Volume 25
Number 3/4
2004

[Table of Contents]

Inequalities in health and health services delivery: A multilevel study of primary care and hypertension control

Paul J Veugelers, Alexandra M Yip and Frederick Burge

Abstract

Delivery of health services is an important determinant of health. Restricted availability and access may result in health inequalities. To determine the extent of geographic variation in the delivery of health services and its effect on the health of community residents in terms of under-diagnosis and under-treatment of hypertension, we carried out a multilevel study of participants in the 1995 Nova Scotia Heart Health Survey (n=3,094). We used individual level survey data and health status measurements linked to geographical level information to examine the importance of adequate delivery of health services to the diagnosis and treatment of hypertension in the universal health care setting of the province of Nova Scotia. The delivery of primary care services across Nova Scotia varied moderately with physician visit rates ranging from 3.3 to 5.5 visits per resident per year. There were neither substantial nor statistically significant differences in the diagnosis and treatment of hypertension among residents of communities varying in the delivery of health services. We concluded that a geographic variation in the delivery of primary care services is a public health concern that is not consistent with the objectives of universal coverage of health services; however, it was not confirmed to result in health inequalities.

Key words: community health; epidemiology; health services, hypertension; medically underserved area; primary health care; small area comparisons

Introduction

Investigations in the past decade have revealed that contextual factors such as income inequality, neighborhood deprivation, and unemployment rate may affect the health of individuals.^1,2 Studies using multilevel methods have been increasingly employed to examine the independent effect of these contextual factors as distinct from the more widely investigated individual-level risk factors such as age, sex, and lifestyle risk factors. These multilevel studies have shown, although not consistently, that neighbourhood factors, in particular socio-economic characteristics of residential areas, affect health independent of individual-level risk factors.^3-9 Various mechanisms within communities and neighbourhoods have been suggested, including the delivery of health services.^2,7,8,10

Geographic variation in the delivery of health services has been reported in various Western countries, including those with universal access health care systems that remove the basic financial barrier to accessing care.^11-15 Local policies, staffing shortages, physician practice styles, transportation, and health-seeking behaviour may all contribute to this geographic variation in availability and access to care and potentially affect the health of the residents of these communities and neighbourhoods.¹⁶ State, metropolitan, and county level comparisons revealed that lower physician-to-patient ratios were associated with poorer health, emphasizing the potential that inadequate delivery of health services may universally result in inequalities in health.^17-22

To further our understanding of the potential health effects of inequalities in the delivery of health services, we conducted a multilevel study examining whether the diagnosis and management of hypertension were hampered in geographies with restricted availability and access to primary health care. More specifically, we sought to determine whether individuals living in areas with restricted availability and access to primary health care are 1) less likely to be diagnosed with hypertension; 2) more likely to have undiagnosed hypertension; 3) less likely to be on medication for hypertension; and 4) more likely to have untreated hypertension.

Methods

Individual-level information

As our primary objective is to study the independent importance of the context of health services delivery, it is important to adjust for other factors that affect the health of individuals and that may include both individual and other contextual determinants. Individual-level determinants were taken from the 1995 Nova Scotia Heart Health Survey, conducted among non-institutionalized adults aged 18 years and older, stratified by age, sex, and four health regions. The population-based sampling frame consisted of 5,578 people selected from the provincial health registry, of whom 83% were located. Three percent were screened out because of pregnancy, breast-feeding, or mental or physical health precluding participation. Of those remaining, 3,227 (72%) completed an in-home interview, conducted by a trained public health nurse, which consisted of a comprehensive questionnaire on risk factors and two measurements of blood pressure, taken at the beginning and end of the session. Of those interviewed, 2,653 (82%) attended a clinic session, which included measurement of height and weight for the calculation of body mass index (BMI).²³

Individual-level confounders considered include age, sex, smoking status, BMI, presence of selected chronic conditions (diabetes, previous myocardial infarction, peripheral vascular disease, and previous stroke), household income, educational attainment, and self-rated health. Age was analyzed as a categorical variable broken into ten-year age groups. Gross household income on the questionnaire was coded using a response tree as follows: Participants were first asked if income totalled less than or greater than $20,000. If the latter, participants were then asked for further refinement, indicating whether income totalled less than or greater than $40,000. Responders to the first part of the question may have chosen not to elaborate upon the second part. Therefore, some of the responses were grouped ">= $20,000", an amount that overlaps with the groups "$20,000-39,999" and ">= $40,000", although participants in each of these groups are distinct. Educational attainment was categorized by highest diploma or degree obtained, as follows: "did not complete secondary", "completed secondary", and "completed post-secondary diploma or degree".

We excluded from the analysis 133 participants (4.1%) with missing information on smoking status, chronic conditions, education, self-rated health, or the outcomes of interest. As BMI and household income had large proportions of missing data (BMI was available only for those who attended the clinic session, while income was an elective question), these participants were retained in the analysis as a missing category for each variable.

Contextual-level measures

In previous work, we subdivided the province of Nova Scotia (populated by approximately 940,000 residents) into 64 functional geographies or contexts, based on Census Consolidated Subdivisions in rural areas and on Enumeration Area groupings ("neighbourhoods") in urban areas.²⁴ In the present study, at the level of these 64 geographies, we considered three contextual covariates: 1) local socioeconomic conditions, 2) local health status, and 3) the covariate of interest - access and availability of health services. Local socioeconomic status was considered as a potential contextual-level confounder because the importance of its contextual effect on the health of individuals is becoming established and its substantial variation throughout the 64 geographies has been demonstrated.²⁴ The measure of local socio-economic status, average household income of these geographies, was drawn from the 1996 Canada Census. Also, because local health will determine local demand for health services, the second contextual covariate - local health status - was considered as a potential contextual confounder. The measure of local health, local life expectancy for the year 1995, was calculated from mortality as reported in provincial vital statistics and from projected population counts based on the Canada Census.^25,26

In state, metropolitan and county level comparisons, researchers have used physician-to-patient ratios as a proxy for access and availability of primary health care.^17-22 As this is a small area comparison and some physicians may have more than one practice in different areas, or may also be working part time or long hours, we evaluate the number of physician visits per resident rather than the number of physicians per resident.²⁷ The number of physician visits per resident was derived from 1995 physician claims on the provincial health insurance plan which captures all health services of Nova Scotia residents.²⁷ To overcome differences in age and gender throughout the 64 geographies, we standardized the number of primary care physician visits per resident in this respect. For each of the 64 geographies, we also generated empirical Bayesian estimates for number of primary care physician visits and for life expectancy, to overcome over-dispersion - the phenomenon of unstable small area estimates resulting from varying population sizes.²⁸ Hierarchical or multilevel methods generate such empirical Bayesian estimates and allow for spatial smoothing.^28,29 Here we considered a multilevel model whereby information of the 64 geographies and their adjacent geographies were pooled (level 1) resulting in robust estimates of the geography-specific life expectancies (level 2). In this model, at level 1, we weighted proportional to the inverse of the standard error of the life expectancy estimates. This methodology is described in more detail elsewhere.²⁴

Data linkage

The linkage and analysis of data for this study were approved by the Health Sciences Human Research Ethics Board of Dalhousie University, Halifax, Nova Scotia, Canada. The original sampling frame of the survey contained residential postal codes, allowing linkage to one of the 64 geographies and its information on local socio-economic status, local life expectancy, and local health care use.³⁰ Nine survey participants (0.3%), however, did not have valid postal codes and therefore could not be linked. They were excluded from the analysis, resulting in an effective study sample of 3,094 participants.

Statistical approaches

The independent importance of contextual characteristics to individuals' health was examined using multilevel logistic regression and hierarchical software.²⁹ The contextual characteristics, local socio-economic conditions, local health status and the covariate of interest, access and availability of health services, were considered as level 2 covariates. Individual characteristics and health outcomes were considered as level 1 variables. We examined the following four health outcomes: 1) previously diagnosed hypertension, 2) presence of undiagnosed hypertension, 3) self-reported use of antihypertensive medication, and 4) presence of untreated hypertension. The presence of hypertension was defined as an average measured diastolic blood pressure of 90 mmHg or greater, in keeping with the Canadian clinical practice guidelines in effect at the time of the survey.³¹

To demonstrate a 25% reduction in the number of individuals diagnosed with hypertension in geographies having the lowest levels of primary health care use (more than 5% below provincial average), the statistical power needed was calculated to be 94% for an alpha level of 0.05. Statistical analyses were conducted using S-Plus and HLM.²⁹

Results

Table 1 presents the individual and contextual characteristics of the 3,094 Nova Scotia Heart Health Survey participants who had complete information and were included in the present analyses. Between 5% and 6% of participants reported having been previously diagnosed with diabetes, myocardial infarction, or peripheral vascular disease, and 2.4% with stroke. Of the 3,094 participants, 868 (28%) had been previously diagnosed with hypertension. Of the remaining 2,226 participants, 148 (6.6%) had undiagnosed hypertension, that is, on the day of the participants' interview their average measured diastolic blood pressure was 90 mmHg or greater. Among all 3,094 participants, there were 457 (14.8%) on antihypertensive medication at the time of interview. Of the remaining 2,637 participants, 228 (8.6%) had untreated hypertension, that is, they were not currently taking medication and their diastolic blood pressure was 90 mmHg or greater.

The average number of visits to primary care physicians across the 64 Nova Scotia geographies ranged from 3.3 to 5.5 visits per resident per year (Figure 1). Of all participants, 19 percent (588) lived in areas where the age-sex standardized estimates of primary care physician use was less than 95% of the provincial average, 59.5% (1,842) lived in near-average primary care physician use areas, and 21.5% (664) lived in areas where primary care physician use was estimated to be more than 5% above the provincial average (Table 1).

FIGURE 1
Availability and access to primary care physicians by geography within Nova Scotia, Canada

TABLE 1
Characteristics of Nova Scotia Heart-Health Survey participants (n = 3,094)

Table 2 presents the independent associations of the contextual factors with the health outcomes of interest, adjusted for the contextual factors and individual differences with respect to age, sex, smoking, BMI, chronic conditions, household income, educational attainment, and self-rated health. The likelihood of being diagnosed with hypertension was neither substantially nor significantly different among geographies that varied with respect to income, life expectancy, or primary care services use: Relative to areas with the lowest levels of services use (less than 95% of the provincial average), the odds ratios for areas with middle levels (within 5% of the provincial average) or increased levels of services use (more than 5% higher than the provincial average) were only marginally elevated. Undiagnosed hypertension was less likely to occur in high-income areas (average income $45,000), areas with near average and high life expectancy, and areas with reduced use of primary care. However, these differences were also not statistically significant. Moreover, the likelihood of using antihypertensive medication or having untreated hypertension were neither substantially nor significantly different among residents of areas that varied with respect to income, life expectancy, or primary care services use (Table 2).

The associations presented in Table 2 were not substantially different from estimates of models in which we had adjusted for individual level confounders only. We also considered covariates as tertiles and as continuous variables rather than categorical variables. Model modifications, in this regard, did not substantially alter the observed associations between local health services use and the outcomes of interest.

TABLE 2
Contextual differences in hypertension diagnosis and management independently of individual-level and other contextual-level differences

Discussion

Cardiovascular disease is the leading cause of death in Canada and in most Western countries. Hypertension is an important risk factor for cardiovascular disease, but, unfortunately, awareness and control of this condition remains relatively low, making it an important public health concern.^32,33 Similarly, a shortage of primary care practitioners is a major public health concern in many Western countries - particularly in rural and deprived areas.^34-36 Availability and access to primary care may affect the diagnosis and management of hypertension. We therefore hypothesized that residents of areas with restricted availability and access would be 1) less likely to be diagnosed with hypertension, 2) more likely to have undiagnosed hypertension, 3) less likely to be on antihypertensive medication, and 4) more likely to have untreated hypertension. The present study could not confirm any of these hypotheses.

