AMBULATORY PEDIATRICS Volume 3, Number 6324Copyright q 2003 by Ambulatory Pediatric Association November–December 2003

Primary-Care Visits and Hospitalizations for Ambulatory-Care–SensitiveConditions in an Inner-City Health Care System

John F. Steiner MD, MPH; Patricia A. Braun, MD; Paul Melinkovich, MD; Judith E. Glazner, MS;Vijayalaxmi Chandramouli, MS; Charles W. LeBaron, MD; Arthur J. Davidson, MD, MSPH

Objective.—Hospitalizations for ambulatory-care–sensitive conditions (ACSCs) are a marker for access barriers forchildren and a possible outcome measure for primary-care interventions. We assessed the relationship between primary-care utilization and subsequent ACSC hospitalization among inner-city children.

Methodology.— We conducted a nested, case-control study of children born in 1993 in Denver Health (DH), a ‘‘safety-net’’ delivery system in Denver, Colo. Utilization of preventive care and other primary-care services was comparedbetween children hospitalized for ACSCs and nonhospitalized children, who were matched by age and duration of care.Comparisons were adjusted for demographics, payer, and chronic health conditions.

Results.—Of 2531 children, 115 (4.5%) were hospitalized for ACSCs. Sixty-eight percent were Hispanic, and 78%were enrolled in Medicaid. Children with ACSC hospitalization and nonhospitalized children made a similar number ofpreventive-care visits (2.7 6 2.0 vs 3.0 6 2.1 visits, P 5 .30) and other primary-care visits (4.4 6 4.6 vs 3.6 6 4.6,P 5 .16) between birth and hospitalization (for cases) or the same time period (for controls). After multivariate ad-justment, each additional preventive-care visit (odds ratio 5 0.87; 95% confidence interval: 0.67–1.12) was associatedwith a nonsignificant reduction in the risk of hospitalization for ACSC.

Conclusions.—Because ACSC hospitalizations are uncommon and the association between primary care and subse-quent hospitalization is weak, a reduction in ACSC hospitalizations may not be a feasible outcome measure for inter-ventions to increase the rate of preventive- or primary-care visits for underserved children within individual deliverysystems.

KEY WORDS: child health services; pediatric hospitalizations; preventive care; primary health care

Ambulatory Pediatrics 2003;3:324 328

In the early 1990s, Billings et al1 developed a list ofambulatory-care–sensitive conditions (ACSCs) toevaluate the impact of socioeconomic status on bar-

riers to care and to evaluate programs to improve accessto care. This report defined ACSCs as ‘‘diagnoses forwhich timely and effective outpatient care can help to re-duce the risk of hospitalization.’’1 A 1993 report of theInstitute of Medicine (IOM) endorsed the use of ACSCsas an evaluation tool.2 Subsequent investigators modifiedthe IOM list of ACSCs to focus on conditions that affectchildren and proposed similar lists associated with avoid-able hospitalizations.3–5 Although the details vary, ACSCsin children typically include asthma, gastroenteritis anddehydration, and upper and lower respiratory tract infec-tions.3–5 Studies using population-level data have found ahigher incidence of ACSC hospitalizations or avoidablehospitalizations in geographic areas with markers of re-

From the University of Colorado Health Sciences Center (DrSteiner, Ms Glazner, and Ms Chandramouli) and the Department ofCommunity Health Services, Denver Health (Drs Braun, Melinkov-ich, and Davidson), Denver, Colo; and the Centers for Disease Con-trol and Prevention (Dr LeBaron), Atlanta, Ga.

Presented in part at the annual meeting of the Ambulatory Pedi-atric Association, Boston, Mass, May 14, 2000.

Address correspondence to John F. Steiner, MD, MPH, Universityof Colorado Health Sciences Center, Mailstop F-443, PO Box 6508,Aurora, CO 80045 (e-mail: john.steiner@uchsc.edu).

