key: cord-0832166-61219o43
authors: CANIGLIA, Ellen C.; MAGOSI, Lerato E.; ZASH, Rebecca; DISEKO, Modiegi; MAYONDI, Gloria; MABUTA, Judith; POWIS, Kathleen; DRYDEN-PETERSON, Scott; MOSEPELE, Mosepele; LUCKETT, Rebecca; MAKHEMA, Joseph; MMALANE, Mompati; LOCKMAN, Shahin; SHAPIRO, Roger
title: Modest reduction in adverse birth outcomes following the COVID-19 lockdown
date: 2020-12-24
journal: Am J Obstet Gynecol
DOI: 10.1016/j.ajog.2020.12.1198
sha: 7218877b2437a71485c3bd2f49a40b29330512d5
doc_id: 832166
cord_uid: 61219o43

Background Widespread lockdowns imposed during the COVID-19 crisis may impact birth outcomes. Objective To evaluate the association between the COVID-19 lockdown and the risk of adverse birth outcomes in Botswana. Study Design In response to COVID-19, Botswana enforced a lockdown that restricted movement within the country. We used data from an ongoing nationwide birth outcomes surveillance study to evaluate adverse outcomes (stillbirth, preterm birth, small-for-gestational-age [SGA], and neonatal death) and severe adverse outcomes (stillbirth, very preterm birth, very SGA, and neonatal death) recorded pre-lockdown (January 1 – April 2), during lockdown (April 3 – May 7), and post-lockdown (May 8 – July 20). Using difference-in-differences analyses, we compared the net change in each outcome from the pre-lockdown to lockdown periods in 2020 relative to the same two periods in 2017-2019, and the net change in each outcome from the pre-lockdown to post-lockdown periods in 2020 relative to the same two periods in 2017-2019. Results 68,448 women delivered a singleton infant in 2017-2020 between January 1 and July 20 and were included in our analysis (mean [interquartile range] age of mothers, 26 [22,32] years). Across the included calendar years and periods, the risk of any adverse outcome ranged from 27.92% to 31.70% and the risk of any severe adverse outcome ranged from 8.40% to 11.38%. The lockdown period was associated with a 0.81 percentage point reduction (95% CI, -2.95%, 1.30%) in the risk of any adverse outcome (3% relative reduction) and a 0.02 percentage point reduction (95% CI, -0.79%, 0.75%) in the risk of any severe adverse outcome (0% relative reduction). The post-lockdown period was associated with a 1.72 percentage point reduction (95% CI, -3.42%, -0.02%) in the risk of any adverse outcome (5% relative reduction) and a 1.62 percentage point reduction (95% CI, -2.69%, -0.55%) in the risk of any severe adverse outcome (14% relative reduction). Reductions in adverse outcomes were largest among women with HIV and among women delivering at urban delivery sites, driven primarily by reductions in preterm birth and SGA. Conclusions Adverse birth outcomes decreased from the pre-lockdown to post-lockdown periods in 2020, relative to the change during the same periods in 2017-2019. Our findings may provide insights into associations between mobility and birth outcomes in Botswana and other low- and middle-income countries.

Using a birth outcomes surveillance study in Botswana including over 119,000 births, we found 31 a modest reduction in adverse birth outcomes following the COVID-19 lockdown. The post-lockdown period in 2020 was associated with a 1.72 percentage point reduction 40 (95% CI, -3.42%, -0.02%) in the risk of any adverse outcome (stillbirth, preterm birth, small-41 for-gestational-age [SGA] , and neonatal death) and a 1.62 percentage point reduction (95% 42 CI, -2.69%, -0.55%) in the risk of any severe adverse outcome (stillbirth, very preterm birth, 43 very SGA, and neonatal death). Reductions in adverse outcomes were largest among women 44 with HIV and among women delivering at urban sites, driven primarily by reductions in 45 preterm birth and SGA.

Our data provide the first evaluation from Sub-Saharan Africa of the impact of a COVID-19 48 lockdown on birth outcomes, and suggest a modest reduction in preterm birth and SGA 49 following the lockdown period.

Widespread lockdowns imposed during the COVID-19 crisis may have impacted birth outcomes 89 worldwide, but the magnitude and direction of these effects remain uncertain. A hospital in 90 Ireland reported a 73% decrease in the incidence of very low birthweight infants from January to 91 April 2020 compared with the same period in the previous two decades, 1 a study in Denmark 92 found a 90% decrease in the incidence of preterm birth during the lockdown period from March 93 12 to April 14 compared with the same period during the previous five years, 2 and a study in The 94 Netherlands found reductions in the incidence of preterm birth across various time windows 95 surrounding the implementation of COVID-19 mitigation measures (e.g., an odds ratio of 0.77 96 comparing 2 months after and 2 months before March 9th). 3 In the United States, one hospital in 97 Nashville estimated there were 20 percent fewer infants in the neonatal intensive care unit in 98 March than during that month in previous years. 4 Still, many hospitals around the world reported 99 no differences in preterm births during the lockdown 4 and there is concern that lockdown 100 restrictions could also lead to increases in more severe outcomes such as stillbirth and neonatal 101 death. A study using data from a London hospital found a higher incidence of stillbirth during 102 the COVID-19 pandemic period compared with the period immediately prior to the pandemic, 103 but no difference in preterm birth. 5 A study in nine hospitals across Nepal found a higher 104 incidence of stillbirth, neonatal mortality, and preterm birth during the 9.5-week lockdown 105 compared with the 12.5 weeks prior to the lockdown. 6 Finally, a study using data from four 106 hospitals in western India found a higher incidence of stillbirth during the 10 weeks following 107 the lockdown compared with the 10 weeks prior to the lockdown. 7 The mechanisms underlying 108 all of these reported findings are speculative, and in most cases need to be considered in the To date there have been no studies on the impact of the COVID-19 lockdowns on adverse 112 pregnancy outcomes in Sub-Saharan Africa, a region with one of the greatest burdens of adverse 113 pregnancy outcomes, and risk factors that are often distinct from those in high-income countries.

Despite having only three reported SARS-CoV-2 cases at the time, 9 Botswana announced a state 115 of emergency due to COVID-19 on March 31, 2020 10 and a nationwide lockdown commenced at 116 midnight on April 2, 2020. 11-13 After the initial 28 day period, the lockdown was extended until 117 May 7, 2020. 14 Movement restrictions were gradually lifted between May 8 and May 22, 2020. 15-118 19 While SARS-CoV-2 swept through South Africa, infecting 364,328 people as of July 20th, 119 2020 20 , COVID-19 largely spared Botswana during this early phase of the pandemic; through 120 July 20 th there were 522 cases of SARS-CoV-2 and no confirmed cases in pregnancy. 20 This 121 provides a unique opportunity to isolate the impact of the lockdown from any direct impact of 122 SARS-CoV-2.

The Tsepamo Study has been conducting birth outcomes surveillance at delivery hospitals 124 throughout Botswana since August 2014 and includes data from more than 119,000 births. In this 125 analysis, we used Tsepamo data to estimate the risk of adverse birth outcomes prior to (January 1 126 -April 2), during (April 3 -May 7), and after (May 8 -July 20) the COVID-19 national 127 lockdown in 2020, and compared these risks with the same three time-periods in 2017-2019. We January 1 and July 20 were included in our analysis (in Botswana, pregnancies that end before 142 24 weeks are considered miscarriage and admitted to the general medical wards). We defined 143 January 1 -April 2 as the period prior to the lockdown ("pre-lockdown"), April 3 -May 7 as the 144 period during the lockdown ("lockdown"), and May 8 -July 20 as the period following the 145 lockdown ("post-lockdown"). We compared the lockdown year, 2020, with the previous three 

Preterm birth was defined as a birth at less than 37 weeks' completed gestation and very preterm 160 was a birth at less than 32 weeks' completed gestation. Tertiary outcomes were birth at less than 161 34 weeks' completed gestation 27 and continuous gestational age at delivery. Gestational age was 162 calculated at the time of delivery by the midwife using the estimated date of delivery determined 163 during antenatal care, typically using reported LMP. SGA was defined as less than the 10 th 164 percentile and very SGA as less than the 3 rd percentile of birthweight by gestational age 165 according to the Intergrowth-21 norms. 28,29 Neonatal deaths included deaths within 28 days of 166 birth among infants who had never left the hospital.

Statistical analysis 168 We used a difference-in-differences analysis to assess the relationship between the lockdown and 169 each outcome. That is, we compared the change in each outcome from the pre-lockdown to 

In sensitivity analyses we adjusted our estimates for individual-level demographic variables 183 (HIV status, calendar year of delivery, age, occupation, education, parity, gravity, marital status, 184 delivery location, smoking status, and use of alcohol), and extended the lockdown period through 185 May 21 to include the two-week period where restrictions were gradually lifted.

