By Eric Crampton 23/05/2019 2


Awww, nutbunnies.

In April, Paula Penfold and Eugene Bingham reported that 2500 women had had their requests for abortion turned down over the last decade.

Me, and a few others, immediately started imagining some rather interesting research that could be done if those records could be linked up in the Integrated Data Infrastructure (IDI). So I started making enquiries. It’s not a study I’d ever do as part of the day-job, but if a few queries from me could help the academics who would actually do the studies and tell me interesting things about the world I didn’t know before – well, I was happy to poke around to see whether that data could be linked up in IDI.

It can’t be. Or, at least, not easily. Consulting physicians provide to the government reporting on approvals and rejections, case-by-case, but with an anonymous ID number attached to each case where only the consulting physician has the key. The government can request more detail, and the physician can then check back in the physician’s notes. But the government itself doesn’t have the keys for linking the records to the de-identified records held in IDI – like by health ID number. It just can’t be done without getting physicians to agree to sharing more of their notes up the system – and that’s a far bigger battle.

It’s a darned shame though. Linked up in IDI, here’s the kinds of things that friends back in academia could have looked at:

  • Outcomes (across a wide range of indicators) for women who proceed with termination compared with:
    • statistically comparable women who never requested termination;
    • statistically comparable women who requested termination, were approved for termination, but then changed their minds (a birth happened soon enough after the request was approved);
    • statistically comparable women whose request for termination was declined.
  • Outcomes (across a wide range of indicators) for children whose mothers never sought termination compared with:
    • children of statistically comparable women who requested termination, were approved, but changed their minds;
    • children of statistically comparable women whose requests for termination were denied;
    • the children of statistically comparable women who did not give birth after being approved for termination (you can’t tell whether the termination proceeded or whether a miscarriage otherwise obtained) but who did give birth at a later date.
  • Lifetime parity among women among women whose requests were approved compared to those whose requests were not – how much of this is a decision about the timing of children rather than the number.
I’m sure you can think of a lot more research that looks to be near-impossible.
Many thanks to the patient communications advisor at the Ministry of Justice who walked me through how the system operates. I think she had to do a fair bit of running around to sort it all out. So big plaudits for public sector helpfulness on this one. No mucking around treating things as OIA requests and then long turnarounds on questions of clarification, just helpful replies. Very happy on that front. A++++ recommendation.

2 Responses to “Data termination”

  • One thing you’d want to add is deferrals. We once had a doctor on this forum who said he didn’t approve abortions.* My impression—just an impression (and recollection!)—was that doctors can decline to do something on religious grounds, but they’re obliged to defer the patient to a colleague.

    You’d hope that it’s well explained that it’s a deferral, not a denial, and looking at the impact of deferring would be another thing to do.

    ————
    * He also once wrote that rubella was a mild illness, which ‘inspired’ me to write a post on the impacts of rubella on unborn kids during (first-trimester) pregnancy, where it can have substantial effects.

  • That would be awesome to add in, agree, if the data were up to it. A deferral adds a cost. That cost will vary by the number of alternative physicians in the area. And then you have (and I know I sound awful in saying this) a beautiful source of variation that lets you test things. It is horrible for the women subject to it, but it also lets you do more in teasing out effects.

    Like, we have this big awful system that generates beautiful natural experiments allowing for testing, but we’re not even getting that benefit out of the big awful system because we’re not collecting the data in ways that would allow for it.