Although geographic factors for cardiovascular outcomes have been addressed in various studies using individual level information,^3,37-40 the importance of adequate delivery of primary health care has only been addressed in studies using geographically aggregated information.^17-22 With respect to the latter, in the US, state and metropolitan level comparisons revealed higher life expectancy and reduced mortality in geographies with a high primary care physician-to-population ratio independent of socio-economic differences.^18,20 Based on these observations, the authors had suggested that an increase in the number of primary care physicians would be an effective and feasible means of addressing adequacy of primary care availability and access in order to improve population health.²⁰ Gulliford observed a similar association in England, but indicated that 'in England the association between supply of primary care doctors and mortality might be largely explained by confounding'.²¹ Unlike the US, the UK and Canada have universal access health care systems through which primary care availability and access to all residents are addressed. As a consequence, local differences in the use of primary care services are expected to be smaller in settings with universal health care and likely account for the varying study findings. This seems consistent with findings of an international comparison within the Organization for Economic Cooperation and Development (OECD) countries demonstrating better health outcomes in countries with better primary care systems.⁴¹

To overcome ecological bias in studies of primary health care, Gulliford had suggested the conduct of studies with 'data collected at the individual as well as area level'.²¹ The present study did collect data at both levels and analyzed the data as distinct individual level and contextual level factors. The contextual level factor of interest, the delivery of primary health care services, may affect the health of individuals in various ways. For example, residents of geographies with serious shortages of primary care physicians may experience difficulties in finding a regular doctor, difficulties in scheduling appointments, shorter consultation times, and greater emphasis on acute rather than preventive care. Other factors that may vary geographically and affect the health of individuals include local policies, staffing shortages of primary care personnel other than physicians, practice styles, transportation, and health-seeking behaviour.¹⁶ The present study also differs from previous work in that it compared small areas to serve the informational needs of health policy makers operating at a community or neighbourhood level.²⁴ Because physicians may have more than one practice or work part-time or long hours, the evaluation of physician visit rates may provide more accuracy in small area comparisons than physician-to-patient ratios that have been applied in comparisons at the state, metropolitan, or district health authority level.^17-22

Of the 64 geographies, 28 deviated more than 5% and six deviated more than 10% from the provincial average. We are not aware of any other small area studies of physician visit rates as a point of reference for our observations, which hampers us in our judgment of whether the observed variation is modest or large. We expressed geographic variation in terms of percentage deviation from the provincial average. One may perceive the percentage deviation as relatively modest; however, in terms of absolute numbers of health care services and costs they represent large differences. In addition, even modest geographic variation in family physician services use may result in a disproportionately larger variation in preventive health services as physicians in understaffed areas may aim to maintain the level of acute services and selectively reduce the number of preventive services, such as the management of hypertension. The present study demonstrated that the latter is not a concern. The present study also provides a point of reference for future studies on geographic variation in health services and their impact on health. Such future analyses may include an updated analysis in Nova Scotia: This is particularly relevant as understaffing of primary care personnel has worsened since 1995. These analyses may also examine other health outcomes amenable to primary care intervention, such as participation in breast, cervical, and prostate cancer screening. Furthermore, future analyses should include the evaluation of the health impact of geographic differences in availability and access to specialist and hospital services. This is pertinent as we previously demonstrated that geographic variation in the use of specialist and hospital services exceeded that for use of family physicians,²⁷ and that this variation has a substantial potential to affect health outcomes among residents.⁴² All these future studies may follow the multilevel approaches as applied in the present study.

In summary, we observed moderate geographic variation in availability and access to primary care in Nova Scotia, but we did not confirm consequent disparities in the diagnosis and management of hypertension among provincial residents. Nevertheless, variation in the delivery of health services does not appear consistent with the objectives of universal coverage of health services.

Acknowledgements

The authors gratefully acknowledge the contributions of Dr. Brian Ashton and Dr. Jeffrey Hoch. The results and conclusions presented are those of the authors. No official endorsement by the Nova Scotia Department of Health is intended or should be inferred. Data access and linkage were provided by the Population Health Research Unit, Department of Community Health and Epidemiology, Dalhousie University.

Financial support for the present study is provided through funding by the Canada Foundation for Innovation, the Nova Scotia Health Research Foundation and a Canadian Institutes of Health Research Investigator Award to Dr. Veugelers.

References

1. Mackenbach JP. Income inequality and population health. Br Med J 2002;324:1-2.
2. Ross NA, Wolfson MC, Dunn JR, Berthelot GM, Kaplan GA, Lynch JW. Relation between income inequality and mortality in Canada and in the United States: Cross sectional assessment using census data and vital statistics. Br Med J 2000;320:898-902.
3. Diez-Roux, AV, Nieto FJ, Muntaner C, Tyroler HA, Comstock GW, Shahar E et al. Neighborhood environments and coronary heart disease: a multilevel analysis. Am J Epidemiol 1997;146:48-63.
4. O'Campo P, Xue X, Wang MC, O'Brien Caughy M. Neighborhood risk factors for low birthweight in Baltimore: A multilevel analysis. Am J Public Health 1997;87:1113-8.
5. Kennedy BP, Kawachi I, Glass R, Prothrow-Stith D. Income distribution, socioeconomic status, and self rated health in the United States: Multilevel analysis. Br Med J 1998;317:917-21.
6. Reijneveld SA, Schene AH. Higher prevalence of mental disorders in socioeconomically deprived urban areas in the Netherlands: Community or personal disadvantage? J Epidemiol Community Health 1998;52:2-7.
7. Yen IH, Kaplan GA. Neighborhood social environment and risk of death: Multilevel evidence from the Alameda County Study. 1999 Am J Epidemiol 1999;149:898-907.
8. Veugelers PJ, Yip AM, Kephart G. Proximate and contextual socioeconomic determinants of mortality: Multilevel approaches in a setting with universal health care coverage. Am J Epidemiol 2001;154:725-32.
9. Lochner K, Pamuk E, Makuc D, Kennedy BP, Kawachi I. State-level income inequality and individual mortality risk: A prospective, multilevel study. Am J Public Health 2001;91: 85-391.
10. Kawachi I, Subramanian SV, Almeida-Filho N. A glossary for health inequalities. J Epidemiol Community Health 2002;56:647-52.
11. Politzer RM, Harris DL, Gaston MH, Mullan F. Primary care physician supply and the medically underserved. A status report and recommendations. JAMA 1991;266:104-9.
12. Roos NP, Fransoo R, Carriere KC, Frohlich N, Bogdanovic B, Kirk P et al. Needs-based planning: the case of Manitoba. CMAJ 1997; 157:1227-8.
13. Gravelle H, Sutton M. Trends in geographical inequalities in provision of general practitioners in England and Wales. Lancet 1998; 352: 1910.
14. Gravelle H, Sutton M. Inequality in the geographical distribution of general practitioners in England and Wales 1974-1995. J Health Serv Res Policy 2001;6:6-13.
15. Johnston G, Wilkinson D. Increasingly inequitable distribution of general practitioners in Australia, 1986-96. Aust N Zealand J Public Health 2001;25:66-70.
16. Field KS, Briggs DJ. Socio-economic and locational determinants of accessibility and utilization of primary health-care. Health Soc Care Community 2001;9:294-308.
17. Vogel RL, Ackermann RJ. Is primary care physician supply correlated with health outcomes? Int J Health Serv 1998;28:183-96.
18. Shi L, Starfield B, Kennedy B, Kawachi I. Income inequality, primary care, and health indicators. J Fam Pract 1999;48:275-84.
19. Roetzheim RG, Gonzalez EC, Ramirez A, Campbell R, van Durme DJ. Primary care physician supply and colorectal cancer. J Fam Pract. 2001;50:1027-31.
20. Shi L, Starfield B. The effect of primary care physician supply and income inequality on mortality among blacks and whites in US metropolitan areas. Am J Public Health 2001;91:1246-50.
21. Gulliford MC. Availability of primary care doctors and population health in England: Is there an association? J Public Health Med 2002;24: 252-4.
22. Campbell RJ, Ramirez AM, Perez K, Roetzheim RG. Cervical cancer rates and the supply of primary care physicians in Florida. Fam Med 2003;35:60-4.
23. Nova Scotia Department of Health & Heart Health Nova Scotia. The Nova Scotia Health Survey 1995. Province of Nova Scotia & Heart Health Nova Scotia, 1996.
24. Veugelers PJ, Hornibrook S. Small area comparisons of health: Applications for policy makers and challenges for researchers. Chronic Dis Can 2002;23:100-10.
25. Chiang CL. The life table and its construction: Introduction to stochastic process in Biostatistics. New York: John Wiley & Sons, Inc, 1968.
26. Veugelers PJ, Guernsey JR. Health deficiencies in Cape Breton County, Nova Scotia, Canada, 1950-1995. Epidemiology 1999;10:495-9.
27. Veugelers PJ, Yip AM, Elliott D. Geographic variation in health services use in Nova Scotia. Chronic Dis Can 2003;23:100-10.
28. Best, N. Bayesian ecological modelling. In: Lawson A, Biggeri A, Böhning D, Lesaffre E, Viel J-F, Bertollini R. Disease mapping and risk assessment for public health. John Wiley & Sons. Toronto, 1999.
29. Raudenbush S, Bryk A, Cheong YF, Congdon R. Hierarchical linear and nonlinear modeling; Chapter 1: Conceptual and statistical background for two-level models. Lincolnwood (Il): Scientific Software International, Inc. 2000.
30. Wilkins R. Geocodes/PCCF version 2 user's guide. Automated geographic coding based on the Statistics Canada postal code conversion file. Ottawa: Health Statistics Division, Statistics Canada, 1997.
31. Logan AG. Screening for hypertension in young and middle-aged adults. In: Canadian task force on the periodic health examination. The canadian guide to clinical preventive health care, pp.636-648. Ottawa: Minister of Supply and Services Canada 1994.
32. Campbell NRC, Fodor JG, Chockalingam A. Hypertension recommendations: Are they relevant to public health? Can J Public Health 2001;92:245-6.
33. Joffres MR, Ghadirian P, Fodor JG, Petrasovits A, Chockalingam A, Hamet P. Awareness, treatment, and control of hypertension in Canada. Am J Hypertension 1997;10:1097-102.
34. Knapp KK, Paavola FG, Maine LL, Sorofman B, Politzer RM. Availability of primary care providers and pharmacists in the United States. J Am Pharmaceutical Association 1999; 39:127-35.
35. Stearns JA, Stearns MA, Glasser M, Londo RA. Illinois RMED: A comprehensive program to improve the supply of rural family physicians. Fam Med 2000;32:17-21.
36. Talbot Y, Fuller-Thomson E, Tudiver F, Habib Y, McIsaac WJ. Canadians without regular medical doctors. Who are they? Can Fam Physician 2001;47:58-64.
37. Alter DA, Naylor CD, Austin P, Tu JV. Effects of socioeconomic status on access to invasive cardiac procedures and on mortality after acute myocardial infarction. N Engl J Med 1999;341: 1359-67.
38. Tydén P, Engström G, Hansen O, Hedblad B, Janzon L. Geographical pattern of female deaths from myocardial infarction in an urban population: Fatal outcome out-of-hospital related to socio-economic deprivation. J Intern Med 2002;250:201-7.
39. Tydén P, Hansen O, Engstrom G, Hedblad B, Janzon L. Myocardial infarction in an urban population: worse long term prognosis for patients from less affluent residential areas. J Epidemiol Community Health 2002;56:785-90.
40. Alter DA, Naylor CD, Austin PC, Chan BT, Tu JV. Geography and service supply do not explain socioeconomic gradients in angiography use after acute myocardial infarction. CMAJ 2003;168:261-4.
41. Macinko J, Starfield B, Shi L. The contribution of primary care systems to health outcomes within Organization for Economic Cooperation and Development (OECD) countries, 1970- 1998. Health Serv Res 2003;38:831-65.
42. Veugelers PJ, Yip AM. Socio-economic disparities in health care use: does universal coverage reduce inequalities in health? J Epidemiol Community Health 2003;57:424-8.

Author References

Paul J Veugelers, Department of Community Health and Epidemiology, Faculty of Medicine, Dalhousie University, and Department of Public Health Sciences, Faculty of Medicine & Dentistry, University of Alberta, Edmonton, Alberta, Canada

Alexandra M Yip, Department of Community Health and Epidemiology, Faculty of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada

Frederick Burge, Department of Family Medicine, Faculty of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada

Correspondence: PJ Veugelers, Department of Public Health Sciences, Faculty of Medicine & Dentistry, Room 13-106D, Clinical Sciences Building, University of Alberta, Edmonton, Alberta, Canada T6G 2G3; Fax: (780) 492-0364; E-mail: paul.veugelers@ualberta.ca

[Previous] [Table of Contents] [Next]