Published online October 1, 2004
PEDIATRICS Vol. 114 No. 4 October 2004, pp. e400-e408 (doi:10.1542/peds.2004-0891)
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow P3Rs: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when P3Rs are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow E-mail this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My File Cabinet
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via CrossRef
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lorch, S. A.
Right arrow Articles by Silber, J. H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Lorch, S. A.
Right arrow Articles by Silber, J. H.
Related Collections
Right arrow Office Practice

ELECTRONIC ARTICLE

Equivalent Lengths of Stay of Pediatric Patients Hospitalized in Rural and Nonrural Hospitals

Scott A. Lorch, MD, MSCE*,{ddagger},§, Xuemei Zhang, MS*, Paul R. Rosenbaum, PhD||, Orit Evan-Shoshan, MS* and Jeffrey H. Silber, MD, PhD*,{ddagger},§,¶

* Center for Outcomes Research
{ddagger} Pediatrics
Anesthesia and Critical Care Medicine, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania
§ Leonard Davis Institute of Health Economics, Philadelphia, Pennsylvania
|| Department of Statistics, the Wharton School, Philadelphia, Pennsylvania


   ABSTRACT
TOP
 ABSTRACT
METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Background. Many children receive their care at local hospitals outside of a large urban area. There may be differences in the length of stay (LOS) between children hospitalized in rural versus urban hospitals. This study compared the differences in LOS, conditional LOS (CLOS), odds of prolonged stay, and 21-day readmission rates for children with 19 medical conditions and 9 surgical procedures admitted to rural, community, and large urban hospitals.

Methods. Discharge records for the hospitalizations of children 1 to 17 years of age were obtained from the New York Department of Public Health Statewide Planning and Research Cooperative System and the Pennsylvania Health Care Cost Containment Council for April 1996 to July 1998. The 19 medical and 9 surgical conditions were identified with the principal condition and procedure codes. Hospitals were classified into 1 of 5 geographic categories on the basis of United States rural-urban continuum codes, ie, large urban, suburban, moderate urban, small urban, or rural. LOS was defined as the period of time between hospital admission and discharge. Readmission rates were calculated for 21 days after discharge from the hospital. A prolonged stay for each condition was defined as any admission lasting beyond the prolongation point, or the day at which the rate of discharge began to decline, as determined with the Hollander-Proschan statistic. This aspect of LOS describes the ability of providers to treat uncomplicated cases of that specific principle diagnosis. CLOS, as a marker for the management of complicated cases, was defined as the LOS beyond the prolongation point. Cox and logistic regression models were developed to describe the geographic effects on the 4 outcome variables, after severity adjustment with 32 demographic and 11 comorbidity variables and adjustment for hospital clustering.

Results. Medical (N = 114 787) and surgical (N = 29 156) admissions to rural hospitals (N = 12 367) had similar outcomes, compared with all geographic categories except the large urban category. Medical patients admitted to rural hospitals had a shorter LOS (12% increase in discharge rate), a shorter CLOS (12% increase in discharge rate), and lower odds of prolonged stay (odds ratio: 0.80), compared with those in large urban hospitals. Surgical patients admitted to rural hospitals had a shorter LOS (12% increase in discharge rate) and lower odds of prolonged stay (odds ratio: 0.81), compared with those in large urban hospitals. For individual conditions, rural hospitals in general had similar or improved LOS, compared with all other hospitals in the 2 states. The addition of hospital-level variables failed to change the results of the primary models.

Conclusions. In their treatment of pediatric hospitalized patients, rural hospitals were not significantly different from hospitals in all geographic regions other than large urban areas. Rural hospitals appear to deliver similar care, compared with nonrural hospitals, for many of the common pediatric conditions included in this study. Additional research is needed to apply these results to other regions or states with different geographic distributions of hospitals and children, in order to determine the overall impact on the regionalization of pediatric care.