Received for publication February 11, 2003; accepted June 17,2003.

duced access to care, such as lower socioeconomic sta-tus,1,2,4,6–8 fewer primary-care physicians,9,10 or low ratesof private health insurance.4,8,11 Hospitalizations forACSCs were less common among Medicaid-eligible chil-dren with more preventive-care visits, whereas avoidablehospitalizations (defined as using prior ambulatory careand pharmacy services as well as hospital discharge data)were even more strongly associated with primary care.3

On the basis of such findings, researchers have suggestedthat the rate of ACSC hospitalizations can be used to as-sess policy initiatives such as the State Children’s HealthInsurance Program or to identify problems in access orquality of care.2,12–14

The incidence of ACSC hospitalizations is a potentialoutcome measure for interventions to increase the numberof primary-care visits or to improve the effectiveness ofcare delivered in those encounters. To be useful in suchstudies, ACSC hospitalizations must be sufficiently com-mon, and the relationship between the quantity or thequality of primary-care visits and subsequent hospitaliza-tions must be sufficiently strong so that successful inter-ventions can lead to a statistically detectable decline inhospitalization rates. To date, no study has assessed theincidence of ACSC hospitalizations within a single ‘‘safe-ty-net’’ delivery system or has evaluated whether a greaternumber of preventive-care visits or other primary-carevisits is associated with a reduction in subsequent hospi-talizations in such a setting. Although the number of pre-ventive-care visits is an imperfect measure of the timeli-

AMBULATORY PEDIATRICS Primary-Care and Ambulatory-Care–Sensitive Hospitalizations 325

ness or effectiveness of primary care, increasing the ex-posure of disadvantaged children to anticipatory guidanceand other preventive services may also be helpful in im-proving quality. In the planning phase of an interventiontrial, we conducted a nested, case-control study to assessthe relationship between primary-care utilization and sub-sequent ACSC hospitalizations in a cohort of childrenborn in an inner-city health care system in Denver, Colo.We hypothesized that children with hospitalizations forACSCs would have fewer preventive-care visits beforetheir hospitalizations than would age-matched, nonhospi-talized children during the same time period. This studywas approved by the University of Colorado Multiple In-stitutional Review Board.

METHODS

Study Setting

We conducted our study in Denver Health (DH), aninner-city health care system with 1 hospital and 12 fed-erally qualified community health centers in low-incomeneighborhoods in Denver, Colo.15 In 1997, the system pro-vided services to more than 120 000 individuals, 32% ofwhom were under age 18. For children under 36 monthsof age, the community health centers used AmericanAcademy of Pediatrics guidelines for preventive-care vis-its at 2 weeks; 2, 4, 6, 9, and 12 months; 15 to 18 months;and 24 months. Although patients seen in the DH systemcould receive care from other institutions in the Denverarea, their access to those systems was limited by theirlack of insurance coverage.

Study Population and Design

Planning for this study began in early 1997. Becauseearlier studies1,2,6,9,11 had not reported the incidence ofACSC hospitalization in a birth cohort of children and hadnot estimated the strength of association between preven-tive-care visits and subsequent ACSC hospitalizations, welacked information necessary for an initial sample-sizecalculation. To calculate an incidence rate and to estimatethe relationship between preventive care and subsequenthospitalization, we conducted a nested case-control studyin the cohort of children born in the DH system betweenJanuary 1, 1993, and December 31, 1993, who had 1 ormore subsequent encounters in DH before age 36 months.In a nested case-control design, all individuals with theoutcome of interest (hospitalization in the first 3 years oflife) were identified from the 1993 DH birth cohort, andcontrols without that outcome were identified from thesame cohort.16 Each child with a hospitalization for anycause after 14 days of age was designated as a case. Weeliminated all hospitalizations before 14 days of age, be-cause these admissions were unlikely to be affected byprimary care. To assure comparable follow-up, wematched each case to a nonhospitalized control from thecohort on the basis of the duration that the child receivedcare in DH. If multiple potential controls were available,we chose the control with the birth date closest to the case.We defined an index date for each control as the date onwhich the control was the same age in days as the

matched case was on the date of first hospitalization. Forexample, if a case were admitted to the hospital on his180th day of life, the index date for the matched controlwould also be his 180th day of life. This strategy enabledus to compare utilization of services for cases and controlsover the same time span.