Finally, we plotted the weekly risk of the primary outcomes over a 28-week period (January 3 - 

A total of 68,448 women delivered a singleton infant in 2017-2020 between January 1 and July 194 20 and were included in our analysis. Table 1 shows the number of births during the pre- Shelter-in-place adherence: Figure 1 shows the 24-hour staying-put percentage from February 28 210 to July 24 2020, by region in Botswana. Staying-put percentage increased from 10% to 40-50% 211 when the nationwide lockdown was instituted on April 3, gradually decreased following the 212 phased relaxation of extreme social-distancing measures beginning on May 8, and was consistent 213 with pre-lockdown levels by June 5. Changes in staying-put percentage over time were 214 consistent across the country.

Birth outcomes: Table 2 shows the net change in the risk of each outcome from the pre- Table 1 ).

The reduction in both primary outcomes during the lockdown period was larger among women 233 with HIV and among women with salaried employment (Table 3) [95% CI, -5.01%, -0.85%] or 21% relative reduction for any severe adverse outcome). In a post-242 hoc analysis, we calculated that for a woman with HIV delivering at an urban delivery site (9.3% 243 of study population), there was a 6.31 percentage point reduction (95% CI, -14.21%, 1.59%) in 244 the risk of having any adverse outcome (16% relative reduction) and a 2.17 percentage point 245 reduction (95% CI, -7.88%, 3.55%) in the risk of having any severe adverse outcome (13% 246 relative reduction) during the lockdown period. In this same subgroup, there was a 3.43 247 percentage point reduction (95% CI, -9.64%, 2.77%) in the risk of having any adverse outcome 248 (9% relative reduction) and a 3.52 percentage point reduction (95% CI, -8.01%, 0.96%) in the 249 risk of having any severe adverse outcome (22% relative reduction) during the post-lockdown 250 period.

Adjusting for individual-level demographic variables and extending the lockdown period by two 252 weeks had no material impact on our estimates (data not shown). 

Our findings are consistent with some previous studies that found decreases in the risk of low 279 birth weight 1 and preterm birth 2,3 following COVID-19 lockdowns, although the magnitude of 280 our findings was smaller. Our results differ from studies in London, Nepal, and India that found decreasing exposure to air pollution. 1,2,4 In Botswana, we found that the shelter-in-place order 299 successfully led to more people staying-put, which could have reduced physical labor, exposure 300 to infections and air pollution, and some sources of stress. The food insecurity mitigation 301 strategy implemented in Botswana could have increased nutritional support during the lockdown, 302 but its impact remains unknown. It is also possible that the lockdown led to a reduction in 303 preterm iatrogenic delivery. In addition, we did not find any evidence that the shelter-in-place 304 order negatively impacted access to medications for the 23% of women living with HIV. We saw 305 greater reductions in adverse outcomes among women delivering at urban delivery sites, women 306 with HIV, and women with salaried employment, suggesting that the lockdown could have 307 impacted the daily lives of these women to a larger extent. While the greater reduction in adverse 308 outcomes among women with HIV could be due to these women being more likely to deliver at 309 J o u r n a l P r e -p r o o f 15 urban deliver sites, it is also possible that sheltering in place directly impacted adverse outcomes 310 in this population, for example through reducing inflammation. It is possible that stay-at-home 311 orders had less of an impact on women in rural areas and women without salaried employment 312 because these women may have continued physical labor such as farming during the lockdown 313 period. It is also possible that the stay-at-home order increased stress 44 , anxiety, and 314 undernutrition (despite mitigation strategies), especially among those who were food insecure 315 and economically disadvantaged. The reduction in adverse outcomes was greater (albeit modest) 316 in the post-lockdown period and negligible in the lockdown period. A plausible explanation for 317 this finding is that the lockdown had a delayed effect on pregnancy outcomes, related to factors 318 in the second trimester or early in the third trimester. 

Further studies are needed to identify both the mechanism and the gestational window for 322 potential benefits related to decreasing movement during pregnancy, and factors associated with 323 pregnancy outcomes during pandemics. While our findings may not be generalizable to other 324 settings with different distributions of risk factors for adverse birth outcomes (such as maternal 325 nutrition, age, and HIV prevalence 45 ), they may also provide insight into potential interventions 326 to reduce unknown causes of adverse outcomes. 

Difference-in-differences analyses rely on the assumption that the trend in adverse outcomes in 330 2017-2019 would be parallel to the trend in adverse outcomes in 2020 in the absence of the 331 lockdown. 46 Our finding that the weekly trend in adverse outcomes during the pre-lockdown 332 16 period was similar in 2017-2019 compared with the same period in 2020 provides support that 333 this "parallel trends" assumption may approximately hold. In addition, we found little variation 334 in the demographic characteristics of women delivering throughout the study period. However, 335 the parallel trends assumption would not be met if other changes occurred in Botswana at the 336 same time as the lockdown that also impacted adverse outcomes. Difference-in-differences 337 analyses also require an assumption of "strict exogeneity" that the choice to impose a lockdown 338 was not determined by the pre-lockdown risk of adverse outcomes. 46 Since the lockdown was 339 imposed exclusively to stop the spread of COVID-19, this assumption is likely to hold.

