Analyzing the Results of Demonstration Studies - Evaluation Methods in Biomedical Informatics - page 235

TABLE 8.6. Repeated measures design.
Intervention

Time Reminders Control

Baseline G1 G2

After installation G1 G2

Repeated Measures Designs

In the complete factorial design and the nested design, each participant is
exposed to only one combination of the independent variables and thus
appears only once in a table representing the design. By contrast, in
repeated measures designs, each participant appears in two or more cells
in the table and is, in effect, reused during the study. Thus in a repeated
measures design, participants are said to be employed as their own con-
trols. The hypothetical study discussed in Chapter 7 (see Simultaneous
Randomized Controls), provides a perfect example of a repeated measures
design. Each participant, a physician, is randomly assigned to one of two
groups (G1 or G2), and all the postoperative infection rates of their patients
are measured before and after any antibiotic reminders are issued. Using
our design notation, this study is illustrated in Table 8.6.
Note that each group, and thus each participant in each group, appears
twice in this design. In the terminology of experimental design, time is a
“within subject” variable because the same participants appear at multiple
levels of that variable: before and after installation. Reminder delivery
method is a “between subject” variable because each participant appears
in only one level of that variable: reminder or control.

Self-Test 8.1
Using the notation developed in the previous section, diagram the studies
described in scenarios 1, 3, 4, and 5 of Self-Test 7.1. Scenario 4, as worded,
can be interpreted two ways. For purposes of this exercise, treat it as a
nested design with care providers as the participants and clinical services
as the unit of randomization.

Logic of Analysis of Variance (ANOVA)
Statistical methods using analysis of variance (ANOVA), discussed briefly
here, exist specifically to analyze results of studies with continuous depen-
dent and discrete independent variables, including all the variants on the
designs discussed in the previous section.
Recognizing that study design and ANOVA are the topics of entire text-
books, 12

we seek here to establish the basic principles using the results of