HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) All Versions of this Article: AJPH.2005.063057v1 96/7/1243    most recent Submit a response View Shopping Cart Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Silveira, M. J. Articles by Feudtner, C. Search for Related Content PubMed PubMed Citation Articles by Silveira, M. J. Articles by Feudtner, C. Related Collections Epidemiology Geography Health Policy Other Health Service Delivery

Maria J. Silveira, MD, MA, MPH, Laurel A. Copeland, PhD, MPH and Chris Feudtner, MD, PhD, MPH

Maria J. Silveira is with the VA Health Services Research and Development Center of Excellence, VA Ann Arbor Healthcare System, Ann Arbor, and the Division of General Medicine and the Bioethics Program, University of Michigan, Ann Arbor. Laurel A. Copeland is with the South Texas Veterans Health Care System VERDICT Research Program, San Antonio, and the Department of Psychiatry, University of Texas Health Science Center, San Antonio. Chris Feudtner is a member of the Pediatric Advanced Care Team, The Children’s Hospital of Philadelphia, Philadelphia, Pa, the Pediatric Generalist Research Group, Department of Pediatrics, Center for Bioethics, Leonard Davis Institute, Philadelphia, and Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania, Philadelphia.

Correspondence: Requests for reprints should be sent to Maria J. Silveira, 300 North Ingalls Building, Room 7C27, Box 0429, Ann Arbor, MI 48109-0429 (e-mail: mariajs{at}umich.edu).

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION References

 

Objectives. We tested whether local cultural and social values regarding the use of health care are associated with the likelihood of home death, using variation in local rates of home births as a proxy for geographic variation in these values.

Methods. For each of 351110 adult decedents in Washington state who died from 1989 through 1998, we calculated the home birth rate in each zip code during the year of death and then used multivariate regression modeling to estimate the relation between the likelihood of home death and the local rate of home births.

Results. Individuals residing in local areas with higher home birth rates had greater adjusted likelihood of dying at home (odds ratio [OR]=1.04 for each percentage point increase in home birth rate; 95% confidence interval [CI] = 1.03, 1.05). Moreover, the likelihood of dying at home increased with local wealth (OR=1.04 per $10000; 95% CI=1.02, 1.06) but decreased with local hospital bed availability (OR=0.96 per 1000 beds; 95% CI=0.95, 0.97).

Conclusions. The likelihood of home death is associated with local rates of home births, suggesting the influence of health care use preferences.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION References

 
Despite most terminally ill patients’ wishing to die at home, two thirds die elsewhere.^1–5 Of the many explanations for this inconsistency between desire and practice, the best-supported theory posits that the site of death is determined primarily not by the characteristics of individual patients or providers, but by characteristics of the local health system. As depicted in Figure 1 (as association 1), greater availability of beds in skilled nursing facilities and hospitals correlates with higher rates of deaths in those facilities,^6,7 even after adjustment for individual characteristics such as age, gender, and marital status.⁸ Evidence also exists that individual characteristics may serve as logistical constraints on the ability to act on a wish to die at home, as people who are older, more functionally impaired, living alone, and suffering from more disabling diagnoses die more often in hospitals.^9–11

View larger version (23K): [in this window] [in a new window]   FIGURE 1—— How local preference for hospital care may confound the relation between hospital bed capacity and place of death.

Relatively homogenous and coherent sets of these values may be shared within local communities, either through a process of person-to-person dissemination of values or by people preferentially migrating to communities that manifest particular values. Community-level values regarding health care preferences may influence both choices to die at home and the availability and type of health services. Individuals operating within a culture that values hospitalization might seek or demand more hospital beds, or choose to reside in regions where more hospital beds were available, and thereby create an inflated association between bed availability and hospital deaths.

