Pediatric Annals

Variations in Pediatric Hospitalization Rates: Why Do They Occur?

James M Perrin, MD

Abstract

Numerous studies have demonstrated significant variations in rates of hospitalization and the use of other technologies in the management of many health conditions.1 Hospitalization rates for many adult health conditions vary as does the use of certain surgical procedures, such as coronary artery bypass or hysterectomies. Explanations proposed for these observed differences include variations in access to services (eg, primary care and surgical care) based on geography, socioeconomic factors, or insurance; in the epidemiology of illness in different areas; in community resources (eg, excess hospital beds); in physician behavior; and in child and family health behaviors. Some variations may be simply random, although the sizable differences observed for many health-care phenomena make random variation an unlikely explanation. For example, children with insurance (especially private insurance) undergo some surgical procedures such as tonsillectomies and adenoidectomies more frequently than children without insurance, although poorer children more frequently undergo other procedures, such as shunt revisions, often associated with significant prematurity.2

Work by Wennberg et al3 demonstrated significant variations between Boston and New Haven in frequency of hospitalizations for adults. We used hospital discharge data from 1982 to examine whether similar variations existed for hospitalization for children, and we added Rochester, New York to the communities analyzed.4 These studies documented extensive variations in hospitalization rates for children in these urban communities, all characterized by having a broad array of child health services, including major teaching hospitals. Insofar as the sample was limited to children residing in these three communities, different rates of referral from more distant areas do not explain these results.

The most striking findings from the 1982 studies were that for all medical conditions examined, children residing in Boston were hospitalized 2lf2 to 6 times as frequently as children in Rochester, with New Haven intermediate for almost every condition (Table 1), For medical conditions examined, Rochester children never experienced higher rates of hospitalization. The story for surgical conditions is more mixed, with substantially more variation. Rochester children had higher reported rates of hospitalization for tonsillectomies and adenoidectO' mies, although different methods of counting and use of outpatient surgery in the three cities in 1982 complicate the interpretation of these findings. The 1982 studies identified a few conditions with only minimal variation in hospitalization rates (Table 2). The three conditions with very tow (and statistically insignificant) variation were appendicitis, bacterial meningitis, and femoral fractures. These three conditions provide physicians little discretion with respect to admission. Essentially, all cases are currently treated in hospitals. Of interest are the findings regarding aseptic meningitis, again indicating significant variation in rates across communities.

Table

Different Rates of Illness

Boston children may experience significantly higher rates of severe illness and thus require hospital ization more frequently than children in New Haven or Rochester. For example, rates of asthma might be higher in an older city such as Boston, with high rates of air pollution and infestation with potential allergens such as cockroaches.7'8 If Boston children have more severe illness, then one would predict that children admitted in Boston would have similar or worse levels of admission severity in comparison with Rochester and New Haven children. Recent data, however, indicate that Boston children are consistently less severely ill at the point of admission for all conditions studied than are children in New Haven and Rochester.

For example, Boston children with asthma evidence substantially higher oxygen saturation rates than do children in New Haven or Rochester.9 In general, when analyses control for socioeconomic status, rates for children in New Haven approach those of children, in Rochester, and those for Boston children become even more extreme. Thus, higher rates of severe…

Numerous studies have demonstrated significant variations in rates of hospitalization and the use of other technologies in the management of many health conditions.1 Hospitalization rates for many adult health conditions vary as does the use of certain surgical procedures, such as coronary artery bypass or hysterectomies. Explanations proposed for these observed differences include variations in access to services (eg, primary care and surgical care) based on geography, socioeconomic factors, or insurance; in the epidemiology of illness in different areas; in community resources (eg, excess hospital beds); in physician behavior; and in child and family health behaviors. Some variations may be simply random, although the sizable differences observed for many health-care phenomena make random variation an unlikely explanation. For example, children with insurance (especially private insurance) undergo some surgical procedures such as tonsillectomies and adenoidectomies more frequently than children without insurance, although poorer children more frequently undergo other procedures, such as shunt revisions, often associated with significant prematurity.2

Work by Wennberg et al3 demonstrated significant variations between Boston and New Haven in frequency of hospitalizations for adults. We used hospital discharge data from 1982 to examine whether similar variations existed for hospitalization for children, and we added Rochester, New York to the communities analyzed.4 These studies documented extensive variations in hospitalization rates for children in these urban communities, all characterized by having a broad array of child health services, including major teaching hospitals. Insofar as the sample was limited to children residing in these three communities, different rates of referral from more distant areas do not explain these results.

