OBJECTIVE: To compare patient reports about hospital care between western New York State and southern Ontario using a random intercept model. METHOD: Cross-sectional survey of 3923 patients who received medical or surgical care between August and October 2004 at 28 hospitals (14 hospitals per jurisdiction). Thirty-five questions were combined to calculate eight indicators with scores ranging from 0 to 100 (best care experience). For each indicator, a model was built where the region (western New York vs. southern Ontario) was included as a fixed effect with hospital as random within region. A number of patient characteristics were also included as fixed effects. RESULTS: The effect of the region was statistically significant (P < 0.05) only for the models predicting the 'continuity and transition', 'involvement of family' and 'physical comfort' indicator scores. The differences were 10.66, 4.05 and -3.23 points, respectively. In all three models, the random intercepts were not statistically significant, indicating that the differences above did not vary by hospitals. The model predicting 'overall impression' scores, however, showed a random intercept statistically significant (P = 0.026). The individual-level explained proportion of variance ranged from 5.68 to 11.22%, and the hospital-within-region-level explained proportion of variance ranged from 2.19 to 52.28%. CONCLUSION: The difference observed on the 'continuity and transition' indicator might be the only one somewhat meaningful, and might be explained by health maintenance organization reimbursements' mechanisms and hospital quality improvement initiatives available in western New York, as well as by the fact that occupancy rates in western New York border the 60% compared with the 95% in southern Ontario.
OBJECTIVE: To compare patient reports about hospital care between western New York State and southern Ontario using a random intercept model. METHOD: Cross-sectional survey of 3923 patients who received medical or surgical care between August and October 2004 at 28 hospitals (14 hospitals per jurisdiction). Thirty-five questions were combined to calculate eight indicators with scores ranging from 0 to 100 (best care experience). For each indicator, a model was built where the region (western New York vs. southern Ontario) was included as a fixed effect with hospital as random within region. A number of patient characteristics were also included as fixed effects. RESULTS: The effect of the region was statistically significant (P < 0.05) only for the models predicting the 'continuity and transition', 'involvement of family' and 'physical comfort' indicator scores. The differences were 10.66, 4.05 and -3.23 points, respectively. In all three models, the random intercepts were not statistically significant, indicating that the differences above did not vary by hospitals. The model predicting 'overall impression' scores, however, showed a random intercept statistically significant (P = 0.026). The individual-level explained proportion of variance ranged from 5.68 to 11.22%, and the hospital-within-region-level explained proportion of variance ranged from 2.19 to 52.28%. CONCLUSION: The difference observed on the 'continuity and transition' indicator might be the only one somewhat meaningful, and might be explained by health maintenance organization reimbursements' mechanisms and hospital quality improvement initiatives available in western New York, as well as by the fact that occupancy rates in western New York border the 60% compared with the 95% in southern Ontario.