METHODS

We analyzed Louisiana 2003–2007 birth records–Medicaid-linked data. LBW was defined as birth weight <2500 g and PTB as birth at <37 weeks' gestation. The clinical estimate for gestation was used; if it was missing, gestational age was imputed by sex and weight.Reference Taffel, Johnson and Heuser21Reference Kotelchuck22

Adequacy of prenatal care was calculated using the Kotelchuck index, based on reported initiation and number of prenatal care visits,Reference Kotelchuck23 and categorized into inadequate, intermediate, adequate, and adequate-plus levels. Cesarean section (either primary or repeat) was separated from vaginal deliveries.

Region of residence was classified according to where mothers reported residing at birth, not where the birth took place. Louisiana is divided into 9 health regions. Region 1 consists of Orleans, Jefferson, Plaquemines, and St Bernard parishes; this is the area that was the most strongly hit by the hurricane and subsequent flooding. We examined birth outcomes (LBW and PTB) among residents of the state and this region for the 2 years before and after the hurricane, and compared birth outcomes for women who delivered inside their region of residence to those who delivered outside. We also examined Orleans Parish alone (city of New Orleans).

Frequency and rate (percentage) of birth outcomes and odds ratios (ORs) and 95% confidence intervals (CIs) were calculated comparing the year before and after and 2 years before and after Hurricane Katrina. Multivariable logistic models were created to compare birth outcomes before and after the hurricane, adjusting for demographic (race, education, Hispanic ethnicity, marital status) and health behaviors and biological risk factors (multiple gestation, parity, time since last birth or termination [interpregnancy interval], tobacco). There were 4343 (1.7%) records with missing data on at least 1 of the covariates, most commonly interpregnancy interval. These observations were omitted from multivariable regressions.

Next, we examined the effects of the storm by racial/ethnic group. All of the analyses were also performed separately for African American, white, and Hispanic women (the major racial/ethnic groups in Louisiana) by stratification and modeling using interaction terms. We compared the racial disparity (non-Hispanic black to non-Hispanic white; Hispanics were omitted from this calculation) in the outcomes before and after the storm, determining whether it differed across the 2 time periods.

We also wanted to put boundaries on the possible effects of misreporting of state of residence. Women who did not report residing in Louisiana at the time of the birth were eliminated from the analysis, but those who gave birth outside the state but reported residing in Louisiana were included (n = 2774) . A total of 4199 fewer births were recorded in Louisiana in the year after Katrina compared to the year before. It is likely that many women who gave birth outside the state did not report themselves as Louisiana residents; therefore, we wanted to estimate the maximum possible effect of Katrina on birth outcomes. For instance, if every birth out of state was PTB or LBW, this would cause us to underestimate the effect of the storm. (The other possibilities for the decline in births are reduced fertility and fecundability, and we do not have data to address those, nor would changes in these affect the rates of the outcomes in the births that did occur.)

To determine the amount of underestimation, we needed to estimate the total number of births that occurred. We chose to assume that the number of births to Louisiana women in the year after Katrina was the same as the year before Katrina. Louisiana population growth had been small in the year before (<0.2%24), and because our goal was to provide an upper bound for the effects, a decrease in births would lead to estimates lower than our upper bound. Finally, we needed to estimate the risk status of the 4199 births that we hypothesized were not reported as attributed to Louisiana residents. Because we had no way of knowing this, we tested the effect of different assumptions, assuming these births had a similar risk of PTB/LBW to women in the state overall and assuming they had greater risk. Before Katrina, the overall risk of PTB in the state was 14% and LBW was 11%. We hypothesized different risks in the 4199 births ranging from 2% to 25%, then added that “population” to the known data, determining what the overall rates of PTB/LBW would have been in the year after Katrina under those assumptions. This assumption allowed us to put boundaries on the maximum effect of the storm. This analysis of deidentified data was approved by the Tulane institutional review board.