The most striking findings from the 1982 studies were that for all medical conditions examined, children residing in Boston were hospitalized 2lf2 to 6 times as frequently as children in Rochester, with New Haven intermediate for almost every condition (Table 1), For medical conditions examined, Rochester children never experienced higher rates of hospitalization. The story for surgical conditions is more mixed, with substantially more variation. Rochester children had higher reported rates of hospitalization for tonsillectomies and adenoidectO' mies, although different methods of counting and use of outpatient surgery in the three cities in 1982 complicate the interpretation of these findings. The 1982 studies identified a few conditions with only minimal variation in hospitalization rates (Table 2). The three conditions with very tow (and statistically insignificant) variation were appendicitis, bacterial meningitis, and femoral fractures. These three conditions provide physicians little discretion with respect to admission. Essentially, all cases are currently treated in hospitals. Of interest are the findings regarding aseptic meningitis, again indicating significant variation in rates across communities.

Table

TABLE 1Discharge Rates for Children Ages O- 1 4 Years for Selected Medical Conditions, 1 982

TABLE 1

Discharge Rates for Children Ages O- 1 4 Years for Selected Medical Conditions, 1 982

These earlier studies lacked satisfactory soctoeconomic data on the children admitted. Available data, based on hospital discharge information, did not even provide expected source of payment, which could have been used as a proxy for socioeconomic status. Rochester, New Haven, and Boston vary substantially in economic status (Table 3), as well as in health resources available in the communities, with Rochester having fewer hospital beds per population. Several studies document higher rates of hospitalization among poor children,5·6 and the 1982 findings could mainly reflect differences in socioeconomic status among the three communities. On the other hand, the data in Table 3 indicate that New Haven and Boston have similar socioeconomic status, although their hospitalization rates remain quite divergent.

THE CAUSES OF VARIATION

The general findings in Table 1 are remarkably consistent across medical conditions. What might explain these variations? Lesions at several points in the process of health care and hospital use could be responsible. Based on the findings from the 1982 data, we began a series of investigations using hospital medical record data rather than discharge abstract data. Medical records provide substantially more information about the clinical status of children before and during hospitalization along with more detail on socioeconomic status and insurance coverage. We examined five major condition groups in these studies: asthma, abdominal pain (including appendicitis), bacterial and nonbacterial meningitis, toxic ingestion, and head injury. Thus, as before, conditions include ones both with and without substantial physician discretion regarding use of hospitalization. The sections that follow describe possible causes for variation. Insofar as study analyses are incomplete, most sections speculate on causes, although some primary data are included.

Poverty and Hospitalization

Several studies have indicated that poor children are hospitalized more frequently than children from higher income families. Insofar as Rochester is a relatively wealthy community, its lower rates of hospitalization could reflect mainly socioeconomic status differences. Current data do support the notion that poor children have higher rates of hospitalization. In all three communities, children living in census tracts with lower mean incomes had higher rates of hospitalization than those in higher income tracts. However; variations among communities were much higher for higher income areas than for poorer areas. Poorer children in Boston were hospitalized only about 1.5 times as frequently as those in Rochester, while middle -income children in Boston faced rates 3 to 4 times those of equivalent Rochester children.

Table

TABLE 2Conditions With Minimal Variation In Hospitalizaron Rates

TABLE 2

Conditions With Minimal Variation In Hospitalizaron Rates

Table

TABLE 3Demographic Characteristics of Boston, New Haven, and Rochester, US Census, 1 990

TABLE 3

Demographic Characteristics of Boston, New Haven, and Rochester, US Census, 1 990

Different Rates of Illness

Boston children may experience significantly higher rates of severe illness and thus require hospital ization more frequently than children in New Haven or Rochester. For example, rates of asthma might be higher in an older city such as Boston, with high rates of air pollution and infestation with potential allergens such as cockroaches.7'8 If Boston children have more severe illness, then one would predict that children admitted in Boston would have similar or worse levels of admission severity in comparison with Rochester and New Haven children. Recent data, however, indicate that Boston children are consistently less severely ill at the point of admission for all conditions studied than are children in New Haven and Rochester.

For example, Boston children with asthma evidence substantially higher oxygen saturation rates than do children in New Haven or Rochester.9 In general, when analyses control for socioeconomic status, rates for children in New Haven approach those of children, in Rochester, and those for Boston children become even more extreme. Thus, higher rates of severe illness among Boston children do not appear to explain different hospital rates.

Emergency Department Utilization

Different rates of use of emergency departments could partly explain the variations in hospitalization rates. Emergency department physicians could have target rates of admission (ie, percentages of children in the emergency department), along with differential thresholds of clinical severity. If Boston children use emergency rooms significantly more frequently for acute illnesses, then they might experience higher rates of admission. Current study data provide no information on whether similar or higher numbers of Boston children use emergency departments for acute illness. On the other hand, the data do indicate differences in the use of primary care in the three cities.

Primary Care Access and Utilization

Differential access to primary care or how families use primary care services also could explain some variations. Proponents of primary care suggest that adequate services keep children healthier and diminish their need for hospitalization. Adequate community primary care provides satisfactory alternatives to hospitalization for many clinical problems. If children in Rochester have more access to primary care or more regular sources of primary care, their parents may have less cause to depend on, emergency departments for routine care. Emergency department physicians in turn, knowing the child's usual source of care, may choose to avoid in-hospital care for an illness episode. Recent work by Newacheck et al,10'11 using the National Health Interview Survey, a representative sample of the US population, indicates that most children have an identifiable source of ongoing care. However, sources of care vary substantially by socioeconomic status, with children from poorer families relying far more often on emergency departments or hospital-based clinics for their primary care services.10'1 ' Partly, the use of hospital-based clinics reflects somewhat higher rates of severe illness among poorer children, but much (indeed the majority) reflects instead lack of access to traditional, communitybased, private practice physicians or health maintenance in many poorer communities.

Our studies, like those by Newacheck and colleagues, indicate high rates of identified primary care sources for children hospitalized in Boston, New Haven, and Rochester. At least 85% of children had some primary care source documented in the child's medical record, although the types of primary care source varied substantially among communities for these hospitalized children. Approximately two thirds of children hospitalized in Rochester had a community private practice as the usual source of care, while almost 50% of children hospitalized in Boston had a neighborhood health center as the usual source of care. Although sources of care varied some by socioeconomic status, they varied more by city. Presence or absence of primary care services appears similar across these three communities and probably explains little if any of the variation in rates of hospitalization. More variation likely follows differences in patterns of use of primary care.

Children and their families may experience different availability of primary care, or they may decide to use primary care sources differently. A practice may provide little or no night and weekend coverage, an incentive for families to use emergency departments for illness care. Geography may make the use of distant primary care sources at inconvenient hours difficult for families. Families also may differ systematically in their expectations of how primary care providers can help them when their child is ill. Although the current studies provide little information on the policies of the specific practices that see children in diese three communities, certain evidence points to the notion that Boston families face less accessible primary care or that they use it in ways different from their counterparts in New Haven and Rochester.

Three variables from the medical record audit described primary care use and attachment. First, for relatively acute conditions that often include primary care visits prior to hospitalization (asthma, abdominal pain, and meningitis), Boston children had substantially lower rates of a primary care visit in the 72 hours prior to admission than did children in Rochester and New Haven. Again, when variables reflecting socioeconomic status were added to the analyses, rates for New Haven and Rochester children converged, while those for Boston diverged even further. Primary care physicians often refer their patients to emergency departments when they believe the child needs more extensive work-up, a procedure not available in an office setting, or hospitalization. Families, however, may view their primary care providers as offering little help in the management of acute illness, and they may decide instead to go directly to an emergency department without initial contact with their primary care physician. In the current studies, Rochester children were far more likely than those in Boston to have their primary care practice refer them to the emergency department. New Haven children again were intermediate on this variable.

Hospitals vary substantially with respect to policies regarding attending physicians. The attending at times comes from the child's primary care setting; in other settings, a hospital-based (or other) physician takes responsibility for inpatient care. The involvement of the child's primary care physician likely improves continuity of care, both before hospitalization and especially in the period after hospitalization. This link may diminish use of emergency departments and hospital services in the future. We examined the relationship between the child's attending physician and the usual source of primary care, excluding abdominal pain and head injury (conditions for which surgeons often serve as the attending physicians). Again, children hospitalized in Rochester had much higher rates of having their attending physician be a member of their primary care group.

For all three primary care variables, despite having comparable rates of identified primary care sources, Boston children demonstrated less evidence of primary care use or attachment. Whether these variations explain variations in hospital rates cannot be determined directly from these studies. Yet, the consistency of findings suggests that combinations of types of usual source of care and availability of primary care services may play important roles in determining whether children are hospitalized or receive more illness care in home and community settings.