We examined the relation between local values and place of death depicted by association 3 in Figure 1. Because no direct measure was available, we used a proxy measure for local health care values. We sought a population-based parameter of health service use that, although having no clear relation to the proportion of deaths occurring at home, would also be affected by any local preference for or against medical care. We posited that the proportion of births occurring at home in a local area would serve as surrogate marker for a local preference against hospital care (depicted as association 2 in Figure 1). This model of the relation between home birth rates and preferences about hospital care accords well with findings from a detailed study of the Netherlands, where the high rates of home births and home deaths are at least in part because of cultural preferences for these modes of care.¹²

Accordingly, we hypothesized that in geographic regions with higher proportions of home births, a death would be more likely to occur at home, even after adjustment for the decedent’s demographic factors and for the local hospital bed supply. We tested this hypothesis by examining the association between home births and home deaths in Washington state during the decade spanning 1989 through 1998.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION References

 
Research Design and Outcome of Interest
We conducted a retrospective population-based cohort study, the outcomes of interest being the person’s place of death and the local rates of home deaths. For analyses done at the level of the individual (henceforth referred to as individual analyses), the unit of analysis was the decedent. Home deaths were defined as those deaths occurring at the person’s primary residence. They were compared with deaths occurring elsewhere (e.g., hospital, nursing home, hospice facility). For analyses done at the group level (henceforth referred to as the group analyses), the outcome was the rate of home deaths per zip code; this dependent variable was used in models that used zip code as the unit of analysis.

Study Population and Data Source
We examined the death certificates of 351110 residents of Washington who died nontraumatic deaths after their 18th birthday during the interval from 1989 through 1998; individuals dying from injury or poisoning were excluded. We designed our selection criteria to identify those subjects most likely to have chosen both their place of residence and place of death. The analytic sample represented 86% of all deaths in Washington during that decade. We also examined Washington state birth certificates for the same period.

Predictor of Interest
The primary predictor variable was the rate (proportion) of home births in a residential zip code serviced by the US Postal Service in 1990. We omitted zip codes that consisted solely of post office boxes.

We determined home birth rates using birth certificate data for 767082 live births occurring in 484 residential Washington zip codes during 1989 through 1998. Home births occurred at the mother’s primary residence, regardless of later transfer to a hospital. Home birth rates were determined by dividing the total number of home births by the total number of births for each zip code and multiplying by 100.

Because the demographics of decedents and new mothers differed greatly, and because maternal demographic characteristics partially reflect the characteristics of the community in which the decedent lived, we adjusted for maternal demographics (year of birth, mother’s marital status categorized as married and nonmarried, mother’s age, and mother’s ethnicity categorized as American Indian, Black, Hispanic, Asian, other non-White, or White), so that these effects would not be confounded with the effect of the cultural value we sought to capture in home birth rates.

A total of 744319 birth records (95.7%) had complete data with the exception of mother’s educational level, which was excluded because of missing data (Washington state did not begin gathering this information until 1992; thus, 38% of birth certificates did not list the mother’s education). These factors were aggregated to the zip code level to permit linkage to individual decedent records by the zip code.

Covariates
Covariates included decedent, maternal, and geographic area factors. Decedents’ covariates in the model included year of death, race/ethnicity (categorized as White or non-White), age, gender, marital status (married or nonmarried), and leading causes of death (International Classification of Diseases, Ninth Revision, Clinical Modification codes for infectious diseases, 001–139; malignancy, 140–239; diabetes and complications, 250–250.9; cerebrovascular disease, 430–438.9; circulatory, 390–430, 440–459, 785.5–785.59; pneumonia, 480–487.6; chronic obstructive pulmonary disease, 490–496; gastrointestinal and liver, 520–579; renal and genitourinary, 580–599; injury and poisoning, 800–994.9).¹³ Geographic factors included median household income per zip code (1990 Census figures), population per zip code (1990 Census figures), and hospital bed availability.

Hospital bed availability was determined from 1990 American Hospital Association statistics counting the number of staffed hospital beds in each of 63 hospital service areas (HSAs) in Washington state. Briefly, HSAs are groupings of contiguous zip codes whose residents preferentially use a certain hospital. The number of hospital beds per HSA was modeled with population as 2 separate variables in our analyses, rather than as a ratio, to avoid autocorrelation.⁸ HSAs for Washington have previously been described.^7,14,15 HSAs can be linked to their component zip codes and then, by zip code, with individual-level records. In this study, the mean population per zip code (in 10000s) was 1.02 (SD= 1.19; range = 0 to 5.06). The mean value of the median household income (in $10000s) per zip code was 2.84 (SD= 0.87, range = 0 to 6.16). The mean number of hospital beds per HSA (in 1000s) was 0.58 (SD= 0.89, range = 0.01 to 3.32).

