Spatial Accessibility of Primary Care Pediatric Services in an Urban Environment: Association With Asthma Management and Outcome
BACKGROUND. Disadvantaged urban children with asthma depend heavily on emergency departments (EDs) for episodic care. We hypothesized that among an urban population of children with asthma, higher spatial accessibility to primary care pediatric services would be associated with (1) more scheduled primary care visits for asthma, (2) better longitudinal asthma management, and (3) fewer unscheduled visits for asthma care.
METHODS. We enrolled children aged 12 months to 17 years, inclusive, who sought acute asthma care in an urban pediatric ED. Eligibility criteria included a history of unscheduled visits for asthma in the previous year. We collected comprehensive data on each participant's asthma medical management and prior health care utilization. In addition, we calculated each participant's spatial accessibility to primary care pediatric services, reported as a provider-to-population ratio at their place of residence. Patients then were stratified by their spatial accessibility to care and compared with respect to measures of medical management and health care utilization.
RESULTS. Among the 411 eligible participants, the spatial accessibility of primary care ranged from 7.4 to 350.2 full-time pediatric providers per 100000 children <18 years of age, with a mean of 57.7 ± 40.0. Patients in the middle and highest tertiles of spatial accessibility made significantly more scheduled visits for asthma care than patients in the lowest tertile. There were no differences among tertiles of accessibility with respect to asthma management or with respect to unscheduled visits for asthma care.
CONCLUSIONS. Within this highly urban, largely disadvantaged and minority population of children with chronic asthma, patients with higher spatial accessibility to primary care services made significantly more scheduled visits for asthma care.
Expert guidelines from the National Asthma Education and Prevention Program of the National Heart, Lung, and Blood Institute (NHLBI)1 and the American Academy of Allergy, Asthma and Immunology2 stress the importance of longitudinal asthma care with a primary care provider (PCP). Despite these recommendations, pediatric ED visits for asthma have not decreased,3 and many urban children with asthma and their families view emergency departments (EDs) as their primary source of asthma care.4,5 Pediatric ED visits for asthma are concentrated disproportionately among minority, disadvantaged, and urban children,6,7 and the ED recidivism of this population for asthma care is high.8,9 Specific objectives of Healthy People 201010 include substantial reductions in the rate of ED visits for childhood asthma, together with an increase in the proportion of persons with asthma who receive appropriate asthma care according to the NHLBI guidelines.
Although relevant to asthma care, Starfield11,12 and others have focused more generally on the role of local availability of primary care services in the maintenance of population health. Maldistribution of health care providers was first appreciated nearly 40 years ago,13 and the Institute of Medicine recently noted it as an ongoing issue potentially related to racial and ethnic disparities in health.14 Although the importance of local spatial accessibility (SA) to primary care is generally acknowledged, little is known about its role in specific health outcomes such as the frequency with which families with asthmatic children visit EDs or PCPs seeking episodic or longitudinal care.
Conceptually, proximity to primary care services should facilitate scheduled visits for asthma care. Such visits, in turn, afford the opportunity to improve guideline-based longitudinal asthma care, which has been demonstrated to improve asthma outcomes.15,16 We therefore conducted analyses to test the hypotheses that higher SA to primary care pediatric services among a population of urban and largely disadvantaged children would be associated with (1) more scheduled primary care visits for asthma, (2) better longitudinal asthma management, and (3) fewer unscheduled visits for asthma care.
This is a secondary analysis of data collected as part of a randomized clinical trial that recruited children who sought acute asthma care in the ED at Children's National Medical Center (CNMC) in Washington, DC, from April 2002 until January 2004. Trained research assistants identified subjects among those presenting with respiratory complaints. For this analysis, inclusion criteria included (1) age between 12 months and 17 years, inclusive, (2) prior physician-diagnosed asthma, (3) ≥1 other unscheduled visit for asthma in the previous 6 months and/or ≥1 hospitalization for asthma in the previous 12 months, (4) a parent/guardian available for interview, (5) residency in an urban community, defined as any address in Washington, DC, or any address in a contiguous Maryland county with a 3-mile population density at least equivalent to that of participants from Washington (4803 persons per square mile minimum), and (6) needed ≥3 doses of nebulized albuterol in the ED at the time of enrollment.
