Background to problem with pregnancy history variables
There are 29 variables relating to each baby ever carried, and 10 others relating to
each pregnancy. These variables are structured in such a way as to accommodate 8
possible pregnancies, with up to 5 babies born as a result of each. So each of the 29
‘baby’ variables are replicated forty times (e.g. birth/termination date prege, prege2,
prege3,..., prege40), and each of the ten ‘pregnancy’ variables are replicated eight
times (e.g. number of babies carried pregnum, pregnuιn2.... , pregnum8).
The convention is that the most recent pregnancy is entered first, and the others
follow in reverse chronological order. For example the variable prege refers to the
date of birth (or other outcome) of the 1st or only baby resulting from the most recent
pregnancy, the variables prege2-prege5 would refer to any other babies from that
same (multiple-birth) pregnancy; then prege6 refers to a baby resulting from the
second most recent pregnancy, prege11 from the third most recent.... etc. We use the
terminology ‘outcome date’ rather than birth date, as in the case of non-livebirths, the
date of miscarriage/stillbirth/abortion is entered.
An apparent error in the setting up of the CAPI instrument for data collection, or a
misunderstanding of the nature of the ‘number of babies carried’ question resulted in
the following unforeseen problems:
(A) In a number of both NCDS and BCS70 cases, babies from separate pregnancies
were entered as if they were part of a multiple birth: their outcome dates, though
clearly more than 9 months apart, were entered, for example, as prege, prege2,
prege3,. etc., instead of prege, prege6, prege11,. etc. In these cases the variable
pregnun (or in some cases pregnun2-8) appears to have been filled in incorrectly,
probably as a result of construing the question ‘how many babies were carried during
this pregnancy?’ erroneously as ‘how many babies have you ever carried?’ This
pregnun error is probably what caused the automatic CAPI routing to bunch these
babies wrongly. See Appendix 1 for a listing of these cases.
(B) Furthermore, in most of these cases, the outcome dates were also in the wrong
order, with the oldest first, and the most recent last, or even an irregular order. See
Appendix 2.
(C) This reversal or disruption of chronological order also occurred in a number of
other cases where there was otherwise no confusion between single and multiple
pregnancies. See Appendix 3.
(D) In the case of NCDS respondents, the intention was only to update pregnancy
history information since the last full NCDS sweep (NCDS5, 1991), and so the CAPI
instrument was set up to enquire about all pregnancies since March 1991. However,
for many cases a complete pregnancy history seems to have been entered, with dates
prior to March 1991 included. See Appendix 4.
In trying to correct the above errors systematically, a number of factors introduced
complications: