The Doctor of Nursing Practice (DNP) final project is a student-led project that demonstrates the DNP students' specialty and establishes the groundwork for future scholarship. Many DNP final project examples are provided by the American Association of Colleges of Nursing (AACN) in their 2006 guide (AACN, 2006). These examples are culminated by the AACN statement that “the theme that links these forms of scholarly experiences is the use of evidence to improve either practice or patient outcomes” (2006, p. 20). Although there seems to be a consensus in faculty perspectives about the necessity of collecting data to show evidence, there seems to be a split in whether faculty believe students should conduct statistical tests or not (Nelson, Cook, & Raterink, 2013). In my first year in a college of nursing, I noticed similar faculty perspectives. These perspectives, I believe, have led to ambiguity in what statistical tools and software should be used when an analysis is appropriate.
A paired sample (repeated measures) design is a typical DNP project setup (Kuerban, 2017). It is a relatively straightforward design and is feasible for a DNP student to implement in a limited amount of time. In a paired sample design, a collection of subjects are observed over time under two or more treatment conditions and compared. In many DNP projects, there are two treatment conditions: pre- and postintervention (Kuerban, 2017). For this design, two statistical tests are often used to test for a statistically significant difference. A paired t test (McDonald, 2009) is used when the observed measurements are an interval or ratio (e.g., Likert scale questionnaire) and measured on the same subject over two time points. A McNemar's test (Morrison, 2010) is used when the observed measurement is two-group categorical (e.g., correct or incorrect response) and measured on the same subject over two time points. Other statistical tests (e.g., Wilcoxon signed-rank, Cochrane's Q) do exist, which some may prefer over the commonly used t test and McNemar's test.
During my first year as an assistant professor and biostatistician in the College of Nursing at the University of Cincinnati, I consulted with more than 20 DNP students about their quality improvement projects. Approximately 90% of the DNP final projects were paired sample designs. Many of the students were unsure about what statistical test they needed to perform, how to set up their data, and what software they needed to purchase to conduct it. After educating them about the t and McNemar's test, how to set up their data, and the software they could use, the most dreadful question lingered: “How do I interpret my statistical result?” Despite my explanation of what the test was doing and, in layman's terms, what it meant for their question of interest, they still felt most comfortable when I gave them a working example of how an interpretation might be written. This pattern was very consistent over all my consultations with DNP projects that required a statistical test.
As a statistician, patterns are something I like to think I am good at detecting. This pattern, fortunately, fell right into my lap and I knew I could automate much of what I was doing without diminishing the education I was providing. After discussing this with the DNP program committee and the director of the DNP program, it was obvious that the students needed a tool for the statistics their paired sample designs demanded. Fortunately, I had numerous opportunities developing such tools in my previous positions. I hoped to make two applications (one for the t and McNemar's test) that would allow students to learn about the possible statistical tests, upload their deidentified data, select the appropriate statistical test, and interpret (in words) the statistical result.
These two applications are now in production and can be found for free on the University of Cincinnati College of Nursing's public server (t test: http://modeling.nursing.uc.edu/statspicker/ttest.Rmd; McNemar's test: http://modeling.nursing.uc.edu/statspicker/chimc.Rmd). The applications are accompanied by YouTube videos explaining their use and can be found under the “About This” application tab on each application. Both applications explain in a short and abbreviated manner what the t and McNemar's tests consist of. Additionally, users can download an example data set to their computer and see how the file should be set up before they upload it. When uploading their own CSV-formatted data (or the example file), the application shows the user the first three rows of their file. Other rows can be viewed by progressing through the pages on the bottom of the “What your data look like?” tab. Next, users select “What type of test?” they wish to perform. On the t-test application the user can select the variance assumption they wish to make, as well as whether they want to complete a one- or two-sided test. Finally, both applications allow the user to select the confidence level of the test they want to perform and which two variables in their data set they would like to test. The variable drop-down boxes are automatically populated based on the column names from their uploaded file. After the choices are selected, the test automatically runs and displays the summary statement on the “Output and Summary Statement” tab on the bottom half of the application. Here, a simple statistical statement is produced and can be copied and pasted into a document for further editing by the user.
It is important to note that this application is not a replacement for a traditional statistics class, nor should it be used without the proper instruction and oversight by a mentor. It is also important to consider what sample size was used when interpreting the results from the application. In future versions of these applications I hope to add data visualizations and more information to educate and inform the user about the methods and results. Also, I plan to create a sample size and power calculation application that can be used at the beginning stages of the student's project design.
In my consultations since creating these applications, I have noticed that many of the students' questions and troubles are quickly resolved when they find out that there is an easy-to-use application that gives them an overview of the method, example data set, instructional video, and statistical statement. My hope is that this application will remove some of the angst and ambiguity that DNP students often face when thinking about what statistical test and software they should use while alleviating some of the repetition that is often found in statistical consultations for DNP projects.
Joshua Willian Lambert, PhD
University of Cincinnati College of Nursing
- American Association of Colleges of Nursing. (2006). The essentials of doctoral education for advanced nursing practice. Retrieved from https://www.aacnnursing.org/Portals/42/Publications/DNPEssentials.pdf
- Nelson, J.M., Cook, P.F. & Raterink, G. (2013). The evolution of a doctor of nursing practice capstone process: Programmatic revisions to improve the quality of student projects. Journal of Professional Nursing, 29(6), 370–380.
- Kuerban, A. (2017). Statistic methods used in scholarly projects of doctor of nursing practice graduates. International Journal of Nursing & Clinical Practices, 5, 276–277.
- McDonald, J.H. (2009). Handbook of biological statistics (3rd ed., pp. 173–181). Baltimore, MD: Sparky House.
- Morrison, M.A. (2010). McNemar's test. In Salkind, N.J. (Ed.), Encyclopedia of research design (pp. 80–83). Thousand Oaks, CA: Sage.