A Conceptual Multivariate Modelling Approach for Integrating Epidemiologic and Psychosocial Determinants to Examine the Epidemiology of Diseases in Under-served Populations

Abstract

Epidemiologic research involves the study of a complex set of host, environment and causative agent factors as these interact to impact health and diseases in any population. A conceptual multivariate modelling approach for integrating epidemiologic and psychosocial determinants to examine the epidemiology of chronic and infectious diseases in under-served populations in the USA was developed. Our approach relies upon systems analysis, i.e. integrating concepts and methods in epidemiology with mathematics and statistics along with computational methods and tools to rigorously examine the dynamics of diseases such as the Human Immunodeficiency Virus/Acquired Immunodeficiency Syndrome (HIV/AIDS) in a community. We used a multifactorial and quantitative epidemiologic (static) model that interrelates multiple determinants including biomedical, behavioral, and socioeconomic factors to analyze morbidity and mortality due to HIV/AIDS. The research involved participation of the community in the collection of socioeconomic, demographic, environmental, epidemiologic and biomedical data. In collaboration with Montgomery AIDS Outreach (MAO), a community based Organization in Montgomery, Alabama; blood samples were collected and tested using Orasure HIV testing kits to establish infection status with HIV/AIDS. Using these models, evaluations of various intervention scenarios with the objective of recommending effective strategies to minimize the risk of new HIV infections and/or manage existing diseases in a community can be generated.

Introduction

The Multifactorial Nature of the Health Disparity Problem

Epidemiologic modelling involves the study of a complex set of host, environment and causative agent factors as these interact to impact health and diseases in any population. Applying systems analysis methods in epidemiologic research to examine the health disparity disease challenges in the USA appears to provide a promising avenue for quantitatively deciphering the web of multiple determinants that have a direct or indirect influence on prevalence and incidence of diseases. Such an approach could provide the clues to devise intervention strategies for minimizing and eventually eliminating the disproportionate diseases burden that minority subpopulations face in the USA.

The Health Disparity Profile in the USA

Health disparities are defined as differences in the incidence, prevalence, mortality and burden of diseases and other adverse health conditions that exist among population groups. The Department of Health and Human Services (DHHS) has identified the six top priority areas for tackling the health disparity challenge [5], ([9], [10], [11]). These are: infant mortality, cancer, cardiovascular diseases, diabetes, HIV/AIDS infections and immunizations. There are stark reminders that the challenges posed by disease burden and health risks in minority communities in the USA are monumental. For example, although the Nation's infant mortality rate is down, the infant death rates among African Americans are still more than double that of whites [8]. Cancer is one of the most devastating health disparity diseases that afflict African Americans disproportionately ([1], [2], [7]).

The death rate for all cancers is 30 percent higher for African Americans than for whites; for prostate cancer, it is more than double that for whites. Cardiovascular diseases (CVD), primarily coronary heart disease and stroke, accounts for 32.4% of deaths among African American men and 41.6% of deaths among African American women ([3], [11], [13]). Hispanics living in the USA are almost twice as likely to die from diabetes as are non-Hispanic whites. With respect to HIV/AIDS in the USA, data from prevalence surveys and from HIV/AIDS case surveillance continue to reflect the disproportionate impact of the epidemic on racial/ethnic minority populations as well as women and children [4]. Researchers estimate that 240,000 - 325,000 African Americans (about 1 in 50 African American men and 1 in 60 African American women) are infected with HIV. From the beginning of the epidemic through December 2005, an estimated 211,559 African American with AIDS died [4].

Given the sober statistics given above, the most powerful nation in the world, with its wealth of resources and expertise and with the moral and ethical imperatives at stake cannot afford to ignore the health disparity dilemma. We propose that a systems based epidemiologic research could provide the framework for addressing health disparities in a multidisciplinary and integrative manner. We are interested in teasing out and quantifying the contributions of socio-behavioral, biomedical, environmental and nutritional factors to health disparities using quantitative, multifactorial and community-based epidemiologic research designs.

The specific objective

this study relies upon systems analysis for developing multivariate epidemiologic models that integrate multiple determinants (biomedical, behavioral, and socioeconomic factors) to analyze morbidity and mortality due to HIV/AIDS, and others in which racial and ethnic minorities experience serious disparities in health access and outcomes.

Hypothesis

Multifactorial and quantitative epidemiologic studies that rely upon systems analysis to interrelate (biomedical, behavioral, and socioeconomic factors) determinants can be developed to tease out the quantitative contributions of factors upon the health disparity problems under study.

Procedure

A Systems Analysis framework for representing health disparities and its determinants

Figure 1 provides a diagrammatic view of the multiple factors as these impact the six health disparity problems identified by DHHS. On the right hand side of the figure, summaries of interventions for mitigating these disparities are presented. On the left side are the major causal determinants that include: Race and Ethnicity, Income and Education, Location, Access to health care, Psychosocial, Environment, Nutrition, Food safety and Genetics.

Figure 1. Factors that affect Health Disparities and Strategies for interventions.

Procedure to collect the comprehensive multifactorial epidemiologic data

A Comprehensive Health Survey Questionnaire was developed using Figure 1 as a model. The study population was defined and an appropriate sample size to carry out a baseline survey in Alabama was established. Informed consent forms were developed to ensure privacy. Information collected from the community (via questionnaires), including data from medical records and death certificates were coded such that the names of individuals and addresses could not be identified or traced. Basically, all personal identifiers were stripped from the paper record as well as from the electronic data in the study database.