Exclusion criteria included (1) significant medical comorbidities affecting the cardiorespiratory system, (2) a visit to an allergist or a pulmonologist in the previous 6 months, (3) ≥2 of the following: a current written asthma medical action plan, current use of >1 controller medication, or a scheduled visit for asthma care with their PCP in the previous 2 weeks, (4) enrollment in another asthma research study, (5) unavailability for telephone follow-up, or (6) primary language other than English or Spanish.
The CNMC Institutional Review Board approved this study. Legal guardians provided informed consent, and participants >6 years of age provided assent.
After informed consent was obtained, the research assistants obtained baseline data from each subject's parent or guardian that assessed demographics, insurance status, past asthma history, home exposure to potential asthma irritants and allergens, medication usage, health care utilization, and asthma classification by the criteria of the NHLBI.1
SA to primary care services was calculated as the ratio of full-time equivalent (FTE) pediatric PCPs to child population in the immediate vicinity of the patient's residence. This method has been described previously.17,18 Briefly, we created 2 density maps of our study region. One represented provider density, and the other represented child-population density. We then divided the provider-density layer by the population-density layer to obtain a layer of provider-to-population ratios at all points in our study region. From this, we determined the ratio at each patient's residential address, geo-coded to latitude and longitude, and defined that ratio as each participant's SA to pediatric primary care. Higher SA denoted greater immediate proximity to sources of primary pediatric care. All spatial calculations were performed by using ArgGIS 9.0 (ESRI, Inc, Redlands, CA).19
The population-density layer was created from block group centroids weighted by the number of children <18 years of age living in the block group (using 2000 US census data) and smoothed with a Gaussian kernel density function to a 1-mile distance.20,21 These procedures closed gaps between block groups without excessive loss of neighborhood resolution.
The pediatric provider-density layer was derived from a unique, comprehensive, and highly detailed database of provider locations built for this project. Research has shown that the commonly used enumerations of physicians derived from the membership lists of the American Medical Association or the American Osteopathic Association do not adequately represent the amount of primary care available in a study region.22 Therefore, we assembled a study-specific list by cross-referencing provider lists of physicians, physician assistants, and nurse practitioners from 4 Medicaid managed-care organizations with a list of pediatric PCPs maintained by CNMC. In addition, telephone books from the Washington, DC, metropolitan region were examined to identify additional care sources. Finally, key informant interviews were performed with selected prominent local PCPs to identify practitioners and point sources of care.
Research assistants then contacted each point source of care and identified each family practice or pediatric physician, nurse practitioner, or physician assistant who worked at that address. Research assistants also recorded the number of hours per week that each provider devoted to primary care services. Using previously published data as guidelines,23 adjustment was made for type of training (pediatrics [weight: 1.0] versus family practice [weight: 0.25]), training level (attending [weight: 1.0] versus resident/fellow [weight: 0.35]), and type of provider (physician [weight: 1.0], nurse practitioner [weight: 0.75], or physician assistant [weight: 0.75]). Finally, a total number of pediatric PCPs was generated for each point source of care and expressed as a number of FTE providers (based on a 40-hour work week treating pediatric patients).
This FTE value was used as a weighting factor in converting care locations to a map density layer. Using methods developed and published earlier,17,18 we used a 3-mile smoothing radius in the provider-density calculation. The use of this distance is based on a survey showing how far black city residents typically travel to community clinics24 and on recent unpublished data from our group showing how far patients typically travel to health care clinics operated by our institution.
Initially, analyses of means and frequencies were conducted to describe study participants' baseline sociodemographic and clinical characteristics and to compare their mean SA by these characteristics. All subsequent analyses were adjusted for baseline differences among tertiles of SA with respect to age, gender, race/ethnicity, insurance status, exposure to environmental tobacco smoke, cockroach exposure, and NHLBI severity classification.
Analyses directed at comparing differences in measures of longitudinal asthma care by SA, grouped in tertiles, were implemented by using multiple logistic-regression models. These models included each specific outcome as the dependent variable, binary indicators of SA tertile, and covariables and yielded estimates of each measure with odds ratios (ORs) and 95% confidence intervals (CIs). In each case, the lowest tertile of SA was used as the reference group. Each estimate and OR is reported twice: unadjusted and then adjusted by the SA of each participant to any ED and specifically to the ED at CNMC. The latter adjustment was added because CNMC receives >85% of all pediatric asthma visits to EDs in the District of Columbia. As such, it provides a unique point of access for families that we wished to hold constant.