RESULTS

There were 128 624 births to Louisiana women in the 2 years before Hurricane Katrina and 126 041 in the 2 years after Katrina. In region 1, the corresponding numbers were 28 287 and 17 955, and in New Orleans, 13 313 and 5698. The year-by-year data are presented in Tables 1 and 2. The demographic profile of mothers in Louisiana changed across the years studied (Table 1). When comparing the year pre-Katrina to the year post-Katrina, proportions of non-Hispanic white women giving birth were greater during the post period (state: pre- 54.1%, post- 56.5%; region 1: pre- 36.0%, post- 48.8%; New Orleans: pre- 17.2%, post- 31.7%). In addition, there were fewer teen births (state: pre- 14.9%, post- 14.1%; region 1: pre- 13.9%, post- 11.9%; New Orleans: pre- 16.0%, post- 13.2%), greater proportions were married (state: pre- 51.2%, post- 51.5%; region 1: pre- 43.9%, post- 51.6%; New Orleans: pre- 32.8%, post- 45.5%), and fewer had less than a high school education (state: pre- 22.1%, post- 20.7%; region 1: pre- 22.5%, post- 18.5%; New Orleans: pre- 23.7%, post- 17.1%). The proportion of Hispanic women rose substantially in region 1 (from 5.9% to 8.6%) and New Orleans (from 2.4% to 5.3%). Medicaid-funded births did not show a consistent trend; they rose in the state as a whole, but declined in the year after Katrina in region 1 and New Orleans.

TABLE 1 Demographic and Social Changes in Women Giving Birth in Louisiana Before and After Hurricane Katrina

TABLE 2 Perinatal Outcomes Before and After Hurricane Katrina, State of Louisiana

For the state as a whole, rates of LBW rose in the 2 years after Katrina compared with the 2 years before, but preterm birth did not (P = .65; Table 1). Adjustments for covariates did not eliminate the LBW association; however, even before Katrina, LBW had been rising (data not shown). In region 1, rates of LBW declined and PTB declined. In New Orleans, both LBW and preterm birth fell. Rates of cesarean section and inadequate prenatal care rose after Katrina for the state, the region, and Orleans Parish. There was a particularly sharp rise in inadequate prenatal care in the year after Katrina. Results were similar when data were limited to singleton births only (data not shown).

The rise in LBW in the state was strongest for non-Hispanic black women, as was the fall in PTB in region 1 and New Orleans (Table 3). The rise in inadequate prenatal care was seen in all of the racial/ethnic groups, but it was strongest in non-Hispanic whites and Hispanics. The racial disparities in LBW did not change after Katrina. The racial disparity was lower in the years after Katrina in New Orleans for PTB only. The racial disparity in inadequate prenatal care was reduced at all levels after Katrina. These patterns were also seen when disparities were examined on an absolute instead of a relative scale (data not shown).

TABLE 3 Perinatal Outcomes Before and After Katrina, by Racial/Ethnic Group, State of Louisiana

Louisiana residents who gave birth outside their home region in the year after Katrina were at greater risk of LBW (adjusted OR [aOR] 1.11, 95% CI 1.02–1.21) and PTB (aOR 1.10, 95% CI 1.03–1.18) compared with those who gave birth in their home region. A similar pattern had been seen in the year before the storm (aOR for LBW 1.28, 95% CI 1.17–1.40; for PTB 1.17, 95% CI 1.07–1.27). Region 1 residents who gave birth outside their home region were not at increased risk and New Orleans residents were at reduced risk both before (aOR for LBW 0.75, 95% CI 0.59–0.94; aOR for PTB 0.64, 95% CI 0.51–0.80) and after Katrina (aOR for LBW 0.73, 95% CI 0.53–1.00; aOR for PTB 0.66, 95% CI 0.50–0.88). There were no significant differences in these associations by race/ethnicity.

Finally, we examined the possible effects of the population reduction on LBW and PTB. We examined the effects of the “missing” births—the number of births declined in the year after Katrina. For the state as a whole, to have a true OR of ≥1.10 comparing the 2 years after to the 2 years before Katrina, the risk in this “missing” population would be required to be at high risk (LBW and PTB>25%), substantially higher than the 2004 risks of 11% and 13.4%. For region 1, the number of “missing” births is large (n = 10 332) relative to the total number of births (n = 17 955) and could have a significant influence on the results. To hypothesize a truly increased risk of ≥10%, however, >15% of the women would had to have given birth to LBW babies and >20% preterm (see Supplementary Table S1, http://www.dmphp.org/misc/harville.pdf).

COMMENT

Hurricane Katrina was not associated with an increased risk of LBW and PTB in those areas most affected, and in fact, some areas had reduced risks of some poor birth outcomes.Reference Hamilton, Sutton, Mathews, Martin and Ventura20 Our analysis indicates that this was somewhat, although not completely, due to changes in the risk profile of the population. After Katrina, the population giving birth was more likely to have characteristics associated with lower risk: more educated, less likely to be teenaged, more likely to be married, and more likely to be non-Hispanic white or Hispanic.Reference Tucker and McGuire25Reference Lang, Lieberman and Cohen26 Medicaid-funded births increased in the state as a whole but initially decreased in region 1 and New Orleans. Medicaid coverage was extended to cover many victims of Katrina,27 which makes interpreting the patterns difficult. Population changes partially accounted for the reduction in LBW in region 1 and fully for the reduction in LBW in New Orleans. For PTB, population changes only partially accounted for the reduction in New Orleans and did not account for the reduction in region 1. Rates of cesarean section rose across the state, region, and parish and were not substantially affected by adjustment for population changes. Population shifts partially accounted for the increased proportion of inadequate prenatal care in region 1, but they caused an underestimation of the likelihood of inadequate prenatal care in New Orleans.

On balance, one would expect that women displaced from the New Orleans area would have had the most severe experiences of the hurricane, but there was no evidence for increased risk in displaced women within Louisiana. Our data are similar to those of Rich-Edwards et al and Endara et al, which addressed the terrorist attacks of September 11, 2001, in not finding a large effect due to disaster.Reference Rich-Edwards, Kleinman, Strong, Oken and Gillman28Reference Endara, Ryan, Sevick, Conlin, Macera and Smith29 If the women whose births were not recorded were at high risk, then an overall small increased risk in LBW/PTB due to the storm is plausible but far from proven. Studies of stress and pregnancy often report odds ratios of 1.5 to 3.0 with increased stress,Reference Paarlberg, Vingerhoets, Passchier, Dekker and Van Geijn30Reference Hobel, Goldstein and Barrett31 and our data are not compatible with that effect size. Another possibility for the lack of major increases in risk would be an increase in miscarriage or reduction in fertility. We do not have data to address this question.

The effects of the hurricane on birth outcomes did not vary substantially by race and, if anything, African American women, the group hardest hit in New Orleans,Reference Logan1 had the greatest reduction in PTB after the hurricane. Racial disparities were not exacerbated. Although these results largely indicate minimal differential effects on maternal and child health, 2 aspects of the context need to be remembered. One, rates of LBW, PTB, and inadequate prenatal care remained unacceptably high in African Americans and were close to double that of non-Hispanic whites. Two, people with the fewest resources (and likely the highest health risk) were also those with the least say in where they evacuated and whether they could return.Reference Logan1 It is likely that the highest-risk group of African Americans was not able to return to New Orleans, indicated by the large reduction in PTB in the area. It is possible that this is reflected in the increase in LBW among non-Hispanic black women in the state as a whole, but not in region 1 or New Orleans. There is substantial heterogeneity of resources within racial/ethnic groups.

The quality of vital statistics data are variable. Some birth outcomes (eg, birth weight) were recorded accurately, whereas other complications tended to be underreported.Reference Reichman and Schwartz-Soicher32Reference Vinikoor, Messer, Laraia and Kaufman33 Birth certificate recording may have been less accurate after the storm than before it. Women who were separated from their usual health care providers may have been more likely to have improperly dated pregnancies; however, this usually produces higher rather than lower rates of PTB,Reference Wingate, Alexander, Buekens and Vahratian34 and it is difficult to understand how these issues would have affected birth weight data. Cesarean section should be accurately reported because it occurs close in time and normally in the same place as the completion of the birth certificate. Prenatal care is the outcome that is most vulnerable to problems in reporting; however, many women probably missed a prenatal visit or 2 or postponed initiation of prenatal care. In general, women who move residences during pregnancy are more likely to initiate prenatal care late (or be recorded as having initiated prenatal care late).Reference Fell, Dodds and King35 Nonetheless, vital statistics data are reliable for many outcomes, collected systematically, and allow for examination of large populations and the detection of small effects.

CONCLUSIONS

We found that Hurricane Katrina had significant effects on the population giving birth and on obstetric health care, even after adjusting for demographic and risk profile changes. We did not, however, find that Hurricane Katrina had major effects on birth outcomes such as LBW or PTB. In addition, we did not find that it exacerbated racial disparities, largely due to similar effects across population groups. The interpretation of research on the effect of disaster on pregnancy needs to be tempered by a clear understanding of population shifts. Future research should focus on identifying particularly high-risk women, as well as trying to determine the effects of disaster on fertility and spontaneous abortion. The population data suggest that the largest concerns for clinicians and disaster planners should be ensuring normal care for most women and focusing care for the smaller group of high-risk women, rather than preparing for an enormous increase in adverse birth outcomes.