The application of generalizability theory to blood pressure resting levels and mental stress responses

Background: Blood pressure measurements taken in the clinic or laboratory are assumed to generalize to the world outside. This is true both of casual blood pressure measurements and of changes in blood pressure responses to stress. Such generalizability is crucial to the usefulness of blood pressure measurements as predictors of long-term disease. In previous generalizability studies, several factors differed between clinic/laboratory and field, making it difficult to interpret the poor laboratory-life associations. The present study varied only one parameter between the laboratory and the field setting. Subjects and methods: Twenty-four women were studied on four occasions: twice in the laboratory, once in a classroom, and once at home. After a resting baseline, the subjects performed a mathematics task, while blood pressure and heart rate were monitored. Only the setting was varied across sessions. Results: Test-retest correlations were 0.81 for systolic blood pressure levels (SBP) and 0.61 for diastolic blood pressure levels (DBP). Generalizability (G) coefficients for blood pressure levels were approximately the same as the reliabilities (0.82, SBP; 0.59, DBP), indicating that the change in location did not affect resting levels. However, for change scores, the reliabilities were higher than the G coefficients. Test-retest correlations were moderate: 0.68 (SBP) and 0.62 (DBP). G coefficients were 0.47 (SBP) and 0.36 (DBP), indicating that the generalizability of change scores suffered due to the change in test location. Conclusions: A minor variation in procedure, such as a change in setting, has little effect on the generalizability of blood pressure resting levels, but a substantial effect on stress-response changes. Other laboratory-field differences may have an even greater impact on generalizability. © Rapid Science Publishers.

Duke Scholars

Author Carl F. Pieper Biostatistics & Bioinformatics, Division of Biostatistics