Analyses then compared the rates of health care utilization and relative risks (RRs) with 95% CIs by tertile of SA. Again, the lowest tertile of SA was used as the reference group. Depending on the degree of dispersion in the rates of utilization, these models were based on either Poisson or negative binomial regression analyses. The latter were used whenever the degree of dispersion exceeded that accommodated by the Poisson model. Each model included baseline covariables as described above, which are similarly reported unadjusted and then adjusted by the SA to all EDs and to CNMC. All analyses were conducted by using procedures in Stata 8.0 (Stata Corp, College Station, TX)25 and SAS 9.1.3 (SAS Institute, Inc, Cary, NC).26
The research assistants screened 2791 patients over the 22-month enrollment period. Of these, 521 (18.7%) were eligible, and 490 (94.0%) were enrolled. Two were subsequently excluded because of enrollment violations, and another 77 were excluded from this analysis because they lived at a nonurban address as defined above. These exclusions yielded a final sample size of 411 participants.
Baseline demographics are listed in Table 1. Of note, 60.1% of participants were <6 years of age, 62.5% were male, 86.1% were black, 77.4% were publicly insured or uninsured, and 57.5% had made >4 unscheduled visits for asthma care in the previous 12 months. Also of note is that 387 (94.2%) of 411 patients reported a specific PCP.
SA of primary care pediatric services ranged from 7.4 to 350.2 FTE providers per 100000 children <18 years of age with a mean of 57.7 ± 40.0. The geographic distribution of providers is depicted as a density map in Fig 1. As demonstrated in Table 1, black children had a significantly lower mean SA to pediatric primary care (55.9 FTE providers per 100000) than a reference group of white, Hispanic, and other children (69.2 FTE providers per 100000). There were no significant differences in SA to primary care pediatric services with respect to age, gender, insurance type, or history of scheduled and unscheduled health care utilization for asthma over the previous 12 months.
The relationship of SA to measures of asthma management is examined in Table 2. Before adjustment for spatial proximity to EDs and to CNMC, when compared with the lowest tertile of SA, significantly more patients in the middle tertile reported use of a written asthma plan, and significantly fewer patients in the highest tertile reported use of leukotriene antagonists in the past month. Neither association remained significant after controlling for proximity to EDs and to CNMC. There were no significant differences in reported rates of spacer use whenever using a metered-dose inhaler, influenza vaccination in the previous year, or use of inhaled corticosteroids in the last month.
Table 3 examines the relationship of SA to health care utilization. Patients in the middle and highest tertiles made significantly more scheduled visits for asthma care than patients in the lowest tertile, although only the association in the middle tertile remained significant after controlling for spatial proximity to EDs and to CNMC. There were no differences in unscheduled visits among tertiles both before and after controlling for ED proximity.
The range of SA to primary care among our study participants was large, varying by a factor of nearly 50, and differed widely within our study area, as depicted in Fig 1. Our participants lived in areas well above and below published guidelines and benchmarks for SA to pediatric primary care, which range from 41.2 to 57.5 pediatric PCPs per 100000 children.23,27,28 Note that these estimates are for PCPs, not for FTE providers as our more-textured analysis produced.
It seems reasonable to hypothesize that SA to primary care should facilitate primary care visits and thereby health outcomes; however, this issue has never been investigated adequately. In fact, there is considerable debate regarding the adequacy of the nation's supply of PCPs and specialty providers in relation to population. Some argue that we face critical shortages and geographic distribution imbalances that, if not corrected, will adversely affect the nation's health.29,30 Others believe that these projections are based on historical provider-to-population ratio benchmarks with little or no evidence derived from actual population health status or medical outcomes.31,32 Notwithstanding published recommendations, it is not known what the optimal provider-to-population ratios are for any medical condition or any specific population.
