All data will be inserted into an Excel worksheet with the participant’s identity listed in the rows and the number of the question asked listed in columns. Answers to each of the questions should be “never”, “sometimes”, “usually”, and “always”; for simplicity in coding the responses, an answer of never will be listed as 0, an answer of sometimes will be listed as 1, an answer of usually will be listed as 2, and an answer of always will be listed as 3. If the participant failed to respond to the question, an answer of 9 will be listed to represent the nonresponse. The leadership style of the department administrator will also be listed next to each participants name or identifier.
To blind the researcher against the identity of the OB/GYN staff who participated in the study and protect against bias, the names of the employees were removed and assigned a random number between 1 and 100 using the RANDBETWEEN function. At this point employee names and all identifiers will be removed from the Excel spreadsheet containing the patient data.
To determine the pattern of responses amongst the employees who answered the survey questions, the percentage of each response for each question will be calculated. If there is no underlying trend for each question, this information will be dismissed. If there is a significant response pattern, it would be useful to create a bar graph that graphically displays the employee response. Furthermore, the percentage of each response per person will be calculated in order to determine whether the nursing staff surveyed answered questions honestly or just selected the same answer to each question in order to rush through the survey; although all answers will be treated as a part of the ultimate analysis, it will be useful to determine whether the results will be biased. If the results are biased due to this kind of question response, this issue will be treated in the discussion section.
After this primary analysis of the data has been completed, a linear regression will be completed for each question asked that compares the leadership style with the staff responses. All non-responses will be excluded for this particular analysis. The main focus of this study will be to generate the correlation coefficient between the employee responses to each question. A correlation coefficient of 0.4 or greater will be considered significant and warrant further analysis.
A secondary analysis of the data will involve the dichotomization of staff responses to each question. An answer of never will be coded as 0, while responses for sometimes, usually, and always will be coded as 1. Surveys that were returned without an answer to the question will be randomized to either group to prevent against bias. After dichotomization, any event that never happened will be considered 0 and events that happen at least sometimes will be grouped together as well. Student t-tests will be used to compare the mean response between each group. The strength of this association will then be measured using odds ratios.
The main goal of this data analysis will be to determine if there is a certain relationship between the staff response patterns and the leadership style of the department administrator. These tests will all be run a second time to compare the leadership style of the department administrator and the responses for each individual employee.
Interpretation of Results
There will be a positive correlation between staff member response and leadership if a majority of the staff responses are associated with a specific leadership style. The student t-test will be examined by looking at the t value and ascertaining whether the test result is statistically significant p-value. If there is an association determined, the strength of this association will be measured by conducting a test of odds ratio. The p-value will also be examined in this situation.
There will be a discussion in this paper that analyzes the power of the study after it was conducted and this will be used to determine whether enough patients were recruited to correctly find a significant effect of leadership in this study. The purpose of this discussion will be to consider whether it would be useful to repeat this study asking the same question using either a different experimental design, a greater amount of participants, or a different cohort.
Addition information regarding patient health status and quality of care may be retrieved from the records. If so, a multivariate regression of this analysis will be conducted to determine the association between leadership style, employee response to the questionnaire, and patient health.