Study Measures

The registration and billing systems of DH providedinformation about patient demographics, insurance, utili-zation, and chronic health conditions. Preventive-care vis-its were identified by International Classification of Dis-eases, Ninth Revision, Clinical Modification (ICD-9-CM)code V20.2, or an internal billing code. We identified chil-dren as up-to-date with preventive care according to thedefinition of Hakim and Bye.5

We assessed the completeness of DH claims data bylinking DH claims files with the billing claims of TheChildren’s Hospital of Denver, the most common alter-native-care site for children in the DH system for 1995and 1996. In those 2 years, 259 children in the 1993 DHbirth cohort (14.1%) utilized both institutions. Among thissubgroup, 2.0% of primary-care visits and 6.9% of hos-pitalizations took place at The Children’s Hospital of Den-ver. We thus concluded that the DH claims were suffi-ciently complete to provide a meaningful estimate of out-patient and inpatient utilization.

We identified hospitalizations for ACSCs by the prin-cipal diagnosis at hospital discharge (consistent with ear-lier studies) and the list of conditions proposed in a recentstudy of ACSCs in children.4 The 6 categories of ACSCsare shown in Table 1. We identified chronic health con-ditions by using inpatient and outpatient ICD-9-CM codesdefined by the National Association of Children’s Hospi-tals and Related Institutions Classification of Congenitaland Chronic Health Conditions.17

Data Analysis

We used the Wilcoxon rank sum test to compare con-tinuous variables and the chi-square test for dichotomousvariables. We used conditional logistic regression, whichaccounted for the matching of cases and controls, to assesswhether relationships between preventive-care visits orother primary-care visits and hospitalizations persisted af-ter multivariate adjustment. Separate models were con-structed with ACSC hospitalizations and all-cause hospi-talizations as dependent variables. Gender, race, payer sta-tus, and the presence of a chronic health condition wereincluded as covariates in all models.

RESULTS

In 1993, 2653 children were born at the DH MedicalCenter. One hundred twenty-two (4.6%) children had nosubsequent DH encounters. Of the remaining 2531 chil-dren, 314 (12.4%) had 1 or more hospitalizations (a totalof 394 hospitalizations) after 14 days of age and beforeleaving the DH system or reaching 36 months of age. Onehundred fifteen (4.5%) children were hospitalized at leastonce for ACSCs (135 ACSC hospitalizations in total). Of

AMBULATORY PEDIATRICS326 Steiner et al

Table 1. Causes of Hospitalization, 1993 Birth Cohort, Denver Health*

Principal Discharge Diagnosis n Total Hospitalizations, %

Ambulatory-care–sensitive conditions (135 hospitalizations in 115 children)†

Upper respiratory infections (eg, otitis media, sinusitis, pharnygitis)Gastroenteritis and dehydrationPneumoniaAsthmaSeizuresCellulitis

33282724149

8.47.16.96.13.62.3

Other reasons for hospitalization (259 hospitalizations in 223 children)‡

BronchiolitisInjuriesViral infections not otherwise categorizedBacterial infections not otherwise categorized (eg, urinary tract infections)CroupAll other causes

5235272517

103

13.28.96.96.44.3

25.9

*n 5 394.†Categories from International Classification of Diseases, Ninth Revision, codes for ambulatory-care–sensitive conditions as defined by

Parker and Schoendorf.4

‡Twenty-four children had hospitalizations for both ambulatory-care–sensitive conditions and other reasons.

