Exploring Health Disparities in Integrated Communities
The Southwest Baltimore Health Survey
PI: Thomas A. LaVeist, PhD
Department of Health Policy and Management
Johns Hopkins Bloomberg School of Public Health
(revised April 2, 2007)
Click here to read "Exploring Health Disparities in Integrated Communities: Overview of the EHDIC Study"
The purpose of this study is to investigate the level and cause of health disparities in a racially integrated community in southwest Baltimore. This community is comprised of U.S. Census tracts 1902 and 1903.
- To determine hypertension prevalence among adult residents of these census tracts
- To determine healthcare utilization patterns and to correlate psychosocial factors
- To conduct a community health needs assessment
Researchers who specialize in minority health have to traverse a series of compromises in conducting their research. These compromises stem primarily from the problem of determining whether to use a race comparative or race specific study design. Race/ethnic comparative data sources often have small numbers of minority respondents compared to whites, and usually do not include the range of variables that are of greatest interest to understanding social and behavioral aspects of minority health. However, race/ethnic specific data sources (such as the National Panel of Black Americans, the Three Generation Study of Mexican Americans, Cardiovascular Disease in American Indians, or Honolulu-Asia Aging Study) lack comparison groups that aid in the assessment of the generalizability of findings.
Additionally, in either case, data sources often do not contain sufficient numbers of low-income urban Caucasians or middle-income African Americans and other minorities to make true racial comparisons. As a result, much of what is published on race disparities in health are based on samples of low-income African Americans compared with middle-income whites.
There is a long history of conducting large-scale epidemiologic longitudinal studies. Some of the best-known studies of this type are: Framingham, MA; Alameda County, CA Study; the Tecumseh, MI; and the Washington County, MD study. These large-scale epidemiologic studies have produced much of what health researchers understand about risk factors for disease and premature death. Yet, these very influential studies tend to be located in rural areas or small towns and do not include African Americans. Or, if African Americans are included, the numbers of respondents are typically insufficient to report in most analyses.
The Jackson Heart Study, supported by the National Heart Lung and Blood Institute is a large-scale study of African Americans, but here again the lack of a comparison group makes it impossible to examine racial disparities (Sempos et al. 1999). Other studies (such as the North Manhattan Study, Texas Community Surveillance, Greater Cincinnati/North Kentucky Stroke Study) have a better racial distribution, but their focus on specific health conditions and lack of social and community indicators make them less useful for the study of racial disparities.
The Principal Investigator of our application conducted an assessment (LaVeist, 1995) of federal data sets to determine the ability of the federal data portfolio to conduct research on minority health. This study found that few data sources were able to support the type of studies that are needed to explore the etiology of race disparities in health. As a result of inadequate data sources many published studies suffer mischaracterizations race associated differences in health outcomes as resulting from biological factors without specifying such differences, or confound race with socioeconomic status (LaVeist, 1994).
Advancing minority health research will require the creation of new data sources that have the following characteristics: longitudinal, socio-economically equivalent across ethnic groups, set in an urban and/or suburban area (which is where most Americans live), allows for racial comparisons, and includes a wide range of behavioral, social and community as well as biological variables.
To address these limitations, we propose to do a comprehensive study of southwest Baltimore, the Southwest Baltimore Health Survey. Southwest Baltimore is a racially integrated community where blacks and whites share the same environment and a low socio-economic status. This study should be able to address health disparities without the confounding factors of class, income, and environment. This study will serve as an important resource that can be used to develop sub-studies conducted by faculty, fellows and students of the Hopkins Center for Health disparities Solutions. This multi-disciplinary study would combine behavioral, social and biological measures to assess their combined and individual effects on morbidity and mortality.
We will attempt to enroll every adult (3,555) resident of the two census tracts into the study. We estimate that we will be able to successfully enroll 40% (1,422) of the targeted adults after adjusting for refusals and those who we cannot locate. The field period will begin with a health fair to “kick-off” the new initiative and introduce it to the community. The health fair will take place on a weekend in April, which is National Minority Health Month. To promote awareness of the project, we will distribute flyers, T-shirts, and maintain a website devoted to the project. To accomplish the goal of enrolling, screening and interviewing respondents within a six-month field period we will use three methods of recruitment. We will:
- Maintain a centralized location with “drop-in” hours for residents to drop-in and participate in the study
- Conduct door-to-door screenings, setting up appointments if necessary
- Conduct one health fair per month to promote awareness of the project and conduct patient enrollment on-site
We estimate that these strategies will allow us to reach up to 90% of the adults in the census tract.
