excluded group vs 73.6 per cent study group). We are
doubtful that most of these 353 names led to an actual
Massachusetts birth. Less than 10 per cent of these women
had locatable birth certificates. Abortions, fraud, computer
errors, and out-of-state moves are the more likely unrecorded
realities for these names.t Women do not always inform
the WIC program of their reasons for discontinuance. Nevertheless,
one cannot rule out of the possibility that the
administratively excluded names may have had specific
characteristics which would bias the overall study results.
The birth certificates for 191 women who were in the
WIC program prenatally could not be located. Again, little
epidemiologic information is available about these women.
Racially (based on their WIC records) they are similar; their
duration of participation in WIC is essentially the same as
the WIC group. No fetal deaths were located among this
group. One can not rule out the possibility that they may
have had specific characteristics which would bias the
overall results. Five WIC women with birth certificates were
not matched to controls.
Overall, we estimate that at least 95 per cent of the WIC
prenatal participant population were included in the study
which represents the largest and most comprehensive series
on WIC prenatal participants to date.
Establishing the existence and magnitude of a WIC
program effect also depends critically on the comparability
of the WIC and matched control groups. Unfortunately,
there are inherent limitations to the conclusions that can be
drawn from a retrospective cohort study in which the
exposure (WIC) group is self-selected and the control group
is derived by a post-hoc matching procedure. A more ideal
randomized case control study would pose serious ethical
dilemmas. Since many known confounding factors have
been controlled, we believe that the statistical differences
between the WIC and control groups are a function of WIC
participation; however, additional confounding factors may
also be characteristic of the WIC or control populations and
account for any birth outcome differences noted.
The Massachusetts birth certificates do not provide
specific information on maternal pre-pregnancy weight or
height, maternal weight gain, maternal smoking habits, or
maternal morbidity. Any of these factors, if unevenly distributed,
may be sufficient to distort the overall outcomes. WIC
participants may be more strongly motivated to improve the
prenatal health of their future offspring than are the control
women. Such a motivational difference could cause both an
improvement in pregnancy outcome and a desire to enroll in
the WIC program. The findings of earlier and more frequent
prenatal care visits may be supportive of this view. The
increase in prenatal care may also be the cause of the
improved birth outcomes, and not simply another consequence
of WIC participation. The lack of prenatal care
improvements among Hispanic origin women who show
enhanced birth outcomes argues somewhat against this
interpretation. The present study design does not lend itself
to a study of prenatal care, nutrition supplementation, or
nutrition counseling independently of each other.
Although these alternative explanations for the birth
outcome differences tend to suggest that the attributed WIC
program effects may be over-estimated, an under-estimation
may be just as likely. The WIC population could be financialt
Women who delivered prematurely, even shortly after joining the WIC
program, would not be administratively excluded; these women would be
switched to the WIC postpartum program and their birth records included in
this study.