Observed communication between oncologists and patients: A causal model of communication competence

AIM: To investigate and test a causal model derived from previous meta-analytic data of health provider behaviors and patient satisfaction. METHODS: A literature search was conducted for relevant manuscripts that met the following criteria: Reported an analysis of provider-patient interaction in the context of an oncology interview; the study had to measure at least two of the variables of interest to the model (provider activity, provider patient-centered communication, provider facilitative communication, patient activity, patient involvement, and patient satisfaction or reduced anxiety); and the information had to be reported in a manner that permitted the calculation of a zero-order correlation between at least two of the variables under consideration. Data were transformed into correlation coefficients and compiled to produce the correlation matrix used for data analysis. The test of the causal model is a comparison of the expected correlation matrix generated using an Ordinary Least Squares method of estimation. The expected matrix is compared to the actual matrix of zero order correlation coefficients. A model is considered a possible fit if the level of deviation is less than expected due to random sampling error as measured by a chi-square statistic. The significance of the path coefficients was tested using a z test. Lastly, the Sobel test provides a test of the level of mediation provided by a variable and provides an estimate of the level of mediation for each connection. Such a test is warranted in models with multiple paths. RESULTS: A test of the original model indicated a lack of fit with the summary data. The largest discrepancy in the model was between the patient satisfaction and the provider patient-centered utterances. The observed correlation was far larger than expected given a mediated relationship. The test of a modified model was undertaken to determine possible fit. The corrected model provides a fit to within tolerance as evaluated by the test statistic, chi(2) (8, average n = 342) = 10.22. Each of the path coefficients for the model reveals that each one can be considered significant, P < 0.05. The Sobel test examining the impact of the mediating variables demonstrated that patient involvement is a significant mediator in the model, Sobel statistic = 3.56, P < 0.05. Patient active was also demonstrated to be a significant mediator in the model, Sobel statistic = 4.21, P < 0.05. The statistics indicate that patient behavior mediates the relationship between provider behavior and patient satisfaction with the interaction. CONCLUSION: The results demonstrate empirical support for the importance of patient-centered care and satisfy the need for empirical casual support of provider-patient behaviors on health outcomes. (C) 2014 Baishideng Publishing Group Inc. All rights reserved.