Atherosclerosis Risk in Communities (ARIC) Cohort

ARIC Data Book

Cohort, Exam 1

Imputed Ultrasound Data

This section contains details concerning the imputed ultrasound datasets provided on the ARIC Visit 1 data tapes. The topics covered are:

• description of data set contents
• data sets and variable naming conventions
• imputed versus unimputed data

Description of Data Set Contents

Included on the updated data tapes are four data sets containing imputed ultrasound values. The data set names and variables included in each data set are described below.

Data Set and Variable Naming Conventions

Data Sets Containing Imputed Values

Because gender-race specific regression models were used to perform the imputation, a separate data set exists for White Males, White Females, Black Males, and Black Females. Each data set name consists of UBMD (indicating ultrasound) + IM (indicating the data are imputed) + WM, WF, BF, or BM (indicating the specific gender-race group). For example, the data set containing imputed ultrasound data for white males is named UBMDIMWM. Similarly, the data set containing imputed ultrasound data for black females is named UBMDIMBF. A similar pattern holds for the other gender-race groups.

The variables contained within the data sets are summarized in the table below. Most variable names consist of LBIA, RBIA, LOPA, ROPA, LINA, or RINA (indicating location) + DA or WA (indicating the type of statistic) +45 (indicating that the measurement is of the far wall). There are a few other summary variables which have unique names. These are included in the following list.

Imputed versus Unimputed Data

You may want to rerun analyses previously run on unimputed (observed) ultrasound data (using the UBMD4 data set), on imputed data (using the UBMDIMxx data sets, where xx can be BM, BF, WM, or WF). Because of the naming conventions used, this should be a relatively easy task. Note that the data set containing unimputed ultrasound data (UBMD) contains variables of average far wall width, such as LINAAV45 and LBIAAV45. These unimputed variables on the UBMD data set correspond to the imputed variables LINADA45 and LBIADA45, respectively, on the UBMDIMxx data sets. Thus, only the middle component of the variable name must be changed for AV (unimputed average) to DA (imputed average). This logic holds true for all of the site-specific averages.

Use of Weights

The weights are a measure of precision which varies by number of sites observed. Regression estimates, using *DA45 or SUM45_1 as dependent variables, will generally be more precise if weighted regression is used.

ARIC Data Book

Cohort, Exam 1

Appendix A

B-Mode Derived Variable Site Prefixes

Schematic Overview of Carotid Artery B-Mode Ultrasound Measurements

Interfaces:

1. Boundary between the periadventitia and adventitia of the near wall (not measured)
2. Boundary between the adventitia and media of the near wall
3. Boundary between the intima of the near wall and the blood
4. Boundary between blood and intima of the far wall
5. Boundary between media and adventitia of the far wall
6. Boundary between adventitia and periadventitia of the far wall (not measured)

Max 23 = B-A; Max 45 = D-C; Min 34 = H-G

The extracranial carotid system is divided into one-centimeter segments: I = internal carotid; II = carotid bifurcation; III = common carotid. A maximum of eleven measurements is made by URC readers on each arterial wall interface, in each arterial segment. These measurements are placed equidistant at 1 millimeter intervals, represented by the eleven points placed on interface B2 on the internal carotid. Also shown on this schematic is the definition of a maximum and a minimum wall thickness variable.

ARIC Data Book

Cohort, Exam 1

Ultrasound data

Imputed, white female