Our study has important limitations. First, our analysis only captures women delivering at a 342 hospital included in the surveillance study. If women were more likely to deliver at home or at a 343 local hospital not included in the surveillance study following the lockdown, our results could be 344 biased. We found that the proportion of births during the post-lockdown period in 2020 was 345 slightly lower than the proportion of births during the same period in 2017-2019; however, it is 346 unlikely that this ~1% decrease would explain our findings. Second, our analysis only captures 347 births after at least 24 weeks gestation. If the risk of miscarriage changed during the lockdown 348 period, we would not be able to capture this. Third, we were not able to assess individual-level 349 mobility. While staying-put percentage increased during the lockdown period, we were not able 350 to evaluate the relationship between individual-level mobility and adverse outcomes. In conclusion, we found a 1.72 percentage point reduction (5% relative reduction) in any adverse 354 outcome and a 1.62 percentage point reduction (14% relative reduction) in any severe adverse 355 outcome from the pre-lockdown to post-lockdown periods in 2020, relative to changes during the 356 same two periods in 2017-2019. We found no meaningful differences in adverse birth outcomes 357 from the pre-lockdown to lockdown periods. The greatest impact was on preterm birth and SGA, 358 and among women with HIV and those delivering in urban areas. While these reductions were 359 modest, they may provide insights into identifying potential interventions to reduce adverse birth 360 outcomes in Botswana and in other low-and middle-income countries throughout the world. Table 1 . Characteristics of women giving birth in Botswana during the pre-lockdown (January 1-April 2), lockdown (April 3-May 7), and post-lockdown (May 8-July 20) periods in 2020 and in the same calendar periods in 2017-2019.

Year Pre-lockdown period (Jan 1-April 2) Table 2 . Risk difference and difference-in-differences (95% CI) of each adverse birth outcome during the pre-lockdown (January 1-April 2), lockdown (April 3-May 7), and post-lockdown (May 8-July 20) periods in 2020 and in the same calendar periods in 2017-2019.

Pre-lockdown period (Jan 1-April 2), Risk (n/total) Table 3 . Difference-in-differences (95% CI) of the composite adverse birth outcomes during the pre-lockdown (January 1-April 2), lockdown (April 3-May 7), and post-lockdown (May 8-July 20) periods in 2020 and in the same calendar periods in 2017-2019, by key subgroups.

Difference-in-differences (95% CI) Pre-lockdown period risk Lockdown vs. pre-lockdown* Post-lockdown vs. prelockdown** Any adverse outcome Overall 31.61% -0.81% (-2.95%, 1.30%) -1.72% (-3.42%, -0.02%) Women with HIV 37.69% -3.51% (-9.40%, 2.38%) -3.86% (-6.32%, -1.39%) Women without HIV 29.58% 0.09% (-1.91%, 2.10%) -0.98% (-3.12%, 1.17%) Urban a delivery sites 33.79% -0.60% (-4.33%, 3.13%) -3.37% (-6.30%, -0.44%) Rural b delivery sites 30.23% -0.88% (-3.74%, 1.98%) -0.83% (-2.17%, 0.52%) Nulliparous women 33.43% -1.39% (-4.21%, 1.43%) -0.64% (-2.98%, 1.71%) Parous women 30.39% -0.47% (-3.14%, 2.20%) -2.30% (-4.48%, -0.12%) Women with salaried employment 27.36% -2.41% (-5.99%, 1.17%) -2.45% (-5.80%, 0.90%) Women without salaried employment 33.73% -0.03% (-2.12%, 2.06%) -1.40% (-3.61%, 0.80%) Women with HIV delivering at urban a site 39.67% -6.31% (-14.21%, 1.59%) -3.43% (-9.64%, 2.77%)