An important area not examined in the current Boston, New Haven, and Rochester studies is whether physicians in these communities view the value of hospitalization differently. Rochester physicians may view hospitals as inherently bad places for children to be, services of last resort when all else foils. Rochester physicians may work hard to keep most children out of hospitals. Physicians' values in Boston (or New Haven) may differ substantially, with the view that hospitals serve important and beneficial functions and that they may be quite satisfactory sources of care for acute illnesses (despite their relatively high cost), or that the psychological and other risks of hospitalization are outweighed by the benefit of close and careful observation.

CONCLUSIONS

Costs of hospitalization account for a sizable proportion of all expenditures for child health (approximately 45%), despite the tact that relatively few children are hospitalized in any given year.12 Thus, better understanding of the processes that underlie use of hospitalization and exploration of major variations in hospital rates across communities can provide opportunities for diminishing the use of this expensive resource. Simply diminishing hospital use (eg, by closing hospital beds or providing other barriers such as complex authorization requirements) without improving access to effective, community-based, primary care will not serve the needs of children and their families well.13

The consistent findings of variations in hospitalization rates support the need for more intensive study of aspects of physician behavior, household behavior, and access and organization of services. Current studies suggest a central role for primary care services in preventing unnecessary hospitalization. The findings also support opportunities for new interventions, including means of assuring better access to community services and maximizing the use of primary care rather than having households rely on emergency departments and hospitals for much acute care.

Acknowledgments

The author appreciates the efforts of his colleagues on the recent Rochester-New Haven-Boston studies: Sheila Bloom, Dianne Finketstein, Peter Greenspan, Charles Homer, John Leventhal, Lance Rodewald, Peter Szilagyi, and Susan Yazdgerdi.

REFERENCES

1. Wennberg J, Gittelsohn A. Variations in medical care among small areas. Sd Am. 1982;246:tZO-134.

2. Perrin JM, Valvona J, SIoan FA. Changing patterns of hospitalización for children requiring 5UTgCTY. ftdiomcs. 19T6;7 7: 587-592.

3. Wennberg JE1 Freeman JL, CuIp WJ. Are hospital services rationed in New Haven or overutitued in Boston? Lancet. 1987:1: 1 185-?89.

4. Perrin JM, Homer CJ. Berwick DM, Woolf AU Freeman JL, Wennberg JE. Variations in rates of hospitalization of children In three urban comminuties. N Engl J Med. t989i320:l 183-1187.

5. Wissow LS, Ginelsohn AM, Szklo M, Starfield B, Mussman M. Poverty, race, and hospitaliiation for childhood asthma. AmJ Public Heal*. 1988;78:777-782.

6. Starfield B, Newacheck R Children's health status, health risks, and use of health services. In: Schlesinger M], Eisenberg L, eds. Children in a Changing Health Sjitem: Assessments and Proposals for Rtf am Baltimore, Md: Johns Hopkins University Press; 1990:3-26.

7- Halfen N1 Newacheck PW. Asthma, poverty, and utilization of health servirti, Pediomcs. 1993;91:56-61.

8. Weitsman M, Gortmaker S, Soboi A. Racial, social and environmental risks for childhood asthma. AmJ Ws Child. I990;144:!189-1194.

9. Hornet CJ, Perrin JM, Bloom S, et al. Variations in asthma care across three communities. Am J Dis Chad. 1993;H7:455.

10. St Peter RF, Newacheck PW, Halfon N. Access to cate for poor children: separate and unequal? JAMA. 1992; 267:2 760-2764.

11. Stoddard JJ, St Peter RF, Newacheck PW. Health insurance status and ambulatory care for children. N Engt J Med. 19M-3ÎO-. WIl- HZS.

12. Lewit EM, Monheit AC. Expend itures on health care for children and pregnant women. Future o/ Children. 1992 ;2:95- 109.

13. Wise PH, Eisenberg L. What do regional variations in the rates of pediatrie hospitalisation really mean? N Engl J Mid. 1989;320:1209-1211.

TABLE 1

Discharge Rates for Children Ages O- 1 4 Years for Selected Medical Conditions, 1 982

TABLE 2

Conditions With Minimal Variation In Hospitalizaron Rates

TABLE 3

Demographic Characteristics of Boston, New Haven, and Rochester, US Census, 1 990

10.3928/0090-4481-19941201-08

Sign up to receive

Journal E-contents