I am conducting a systematic review and meta analysis on prevalence of X among Y population. From the many studies, I have found that one study has reported a zero prevalence in its study. Therefore, I can not calculate standard error for zero prevalence and faced difficulties to pool the result of included studies. What shall I do to solve this problem? Shall I exclude this study from the meta analysis but including in the systematic review?
Dear Ayelign,
You can't exclude literatures with low or zero prevalence. This is what we call it selection bias or publication bias. And this is strongly against the methods of meta-analysis or summary effect estimates. Thus, I am not sure which software are you using to do the systematic review but I would suggest you can use the raw data rather than the effect sizes to get results pooled. Or if the prevalence is zer know that teh SD or SE will be the same as in incidence or prevalence estimation, nill refers absence of value. That implied the variation from the estimate to get the population parameters is meaningless. Further if you need to know, I would be happy to share with you reference mterials. let me know the statistical software you are doing the analysis.
Good luck
Use software that considers continuity correction (adding values less than 1 case to the zero studies (maybe 0.25 or 0.5).
I would avoid leaving out the study because this risks creating bias. I have used studies with 100% prevalence before without issue. Meta-XL did not have issues with the confidence intervals (Which did not extend beyond 1.0). Perhaps try this program if your program is having issues (It is a free program).
I would like to thank all of you. I am using STATA Software and planned to use 0.5, continuity formula for zero events. If any additional advice, I would welcome.
I would like to thank all of you for your answers. I have taken 0.5 for zero event.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis