I am analyzing data from a baseline survey before implementing health education and endline survey after implementing health education to see the effect of health education on childhood intestinal parasitic infection. I have the prevalence of intestinal parasitic infections at the baseline and endline surveys. I have also confounding variables such as child sex, age and educational status of care givers. I planned to use adjusted prevalence ratio (PR) with 95% CI to see the effect. But, I faced difficulty to do adjusted PR and 95% CI. So, how I can do this either in SPSS or Stata?
Thank you
The data look like as below. It will help you for the discussion.
Baseline Endline
Infection Sex age Edu. Infection sex age edu.
yes male 24 month primary No female >24 no educ.
In SPSS you need to include all variables together while performing the analysis. In case you are performing a regression, just include them all as dependent variabiles, bothyhe variable of interested and those you want to adjustify for. The resulting OR and 95% CI is adjustified. Kndeed, if you perform the same analysis one factor by one, you will get different results. Hope it helped!
I am not sure I am understanding exactly what your issue is: are you looking for a suitable method of analysis for your data or you already know what to do and you are now merely looking for how to do it using stata or SPSS? I think the question needs some clarifications. You should have an intervention variable as well, wich is not showing up in your description of the data unless the the design did not have a control group.... Looking at the design of your study you might want to try a diff-diff analysis using a log-binomial regression with an interaction term between intervention and time (period) variables. By doing so you can derive PR with the glm function in stata if of course my understanding of your design is correct.
The best
Dear Tieba, I have no control group. I collected data before the intervention from let say x population, implement the intervention for that population and collect data from that population after the intervention. The design is uncontrolled before and after intervention study design. I want to see the effect of the intervention. I know that percentage point change and prevalence ratio are the appropriate statistical analysis method. But, I faced problem to do adjusted percentage point change or/and prevalence ratio. If you know the method, please share me. I know that difference in difference is appropriate for controlled before and after intervention studies meaning if the design has intervention and control groups both at the baseline and endline. In my case, I have no intervention and control groups at the baseline and endline.
Zemichael Gizaw Then I think what you can derive here is a point percentage change. However, your ability to attribute the change to the intervention will be very weak because of the uncontrolled nature of the study. You don’t also necessarily have to adjust for sex, age, education unless you think the sex-composition or age-composition of your population have changed between measurements (baseline-endline). You can quickly look for imbalance in your samples between baseline and endline with respect to sex, age, education…That will let you know if any of these is really of concern. You might then want to compute standardized prevalence instead of crude prevalence of infection. You can have a look to how you do standardization in stata here https://www.stata.com/manuals13/rdstdize.pdf
Hope this helps
The best
Tieba M
je ne sais si j'ai bien compris....ton rapport de prévalence. est ce un rapport entre l’enquête finale et initiale. si non pourquoi tu ne considère pas l'évolution entre les deux temps et faire un modèle de poisson avec terme offset ; tu pourras donc avoir l'incidence
I totally agree with Zemichael Gizaw. The participants are their own controls here so no need to worry about confounding unless there are significant differences between baseline and end of study.