3.1.4. Data analysis
The validity testing modeling was fitted with SPSS 22.0 (IBM) using 2-level (unit, hospital) hierarchical linear mixed modeling to determine associations of nursing characteristics, nursing pro- cesses, and physical restraints with the Pressure Ulcer and Fall Rate Quality Composite Index (See Supplemental Data File 3 for SPSS code). The hierarchical analysis was performed using two statistical model specifications varying by hospital level control variables. Unit type was included in both statistical models. The effects of interactions among variables within and between levels were not considered, as effects of interactions was not one of our hypotheses. We only included units in the analysis that had all variables non-missing and did not perform any imputation. The distributional assumptions in the model were not a concern as the number of units was so large that the central limit theorem provides estimates that are normally distributed. The statistical model with the better fit (i.e., lowest Bayesian Information Criteria, [BIC]) was selected for the final statistical inference.
A sensitivity analysis was performed and it was determined that units with less than 30 patients assessed for hospital-acquired pressure ulcers during any quarter should be excluded from all analyses. The model then was re-analyzed and the final are results presented in the results section below. When including units only with 30 or more patients we found two changes relative to the all- inclusive model. First, we found that total nursing hours per patient day became significantly associated with the Pressure Ulcer and Fall Rate Quality Composite Index. Second, mean hospital-acquired pressure ulcer prevention measures in place per patient became significantly associated with the Pressure Ulcer and Fall Rate Quality Composite Index. All the other explanatory variables resulted in the same conclusions.
3.1.4. Data analysisThe validity testing modeling was fitted with SPSS 22.0 (IBM) using 2-level (unit, hospital) hierarchical linear mixed modeling to determine associations of nursing characteristics, nursing pro- cesses, and physical restraints with the Pressure Ulcer and Fall Rate Quality Composite Index (See Supplemental Data File 3 for SPSS code). The hierarchical analysis was performed using two statistical model specifications varying by hospital level control variables. Unit type was included in both statistical models. The effects of interactions among variables within and between levels were not considered, as effects of interactions was not one of our hypotheses. We only included units in the analysis that had all variables non-missing and did not perform any imputation. The distributional assumptions in the model were not a concern as the number of units was so large that the central limit theorem provides estimates that are normally distributed. The statistical model with the better fit (i.e., lowest Bayesian Information Criteria, [BIC]) was selected for the final statistical inference.A sensitivity analysis was performed and it was determined that units with less than 30 patients assessed for hospital-acquired pressure ulcers during any quarter should be excluded from all analyses. The model then was re-analyzed and the final are results presented in the results section below. When including units only with 30 or more patients we found two changes relative to the all- inclusive model. First, we found that total nursing hours per patient day became significantly associated with the Pressure Ulcer and Fall Rate Quality Composite Index. Second, mean hospital-acquired pressure ulcer prevention measures in place per patient became significantly associated with the Pressure Ulcer and Fall Rate Quality Composite Index. All the other explanatory variables resulted in the same conclusions.
การแปล กรุณารอสักครู่..
