We analyzed 5 waves of NHANES data, between the years of 1999-2008, to explore whether an independent relationship exists between respondents’ self-reported source of usual care and the prevalence of selected chronic and cardiovascular events. We examined the prevalence of self-reported Hypertension, Diabetes Mellitus, Hypercholesterolemia, Angina, Coronary Heart Disease (CHD), Myocardial Infarction (MI), Congestive Heart Failure (CHF) and Stroke among individuals receiving primary care across four distinct sites of care, as well as people reporting having no usual site of care.
NHANES is a cross-sectional survey administered to a nationally representative sample of non-institutionalized U.S population. Data collection for patients includes a detailed questionnaire of health status, history, and behaviors, an examination of currently prescribed medications available at the time of interview, and selected lab tests on patient samples when appropriate. Our analysis was limited to adult participants ages 20 or above who identified one primary or no primary site of usual care, yielding a total of 21,778 participants.
Our dependent variables were the prevalence of selected chronic conditions and cardiovascular events. Appendix Table 1
provides the actual wording of the questions used to assess disease status. Survey participants were asked whether a health professional had informed them they had each of our studied conditions. In reference to diabetes, hypercholesterolemia and hypertension, we used a list of currently prescribed anti-diabetic medications, total cholesterol readings of > 200 mg/dl and average blood pressure readings > 140/90 respectively to identify additional cases.
The primary independent variable was participants’ report of their site of usual care. Categories were a composite variable derived from two questions asked in the NHANES questionnaire. The first question asked respondents whether they had a place where they usually went for their health care. If participants indicated that they had a usual site of care, a follow up question prompted them to identify the site as one of the following: Community Health Center/Clinic (CHC), Hospital Outpatient Clinic, Emergency Room (ER), Private Doctor’s Office/Health Maintenance Organization (HMO), or multiple sites. We excluded patients who indicated using multiple sites of primary care from our sample due to the small number of positive respondents (N = 187).
Participant level covariates included age (20 to 29 years, 30 to 44 years, 45 to 54 years, 55 to 65 years, or older than 65 years), race/ethnicity (White, Black, Mexican, other Hispanic, or other race), type of insurance (Medicare, Medicaid, private, uninsured, or other), gender, and income as measured by a poverty to income ratio.[1
] Additional covariates were smoking status (current, former, and never) and number of physician visits in the 12 months prior to the survey
Using patient weights provided by NHANES, we evaluated the demographic distribution of participants at each primary care site. We fit a multivariable logistic regression model to estimate the age-adjusted and fully adjusted prevalence of chronic conditions across sites. Fully adjusted logistic regression models included the primary independent and dependent variables as described previously and the aforementioned covariates. Using the results from these logistic regression models, for each site of care, we calculated fully adjusted proportions (prevalence), which are adjusted to the total population distribution of these potential confounders.
We next modeled the 3-level outcome ‘Global burden of cardiovascular disease’, defined as the proportion of participants at each site with a self-reported history of 0, 1 and 2 or greater of the following conditions: hypertension, diabetes, hypercholesterolemia, or one of several related cardiovascular events (Stroke, MI, CHD, Angina, or CHF). To estimate the global risk of disease prevalence at each site of care, we constructed age-adjusted and fully adjusted multinomial logistic regression models for the 3-level outcome corresponding to the aggregate number (0, 1, ≥2) of observed conditions.
In secondary analyses, we explored patient awareness of selected chronic conditions, since poor management of these conditions may impact the likelihood of cardiovascular events. Lack of awareness was examined for hypertension and cholesterol separately; defined as the proportion of individuals with average blood pressure readings or total cholesterol measures out of normal range (> 140/90 for hypertension, > 200 mg/dl for cholesterol) who reported not being informed they had high blood pressure or high cholesterol by a health professional. We constructed logistic regression models adjusted for all of the previously mentioned covariates other than number of visits and compared participants receiving usual care at private doctor’s offices/HMO’s to other care sites. Finally, to understand how the observed relationship was mediated by number of visits we constructed additional models including this variable. Results are reported with two tailed-p values significant at the alpha < .05 level and confidence intervals when appropriate. All analyses were performed with SAS 9.2 and SUDAAN 10.0, both of which accounted for the weights, strata, and clusters of the complex survey design.