Overall 11.26% -0.02% (-0.79%, 0.75%) -1.62% (-2.69%, -0.55%) Women with HIV 13.78% -1.05% (-3.94%, 1.85%) -2.26% (-4.14%, -0.38%) Women without HIV 10.29% 0.49% (-1.04%, 2.01%) -1.33% (-2.37%, -0.28%) Urban a delivery sites 14.08% -0.15% (-2.79%, 2.50%) -2.93% (-5.01%, -0.85%) Rural b delivery sites 9.48% 0.02% (-0.69%, 0.74%) -0.98% (-2.28%, 0.32%) Nulliparous women 11.70% 0.71% (-1.34%, 2.75%) -0.96% (-3.26%, 1.35%) Parous women 10.91% -0.33% (-1.57%, 0.91%) -1.93% (-2.94%, -0.93%) Women with salaried employment 10.17% -0.53% (-2.04%, 0.98%) -1.48% (-2.73%, -0.22%) Women without salaried employment 11.81% 0.23% (-0.87%, 1.34%) -1.69% (-3.11%, -0.27%) Women with HIV delivering at urban a site 16.22% -2.17% (-7.88%, 3.55%) -3.52% (-8.01%, 0.96%) *Calculated as the difference between the change in each outcome from the pre-lockdown to lockdown periods in 2020 and the change in each outcome during the same two calendar periods in 2017-2019. ** Calculated as the difference between the change in each outcome from the pre-lockdown to post-lockdown periods in 2020 and the change in each outcome during the same two calendar periods in 2017-2019. 

Reduction in preterm births during the COVID-19 lockdown in 394 Ireland: a natural experiment allowing analysis of data from the prior two decades

Changes in premature birth rates during the 397 Danish nationwide COVID-19 lockdown: a nationwide register-based prevalence proportion study

During Coronavirus Lockdowns, Some Doctors Wondered: Where Are the Preemies? 403 The New York Times

Change in the Incidence 405 of Stillbirth and Preterm Delivery During the COVID-19 Pandemic

Effect of the COVID-19 pandemic response on intrapartum 407 care, stillbirth, and neonatal mortality outcomes in Nepal: a prospective observational study

COVID-19 outbreak and decreased hospitalisation of pregnant 410 women in labour

Republic of Botswana. Government Gazette Extraordinary 2nd April, 2020. Gaborone; 2020. 420 13. Republic of Botswana. Government Gazette Extraodinary 31st March, 2020. Gaborone; 2020. 421 14. Republic of Botswana

Aggregated mobility data could help fight COVID-19

The Late Preterm: A Population at Risk

International standards for newborn weight, length, 448 and head circumference by gestational age and sex: the Newborn Cross-Sectional Study of the 449 INTERGROWTH-21st Project

Easy SAS calculations for risk or prevalence ratios and differences

School collective 455 occupation movements and substance use among adolescents: A school-level panel design

Sas Institute; 2012. 458 33. Bodnar LM, Simhan HN. The prevalence of preterm birth and season of conception. Paediatric 459 and perinatal epidemiology

Time trends and sociodemographic determinants of 461 preterm births in pregnancy cohorts in Matlab

Maternal body composition, HIV infection and other 466 predictors of gestation length and birth size in Zimbabwe

Seasonality in birth weight: review of global patterns and 469 potential causes

Effects on birth weight and perinatal mortality of 473 maternal dietary supplements in rural Gambia: 5 year randomised controlled trial

Seasonal variation of birth weight distribution in Morogoro

Emulating a target trial of antiretroviral therapy regimens 484 started before conception and risk of adverse birth outcomes

5%) 8,316 (17.3%) 17,396 (36.2%) (% of births during

Women living with HIV

8%) 3,380 (40.6%) 7,020 (40.4%) 2020 3,146 (32.7%) 1,124 (31.3%) 2,284 (31.9%) Salaried occupation

Antenatal visits, median (IQR)

56%, 0.62%) -0.01% (-0.90%, 0.87%) -0.93% (-1.54%, -0.33%) -0.04% (-1.03%, 0.94%) -0.96% (-1.87%, -0.05%) Neonatal death 2017-2019

*Calculated as the difference between the change in each outcome from the pre-lockdown to lockdown periods in 2020 and the change in each outcome during the same two calendar

Calculated as the difference between the change in each outcome from the pre-lockdown to post-lockdown periods in 2020 and the change in each outcome during the same two calendar

Data are the average number of Facebook users with location services turned-on that were present in the same 600x600 m grid location over a 24-hour period. Presence in the same location considered as GPS ping in at least three different time blocks of the day. Threshold: at least 300 unique users present

Appendix Table 1 . Mean difference, risk difference and difference-in-differences (95% CI) of gestational age at delivery and birth at <34 weeks gestation during the pre-lockdown (January 1-April 2), lockdown (April 3-May 7), and post-lockdown (May 8-July 20) periods in 2020 and in the same calendar periods in 2017-2019.Pre-lockdown period (Jan 1-April 2) Lockdown period (April 3-May 7)Post-lockdown period (May 8-July 20)