Table 2. Bivariate Comparisons Between Children Hospitalized for an Ambulatory-Care–Sensitive Condition and Matched, NonhospitalizedControls*

CharacteristicHospitalized Children

(N 5 115)Nonhospitalized Children

(N 5 115) P Value

Male, %Hispanic, %Age at last DH visit before third birthday (mo),

mean 6 SD†Age on date of first hospitalization (cases) or index

date (controls)‡Medicaid insurance, %

53.968.7

26.3 6 8.7

10.7 6 10.278.3

53.071.3

26.4 6 8.7

10.7 6 10.270.4

NSNS

NS

NSNS

Chronic health condition present, %No. of preventive care visits before index dateOne or more preventive care visits before index date, %Up-to-date with preventive care on index date, %No. of other ambulatory visits before index date

43.52.7 6 2.0

94.853.2

4.4 6 4.6

14.83.0 6 2.1

97.462.8

3.6 6 4.6

,.001NSNSNSNS

Total no. of primary care visits (preventive 1 otherambulatory visits) before index date

No. of ED visits before index dateProportion with 1 or more ED visits before index

date

7.1 6 6.00.54 6 0.97

36.5

6.5 6 6.10.33 6 0.69

22.6

NS(.06)

.02

*DH indicates Denver Health; ED, emergency department.†Hospitalized children (cases) and nonhospitalized children (controls) matched on duration of care in DH system.‡For nonhospitalized control children, an index date was defined as the same age (in days) as the matched, hospitalized case on the date

of first hospitalization.

ACSC hospitalizations, 55.6% occurred in children lessthan 12 months of age, 27.4% between 13 and 24 monthsof age, and 17.0% between 25 and 36 months of age.Overall, 135 of 394 hospitalizations (34.3%) were due toACSCs. The most common causes of hospitalization areshown in Table 1.

Children hospitalized for ACSCs are compared withtheir matched controls in Table 2. Chronic health condi-tions were more common in children who were hospital-ized for ACSCs (43.5%) than in nonhospitalized controls(14.8%, P , .001). Cases and controls did not differ inthe number of preventive-care visits or other primary-carevisits before hospitalization for cases and over the sametime period for controls.

The 314 children hospitalized for any reason (ACSCs

or non-ACSCs) and the matched controls also differed intheir prevalence of chronic health conditions (30.6% vs14.3%, P , .001). Children hospitalized for any reasonhad a comparable number of earlier preventive-care visits(2.6 6 2.0 vs 2.8 6 2.0 visits, P 5 .41), more earlierprimary-care visits for other reasons (4.0 6 4.7 vs 3.2 64.5, P 5 .03), and were more likely to have an emergen-cy-department visit than were the controls (34.1% vs20.4%, P , .001).

After controlling for gender, ethnicity, payer, and chron-ic health conditions, each additional preventive-care visitwas associated with a slight reduction in the odds of hos-pitalization for ACSC or for any cause (Table 3). Childrenwho were up-to-date with preventive care were less likelyto have an ACSC hospitalization (odds ratio [OR] 5 0.77;

AMBULATORY PEDIATRICS Primary-Care and Ambulatory-Care–Sensitive Hospitalizations 327

Table 3. Multivariate Relationship Between Primary Care Visits and Hospitalizations*

VariableAdjusted† OR (95% CI) for All

Hospitalizations (N 5 314)Adjusted OR (95% CI) for ACSC

Hospitalizations (N 5 115)

Each additional preventive care visit 0.94 (0.83–1.08) 0.87 (0.67–1.12)Each additional primary care visit (preventive visit

or other ambulatory care visit) 1.02 (0.97–1.06) 1.02 (0.94–1.11)Received all recommended preventive visits for age 0.91 (0.57–1.43) 0.77 (0.34–1.75)

*OR indicates odds ratio; CI, confidence interval; ACSC, ambulatory-care–sensitive condition.†Adjusted for gender, ethnicity, payer, and the presence of a chronic health condition.

95% confidence interval [CI]: 0.34–1.75), although thisrelationship did not achieve statistical significance. Thenumber of other primary-care visits was not associatedwith hospitalization.

DISCUSSION

We found that hospitalizations for ACSCs occurred inonly 4.5% of a birth cohort of disadvantaged children overtheir first 3 years of life and comprised 34.3% of all hos-pitalizations. The proportion of hospitalizations due toACSCs was within the range of 27% to 55%, reportedpreviously.4,7,18 Chronic health conditions were more com-mon among children hospitalized for ACSCs, suggestingthat these hospitalizations may have been due to acuteillnesses superimposed on long-term health problems.Children hospitalized for ACSCs did not have signifi-cantly lower earlier utilization of preventive care than didchildren without an ACSC hospitalization, although theORs were consistent with a small benefit (Table 3).