Interviews per day
Days per week
As we are attempting to obtain a census of the adults in the two target tracts, we will not be sampling. All adults, age 18 and above, are eligible to participate. There will be no exclusions for gender, race, ethnicity, etc. According to the 2000 US Census, there are 3,555 adults living in the target area. We will attempt to enroll every adult. However, we anticipate obtaining a 40% completion rate.
Respondents enrolled in the study will be screened for blood pressure and reply to a structured questionnaire. These activities will be conducted using the following format:
- Blood pressure will be measured while the respondent is seated and will occur after the respondent has consented and before the questionnaire is administered.
- Questionnaire administration will occur after blood pressure is taken. Respondents will be asked to complete a questionnaire consisting of validated questions designed to assess health behavior, health services utilization, attitudes values and beliefs, quality of life, mental health (including depression), demographic variables, a medical history and utilization of over-the-counter and prescription drugs. The questionnaire will be administered in a private location – either in the respondent’s home, at the “drop-in” location (space we will rent within the community) or some other appropriate and discrete location if the respondent would prefer. We will catalogue all medications the patient is taking; we will request that patients bring all of the medications they are taking along with them. The interviewer will record the names of each medication. That list will, then be coded to determine the class of each drug.
The entire encounter is estimated to average 45 minutes. We believe that study participants will tolerate a 45-minute interview. This is not out of line with previous studies conducted by the PI. We will not conduct proxy interviews.
The questionnaire consists of established batteries. The only sensitive questions are those regarding use of alcohol. The complete questionnaire is attached.
The study is largely exploratory. As such there are no specific hypotheses, rather there are research questions. To illustrate the type of data analysis that will be conducted I will pose an example of a research question that these data will be able to address and then demonstrate the analytic techniques that would be used to address it. I will then discuss sample size.
One set of analysis might example race differences in the relationship between having a usual source of care and having undiagnosed hypertension.
Hypothesis: Not having a usual source of care will be a stronger predictor of undiagnosed hypertension for African American study participants then it will be for whites. The hypothesis posits that there will be a Black/White difference in undiagnosed hypertension (UDH) among patients who have a usual source of care. One might address this question simply by adding a multiplicative interaction term to a regression model. However, I will describe an alternative approach. The technique to be used to test this hypothesis is similar to the one used in analysis previously published by Dr. LaVeist. The hypothesis will be tested by specifying models in race specific sub-samples and testing for statistically significant differences in the parameters. We will then conduct a formal test for a statistically significant difference in the coefficient for not having a usual source of care (b1) between the African American and White models:
(3) UDH = bb0 + bb1USC bbiXi + ei
(4) UDH = bw0 + bw1USC bwi Xi + ei
Where: UDH is undiagnosed care (binary variable), bb0 is the model intercept for the black sub-sample, bb1 is the parameter estimate having a usual source of care for the black sub-sample, USC is a binary variable indicating whether or not the respondent has a usual source of care, bbi Xi represents a series of covariates and their associated parameter estimates (anticipated confounders or predictors of UDH). And, bw0 is the model intercept for the white sub-sample, bw1 is the parameter estimate having a usual source of care for the white sub-sample, USC is a binary variable indicating whether or not the respondent has a usual source of care, bwi Xi represents a series of covariates and their associated parameter estimates (anticipated confounders or predictors of UDH).
We expect that although the vector of covariates and their parameters represented by bi Xi will be statistically significant, bb1 and bw1 will continue to be statistically significant, meaning that having a usual source of care will be significantly and inversely related with having undiagnosed hypertension. However, there will be a difference in the magnitude of the effect, which the black model having a significantly smaller effect size. To test this expectation we will do the following:
The null hypothesis in this example is that there will be no race difference in the effect of having a usual source of care on having undiagnosed hypertension.
(5) NULL: H0 = bb1USC - bw1USC = 0
A test of this null hypothesis can be achieved by creating a testing statistic T to compare bb1 with bw1.
(6) T = bb1USC - bw1USC
Under the null hypothesis, the testing statistic will be asymptotically normal with mean 0 and covariance s2 (T = bb1USC - bw1USC ~ N(0, s)), where: s is the sum of the covariance estimates for the not having a usual source of care partial likelihood estimates for the model in the Black and White sub-samples ( s2 = s2 b1USA + s2 w1USC), bb1USC and bw1USC are asymptotically normal with means of bb1USC and bw1USC and variances of sb1USC and sw1USC respectively. In practice, these statistics will be produced by STATA for Windows resulting from the model being estimated on the Black and White sub-samples.
