National Meeting 2009

3057 — A Multigroup Invariance Analysis of the Organizational Culture Scale

Li Y (Seattle HSR&D) , Helfrich CD (Seattle HSR&D)

Objectives:
Organizational culture has been implicated in health care quality indicators, such as patient satisfaction, provider team functioning, and CQI implementation. However, these findings depend on an untested assumption that organizational culture has an equivalent meaning and measurement across subgroups within an organization. The objective of this study was to evaluate whether the organizational culture scale of the VA All Employee Survey (AES) provides equivalent measurement across two sets of key subgroups: employees with differing levels of supervisory responsibility, and physicians vs. nurses.

Methods:
We used data from the 2006 AES (n = 112,331), including surveys from 9,976 registered nurses (RNs) and 2,942 physicians who had no supervisory responsibility. The organizational culture scale comprised 14 items measuring the four theoretical dimensions of the competing value framework (CVF): entrepreneurial, hierarchical, teamwork, and rational cultures. Items were measured on a 5-point scale (1 = strongly disagree; 5 = strongly agree). We tested measurement invariance across five supervisory levels, and between RNs and physicians.

Results:
The organizational culture scale demonstrated good reliability; Cronbach’s alpha was .91 overall, ranging from .71 (hierarchical) to .86 (entrepreneurial) for subscales. Fit indices for each of the seven subgroups suggested borderline model-data fit. While the NFI and CFI were both above .95 (i.e., satisfactory model fit), RMSEA was between .10 and .11, and GFI at .90 or lower (i.e., poor model fit). Most item-factor loadings were above .65, indicating the item was a good measure of its corresponding construct. The exception was two items with factor loadings of .36 and .55, both from the hierarchical subscale, measuring formalization and structure, and emphasis on permanence and stability, respectively. Our assessment of the measurement model holding an equal configuration of factors and indicators across subgroups provided adequate fit. However, analyses on the indicators having equal factor loadings across subgroups were not supported. This indicates significant differences among subgroups in how they respond to the organizational culture scale.

Implications:
Our findings suggest that the four dimensions of the CVF might not be equally important to the latent construct of “organizational culture” across supervisory levels or between physicians and nurses.

Impacts:
Culture, as measured by the CVF, may not be an organizational-level construct. Pooling organizational culture data across supervisory levels, and between nurses and physicians, may obscure important differences in the perceived culture among employees.