Seeking advice on my capstone project

#1
Hi,

I'm a graduate nurse practitioner student, and I'm currently attempting to analyze my data for my DNP project- it's something of a capstone research/quality improvement project that we design and implement over the course of a year. I've set my data up in SPSS and have some familiarity with the program (we took data analysis two years ago, so it is not fresh in my mind). I've sought help from my school, but so far no one has been able to help- so much for that tuition!

A little bit about the project: I've recruited 28 subjects from a doctor's office to enroll in a text message reminder program for the flu vaccine. I've taken general demographic data on the subjects (age, gender, ethnicity, insurance status, past receipt of flu vaccine), and separated them into 3 cohorts based on month of enrollment (September and October). I've randomized each cohort into standard care (no text message reminder- something of a "control") and intervention (aka text message reminder) groups, and sent a single reminder message out to the intervention group at the end of each respective month. I want to look at vaccine receipt by intervention group (the main variable of interest), as well as demographic data.

I'm setting up my data in SPSS. I have 28 subjects total, 6 of whom received the vaccine (4 from the intervention group and 2 from the standard care group). I've set up my data in variable view for: vaccine receipt (Yes or No), month of vaccine receipt (Sept, Oct), intervention group (Text message or standard care group), month of receipt (Sept/Oct), demographic factors (age, gender, ethnicity, insurance status), and if the patient had a prior flu vaccine (Yes/No/Unsure). I've coded these values as nominal. I want to first look at vaccine receipt by intervention group (the independent variable and main variable of interest), and then analyze it further by demographic data and the other independent variables to see if there are any correlations.

I know that I will use uni-variate analysis to describe my demographic data, but I'm not sure which test to use to measure the primary outcome of interest: is there a correlation between receiving a text message and receiving the flu shot? Would I use a Chi Square to test this? Or do I compare vaccine receipt for each group with a between-subjects design? I know my sample size is small, which makes meaningful analysis more difficult (most DNP projects are < 30 participants). My knowledge of statistics is pretty minimal, and I've been reading my old textbook and running different analyses in SPSS to try and figure it out on my own, but at this point, I could use some guidance. Thus far, I've been running Chi Square tests- nothing has been significant and I'm not expecting anything to be with a sample size this small.

I have a few other facets to the project (In October, I recorded rationales for non-participation in the study, and will be sending out a post-study 4 question satisfaction survey), but these should be more straightforward for analysis.

Any advice is appreciated!
 

hlsmith

Not a robit
#2
I would use an exact logistic regression model. Your model will struggle given the rare outcome event and comparison of three groups (a v b, a v c, c v b, which requires you to change your alpha value to account for false discovery).

You can also, for simplicity use the Fisher's exact test, though if it is significant, you then have to conduct pairwise comparisons between groups (a v b, a v c, b v c, and correct your alpha level for false discovery). Just a heads up, I would be very surprise if you find anything of note given the sample size, rare outcome and need to control for multiple comparisons.

Thanks.
 
#3
I would use an exact logistic regression model. Your model will struggle given the rare outcome event and comparison of three groups (a v b, a v c, c v b, which requires you to change your alpha value to account for false discovery).

You can also, for simplicity use the Fisher's exact test, though if it is significant, you then have to conduct pairwise comparisons between groups (a v b, a v c, b v c, and correct your alpha level for false discovery). Just a heads up, I would be very surprise if you find anything of note given the sample size, rare outcome and need to control for multiple comparisons.

Thanks.

Thank you SO much! I had used the Fisher's exact test- and it was not significant- but thank you for your guidance- wanted to make sure I'm on the right track.