An Analysis of Healthy Start Prenatal Screening Data: Women Who Choose to Be Screened Versus Those Who Do Not
Since April, 1992 all pregnant women in Florida are to be offered the Healthy Start Prenatal Risk screening. This was mandated by the Healthy Start legislation passed in 1991. In reality, the screening has been offered to approximately 60% of all pregnant women. The purpose of the screening is to identify women who are at greater than average risk for having a poor birth outcome such as a low birth weight or pre-term birth.
The Healthy Start Computer system captures prenatal data for all women who are offered the screening. Periodically, the prenatal data are matched to the latest birth data. The result are records that contain all of the mother's prenatal data with the infant's corresponding birth data where matching records are found. The match is done using the mother's social security number which is recorded on approximately 95% of the prenatal and birth records. Assuming there are no errors in the recording of the social security numbers, this would produce a 90% match rate (.95 X .95 = .9025). Allowing for recording errors, the match rate is probably between 80% and 90%.
These matched records were used to examine the differences between the women who elected to be screened and those who declined to be screened. In table 1, data are presented for the two groups classified according to 12 mutually exclusive categories based on demographic characteristics. In this analysis, LBW means low birth weight and is defined as birth weight below 2500 grams.
In table 1, the overall LBW rate for women who declined the screening is much lower at .0604 (column 6) than the rate for women who accepted the screening which is .0802 (column 5). The risk ratio (column 7) is 1.33. However, the differences in these rates and the risk ratios are generally much lower for the specific demographic categories. This suggests that the large difference in the overall LBW rates is due to different demographic composition of the two groups.
In table 2, the demographic composition is illustrated. The group of women who declined the screening have a much larger proportion of non-black, married, women age 18 to 39 with at least a high school education. In column 5, the percentage of these women is 64.52% for the group that declined the screening. The percentage is much lower at 28.83% for the women who accepted the screening. These women have the lowest LBW rates of the 12 demographic categories so the larger proportion of these women in the groups of women who declined the screening would account for at least part of the difference in the LBW rates.
Table 3 attempts to answer the question, what would happen if the group of women who accepted the screening experienced the same rates of LBW as the group of women who declined the screening. This is done by applying the LBW rates for the women who declined the screening (column 2) to the number of women who consented to the screening in each demographic category (column 1). The result is the adjusted LBW births for women who consented to the screening. This is an estimate of the LBW births that would occur if the women consenting to the screening in each demographic category experienced the same LBW rates as the women who declined the screening in the same demographic category.
The total adjusted LBW births for women consenting to the screening is 6411 (column 3). The actual number is 6579 (table 1 column 1) which means that some improvement would be made if the women who consented to be screened had the same LBW rates as women who declined to the screening. The standardized morbidity ratio (SMR) at the bottom of table 3 shows that risk of LBW for the women who consented to be screened is 1.03 times what it would be if these women had the same LBW rates as the women who declined to be screened. This is much lower than the risk ratio in table 1 of 1.33. The 1.33 reflects the influence of the demographic differences between the two groups and the 1.03 is adjusted so that the demographic influence is removed.
The difference between the adjusted and actual LBW rates is significant at the 95% level, but just barely, as shown by the confidence intervals for the SMR at the bottom of table 3. This difference would not be statistically significant if the confidence interval included 1. Since the interval is 1.01 to 1.04, 1 is not included, but it is close.
In conclusion, the difference in LBW rates for the women who consented to be screened versus women who declined the screening is mainly due to the different demographic characteristics of the two groups. After the differences due to demographic factors are removed, there is a small difference that remains. One possible explanation for the remaining difference is women who know their pregnancies are going well are more likely to decline the screening and women who think they may have problems are more likely to accept the screening. Another theory is the poorer women are more likely to accept the screening. Since low income women tend to have higher LBW rates, this would account for the difference.
Finally, it should be noted that this analysis includes only women who were offered the prenatal risk screening. Table 4 shows that the demographic characteristics are different for women who gave birth in 1993 compared to women who were offered the prenatal screening in 1993. In general, the population of women who were offered the screening has higher proportions of women who are unmarried and teenagers. This supports the theory that prenatal care providers tend to offer the screening to women who appear to the provider to be at risk and tend not to offer the screening to women who the provider believes are not at risk.