Results

An integrated database that includes psychosocial, environmental and biomedical data directly from the study communities was developed using FileMaker Pro software (FileMaker Pro Inc.). Working with a multidisciplinary team of epidemiologists, public health experts and a social scientist, a comprehensive questionnaire was developed and data collected to quantify the impact of psychosocial, epidemiologic, health care and delivery and related determinants of HIV/AIDS. Specific multi-factorial data on behaviors that could be associated with HIV/AIDS transmissions in the Black Belt Counties (BBC) of Alabama were also collected.

We have collected 800 questionnaires on health disparities from four randomly selected BBC in Alabama. The data has been entered into FileMaker Pro database and statistical analysis is underway using StatView and SAS (SAS Inc.). In collaboration with Montgomery Aids Outreach (MAO) Inc., 266 questionnaires from HIV positive clients were collected. Two hundred (200) questionnaires from HIV/AIDS negative clients were collected. OraSure HIV testing kits were used to establish the HIV negative status from blood samples.

A sizeable amount of data has now been collected and multivariate statistical analysis and static modeling is underway in collaboration with a mathematician. The statistical analyses underway include descriptive statistics and exploratory analysis including associations and correlations. The most rigorous analyses involve multiple regression, stepwise regression, logistic regression and discriminant analyses. Multiple regression involving step-wise regression and logistic regression will be used to evaluate the quantitative relationships between sets of independent and dependent variables. Disciminant analysis will be used to classify determinants into at least two groups; those that lead to high levels of disease burden and those that are not so.

Mathematical Formulation

By a multivariate model, we mean a model for a multivariate response vector Y = (Y₁; : : : ; Y_d) with covariate vector x. The parameters are either univariate parameters or dependence parameters, and the approach is aimed primarily at a multivariate non normal response. For a sample of size n, the multivariate data has the form (Y_i; x_i) for the i^th subject, with Y_i = (Y_i1; : : : ; Y_id), where d is the dimension of the response vector. If the Y_ij 's are repeated measures or observed sequentially in time, then more generally we could have a time varying or margin dependent covariate vector. The set of parameters of the model are estimated through a (nonlinear) system of estimating equations, with each estimating equation being a score function (partial derivative of log likelihood) from some marginal distribution of the multivariate model. For instance corresponding to Figure 1 for multiple regression:

$$y_{ij}=\sum _{i=0,k}{a}_{ij}{x}_{ij}+{\epsilon }_j$$

with i= 0, 1, 2, …, k (the i's are factors that contribute to disparities in health; like: demography, psycho-social factors, environmental factors, etc.) and j = 1, 2, 3 …, n (the j's are health disparities; like: Infant mortality, cancer, etc.) and where k is the number of parameters that contribute to disparities in health and n is the number of health disparities observations. see Figure 1. a_ij is the factor loadings. In matrix notation it can be represented as follows:

$$\text{Y}=\text{XA}+\mathbf{\epsilon }\text{where}\text{Y}\sim (\text{XA},{\mathbf{\sigma }}^2\mathbf{\text{I}}),\mathbf{\epsilon }\sim (0,{\mathbf{\sigma }}^2),\text{and}\mathbf{\text{I}}\text{is a}k\times k\text{unit matrix}.$$

In subsequent publications the detailed results of these analyses will be presented.

Discussion

Strategies for reducing health disparities

Create health disparity studies that integrate biomedical, nutritional and socio-behavioral research based in a systems approach

Integrative, multidisciplinary and interdisciplinary team approaches could be applied in every facet of studies for addressing health disparities. This includes a team approach to research, education and training and outreach activities. More specifically, the training and education should be focused on integrative approaches that interrelate multiple disciplines so as to solve societal problems as opposed to esoteric research directed at understanding nature.

Intervention research should be focused on the whole system (whole family and the community), emphasizing the “one medicine/one health approach”. The systems approach to education, research and service builds and promotes teamwork, integrates disciplines, organizations and resources. It recognizes and interrelates the contributions and engagement of federal agencies, the private sector, state and local agencies and the local community as members of a team. Another key point is to integrate biomedical with socioeconomic and behavioral factors. We should be able to tease out and quantify the quantitative and qualitative contributions of socio-behavioral factors on the burden of health disparities. The organization and structural components that need to be accounted for in the systems based studies of health disparities will include developing appropriate databases and links that integrate with the different data sources that already exist. If health disparity related data and information systems are integrated and made accessible to researchers, significant and targeted analyses can be performed and improvements made in intervention strategies.

Access to health care is critical

Factors including shortage of health care providers or lack of facilities, financial barriers (having no health insurance or being underinsured), structural barriers (such as lack of facilities or health care professionals nearby), and personal barriers (cultural differences, language differences, not knowing what to do, or environmental challenges for people with disabilities) are critical factors that contribute to health disparities ([6]). Sound ethical practices coupled with cultural competency and sensitivity in understanding how best to serve the community while providing training and/or when carrying out research to create and maintain trust is also important. Finally, the moral and ethical arguments for eliminating health disparities are quite evident and universal in appeal. Just as importantly, eliminating health disparities and therefore diseases in minority populations has added benefit of improving the overall health of the whole nation including the majority population. It also provides an economic benefit by reducing the overall per capita investment in health care delivery services.