We studied the relationship between the spatial distribution of providers and outcomes for 1 condition, pediatric asthma, in a well-defined population. The data and methods are unique for this purpose. We controlled not only for the usual demographic, insurance status, and comorbidity factors but also for severity of chronic illness and for home exposures known to impact asthma morbidity (environmental tobacco exposure and cockroach allergen). We also built a local health care provider database that is more comprehensive and textured than any used before to our knowledge. Finally, our measure of SA has many advantages over traditional methods such as distance to nearest provider or provider-to-population ratios within small-bordered areas. For example, our method avoids concerns about crossing artificial small-area borders to obtain care, and it also takes into account population demand at each provider location.17
In contrast to these strengths, this study has important limitations. Most relevant is that it is a secondary analysis of a sample that is biased and narrowly defined. For example, participants were deliberately recruited on the basis of a history of prior unscheduled utilization (57.5% of the sample had made >4 unscheduled visits for asthma care in the previous 12 months). It may be that our population was selected to have little sense of partnership in longitudinal asthma care with their PCPs, and therefore the impact of SA to primary care is limited.
Even within this selected population, however, higher SA to primary care pediatric services was associated with an increased likelihood of scheduled primary care visits for asthma within the past 12 months. This is noteworthy, because scheduled visits are advocated by expert guidelines that hope to positively affect the asthma management and health care utilization that we studied.
Findings for the specific aspects of clinical management and unscheduled health care utilization that we investigated were mixed, however, and less compelling. The middle tertile for SA had higher odds of using a written asthma plan in the last month compared with the lowest SA tertile, whereas the highest SA tertile had lower odds of using a leukotriene antagonist in the last month. These findings were not apparent in the final regression models, which were fully adjusted for SA to EDs and distance from CNMC. No other findings were statistically significant. The lack of an association between higher SA and aspects of clinical management and unscheduled health care utilization despite an association between higher SA and increased scheduled PCP visits raises questions about the content and effect of the scheduled visits.
Of course, these negative findings do not disprove the importance of SA to primary care in the management of pediatric asthma. Apart from the usual statistical considerations in hypothesis testing, there may be unique issues in the population we recruited. For example, their socioeconomic disadvantage and urban environment may create important, but unmeasured, barriers to adequate care even in the presence of good SA to care. A similar analysis might yield different results in other populations.
Our analysis has other limitations. First, SA is only 1 determinant of access to a point source of care. Other factors include appointment practices, phone availability, availability of evening and weekend appointments, and adequacy of public transportation. We were not able to quantify these aspects of access. Therefore, local point sources of care could be spatially proximate but not temporally or logistically accessible. Second, accessibility to primary care services is distinct from the quality of those services, and our analysis does not control for practitioner quality or familiarity with asthma care guidelines. Finally, we measured and analyzed accessibility to all potential sources of pediatric primary care services within reasonable distance, not to each participant's usual source of services. It could be that for many of our patients, their actual point source of primary care was geographically quite remote, perhaps more remote than ≥1 EDs. Despite these limitations, the analyses reported here may be the first to link local SA of local providers to measures of clinical management and health care utilization for a specific health condition.
We found that within an urban population of largely minority and disadvantaged asthmatic children with a history of ED recidivism for asthma care, higher SA to primary care pediatric services was associated with an increased number of scheduled primary care visits for asthma. However, there was no association between SA to primary care and multiple measures of asthma management and unscheduled health care utilization. These findings may have implications for programs that stress the role of local community providers as focal points for longitudinal asthma management and as viable alternatives to episodic ED care. In addition, our model of SA may be an important adjustment tool for future models of health care outcomes, and may serve as a tool in efforts to determine optimal provider ratios and geographic distributions for other diseases and conditions.
Support for this study and publication of this article was provided by the Robert Wood Johnson Foundation and grant M01-RR020359 from the General Clinical Research Center Program of the National Center for Research Resources, National Institutes of Health, Department of Health and Human Services.
Technical assistance was provided by the National Program Office (director, Gary Rachelefsky, MD, Allergy Research Foundation Inc; deputy director, Amy Stone, American Academy of Allergy, Asthma and Immunology; and research associate, Suzanne Kennedy, PhD, American Academy of Allergy, Asthma and Immunology). In addition, we acknowledge the database support of Bruce M. Sprague.
- Accepted December 6, 2005.
- Address correspondence to Stephen J. Teach, MD, MPH, Division of Emergency Medicine, Children’s National Medical Center, 111 Michigan Ave NW, Washington, DC 20010. E-mail:
Financial Disclosure: Dr Teach serves on the speakers' bureau for AstraZenca. The other authors have indicated they have no financial relationships relevant to this article to disclose.
