You could be reading the full-text of this article now if you...

If you have access to this article through your institution,
you can view this article in

Intraclass Correlations Among Physiologic Measures in Children and Adolescents

Amorim, Leila D.; Bangdiwala, Shrikant I.; McMurray, Robert G.; Creighton, Dana; Harrell, Joanne

Nursing Research:
doi: 10.1097/01.NNR.0000289497.91918.94
Methods
Abstract

Background: Cluster designs have become increasingly popular in many research areas. Adopting these designs requires special considerations because individuals within the same cluster may be correlated. Therefore, appropriate sample size calculations in these settings require the incorporation of additional information regarding intraclass or intracluster correlations (ICCs), which measure the degree of similarity between individuals within the same cluster or class.

Objectives: To discuss the importance of the ICC in cluster-designed studies and present ICC estimates for selected physiologic characteristics in children and adolescents.

Method: The ICCs for body mass index (BMI), sum of skinfolds (SSF), blood pressure, glucose, insulin, and lipids were determined using baseline data from the Cardiovascular Health in Children and Youth Studies (CHIC II and III). The ICC estimates presented were obtained through the use of mixed models for the entire data set and separately by age groups, gender, and ethnicity.

Results: The estimated ICCs ranged from .0104 for BMI to .1657 for glucose. The differences in the estimated ICCs among the three age groups were most marked for blood pressure, SSF, and glucose and were small for aerobic power, insulin, and lipids. The greatest difference in ICC by gender was in BMI and SSF: The ICC for SSF in girls was 8.2 times larger than that for boys. Caucasians had higher ICC estimates for insulin than did African Americans and other ethnic groups.

Discussion: The magnitude of the ICC varied by the outcome of interest, and factors such as age, gender, and ethnicity also influenced the magnitude of the ICC. The presence of ICCs should be assumed when using cluster designs, and ICCs should be considered when conducting sample size calculations for such studies.

Author Information

Leila D. Amorim, DrPH, is Professor, Department of Statistics, Federal University of Bahia, Salvador, Brazil.

Shrikant I. Bangdiwala, PhD, is Research Professor, Department of Biostatistics; Robert G. McMurray, PhD, is Professor, Department of Exercise and Sport Science; Dana Creighton, MS, is Research Assistant; and Joanne Harrell, PhD, is Frances Hill Fox Distinguished Professor, School of Nursing, University of North Carolina at Chapel Hill.

Accepted for publication April 18, 2007.

This research was partially supported by Grant No. NR-01-1837 from the National Institute of Nursing Research. Dr. Amorim was also supported by a CAPES scholarship from Brazil.

Corresponding author: Leila D. Amorim, DrPh, Department of Statistics, Universidade Federal da Bahia Salvador, Bahia, Brazil (e-mail: lamorim@email.unc.edu).

© 2007 Lippincott Williams & Wilkins, Inc.