Individual-Level Analysis
To analyze individual decedents’ place of death, we first determined bivariate associations between home deaths and categorical variables using ² analyses and between home deaths and continuous variables with Student t tests. We used Mantel–Haenszel ² to assess trends over time. We used logistic regression modeling to test the main hypothesis that local home birth rates would be positively associated with the likelihood of death at home, with adjustment for maternal characteristics (mothers’ average age, percentage married, percentage American Indian, percentage Black, percentage Hispanic, percentage Asian American, percentage other non-White, average year of birth), decedent characteristics (year of death, ethnicity, age, gender, marital status, and cause of death with other as the referent category to highlight the relative effects of specific causes of death), and geographic area characteristics (hospital bed availability, population size, and median household income).

The sample for the individual-level analyses was 349 613 decedents with complete data. For our main analysis, we examined the relation between an individual’s place of death and local home birth rates (zip code rates) with maternal covariates aggregated to the zip code level. To confirm our findings, we repeated the multivariate analyses with all maternal covariates aggregated to the HSA level, examining the relation between the place of death and HSA home birth rates. Because of the large number of predictors and lack of theoretical or empirical guidance from the literature, we did not explore interactions.

Group-Level Analysis
To analyze a local area’s home death rate, we first aggregated the death data to the zip code level by taking the mean age of decedents per zip code and the percentage of decedents characterized by each indicator variable per zip code. We then entered these zip code–level variables into a linear regression estimating the percentage of deaths occurring at home per zip code. In addition, we tested models including median income, population, and hospital bed availability, as described previously. We used zip codes as our geographic unit of analysis because it was the smallest geographic unit available and, thus, was more likely to capture local cultural variation than larger units such as counties. The sample for the group level analyses was 475 zip codes with complete data.

For all data analysis, we used SAS 8.2 (SAS Institute, Cary, NC) and Stata 7.0 (Stata-Corp LP, College Station, Tex).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION References

 
Deaths
In Washington state from 1989 through 1998, 26% of adults who died because of nontraumatic causes did so at home. Dying at home became more common over the decade, increasing from 23% in 1989 to 28% in 1998 (P < .0001). Among zip codes in Washington, the percentage of home deaths during this decade ranged from 0% to 56.3%. As shown in Table 1, individuals who died at home were more likely to be male, married, Black, and less educated than those who did not die at home. Those who died at home were younger than those who died elsewhere (72 vs 76 years, P < .0001) and tended to suffer from cancer more often (40% vs 22%, P < .001).

Individual Likelihood of Home Death
In a multivariate logistic regression model (Table 2), individuals residing in zip codes with higher home birth rates had greater odds of dying at home (adjusted OR = 1.04; 95% CI = 1.03, 1.05; P < .0001), representing a 4% increased likelihood of home death for every percentage increase in the home birth rate. Thus, decedents in a zip code with 10% home births would have roughly a 40% increased risk of dying at home. The pseudo-R² for this model was 6.1%. Within the range of home birth rates among Washington zip codes (0% to 26.3%), the predicted likelihood of home death ranged from 2.7% to 75.3%. Similar results were obtained when median household income, population, and hospital bed availability were removed from the model: the odds ratio for home birth remained 1.04, suggesting that income and bed availability do not affect the relation between home births and home deaths. The relation between home births and home deaths was equally strong when the analysis was limited to decedents with cancer (OR = 1.05; 95% CI = 1.04, 1.06).

Local Likelihood of Home Death
In a linear regression model in which the home death rate per zip code was the primary outcome, home birth rates per zip code were significantly associated with the percentage of deaths at home (Table 3). The effect estimate for zip code home birth rates was 0.41, meaning that on average for each percentage increase in home birth rates, the home death rate increased by four tenths of a percent. The proportion of the variance in zip code home death rates accounted for by this model was 28.6% (adjusted R² = 0.25). Other significant factors were decedent gender, year of death, some causes of death, and median household income. HSA beds and population were not significantly related to the outcome.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION References

 
Research into the geography of death in America has attempted to answer 2 questions. First, why, despite 9 out of 10 Americans preferring otherwise,¹⁶ do half die in institutions?⁵ Second, why are there regional differences in places of death, with higher rates of home death in the western parts of the United States compared with elsewhere?^7,17,18

In the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatment study, patient preference had little association with the place of death.¹⁰ Consequently, research has explored the relation between the place of death and conditions beyond patients’ control. These conditions fall into 2 types: those pertaining to the individual and the home environment (such as age, race, diagnosis, gender, and caregiver status), and those of the local health system (such as the availability of hospital beds, the per capita supply of physicians, and local wealth). Two studies have examined regional differences in the relation between these factors and place of death, both with similar conclusions. Pritchard and colleagues¹⁰ used hospital referral regions as the unit of analysis to show that hospital bed availability is associated with the risk of in-hospital death after control for individual factors.

Most recently, Hansen and others¹⁸ conducted an analysis of deaths nationwide demonstrating that for every additional hospital bed per 1000 population, there is a 4.2% increase in the hospital death rate after adjustment for individual and other systemic factors. They confirmed these results at the state level, where greater hospital bed availability was associated with an increased likelihood of hospital death. On the basis of these findings, they concluded that "although we would like to believe otherwise, where we die is influenced less by our individual personal preferences than by factors outside our direct control."¹⁸ In addition, various authors have suggested that fewer hospital deaths would occur if changes were made in the local health care system, for instance by increasing access to hospice or limiting hospital transfers.^17,18

Although these observational studies demonstrate a relation between hospital bed availability and hospital deaths, they do not explain how or why this relation exists. Indeed, the inference that individuals make decisions regarding their location at the time of death on the basis of the local accessibility of hospital beds is premature and rests on a potential ecologic fallacy: that hospital bed availability is a fundamental driver of place of death rather than an epiphenomenon brought about by a heretofore hidden relation between an as-yet-unspecified causal factor and place of death. The question, couched in epidemiological terms, remains: is bed availability a main effect, a confounder, or an intermediate variable on a causal pathway?

We examined whether hospital bed availability and other systemic factors are best understood as links in the chain of causation between geography and place of death. Moreover, we posited that social factors, namely local preferences regarding medical care, would be the beginning of the chain. Our findings suggest that a local preference regarding medical care, measured by the surrogate marker of home birth, is associated with home death, albeit to a small degree.

The small effect of local home birth rates on the likelihood of home death that we observed has at least 3 potential explanations. First, the capacity of home birth rates to serve as an accurate proxy for the underlying construct of local cultural and social values may be quite limited, leading directly to an underestimation of the proposed relation, assuming it exists. Second, the proposed relation between local values and the rates of home births may be entirely invalid, implying that the observed correlation between home births and home deaths is driven by some unidentified third variable.

Third, rather than revealing a linear chain of causation, our results may instead suggest a "web of causation,"¹⁹ whereby home deaths in certain areas appear to be determined to different degrees not only by the local values but also by individual demographic and clinical factors and by characteristics of the local health system such as bed availability and local wealth, each acting independently and simultaneously. Any such complex causal web could attenuate or inflate an observed relation if any of the key "threads" in the web were omitted from the analytic model. This potential causal complexity may also explain why other studies have shown different results regarding the association between place of death and the decedent’s individual factors, such as gender and age.²⁰

Although we found that hospital bed availability was associated with hospital death at the individual level, the relation became insignificant at the aggregate level. This observation highlights a major challenge in conducting multilevel analyses, namely that correlations between variables can differ markedly in magnitude, and even direction, when analyzed at the individual and aggregate levels.²¹ Accordingly, multilevel analyses and inferences drawn from them should be based on causal models specifying how group-level and individual-level variables interact.¹⁹

Theories regarding place of death will need to be grounded through qualitative research in the real-life experiences of patients, their families, and their communities. For example, we note high rates of home deaths and home births for zip codes on island communities; qualitative research could determine whether home deaths or births represent choice or constraint of choice because of location. Additionally, better theories are needed about what constitutes community values: are they simply a composite of individual-level values of patients, family members, and health care providers, or do they represent more-complex emergent values that influence individuals but exist primarily at the level of communities?

Our study had 2 limitations that warrant discussion. First, having no established measure for local preferences regarding hospital-based medical care, we used home birth rates as a proxy marker because studies suggest that home birth is a culturally driven phenomenon.^12,22–24 As argued previously, to fully appreciate the cultural connection between home births and home deaths, qualitative work is needed.

Second, we used zip codes as our unit of analysis. Zip codes were created to reflect postal routes, not the boundaries of neighborhoods and communities per se. Thus, some might argue that the relation we observed between home births and home deaths at the zip code level might be because of location rather than shared culture. Alternative units of analyses, such as census tracts or school districts, could be equally flawed. Although the ideal unit of analysis for this study—coherent communities—has yet to be geographically defined and developed, any failure of zip codes to map accurately onto cohesive local communities would most likely result in nondifferential misclassification, thereby having biased this study’s estimates of an association of local home birth rates and the likelihood of home death in a conservative direction.

Beyond these methodological issues, the practical question remains: what do we do to bring the experience of dying in America in line with individuals’ stated preference to not die in a hospital? Do we work on structural features of the health care system, such as enhancing access to home care or building more hospices? Or do we intervene on psychosocial features of individuals, such as enhancing empowerment and knowledge of hospice? Our findings, if substantiated, and the considerations discussed earlier argue against a mutually exclusive focus on either systems or individuals. Reducing the proportion of people who die where they had not wanted to die is likely to require programs that address individuals, their society, and its cultural values, and the health system in which they reside. For this reason, we propose that the next step involve individual and community-level qualitative work to understand the causal mechanism behind place of death and identify factors amenable to reform.

	Acknowledgments

 
Members of this project team were supported by the Life Sciences Values and Society Program at the University of Michigan, the Robert Wood Johnson Clinical Scholars Program, The Robert Wood Johnson Faculty Scholars Program, the VA Health Services Research and Development Center of Excellence, and the Agency for Healthcare Research and Quality (grant KO8-HS0000002).

We are indebted to Susan Dorr Goold, Patrick Heagerty, Dick Hoskins, and Tom Koepsell for their encouragement and comments.

Human Participant Protection
The institutional review boards at the University of Washington and the University of Michigan reviewed, approved, and supervised the conduct of this study.

	Footnotes

 
Peer Reviewed

Contributors
M. J. Silveira and C. Feudtner originated and designed the study and co-wrote the article. M. J. Silveira obtained the data and, along with L. A. Copeland, refined and conducted the analyses. All authors contributed interpretation of the analyses and critical review and revision of article drafts.

Accepted for publication July 17, 2005.

	References

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION References

 
1. Townsend J, Frank AO, Fermont D, et al. Terminal cancer care and patients’ preference for place of death: a prospective study. BMJ. 1990;301:415–417.[Abstract/Free Full Text]

2. Fried TR, Bradley EH, Towle VR, Allore H. Understanding the treatment preferences of seriously ill patients. N Engl J Med. 2002;346:1061–1066.[Abstract/Free Full Text]

3. McCormick WC, Inui TS, Deyo RA, Wood RW. Long-term care preferences of hospitalized persons with AIDS. J Gen Intern Med. 1991;6:524–528.[Web of Science][Medline]

4. Committee on Care at the End of Life, Field MJ, Cassel CK, eds. Approaching Death: Improving Care at the End of Life. Washington, DC: National Academy Press; 1997.

5. Flory J, Yinong YX, Gurol I, Levinsky N, Ash A, Emanuel E. Place of death: US trends since 1980. Health Aff (Millwood). 2004;23:194–200.[Abstract/Free Full Text]

6. Wennberg JE, Freeman JL, Shelton RM, Bubolz TA. Hospital use and mortality among Medicare beneficiaries in Boston and New Haven. N Engl J Med. 1989; 321:1168–1173.[Abstract]

7. Dartmouth Medical School Center for the Evaluative Clinical Sciences. The Dartmouth Atlas of Health Care, 1998. Chicago, Ill: American Hospital Publishing, 1998.

