Working with mother and baby dyads in APC and A&E data

[AMJ-coder]

Question / Problem statement

I became concerned when linking mothers (mtokenid) to the APC birth records (tokenid) as I had no babies (i.e.startage 7001:7002).
So I realised:

In the APC records, if you are looking at a birth, both mothers AND babies are both identified by 'token person id' and have separate records. In the A&E records mothers AND babies have separate records and are both identified by the 'token person id'.

How do you suggest is the best way to handle this in a merged dataset where I am looking at the impact of maternal mental health (before or during pregnancy) on the use of emergency department access (during pregnancy) and impact on children's educational outcomes. Just add flags for those contacts?

Compliance checklist

• I confirm I have not entered any information from within the ONS SRS (or only information that has been cleared for release by the ONS SRS output checking team).
• I confirm I have asked only one question in the textbox above.
• I confirm I have checked for similar questions amongst existing Q&As.

[zcabh35]

Howdy

yes, all HES produced datasets do not have the link between mater and progeny; the mother-baby link is something the ECHILD team created using raw HES data. The methodology can be found here on how this was created: https://academic.oup.com/ije/article/53/3/dyae065/7664509. The raw HES data has a tokenpersonid used to identify individuals.

When it comes to longitudinal administrative data analysis, my suggestion is to envision what the final cleaned dataset would look like so that you can conduct your analysis...and then think about how you could wrangle your longitudinal data to fit that.

It sounds like the base of your work is the mother baby link in ECHILD and from that you intend to merge in other data into this dataset where some variables are related to the mother, whilst others variables are related to the child. To reduce confusion between which variables belong to which individual, my suggestion is to rename the columns of the dataframe you intend to merge in prior to merging. You can do this using the following code:

colnames(dataframe) <- paste0("child_", colnames(dataframe))
colnames(dataframe) <- gsub("child_tokenpersonid", "tokenpersonid", colnames(dataframe))

For example, if the columns in the original dataframe are : tokenpersonid, diag_k, birweight
The new columns would be: tokenpersonid, child_diag_k, child_birweight

In terms of the maternal mental health (before or during pregnancy) on ED access during pregnancy, you should probably use the mtokenid to get data related to mental health (e.g. mental health data, or mental health diagnoses in APC etc), create the variables you need, rename the columns appropriately, and then merge them in.

Similarly for the child, you'd need to get their educational outcomes, wrangle that into the format you need, rename the columns appropriately and then merge those in

One mistake people often forget is that a single person might have multiple children, so ensure your derived variables are related to that mother-child (pregnancy specific pair) to ensure the dates are correct.

I hope this helps