Several earlier studies have suggested that ACSC hos-pitalizations or avoidable hospitalizations are associatedwith markers of reduced access to primary care.1,2,4,6–11

These studies are susceptible to the ecological fallacy, abias that can arise when access and social disadvantageare measured at the group level rather than at the individ-ual level.19,20 Three studies have evaluated the relationshipbetween primary care and ACSC hospitalizations oravoidable hospitalizations at the individual level. A studyin Spain found no reduction in ACSC hospitalizationsamong children whose parents reported a recent visit to aprimary-care physician.18 An evaluation in Marylandfound that greater utilization of preventive-care visits wasassociated with a decreased risk of all-cause hospitaliza-tion (OR 5 0.91; 95% CI: 0.89–0.93), ACSC hospital-ization (OR 5 0.83; 95% CI: 0.80–0.85), and avoidablehospitalization (OR 5 0.70; 95% CI: 0.67–0.74).3 A studyof Medicaid enrollees in 3 states found that children up-to-date with preventive care had hazard ratios of 0.52 to0.74 (varying by state) for a somewhat different list ofavoidable hospitalizations, whereas a recent preventive-care visit was associated with hazard ratios between 0.86and 0.90.5 Our findings are consistent with these studies,although the relationships were not significant because ofits smaller size.

In the DH system, ACSC hospitalizations were uncom-mon in the first 3 years of life, and preventive-care visitswere associated with only a small reduction in ACSC hos-pitalizations. These findings suggested that hospitalizationfor ACSC would not be a feasible outcome measure for

our planned intervention to increase the number of pre-ventive-care visits. The necessary sample size for eacharm of an intervention trial necessary to reduce ACSChospitalizations by 20% from a baseline rate of 4.5%, witha 5 .05 (2-sided) and power 5 .90, would be 10 085children, substantially more than the size of the annualDH birth cohort. Any sample-size benefit gained by en-rolling older children would be offset by the reduced rateof ACSC hospitalizations in children over 4 years of age.4

A study that found no differences in rates of ACSC hos-pitalizations among community health centers, despitesignificant differences in access to care, further confirmsthe insensitivity of this outcome measure.14

Our study has several limitations: 1) Because the datawere derived from a single health care system, utilizationoutside DH could have affected the observed associations.Linkage of claims data with the other institution com-monly used by DH patients revealed that few primary-care visits occurred outside the DH system. 2) We couldnot determine whether the content of preventive-care vis-its or other primary-care visits emphasized anticipatoryguidance issues, such as home management of minoracute illnesses, that might prevent subsequent hospitaliza-tions. 3) Because only 53% of children with hospitaliza-tion for ACSCs were up-to-date with preventive-care vis-its, it is possible that the rate of ACSC hospitalizationswould have been reduced if the children had received allrecommended preventive care.5 Pediatric preventive care,however, as currently structured, is not risk- or outcome-based. Further research is necessary to define strategies,such as targeted interventions in children with asthma,that might reduce hospitalizations.21 4) The number ofpreventive-care visits is affected both by the delivery sys-tem and by the willingness of parents to seek preventivecare for their children. This claims-based study could notmeasure parental factors that might influence both the uti-lization of preventive- or primary-care services and theclinical decisions to hospitalize their children.

Health-services researchers have promoted the use ofACSC hospitalizations as an access and quality indicatorfor primary care.12–14 Our study suggests that, althoughACSC hospitalizations may be a marker of access at theregional level, this outcome measure is unlikely to be use-ful in interventions to improve the number of preventive-care visits for children within most delivery systems. Sys-tems wishing to reduce unnecessary hospitalizations mayneed to expand access to timely care through strategiessuch as improved after-hours phone consultation and en-hanced availability of same-day appointments for acute

AMBULATORY PEDIATRICS328 Steiner et al

care, whereas clinicians may need to selectively empha-size issues such as early home management of acute ill-nesses if reduction of hospitalization is a primary goal.

