Introduction to Hierarchical Modeling

Abstract

Abstract

Background:

Hierarchical modeling (HM) is a statistical technique that has gained in popularity in health care research. It has been used for analysis of secondary data, performance profiles or benchmarking studies, and in prospective trials. The technique is used in situations in which traditional regression analysis might lead to incorrect conclusions. Specifically, data drawn from nested settings such as hospital units or hospice providers may be correlated, thus violating an assumption required for ordinary least squares regression.

Objective:

This article provides a description of HM, reviews two recent articles in palliative care that have used the technique, and presents an illustrative case study to further illuminate the potential of the method.

Conclusion:

When used appropriately, HM allows researchers to specify and test hypotheses that would not otherwise be possible, and avoid incorrect conclusions from nested data.