Key Words: pediatric hospitalizations • rural hospitals • LOS

Abbreviations: LOS, length(s) of stay • CLOS, conditional length(s) of stay • MSA, Metropolitan Statistical Area • ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification

More than 20% of children in the United States reside in rural communities,1,2 and rural hospitals constitute 16% to 34% of all hospital discharges. Whether these children receive efficient treatment when hospitalized is not known. The adult literature indicates that there may be differences between rural and urban locations in the management of some adult diseases. Keeler et al3 found that, for 5 medical conditions, patients treated at rural hospitals had excess mortality rates and worse quality of care, compared with patients treated at urban hospitals. Similar results were observed for the management of acute myocardial infarction4,5 and depression.6 These differences may center on the amount of technology needed to treat the patient, how quickly knowledge about the disease changes, and how critical specialist management is for appropriate treatment of the patient.3 However, rural hospitals have been shown to give similar quality of care to adults hospitalized for treatment of community-acquired pneumonia7,8 and for general surgical procedures.9

Potential differences between rural and urban hospitals may influence the efficiency of care received by hospitalized children. The efficiency of a good or service is defined as the resources needed to obtain a given level of benefit from a good or service. The efficiency point is the amount of resources allocated to that service at which the marginal benefits of the service equal the marginal costs. At resource levels lower than the efficiency point, a good or service can produce benefits that are worth more than the additional costs needed to obtain those benefits. At resource levels higher than the efficiency point, the benefits produced, although still positive, are worth less than the additional allocated resources. In heath care, more expensive care may still be "more efficient" because the improvement in patient outcomes outweighs the added costs; in other cases, the most efficient care is the least expensive, although additional expenditures lead to improvements in some patients’ health.10 How efficient rural hospitals are in caring for pediatric patients may influence the health policies of a state or region. If care is as efficient at rural hospitals, then parental preferences and decreased travel costs may justify duplicating at rural hospitals services that larger regional centers also provide. However, there are other conditions for which the improvements of outcomes with higher patient volumes necessitate more regionalization of care, irrespective of nonmedical costs and parental preferences.

There are difficulties in applying this definition of efficiency to clinical medicine. Costs are difficult to quantify; many studies rely on hospital charges or payments to the hospital, which do not represent the true costs to either the hospital or society.11 Although accurately measuring the costs to a third-party payer, payments made to hospitals underestimate the true costs of care to the hospital, which may provide uncompensated care to children without insurance, and to society, which may divert resources away from other services to health care to provide uncompensated care to its citizens.11 Pediatric care also incurs large nonmedical costs, in the form of lost parental wages and travel time.

In lieu of detailed resource utilization by patients,11 studies have examined hospital length(s) of stay (LOS). LOS does not account for geographic variations in hospitalization rates12,13 but does measure another critical step in the treatment of an ill child, ie, the time a child required care in the hospital. LOS is also an effective tool for measurement of hospital performance for conditions for which geographic variations in hospitalization rates do not occur. LOS was 1 outcome measure used in a recent study comparing the delivery of care to pediatric asthma patients in rural and urban hospitals.14

This study was designed to compare differences in the LOS of children hospitalized at rural and nonrural hospitals in New York and Pennsylvania for 19 medical conditions and 9 surgical procedures. Each hospitalization was analyzed by using 4 outcome measures that describe different aspects of the hospital stay. We hypothesized that differences would be found in the care given to children at hospitals in different geographic regions.


   METHODS
TOP
 ABSTRACT
 METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Data Source and Patient Population
We obtained discharge records for all pediatric hospital admissions (patients 1–17 years of age) in Pennsylvania for the period from January 1, 1996, to December 31, 1998, and in New York State for the period from January 1, 1996, to September 30, 1998. Patients admitted to psychiatric or non-acute care hospitals were not included in this study. Both New York and Pennsylvania offer a diverse mixture of large urban hospitals (in cities such as Philadelphia, New York City, and Buffalo), hospitals in moderately large urban areas (such as Syracuse, Albany, and Harrisburg), and many rural hospitals. Data from Pennsylvania were provided through the Pennsylvania Health Care Cost Containment Council. Data from New York were provided through the New York Department of Public Health Statewide Planning and Research Cooperative System data set. Both data sets included hospital admissions linked by patient. For the purposes of this study, we included only admissions between April 1, 1996, and June 30, 1998, to allow calculation of prior hospitalization and readmission rates. The data sets included standard UB92 data elements, with procedures and diagnoses coded with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) nomenclature. Nineteen medical conditions and 9 surgical procedures were identified from the principal procedure and diagnosis codes. We chose those conditions to represent a common set of disease processes for which children are admitted to a variety of hospitals, including both large children’s hospitals and community hospitals. Table 1 shows the medical conditions and surgical procedures included in this study and the ICD-9-CM codes used to identify each diagnosis.


View this table:
[in this window]
[in a new window]
  		TABLE 1. Conditions and Procedures Included in Analysis (N = 143 943)

 
Statistical Methods
Geographic Classification of Hospitals
Each hospital included in the study was categorized into 1 of 5 geographic categories on the basis of United States rural-urban continuum codes15 for the county in which the hospital was located. These categories were based on prior work that compared rural hospitals with not only large urban centers but also community hospitals, to which they may be more similar.16 Hospitals located in the central county of a Metropolitan Statistical Area (MSA) with >1 000 000 people were classified as large urban hospitals. Hospitals located in suburban or ring counties of a MSA with >1 000 000 people were classified as suburban. Moderate urban hospitals were located in MSAs with 250 000 to 1 000 000 people. Small urban hospitals were located in MSAs with <250 000 people. Rural hospitals were hospitals located in counties outside of a MSA.

Outcome Definitions and Modeling
This study analyzed each hospitalization with 4 outcomes, measured as defined in previous work,17,18 ie, (1) LOS (in days), (2) readmission within 21 days after discharge, (3) odds of a prolonged stay, and (4) conditional LOS (CLOS) (in days). LOS was calculated as the number of days between hospital admission and discharge; it provides insight into the resources used by a provider and hospital. Rates of readmission within 21 days were used to assess potentially inadequate care during the hospitalization. Higher readmission rates could also represent inadequate outpatient care associated with the local health care system.

Earlier work17,18 suggested that a hospital stay should be divided into 2 periods, on the basis of the daily rate of discharge for that condition. The daily rate of discharge is the proportion of patients in the hospital at the start of the day who were discharged before the end of the day. We calculated the daily rate of discharge separately for each principal diagnosis, because the dynamics of patient discharge are different for each disease category. The daily rate of discharge typically increases during the first few days of hospitalization and then decreases rapidly (Fig 1). This suggests that 2 distinct processes are at work, ie, routine treatment of patients with uncomplicated conditions who are discharged promptly and treatment of patients with special complications that delay discharge. Prior work showed that patients who remain hospitalized into the period with declining discharge rates are very likely to have complications, often serious ones.17,18 For 3413 appendectomy and 7777 pneumonia admissions, staying beyond the prolongation point, or the day at which the discharge rate begins to decline, had a sensitivity of 88% and a specificity of 81% for identifying a complication associated with an admission for an appendectomy and a sensitivity of 89% and a specificity of 75% for identifying a complication associated with a hospitalization for pneumonia treatment.17 We defined the prolongation day with a statistical test described by Hollander and Proschan.19,20 Let T be a time to discharge measured in days and consider its distribution. We define the technical concept of a "discharge is extended" if Prob(T > t) < Prob(T > t + b|T > b) for all t > 0 and b > 0. This equation indicates that a patient who has already stayed b days is more likely to stay t more days than a randomly selected patient is to stay t days. For example, if t = 7 and b = 14 and the discharge is extended, then a patient who has already been in the hospital for 2 weeks (b = 14) is more likely to stay an additional 1 week (t = 7) than a normal patient is to stay 1 week (t = 7) originally. If discharge is extended, then the longer the patient has been in the hospital, the worse the prospects of discharge become and the greater the chance the patient will remain hospitalized. For historical reasons related to engineering, Hollander and Wolfe20 say that a distribution is not "extended" but rather "new-worse-than-used"; however, "extended" conveys the sense of the term better when speaking of hospital care. Their nonparametric statistical test yields significant P values for distributions that are clearly extended. Discharge is not extended for any patient on the day of admission; however, among patients who have stayed beyond a certain initial period, the remaining time to discharge is extended. We used the Hollander-Proschan test to identify the first day such that stays beyond that day are unambiguously extended (P < .05). That hospital day was defined as the prolongation point for that principle diagnosis.


Figure 1
View larger version (12K):
[in this window]
[in a new window]
  		Fig 1. Development of the prolongation point outcome measure from hospital LOS data.

 
The prolongation point was used in the last 2 outcome measures of this study. The first outcome measure is the probability of having a stay beyond the prolongation point, or a prolonged stay. It describes the ability of providers to treat and discharge patients with less complicated conditions,17,18 with higher odds being associated with less efficient care. The second outcome measure is CLOS, which was defined as the number of days a patient remained hospitalized beyond the prolongation point. CLOS has been shown to measure the provider’s ability to manage complicated cases,17,18 with a longer CLOS being associated with less efficient care. The CLOS statistic also allows assessment of the outcome of a hospitalization when inadequate risk adjustment is possible. Data from prior work17,18 showed that patients who stay beyond the prolongation point are much more likely to have a comorbid medical condition or a prior hospitalization within 3 months of discharge and thus represent a more homogeneous, if more complex, population of patients, compared with all admissions to that hospital.