Dr Joseph’s current address is Department of Pediatrics, University of California, Davis, 2516 Stockton Blvd, Sacramento, CA 95817.
- ↵National Heart, Lung, and Blood Institute; National Asthma Education and Prevention Program. Expert Panel Report 2: Guidelines for the Diagnosis and Management of Asthma. Bethesda, MD: National Institutes of Health; 1997. NIH publication No. 97-4051
- ↵Pediatric Asthma Committee. Pediatric Asthma: Promoting Best Practice Guide for Management of Asthma in Children. Milwaukee, WI: American Academy of Allergy, Asthma and Immunology; 1999
- ↵Mannino DM, Homa DM, Akinbami LJ, Moorman JE, Gwynn C, Redd SC. Surveillance for asthma: United States, 1980–1999. MMWR Surveill Summ.2002;51(1) :1– 13
- ↵Shields AE, Comstock C, Weiss KB. Variations in asthma care by race/ethnicity among children enrolled in a state Medicaid program. Pediatrics.2004;113 :496– 504
- ↵US Department of Health and Human Services. Respiratory diseases. In: Healthy People 2010: With Understanding and Improving Health and Objectives for Improving Health. Vol 2. 2nd ed. Washington, DC: US Government Printing Office; 2000:24
- ↵US National Advisory Commission on Health Manpower. Report of the Commission. Washington: US Government Printing Office; 1967
- ↵Smedley BD, Stith AY, Nelson AR, eds. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington, DC: National Academies Press; 2002
- ↵Lieu TA, Quesenberry CP Jr, Capra AM, Sorel ME, Martin KE, Mendoza GR. Outpatient management practices associated with reduced risk of pediatric asthma hospitalization and emergency department visits. Pediatrics.1997;100 :334– 341
- ↵Adams RJ, Fuhlbrigge A, Finkelstein JA, et al. Impact of inhaled antiinflammatory therapy on hospitalization and emergency department visits for children with asthma. Pediatrics.2001;107 :706– 711
- ↵ArcGIS [computer program]. Version 9.0, Service Pack 3. Redlands, CA: ESRI, Inc; 2004
- ↵McLafferty S, Williamson D, McGuire PG. Identifying crime hot spots using kernel smoothing in analyzing crime patterns. In: Goldsmith V, McGuire PG, Mollenkopf JB, Ross TA, eds. Analyzing Crime Patterns: Frontiers of Practice. Thousand Oaks, CA: Sage Publications; 1999:77– 85
- ↵Longley PA, Goodchild MF, Maguire DJ, Rhind DW. Geographic Information Systems and Science. Chichester, NY: Wiley; 2001
- ↵American Academy of Pediatrics. Physician Workforce: Ratios for Child Health. Elk Grove Village, IL: American Academy of Pediatrics; 1998
- ↵Stata Statistical Software [computer program]. Release 8.0. College Station, TX: Stata Corp; 2003
- ↵SAS Statistical Software [computer program]. Version 9.1.3. Cary, NC; SAS Institute Inc; 2004
- ↵Marder W, Gaumer G. Reexamination of the Adequacy of Physician Supply Made in 1980 by the Graduate Medical Education National Advisory Committee (GMENAC) for Selected Specialties. Cambridge, MA/Rockville, MD: Abt Associates, Inc; 1991
- ↵US Department of Health and Human Services. Report of the Graduate Medical Education National Advisory Committee to the Secretary. Vol II. Washington, DC: Modeling, Research and Data Technical Panel, US Government Printing Office; 1980
- ↵Council on Graduate Medical Education. Tenth Report: Physician Distribution and Health Care Challenges in Rural and Inner-City Areas. Washington, DC: US Department of Health and Human Services, Public Health Service, Health Resources and Services Administration; 1998
- ↵Cooper RA, Getzen TE, McKee HJ, Laud P. Economic and demographic trends signal an impending physician shortage. Health Aff (Millwood).2002;21 :140– 154
- ↵Goodman DC. The physician workforce crisis: where is the evidence? Health Aff (Millwood).2005;W5-108-110
- Copyright © 2006 by the American Academy of Pediatrics