3
$\begingroup$

I am working on a medical research project where I am trying to find the associations between specific maternal factors and neonatal limb ischemia (NLI) to declare the maternal factors as "risk factors for NLI". I am trying to find a way to estimate the percentage of a specific maternal factor being associated with NLI, but I am having a hard time figuring out how to make the estimates.

The data is only on NLI patients. In terms of associations between maternal factors and NLI, the factors include maternal health problems, maternal age, maternal ethnicity, previous pregnancies and births, and maternal substance use (smoking, drinking, drug use, medications).

I have a bit of a background in stats, but I am not sure what the best way is to calculate the estimates. I have a lot of data of 120 patient cases from around the world retrieved from papers. I have made an Excel of all the relevant information regarding each patient. The data is not perfect since it all comes from different papers, but I've tried to make an Excel where the information is organized in a consistent manner. I have an idea of what trends seem to show within my data sample but I don't know how to validate these trends by conducting an estimate.

I have tried to think of ways to do the estimate, but I am not a stats expert. I thought that clustering might be helpful here to group the data, but I'm not sure what to do with the groups and if clustering would even help develop the estimate of the percentage. I am also still learning how to cluster, so I am hesitant to go this way.

Are there any suggestions for simple techniques I should employ for the estimate?

$\endgroup$
3
  • $\begingroup$ Could you please edit the question to define "NLI"? Also, when you edit, please say more about the data that you have collected: in particular, do you only have data on NLI cases, or do you have corresponding information on other cases? How many maternal factors are you evaluating? $\endgroup$
    – EdM
    Aug 20, 2022 at 18:38
  • $\begingroup$ Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. $\endgroup$
    – Community Bot
    Aug 20, 2022 at 19:09
  • $\begingroup$ @EdM thank you for the response, I have made the edits. I listed the maternal factors being looked at and the data is only on NLI cases. $\endgroup$
    – eddie
    Aug 20, 2022 at 21:33

1 Answer 1

2
$\begingroup$

Without data on situations without NLI, you won't be able to answer your question adequately.

For a trivial but important example of the problem with evaluating maternal factors just within NLI data: all mothers of your NLI cases were females. That would be the strongest association with NLI in your data set, even though it's also true of all births without NLI.

Even if you found clusters of less-trivial maternal factors within the NLI cases, you wouldn't know whether those clustered similarly in births without NLI. At the least, you need to find some distributions of values in normal cases for comparison.

Then you could try to examine whether the distribution of one or more maternal factors among NLI cases differs from that in normal cases, although the multi-cultural distribution of NLI cases in your data set will even make that a challenge. You would probably need to find normal cases with the same geographic distribution as your NLI cases.

That said, there's nothing wrong with presenting a summary of multiple case reports as you seem to have accumulated, including the distributions and associations of maternal factors in your data. On that basis alone, however, you can't call those "risk factors for NLI."

$\endgroup$
2
  • $\begingroup$ so the data I have on the 120 cases of NLI will not be enough to declare a specific maternal factor as a risk factor for NLI? I would need to do a comparison with normal healthy births in order to actually extrapolate? I was wondering if there would be a recommended way to graph the data or find a way to better understand and explain the distributions? I made bar graphs and line graphs, but should I try to make histograms? $\endgroup$
    – eddie
    Aug 23, 2022 at 3:07
  • 1
    $\begingroup$ @eddie visualizing the data extensively is critical. Instead of histograms, which use arbitrary choices of cutoffs to make bins, use density plots (sort of a smoothed histogram). "Pairs plots" are convenient ways to examine pairwise associations of multiple variables at once. The ggpairs() function of the GGally package in R is one implementation that allows you to compare both continuous and categorical variables. $\endgroup$
    – EdM
    Aug 24, 2022 at 13:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.