We developed the models for each outcome measure by splitting the sample in half randomly, building the model with 1 half and validating it with the other half. The development and validation halves were combined and final estimates were obtained as described in prior work.18 Cox regression models were developed to estimate geographic effects on LOS and CLOS. Because the Cox model was developed for survival data, its parameters are usually called hazard rates or ratios of hazard rates. For clarity, we defined these same technical quantities as discharge rates and discharge ratios. Longer LOS and CLOS have lower discharge rates and discharge ratios of <1. Logit models were developed to estimate the geographic effects on prolonged stay and readmissions. Odds ratios were reported for these models. Ratios of >1 denote higher odds of a prolonged stay or readmission, compared with the comparison group. We assigned the LOS of children transferred to another hospital for care to the initial admitting hospital. We coded the LOS of the few children who died as the longest stay in the data set. Because the Cox model is semiparametric, only the rank order of the LOS is important (not the actual values). This technique does not allow hospitals to benefit from shorter LOS resulting from the death of the patient. Deaths were coded as prolonged stays regardless of the day the death occurred.

Adjustment for Hospital Clustering
It was important to control for clustering at the hospital level in the statistical analysis. Because of similarities in practice styles and the organization of the hospital, children treated at the same hospital are more likely to receive similar care than are children treated at different hospitals. Without adequate control for these similarities, SEs may be erroneous, depending on correlations between the outcomes for patients in each hospital cluster. This study used robust variance estimations for cluster-correlated data, with unconditioned models for each of the 4 outcome measures.21,22 These techniques control for the clustering of children within a given hospital and modify the SEs obtained with the logistic and Cox regression analyses.

Severity Adjustment
Initial severity adjustment used age, gender, race, insurance status, emergency admission, median income (based on patient zip code),23 distance from the hospital (as a proxy measure of complexity of care),24 and previous hospitalization rates. Missing values were coded as missing or unknown in each model. A state dummy variable was included in the model to account for differences in hospital discharge rates between Pennsylvania and New York. Only hospitals that treated at least 20 medical or surgical patients were included in the model. We accounted for the presence of diabetes mellitus, sickle cell disease, cerebral palsy, seizures, congenital heart disease, acquired immunodeficiency syndrome, chromosomal anomalies, spina bifida, cystic fibrosis, or congenital muscle, nervous system, or respiratory system anomalies by using ICD-9-CM codes in the secondary diagnosis list.18 Indicator variables for each condition or procedure were included to account for disease-specific differences in each outcome measure.


   RESULTS
TOP
 ABSTRACT
 METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
We identified 114 747 medical admissions and 29 156 surgical admissions during the 27 months of the study. A total of 323 hospitals met the inclusion criteria, including 55 hospitals in rural regions, 109 in large urban regions, 76 in suburban regions, 66 in moderate urban regions, and 17 in small urban regions. Table 1 shows the number of children admitted with each of the 19 medical conditions and 9 surgical procedures, according to geographic category. Children with asthma made up 33.4% of the studied medical admissions, and appendectomies made up 50.2% of the studied surgical procedures. We did not find differences in the patient case mixture across geographic categories, with the exception of specialized surgical procedures, such as back surgery and tonsillectomy and adenoidectomy, which were more commonly performed at large urban hospitals.

Table 2 presents the characteristics of children admitted with a medical condition, according to geographic category. Because of the large sample size, any difference between rural hospitals and other geographic categories was statistically significant unless otherwise noted. Children admitted to rural hospitals had similar characteristics, compared with children admitted to suburban, moderate urban, and small urban hospitals. Children admitted to large urban hospitals were more likely to be younger, black, and on public insurance. The percentage of children with any comorbid condition was slightly higher in large urban hospitals than in other areas (6.74%, compared with 4.35-5.60% in the other geographic categories). Table 3 shows the same data for children admitted to hospitals in these geographic categories for a surgical procedure. Children admitted to rural hospitals for a surgical procedure were similar to children admitted to non-large urban hospitals. Children admitted to large urban hospitals for a surgical procedure were more likely to be younger, black, and on public insurance.


View this table:
[in this window]
[in a new window]
  		TABLE 2. Patient Characteristics and Medical Conditions (N = 114 787)

 

View this table:
[in this window]
[in a new window]
  		TABLE 3. Patient Characteristics and Surgical Procedures (N = 29 156)

 
Tables 2 and 3 also report the overall LOS, CLOS, and proportions of children who had prolonged stays or readmissions within 21 days after discharge, stratified according to the 5 geographic categories. These measures were not adjusted for case severity. Rural hospitals performed as well as hospitals in the non-large urban categories in all 4 outcome measures, with average LOS ranging from 2.2 to 2.39 days for medical conditions and from 2.69 to 2.93 days for surgical procedures. For both medical and surgical admissions, large urban hospitals had longer LOS and CLOS and greater proportions of children with prolonged stays or readmissions, compared with the other 4 geographic categories.

With rural hospitals as the reference group, Table 4 shows the outcomes of children treated for medical conditions at hospitals in each of the geographic categories, as defined with rural-urban continuum codes. After adjustment for case severity and demographic characteristics, rural hospitals had similar LOS, CLOS, odds of a prolonged stay, and 21-day readmission rates, compared with hospitals in all regions except for large urban MSAs. Medical admissions to large urban hospitals continued to have a longer LOS (11% decrease in discharge rate, P < .05), longer CLOS (12% decrease in discharge rate, P < .05), and higher odds of a prolonged stay (odds ratio: 1.26; P < .05), compared with rural hospitals. These findings suggest that the longer LOS seen at large urban hospitals occur both among patients with uncomplicated conditions, who may be discharged quickly, and among patients with complicated conditions, who remain in the hospital for longer periods of time. We found similar results for children admitted for surgical procedures (Table 5). Large urban hospitals had a longer LOS (10% decrease in discharge rate, P < .05) and higher odds of a prolonged stay (odds ratio: 1.23; P < .05), compared with rural hospitals.


View this table:
[in this window]
[in a new window]
  		TABLE 4. Outcomes of Medical Conditions According to Site of Treatment

 

View this table:
[in this window]
[in a new window]
  		TABLE 5. Outcomes of Surgical Procedures According to Site of Treatment

 
Table 6 shows the results of the 4 outcome models for each of the 19 medical and 9 surgical conditions. Unlike the combined models, no interaction terms were added to these models, because fewer children were admitted with each condition or procedure. We also grouped all urban hospitals into an urban category and compared this group with rural hospitals. For most of the medical conditions, rural hospitals had similar LOS, compared with nonrural hospitals. The only statistically significant results found longer LOS, longer CLOS, and higher odds of a prolonged stay in nonrural hospitals. Surgical procedures showed more variation in all 4 outcome measures. This variation is likely the result of several conditions, such as cleft palate repair and thyroid surgery, with very few treated patients in non-large urban hospitals (Table 1). Although the lack of interaction terms may reduce the strength of risk adjustment in these models, Table 6 suggests that the results of the grouped medical (Table 4) and surgical (Table 5) models are valid for a number of common pediatric conditions.


View this table:
[in this window]
[in a new window]
  		TABLE 6. LOS for 19 Medical and 9 Surgical Admission Types in Urban Hospitals

 
Because rural hospitals in Pennsylvania and New York are heterogeneous, we next studied whether the addition of hospital-level variables would affect the results of the study. The hospital-level variables included the number of pediatric beds reported to the state health boards, the presence of a pediatric or neonatal intensive care unit, trauma center designation, the total number of beds (both adult and pediatric), and the average yearly number of pediatric admissions. The addition of these variables failed to change the results of any model by >5%. We performed additional analyses to assess the robustness of the primary results. Adding variables for the hospital’s teaching status, being a children’s hospital, or the home zip code of the child failed to change the results of any model.


   DISCUSSION
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
Between 16% and 34% of all pediatric admissions occur at rural institutions.1,2 Rural hospitals delivered care to these children at least as well as did their nonrural counterparts for the 19 medical and 9 surgical conditions identified in this study. With adjustment for demographic variables and comorbid conditions, rural hospitals may deliver slightly more efficient care than large urban hospitals, as measured with the LOS, CLOS, and odds of a prolonged stay. Our study supports the findings of previous work on asthma,14 which found that rural centers had equal LOS and lower charges, compared with urban hospitals. The performance of rural centers is explained additionally by dividing nonrural centers into geographic categories; other regions besides large urban areas had similar LOS, compared with rural areas.

Why have studies of differences in rural and urban hospital care found such differing results? The studies that found inequities in care focused primarily on conditions requiring rapid assessment and treatment4,5 or complex technical and subspecialty skills.3,6 The patients in this study were more similar to those in studies on pediatric asthma14 and community-acquired pneumonia.7,8 These conditions are frequently treated without referrals to large specialty centers and represent typical cases for rural pediatricians or general practitioners. For this study, we assigned the LOS of children transferred to another hospital for their care to the initial admitting hospital. This assignment limited any bias associated with the transfer of patients with complicated conditions to large pediatric referral centers, allowing assessment of rural hospitals within the larger health care system of the state or region.

Care for hospitalized patients does not end when the patient is discharged from the hospital. Adequate outpatient follow-up monitoring and management and provision of a medical home through the consistent use of a primary care physician or medical clinic are critical for preventing hospital readmissions and may influence LOS. In this study, large urban hospitals had more patients who were uninsured. Uninsured children have less access to primary care and outpatient services, both before admission and after discharge from the hospital.25 Prior studies, however, found no differences between uninsured and insured children in hospitalization rates,26–28 LOS,26–28 or the quality of asthma management.29 Another difference could be in the availability of a medical home for children in large urban areas. Ryan et al30 found that adolescents without a medical home had higher emergency department use. Other work25,26,31 has shown decreased access to care and more hospitalizations for chronic conditions when a medical home is not present. Although we controlled for insurance status in our regression models, differences in access to care may partly explain the performance of large urban hospitals, compared with rural centers.

Risk adjustment is an important element of any observational study that compares the treatment of patients in different settings. Methods to adjust for the severity of illness have not been well established in pediatric medicine, compared with adult medicine,32,33 especially for large administrative data sets. The CLOS outcome measure is less sensitive to the problem of inadequate risk adjustment, compared with other outcome measures. CLOS focuses on the efficiency of care after the hospitalization has passed a defined prolongation point. Prior work17,18 showed that these cases are more complicated and thus represent a more homogeneous group of children. If the findings in our study had been the result of rural hospitals caring only for the least complicated cases and if rural hospitals were in fact treating children less effectively, then we would have expected the CLOS for rural hospitals to be longer than those for other geographic categories. Instead, we found that rural hospitals had the same or shorter CLOS, compared with other geographic categories including large urban areas, which contain large pediatric referral centers. These data strengthen our findings that rural hospitals manage many pediatric cases at least as well as do hospitals in other geographic categories.

The policy implications of this research are less clear. Regionalization policies must balance the advantages of centralized care of specialty cases, for which increased volume and the elimination of duplicated services lead to increased efficiency, with increased travel costs and diminished access to care. Studies of neonatal intensive care,34 pediatric intensive care,35 and cardiac surgery36 all suggest that regionalization of highly technical care leads to improved mortality rates at somewhat lower costs. Our study suggests that rural hospitals are capable of serving as the gatekeepers for more common pediatric conditions that lead to hospitalization. One factor in this study was the choice to assign patients to the hospital to which they were initially admitted. The LOS for children who were transferred to a second or third hospital were assigned to the first hospital. This algorithm was used to prevent hospitals from appearing better by transferring all patients with more serious or complicated conditions to a tertiary referral center. Rural and community hospitals were evaluated not only on their direct treatment of children but also on how well they identified children who required more-specialized care at a regional center. Because of this element of the study design, our results indicate that the current level of regionalization in Pennsylvania and New York provides appropriate care to rural children. However, additional efforts to deregionalize pediatric care must assess the ability of smaller hospitals to care for sicker patients whom they currently transfer to more-specialized centers. Such patients may fare more poorly than children currently cared for only at a rural hospital. Also, parental satisfaction with the location and quality of care and the added costs resulting from duplication of specialty services must be addressed by policy makers within a region or state.

We chose to categorize patients according to the location of the hospital rather than the home address of the patient. Many children living in rural communities may receive their care in a more-specialized urban center; such children would not be categorized as rural in this study. The focus of this study was on the hospitals, not individual patients, and use of the home zip codes of the patients to determine rural or urban status in each model failed to change the results. Individual cases were identified with ICD-9-CM codes found in the principal procedure or condition field. Systematic mistakes in this coding could affect the results of this study. Also, without detailed chart abstraction for all 143 943 patients, unmeasured severity or socioeconomic status could be another reason why large urban hospitals had slightly longer LOS, compared with other geographic categories. Although New York and Pennsylvania have a diverse mixture of hospitals in rural and urban areas, the results from these 2 states may not be generalizable to states dominated by 1 large tertiary referral center or with rural areas that are farther away from large urban areas. Finally, large urban MSAs at the county level represent a broad categorization of what constitutes an urban area. Many counties coded as urban have areas of rural land that are misclassified with the MSA designation. For the purposes of this study, however, the overall location of a hospital is representative of the children being admitted for treatment.