8. Fisher ES, Wennberg JE, Stukel TA, et al. Associations among hospital capacity, utilization, and mortality of US Medicare beneficiaries, controlling for sociodemographic factors. Health Serv Res. 2000;34: 1351–1362.[Web of Science][Medline]

9. Polissar L, Severson RK, Brown NK. Factors affecting place of death in Washington State, 1968–1981. J Community Health. 1987;12:40–55.[CrossRef][Medline]

10. Pritchard RS, Fisher ES, Teno JM, et al. for the SUPPORT investigators. Influence of patient preferences and local health system characteristics on the place of death. Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatment. J Am Geriatr Soc. 1998;46:1242–1250.[Web of Science][Medline]

11. Fried TR, van Doorn C, O’Leary JR, Tinetti ME, Drickamer MA. Older persons’ preferences for site of terminal care. Ann Intern Med. 1999;131:109–112.[Abstract/Free Full Text]

12. De Vries RG. A Pleasing Birth: Midwives and Maternity Care in the Netherlands. Philadelphia, Pa: Temple University Press; 2004.

13. National Center for Health Statistics, Centers for Disease Control and Prevention. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM ). Washington, DC: American Medical Association; 1996. Available at: http://www.cdc.gov/nchs/about/otheract/icd9/abticd9.htm. Accessed March 19, 2006.

14. Wennberg J, Gittelsohn A. Small area variations in health care delivery. Science. 1973;182:1102–1108.[Abstract/Free Full Text]

15. Dartmouth Medical School Center for the Evaluative Clinical Sciences. The Dartmouth Atlas of Health Care, 1996. Chicago, Ill: American Hospital Publishing; 1996.

16. Emanuel LL, von Gunten CF, Ferris FD. Gaps in end-of-life care. Arch Fam Med. 2000;9:1176–1180.[Abstract/Free Full Text]

17. Tolle SW, Rosenfeld AG, Tilden VP, Park Y. Oregon’s low in-hospital death rates: what determines where people die and satisfaction with decisions on place of death? Ann Intern Med. 1999;130:681–685.[Abstract/Free Full Text]

18. Hansen SM, Tolle SW, Martin DP. Factors associated with lower rates of in-hospital death. J Palliat Med. 2002;5:677–685.[CrossRef][Medline]

19. Diez-Roux AV. Bringing context back into epidemiology: variables and fallacies in multilevel analysis. Am J Public Health. 1998;88:216–222.[Abstract/Free Full Text]

20. Grande GE, Addington-Hall JM, Todd CJ. Place of death and access to home care services: are certain patient groups at a disadvantage? Soc Sci Med. 1998; 47:565–579.[CrossRef][Web of Science][Medline]

21. Robinson WS. Ecological correlations and the behavior of individuals. Am Sociol Rev. 1950;15: 351–357.[Web of Science][Medline]

22. Simmons R, Bernstein S. Out-of-hospital births in Michigan, 1972–79: trends and implications for the safety of planned home deliveries. Public Health Rep. 1983;98:161–170.[Web of Science][Medline]

23. Hinds MW, Bergeisen GH, Allen DT. Neonatal outcome in planned v unplanned out-of-hospital births in Kentucky. JAMA. 1985;253:1578–1582.[Abstract/Free Full Text]

24. Davis-Floyd RE. The technocratic body: American childbirth as cultural expression. Soc Sci Med. 1994;38: 1125–1140.[Medline]

This article has been cited by other articles:

				J. M. Linton and C. Feudtner What Accounts for Differences or Disparities in Pediatric Palliative and End-of-Life Care? A Systematic Review Focusing on Possible Multilevel Mechanisms Pediatrics, September 1, 2008; 122(3): 574 - 582. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) All Versions of this Article: AJPH.2005.063057v1 96/7/1243    most recent Submit a response View Shopping Cart Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Silveira, M. J. Articles by Feudtner, C. Search for Related Content PubMed PubMed Citation Articles by Silveira, M. J. Articles by Feudtner, C. Related Collections Epidemiology Geography Health Policy Other Health Service Delivery