ACKNOWLEDGMENTSThe authors wish to thank Allison Kempe, MD, MPH, and Simon

J. Hambidge, MD, PhD, for reviewing the manuscript. This researchwas funded by the Colorado Department of Public Health and En-vironment (contract 97-06432).

REFERENCES1. Billings J, Zeitel L, Lukomnik J, et al. Impact of socioeconomic

status on hospital use in New York City. Health Aff (Millwood).1993;12:162–173.

2. Institute of Medicine Committee on Monitoring Access to Per-sonal Health Care Services. In: Millman M, ed. Access toHealth Care in America. Washington, DC: National AcademyPress; 1993:103–124.

3. Gadomski A, Jenkins P, Nichols M. Impact of a Medicaid pri-mary care provider and preventive care on pediatric hospital-ization. Pediatrics. 1998;101:e1.

4. Parker JD, Schoendorf KC. Variation in hospital discharges forambulatory care-sensitive conditions among children. Pediatrics(J Ambul Pediatr Assoc). 2000;106:942–948.

5. Hakim RB, Bye BV. Effectiveness of compliance with pediatricpreventive care guidelines among Medicaid beneficiaries. Pe-diatrics. 2001;108:90–97.

6. Billings J, Anderson GM, Newman LS. Recent findings on pre-ventable hospitalizations. Health Aff (Millwood). 1996;15:239–249.

7. Pappas G, Hadder WC, Kozak LJ, Fisher GF. Potentially avoid-able hospitalizations: inequalities in rates between US socioeco-nomic groups. Am J Public Health. 1997;87:811–816.

8. Shi L, Lu N. Individual sociodemographic characteristics as-sociated with hospitalization for pediatric ambulatory care sen-sitive conditions. J Health Care Poor Underserved. 2000;11:373–384.

9. Parchman ML, Culler S. Primary care physicians and avoidablehospitalizations. J Fam Pract. 1994;39:123–128.

10. Schrieber S, Zielinski T. The meaning of ambulatory care sen-sitive admissions: urban and rural perspectives. J Rural Health.1997;13:276–284.

11. Weissman JS, Gatsonis C, Epstein AM. Rates of avoidable hos-pitalization by insurance status in Massachusetts and Maryland.JAMA. 1992;268:2388–2394.

12. Bierman AS, Steiner C, Friedman B, et al. Ambulatory caresensitive conditions: ready for prime time?[abstract] J Gen In-tern Med. 1999;14(suppl 2):88.

13. Friedman B, Jee J, Steiner C, Bierman A. Tracking the StateChildren’s Health Insurance Program with hospital data: nation-al baselines, state variations, and some cautions. Med Care ResRev. 1999;56:440–455.

14. Starfield B, Powe NR, Weiner JR, et al. Costs vs. quality indifferent types of primary care settings. JAMA. 1994;272:1903–1908.

15. Gabow P, Eisert S, Wright R. Denver Health: a model for theintegration of a public hospital and community health centers.Ann Intern Med. 2003;138:143–149.

16. Kupper LL, McMichael AJ, Spirtas R. A hybrid epidemiologicstudy design useful in estimating relative risk. J Am Stat Assoc.1975;70:524–528.

17. Muldoon JH, Neff JM, Gay JC. Profiling the health serviceneeds of populations using diagnosis-based classification sys-tems. J Ambul Care Manage. 1997;20:1–18.

18. Casanova C, Colomer C, Starfield B. Pediatric hospitalizationdue to ambulatory care-sensitive conditions in Valencia (Spain).Int J Qual Health Care. 1996;8:51–59.

19. Ricketts TC. Access and ambulatory care sensitive conditions.J Rural Health. 1997;13:275.

20. Piantadosi S, Byar DP, Green SB. The ecological fallacy. Am JEpidemiol. 1988;127:893–904.

21. Lieu TA, Quesenberry CP, Capra AM, et al. Outpatient man-agement practices associated with reduced risk of pediatric asth-ma hospitalization and emergency department visits. Pediatrics.1997;100:334–341.