Rural hospitals appear to provide competent care for children. This study found no differences in the LOS, CLOS, odds of a prolonged stay, or 21-day readmission rates for children treated at rural versus nonrural hospitals. Many childhood hospitalizations are for treatment of common ailments that can be managed at least initially at the local level. Rural hospitals can serve as the gatekeepers to the health care system in states with a diverse mixture of rural and urban hospitals. Additional research is needed to apply these results to other regions or states with different geographic distributions of hospitals and children to determine the impact of these findings on the regionalization of pediatric care.


   ACKNOWLEDGMENTS
 
This work was supported in part by Agency for Healthcare Research and Quality grants R01 HS09983 (J.H.S.) and T32 HD07740-06 (S.A.L.) and a grant from the National Science Foundation (P.R.R.).


   FOOTNOTES
 
Accepted Apr 27, 2004.

Reprint requests to (S.A.L.) Center for Outcomes Research, 3535 Market St, Suite 1029, Philadelphia, PA 19104. E-mail: lorch{at}email.chop.edu


   REFERENCES
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. McManus MA, Newacheck PW. Rural maternal, child, and adolescent health. Health Serv Res. 1989;23 :807 –848[ISI][Medline]
  2. Melzer SM, Grossman DC, Hart LG, Rosenblatt RA. Hospital services for rural children in Washington State. Pediatrics. 1997;99 :196 –203[Abstract/Free Full Text]
  3. Keeler EB, Rubenstein LV, Kahn KL, et al. Hospital characteristics and quality of care. JAMA. 1992;268 :1709 –1714[Abstract]
  4. Thiemann DR, Coresh J, Oetgen WJ, Powe NR. The association between hospital volume and survival after acute myocardial infarction in elderly patients. N Engl J Med. 1999;340 :1640 –1648[Abstract/Free Full Text]
  5. Sheikh K, Bullock C. Urban-rural differences in the quality of care for Medicare patients with acute myocardial infarction. Arch Intern Med. 2001;161 :737 –743[Abstract/Free Full Text]
  6. Rost K, Zhang M, Fortney J, Smith J, Smith GR Jr. Rural-urban differences in depression treatment and suicidality. Med Care. 1998;36 :1098 –1107[CrossRef][ISI][Medline]
  7. Lave JR, Fine MJ, Sankey SS, et al. Hospitalized pneumonia: outcomes, treatment patterns, and costs in urban and rural areas. J Gen Intern Med. 1996;11 :415 –421[ISI][Medline]
  8. Whittle J, Lin CJ, Lave JR, et al. Relationship of provider characteristics to outcomes, process, and costs of care for community-acquired pneumonia. Med Care. 1998;36 :977 –987[CrossRef][ISI][Medline]
  9. Welch HG, Larson EH, Hart LG, Rosenblatt RA. Readmission after surgery in Washington State rural hospitals. Am J Public Health. 1992;82 :407 –411[Abstract/Free Full Text]
  10. Folland S, Goodman A, Stano M. The Economics of Health and Health Care. Upper Saddle River, NJ: Prentice Hall; 2001
  11. Gold M, Siegel J, Russell L, Weinstein M. Cost-Effectiveness in Health and Medicine. New York, NY: Oxford University Press; 1996
  12. Perrin JM, Homer CJ, Berwick DM, et al. Variations in rates of hospitalization of children in three urban communities. N Engl J Med. 1989;320 :1183 –1187[Abstract]
  13. Goodman DC, Fisher ES, Gittelsohn A, Chang CH, Fleming C. Why are children hospitalized? The role of non-clinical factors in pediatric hospitalizations. Pediatrics. 1994;93 :896 –902[Abstract/Free Full Text]
  14. Meurer JR, Kuhn EM, George V, Yauck JS, Layde PM. Charges for childhood asthma by hospital characteristics. Pediatrics. 1998;102 (6). Available at: www.pediatrics.org/cgi/content/full/102/6/e70
  15. Economic Research Service. Rural-urban continuum codes. Available at: www.ers.udsa.gov/briefing/rurality/RuralUrbCon/code93.txt
  16. Cromwell J, Mitchell JB, Calore KA, Iezzoni L. Sources of hospital cost variation by urban-rural location. Med Care. 1987;25 :801 –829[ISI][Medline]
  17. Silber JH, Rosenbaum PR, Koziol LF, et al. Conditional length of stay. Health Serv Res. 1999;34 :349 –363[ISI][Medline]