The test is completed by performing the following standardization procedure:
bb1USC - bw1USC
(7) T* =
s2 b1USC + s2 w1USC
T* will have an asymptotically standard normal distribution with mean 0 and variance 1. The resulting statistic is compared with a normal table with a critical value of .05 (t=1.96). If the statistic is within the critical area of the distribution (that is, the p value is less than .05) then the null hypothesis is not supported and it can be concluded that the effect of not having a usual source of care on having undiagnosed hypertension is different for the African American and White respondents. If the p value is greater than .05 then the null hypothesis is accepted and it can be concluded that the effect of not having a usual source of care is the same for African American and White respondents.
Analysis of Sample Size
As it is our long-term hope that his study can serve as a baseline for a longitudinal study, and there are many different sets of analyses that will be conducted, I will assess sample size in various ways (each of which demonstrate that the study will be adequately powered). First some analysis will be interested in testing for race differences in proportions (such as the percent of black and white respondents who have undiagnosed hypertension). The following table displays the required sample size to detect race differences in different proportions with 90% power and alpha=.05. The table shows that even under these rather high standards, the anticipated sample size (n=1,422) is adequately powered for nearly every scenario.
Summary of Sample Size Power Analysis, 90% Power, Alpha .05
Methods for Dealing With Adverse Events
While the likelihood of adverse events is low, it is possible that we will encounter respondents whose blood pressure is dangerously high or low, or we may encounter respondents who are exhibiting signs of depression. If the answers to the depression measures (we are using the PHQ-9) we will consider that an adverse event. Specifically, if the respondent answers yes to the question, “Over the last 2 weeks have you had thoughts that you would be better off dead, or of hurting yourself in some way?” We will ask the respondent if we can have someone contact them about how they have been feeling. If the respondent agrees, we will call Baltimore Crisis Response (410-837-2647), which is a City Health Department-run hotline for mental health crises. If the respondent does not consent to having use call the hotline, we will give them a flyer with mental health resources (including free services) throughout Baltimore.
A second possible adverse event is high or low blood pressure. Using the guidelines from the American Heart Association, we have established the following protocol for dealing with blood pressure problems among respondents.
Systolic greater than 170
Diastolic greater than 100
Ask respondent if they are on hypertension medication, and if they have taken their pills today.
If respondent has taken medications or is not on medication, call 911.
If respondent has not taken medication today, have them take it.
140 and 170
90 and 100
Ask respondent if they are on hypertension medication, and if they have taken their pills today.
If respondent has not taken medication today, have them take it.
If respondent has taken medication today or if respondent is not on medication, have them call their doctor as soon as possible.
90 and 140
60 and 100
No action needed.
Systolic less than 90
Diastolic less than 60
Have respondent take fluids and call 911.
In the event of child abuse, the interviewers will be instructed to call the Baltimore City Department of Social Services, Protective Services Screening Hotline (410-361-2235). They will also be instructed to call the study director or principal investigator (on their cell phone) and report the event.
Because we are only taking blood pressure, height and weight and asking questions, there are no physical risks of involvement in the study. However, respondents are asked about alcohol use and depression. These questions may cause some discomfort for respondents. All interviews will be conducted in privacy. Contact with the respondent will take approximately 45 minutes. This is not inconsistent with other similar studies. The primary benefit to the respondent is in learning their blood pressure status.
Study participants will be paid $20 for their time.
The consent form will be read to respondents in person or respondents will be given the consent form to read. Respondents will be advised that they will retain the right to refuse to answer any question or questions that they wish. It will also be made clear that their responses will be held confidential. No measurements or interviews will be conducted on those who have not given their written consent.
Confidentially of the subjects will be rigidly protected by coding subject data by assignment of respondent identification number (by the PI). Access to computer files and completed questionnaires will be limited to personnel involved in the data analysis. To the extent possible, all identifying records will be maintained in electronic form only and protected by password access. All records that are not practical to be kept in electronic form will be secured in a locked file cabinet in Dr. LaVeist's office. No data will be reported in any form in which individual's identity could be established. Location of stored data: Johns Hopkins University Bloomberg School of Public Health, 624 North Broadway, Room 441, Baltimore, Maryland 21205. Person responsible: Dr. Thomas LaVeist, Ph.D., (410) 955-3774 or (410) 614-8964 - fax.
After the data are entered into the computer and checked all paper copies will be shredded.