989 through 1994, and readmissions for asthma within 1 week after the initial asthma discharge.
χ2 And t tests were used to test significance among the sociodemographic and health risk variables between the two hospital groups (Table 1). An adjusted mean LOS for each sociodemographic and health risk characteristic was generated using analysis of covariance with a Scheffe adjustment to detect differences between each variable at the UACH and the KCCH (Table 2). Other descriptive statistics for LOS included median, interquartile range, and 95th percentile. Mean LOS was used as the dependent variable in an analysis of covariance to examine the effect of the independent variable of interest, hospital type, adjusting for four other covariates: age, gender, insurance status, and the comorbidity severity score (Table3). All statistical analyses were performed using Statistical Analysis Software (SAS Institute, Cary, NC).
We also examined several transformations of LOS including the log, square, and square root to see if we could improve the distribution of the LOS variable.
Of the original 4583 discharges with a primary diagnosis of asthma during the study period, 80 were excluded because of a primary diagnosis of chronic obstructive asthma (ICD-9 493.2), and 73 because of their transfer status. Two hundred fifty (250) were discharged from hospitals outside of King County. Other exclusions included nonexistent zip codes or zip codes outside of King County (n = 145), children <1 year old (n = 385), and patients with secondary diagnoses that could be a complicating factor in the diagnosis of asthma (n = 61). These secondary diagnoses were cystic fibrosis (n = 3), bronchiolitis (n = 39), respiratory syncytial virus infection (n = 2), bronchopulmonary dysplasia (n = 1), gastrostomies (n = 7), congenital heart defects other than ventricular septal defect (n = 7), and mental retardation (n = 2). Of the remaining 3589 asthma discharges, 2491 were first or only asthma discharges, and, thus, represented individual patients.
Sixty-two percent (62%) of the patients were discharged from the UACH. The number of patients discharged from the 17 community hospitals between 1989 and 1994 ranged from 1 patient (two hospitals) to 306 patients. Seventy-two percent (72%) of all children discharged from the KCCH were from three hospitals; 88% from five hospitals.
Sociodemographic and health risk characteristics for each hospital type are presented in Table 1. Patients discharged from the UACH were significantly more likely to be <10 years old, male, have public insurance, and reside in areas with lower median household incomes. Comorbidity severity scores and the percentage of patients with at least one procedure were not significantly different between the UACH and the KCCH (Table 1). Forty-nine percent (49%) of patients at the UACH and 44% at the KCCH had no secondary diagnoses recorded.
Surprisingly, only 8 patients had one or more procedures recorded on the computerized abstracts; 4 at each hospital type. The most common procedure (n = 5) was other mechanical ventilation. Other procedures included fiber-optic bronchoscopy (3), other bronchoscopy (3), insertion of endotracheal tube (2), and respiratory medication administered by nebulizer (2).
The mean number of hospitalizations during the study period was not significantly different. Over 75% of patients at both the UACH and the KCCH were hospitalized only once for asthma (Table 1). The maximum number of hospitalizations for asthma was 21 at the UACH and 20 at the KCCH (1 patient each). Four patients from each hospital group, were readmitted within 1 week for asthma. Three patients at the UACH and 2 patients at the KCCH were readmitted for another diagnosis within 1 week of their asthma hospitalization.
LOS was skewed with 42% of patients at both the UACH and the KCCH hospitalized for only 1 day. Transformations of the LOS variable did not improve the distribution. Six percent of the UACH and 5% of KCCH patients were hospitalized 5 or more days, the 95th percentile for LOS. Minimum LOS was 1 day at both hospital types. The longest hospitalization was 14 days at the UACH and 11 days at the KCCH.
Median LOS was 2 days overall and 1 or 2 days for most sociodemographic and health risk subgroups. Median LOS was 3 days for patients with comorbidity severity scores of 3 at both hospital groups. Interquartile range was 1 day overall and 1 or 2 days among all subgroups except for patients with a comorbidity score of 3 at the UACH who had an interquartile range of 3 days.
Table 2 summarizes adjusted mean LOS for each sociodemographic and health risk subgroup by hospital type. LOS was longer at the UACH than at the KCCH for five characteristics: female, public insurance, residence in poorer areas, older age, and higher comorbidity scores, although the differences were significant at the <.05 level only for the first three of these characteristics.
Table 3 presents the results for the analysis of covariance. Coefficients represent the effect on LOS. After adjusting for age, sex, insurance status, and comorbidity severity score, LOS was only .06 days longer for patients hospitalized at the UACH. The difference was not significant. Older age, public insurance, and a higher comorbidity severity score all increased LOS significantly.
When we looked at first order interactions (Table 3), two were significant: comorbidity severity score and hospital type and comorbidity severity score and age. This can be interpreted as follows: as comorbidity severity score increased from 0 to 3, LOS increased more for patients hospitalized at the UACH than for those hospitalized at the KCCH, as was suggested by the adjusted mean LOS for the comorbidity severity score in Table 2. The same relationship occurs between increasing comorbidity severity score and older age.
We compared asthma discharges from a pediatric tertiary care referral center with asthma discharges from community hospitals in the same county without major teaching affiliations in pediatrics. Eliminating patients with chronic obstructive asthma and other diseases that complicate the diagnosis of asthma allowed us to compare discharges for the more common, less severe forms of asthma. Major deviations in measurable outcomes such as LOS may thus signify a substantial difference in practice patterns for the routine asthma patient for one of the hospital types.
Because LOS is skewed, we did look at transformations of the variable, including the log, square, and square root of LOS. None of the transformations improved the normality of the distribution, so we used the LOS variable for its readily understandable interpretation. Median LOS, the better statistic for the skewed LOS variable, was similar not only for the two hospital types, but its value (2 days) was also similar to the adjusted mean LOS generated by the analyses of covariance.
Our analysis showed similar LOS statistics for children with asthma, whether hospitalized at the UACH or the KCCHs. Among subgroups where differences in LOS did exist, the values for the UACH were usually longer. Using LOS as a proxy for cost, the analysis suggests that overall the UACH can care for many hospitalized children with asthma at a cost similar to the KCCH, although the UACH may be more expensive for specific subgroups of children.
The largest differences in mean LOS between the UACH and the KCCH occurred in children with sociodemographic and health risk traits that have been found to increase LOS: older age, female sex, public insurance, residence in poorer neighborhoods, and more severe comorbidities,18,19,23 although not all these differences were significant in our study. In general, the subgroups with longer LOS statistics at the UACH are precisely those who are admitted to the UACH in higher proportions. The exception was patients with more severe comorbidities. Patients with higher comorbidity severity scores in our study population were hospitalized at similar rates at each of the two hospital types, but those hospitalized at the UACH had longer LOS. As suggested by the analysis of covariance, there appeared to be a small, but significant interaction between higher comorbidity severity scores and hospitalization at the UACH that increased LOS, although the nature of this interaction is unclear. Whether the longer LOS for these asthma patients reflects an appropriate need for longer hospitalizations or superfluous, unnecessary days of inpatient care at the UACH cannot be determined from the data available to us.
Boys, children with private insurance, and those residing in more affluent neighborhoods had similar mean, median, and interquartile range values for LOS, suggesting that care for these patients with routine asthma was comparable at the UACH and the KCCH. It appears from our data that the UACH treated this group of children at costs similar to the KCCH, despite the perception that teaching hospitals are more costly for the routine patient.
The differences in sociodemographic characteristics between the UACH and the KCCH were expected and conformed to findings in the literature.11,20,25,27 The UACH, like many teaching hospitals admitted a proportionally larger percentage of poor children and children with public funding sources.
The similarities between the two hospitals types for severity of comorbidity were more surprising. Removing patients with chronic obstructive asthma from the analysis, which was not done in other studies, had an effect on these findings, because they often have long hospitalizations and more severe comorbidities. Over the entire range of ICD-9 codes (493.0–493.9), the UACH may hospitalize a sicker group of children, but removing children with chronic obstructive asthma from the analysis allowed us to compare a group of patients with similar illness between the UACH and the KCCH, which was the intent of this study.
The limitations of our study include the lack of data that would help to better describe the patient populations and to find more precise proxies for cost in addition to LOS. Charges, although included in the data set, vary among all hospitals making comparisons difficult and were not available for one of the largest hospital systems in the county. Although we used LOS as a proxy for cost, we could not measure the intensity or quality of the care provided nor could we make judgments on whether longer LOS at the UACH for specific groups was appropriate or inappropriate for the illness that required hospitalization. Procedures were rarely coded in this patient population, and thus, not useful as a proxy for cost.
Using LOS as a proxy for cost, the UACH appeared overall to provide care for children hospitalized with asthma at the same cost as KCCH. The results for some subgroups of children, however, were not comparable. Specifically, children with public funding sources or from poor neighborhoods, who tend to gravitate preferentially to UACHs, and patients with more severe comorbidities were hospitalized longer at the UACH. Whether these children in larger numbers would be hospitalized for significantly shorter periods at the community hospitals is not clear. On the other hand, children with a primary diagnosis of asthma who resided in more affluent neighborhoods and had private insurance had similar lengths of stay whether hospitalized at the UACH or the KCCHs.
This research was funded in part by the Agency for Health Care Policy and Research (AHCPR) Training Grant, Health Services Research Training Program, Department of Health Services, University of Washington, Seattle, WA and by Grant MCJ-9403 from the Maternal and Child Health Bureau (Title V, Social Security Act), Health Resources and Services Administration, Department of Health and Human Services.
We thank Ted Grichuhin and Bill O'Brien for their work preparing the CHARS file, Jon Lindenauer for his help with SAS programming, and Gregory Redding, MD, Gary Hart, PhD, and Allen Cheadle, PhD, for their advice.
- LOS =
- length of stay •
- UACH =
- university-affiliated children's hospital •
- CHARS =
- Comprehensive Hospital Abstract Reporting System •
- KCCH =
- community hospitals in King County •
- ICD-9 =
- International Classification of Diseases, Ninth Revision
- ↵Relman AS. Are teaching hospitals worth the extra costs? N Engl J Med. 1984;310:1256–1257. Editorial
- Wall TC,
- Fargason CA Jr.,
- ., Johnson VA
- Halfon N,
- Newacheck PW
- ↵Hardiman J. Comparison of Inpatient Charges Across Hospitals in King County for Common Pediatric DRGs. Seattle, WA: University of Washington; 1993. Thesis
- Parrish KM,
- Holt VL,
- Connell FA,
- Williams B,
- LoGerfo JP
- Skobeloff EM,
- Spivey WH,
- St Clair SS,
- Schoffstall JM
- ↵The Sourcebook of Zip Code Demographics. 10th ed. Arlington, VA: CACI Marketing Systems; 1995
- Copyright © 1998 American Academy of Pediatrics