Chapter IV Methodology
Question 1. Are there significant differences in quality outcomes among VA
hospitals with different shared governance structure and in VA hospitals without shared governance structure? Analysis for this question required both multivariate analyses of variance (MANOVA) for the three dependent variables of RN satisfaction which were highly correlated. Both shared governance variables were examined. For the two category shared governance “have” or “no” the Independent Samples T-test was utilized, while one-way analysis of variance (ANOVA) was used for the five
category shared governance maturity variable for testing the remaining dependent variables. Each analysis was conducted using the two category independent variable of shared governance and the five category independent variable of shared governance maturity.
Question 1a. Are there significant differences in RN satisfaction with
participation, quality and overall satisfaction in VA hospitals with shared governance structures? A MANOVA was conducted with shared governance as the independent variable and RN satisfaction with participation, quality and overall satisfaction as the dependent variables. This approach minimized the risk of a type one error (Kellar, 2013). Previous examination of correlations indicated that the three dependent variables were significantly positively correlated (see Table 7). The independent variable of shared governance is a nominal measurement scale. The assumptions of normality, homoscedasticity, multicollinierity, and linearity have been met and previously described. No outliers were found in the dependent variables. There is not between group equality with 38 VA facilities having shared governance and 16 VA facilities without shared governance. Results of this MANOVA were not significant and no further analysis were conducted (Roy’s Largest Root = 0.34, F (3, 50) = 1.144, p <.05).
A MANOVA was conducted with shared governance by years or maturity as the independent variable and RN satisfaction with participation, quality and overall as the dependent variables. This approach minimizes the risk of a type one error (Kellar, 2013). Previous examination of correlations indicated that the three dependent
governance years is a nominal measurement scale. The assumptions of normality, homoscedasticity, multicollinierity, and linearity have been met and previously described. No outliers were found in the dependent variables. The results of Levene’s test of equality indicates that there is homogeneity of between-group variance for RN satisfaction with participation (p = .394), quality (p = .453) and overall (p = .712). Box’s M test for equality indicates that the correlations between the dependent variables is equal across all groups (F = 1.367, p = .110). The descriptive statistics indicate that there is a trend in the mean scores of RN satisfaction as shared governance matures the means ranged from (2.50-2.71) with a slight dip in the 1-3 year category to 2.50. This was similar for RN satisfaction with quality. Overall RN satisfaction was similar except the mean dipped again in the >7 years category as well.
The MANOVA resulted in a significant model with the 3 RN satisfaction outcomes of participation, quality and overall (Roy’s Largest Root = 0.263, F (4, 49) = 3.224, p = .02). Shared governance accounted for 21% (partial eta2 = .208) of the group differences. Follow-up univariate tests of between-subjects effects suggest that both RN satisfaction with participation and RN satisfaction with quality differ
significantly with respect to shared governance maturity: participation: [F (4, 49) = 2.701, p = .041]; quality: [F (4, 49) = 2.573, p = .049]. Overall satisfaction was not significant [F (4, 49) = 1.664, p = .173]. Follow-up ANOVAs were conducted with the same results reported in the MANOVA. Tukey’s post hoc analysis were not
significant, however the means do indicate a trend related to years of shared governance maturity as seen in Tables 10-11.
Table 10
RN Participant Satisfaction Tukey HSD Post Hoc Test Results
Shared Governance Years
1.0 .0 2.0 3.0 4.0
RN participant satisfaction
Mean (SD) 2.42 (.22)* 2.50 (.14) 2.59 (.23) 2.64 (.18) 2.71 (.20)* Note. * indicates significant difference [F (4, 49) = 2.701, p = .041]
Table 11
RN Quality Satisfaction Tukey HSD Post Hoc Test Results
Shared Governance Years
1.0 .0 2.0 3.0 4.0
RN quality satisfaction
Mean (SD) 2.78 (.19)* 2.85 (.10) 2.91 (.15) 2.95 (.15) 2.99 (.12)* Note. * indicates significant difference [F (4, 49) = 2.573, p = .049]
Power analysis for Question 1a. The effect size was examined for research question 1 for the univariate dependent variable models and for the overall MANOVA model using G*Power 3.1.9.2 (Faul, Erdfelder, Buchner, & Lang, 2009; Faul,
Erdfelder, Lang, & Buchner, 2007). First, the model with the dependent variable, RN satisfaction with participation, with a partial eta squared of 0.181 the effect size (ES) of 0.47 was considered to have a large effect size (Cohen, 1992). The ES for the model with the dependent variable, RN satisfaction with quality, given a partial eta squared of 0.174 was 0.46 which again is a large effect size. Based on the ES, α of 0.05 and a sample size of 54 the power analysis resulted in 0.755 for the dependent variable of RN satisfaction with participation and 0.732 for the dependent variable of RN
satisfaction with quality indicating that the sample size was not adequate to achieve the preferred power of .80, however it is not exceptionally low. The overall power for the
MANOVA using Roy’s Largest Root was 0.791 which was just below the preferred power of 0.80. Roy’s Largest Root was the most significant of the four multivariate test statistics calculated by SPSS and is likely to be more sensitive to small sample size. However, caution must be exercised in interpreting the results as Roy’s Greatest Root is considered the least robust and may substantially increase the likelihood of a type I error (Olson, 1976).
Question 1b. Are there significant differences in patient satisfaction in VA
hospitals with shared governance structures? An Independent Samples T-Test was conducted with the independent variable of shared governance and the dependent variable patient satisfaction. The results of the Independent Samples T-test (2 sided) of 0.127 was not significant. Levene’s test for equality of variance was significant
(p = 0.027) indicating inequality between groups.
One-way ANOVA was conducted with the independent variable of maturity of shared governance at the VA hospitals and dependent variable of patient satisfaction. The analysis was not significant. F (4, 49) = .882, p = .481.
Question 1c. Are there significant differences in LOS in VA hospitals with
shared governance structures? An Independent Samples T-Test was conducted with the independent variable of shared governance and the dependent variable LOS. The results of the Independent Samples T-test (2 sided) of 0.765 was not significant. Levene’s test for equality of variance was also not significant.
One-way ANOVA was conducted with the independent variable of maturity of shared governance and dependent variable of LOS. The analysis was not significant F (4, 44) = .613, p = .655.
Question 1d. Are there significant differences in hospital acquired MRSA
infections in VA hospitals with shared governance structures? An Independent Samples T-Test was conducted with the independent variable of shared governance and the dependent variable MRSA. The results of the Independent Samples T-test (2 sided) of 0.140 was not significant. Levene’s test for equality of variance was not significant.
One-way ANOVA was conducted with the independent variable of maturity of shared governance at the VA hospitals and dependent variable of MRSA. The analysis was not significant [F (4, 48) = 2.134, p = .091].
Question 1e. Are there significant differences in HAPU in VA hospitals with
shared governance structures? An Independent Samples T-Test was conducted with the independent variable of shared governance and the dependent variable HAPU. The results of the Independent Samples T-test (2 sided) of 0.334 was not significant. Levene’s test for equality of variance was also not significant.
One-way ANOVA was conducted with the independent variable of maturity of shared governance and dependent variable of HAPU. The analysis was significant F (4, 48) = 2.815, p = .035. Results of Tukey’s HSD were not significant (Table 12). In the Tukey’s post hoc test group sizes were unequal and harmonic mean of group sizes were used. Type I error levels are not guaranteed.
Table 12
HAPU Tukey HSD Post Hoc Test Results
Shared Governance Years
0 1-3 3-5 5-7 >7
HAPU
Mean (SD) 1.59 (.92) 1.42 (.55) 2.26 (.83)* 1.37 (.45)* 1.63 (.64) Note. * indicates significant difference [F (4, 48) = 2.815, p = .035]
The effect size was examine for the HAPU dependent variable ANOVA effect using G*Power 3.1.9.2 (Faul et al., 2009; Faul et al., 2007). The ES 0.322 indicated a medium effect, however the power of .40 indicated that the analysis was
underpowered.