  18. Silber JH, Rosenbaum PR, Even-Shoshan O, et al. Length of stay, conditional length of stay, and prolonged stay in pediatric asthma. Health Serv Res. 2003;38 :867 –886[CrossRef][ISI][Medline]
  19. Hollander M, Proschan F. Testing whether new is better than used. Ann Math Stat. 1972;43 :1136 –1146
  20. Hollander M, Wolfe DA. Nonparametric Statistical Methods. 2nd ed. New York, NY: Wiley; 1999
  21. Localio AR, Berlin JA, Ten Have TR, Kimmel SE. Adjustments for center in multicenter studies: an overview. Ann Intern Med. 2001;135 :112 –123[Abstract/Free Full Text]
  22. Williams RL. A note on robust variance estimation for cluster-correlated data. Biometrics. 2000;56 :645 –646[CrossRef][ISI][Medline]
  23. Goodman DC, Stukel TA, Chang CH. Trends in pediatric asthma hospitalization rates: regional and socioeconomic differences. Pediatrics. 1998;101 :208 –213[Abstract/Free Full Text]
  24. Luft HS, Garnick DW, Mark DH, et al. Does quality influence choice of hospital? JAMA. 1990;263 :2899 –2906[Abstract]
  25. Andersen RM, Yu H, Wyn R, et al. Access to medical care for low-income persons: how do communities make a difference? Med Care Res Rev. 2002;59 :384 –411[Abstract/Free Full Text]
  26. Ferris TG, Crain EF, Oken E, et al. Insurance and quality of care for children with acute asthma. Ambul Pediatr. 2001;1 :267 –274[CrossRef][ISI][Medline]
  27. Wood PR, Smith LA, Romero D, et al. Relationships between welfare status, health insurance status, and health and medical care among children with asthma. Am J Public Health. 2002;92 :1446 –1452[Abstract/Free Full Text]
  28. Eisert S, Gabow P. Effect of Child Health Insurance Plan enrollment on the utilization of health care services by children using a public safety net system. Pediatrics. 2002;110 :940 –945[Abstract/Free Full Text]
  29. Finkelstein JA, Brown RW, Schneider LC, et al. Quality of care for preschool children with asthma: the role of social factors and practice setting. Pediatrics. 1995;95 :389 –394[Abstract/Free Full Text]
  30. Ryan S, Riley A, Kang M, Starfield B. The effects of regular source of care and health need on medical care use among rural adolescents. Arch Pediatr Adolesc Med. 2001;155 :184 –190[Abstract/Free Full Text]
  31. Falik M, Needleman J, Wells BL, Korb J. Ambulatory care sensitive hospitalizations and emergency visits: experiences of Medicaid patients using federally qualified health centers. Med Care. 2001;39 :551 –561[CrossRef][ISI][Medline]
  32. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40 :373 –383[CrossRef][ISI][Medline]
  33. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45 :613 –619[CrossRef][ISI][Medline]
  34. Richardson DK, Reed K, Cutler JC, et al. Perinatal regionalization versus hospital competition: the Hartford example. Pediatrics. 1995;96 :417 –423[Abstract/Free Full Text]
  35. Tilford JM, Simpson PM, Green JW, Lensing S, Fiser DH. Volume-outcome relationships in pediatric intensive care units. Pediatrics. 2000;106 :289 –294[Abstract/Free Full Text]
  36. Chang RK, Klitzner TS. Can regionalization decrease the number of deaths for children who undergo cardiac surgery? A theoretical analysis. Pediatrics. 2002;109 :173 –181[Abstract/Free Full Text]
PEDIATRICS (ISSN 1098-4275). ©2004 by the American Academy of Pediatrics




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow P3Rs: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when P3Rs are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow E-mail this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My File Cabinet
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via CrossRef
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lorch, S. A.
Right arrow Articles by Silber, J. H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Lorch, S. A.
Right arrow Articles by Silber, J. H.
Related Collections
Right arrow Office Practice