Many countries imposed national lockdowns in March of 2020 in response to the coronavirus disease 2019 pandemic. In the Netherlands, the lockdown was associated with an immediate decline in preterm birth.1 Subsequent analyses from other countries and from hospitals and states within the United States reported mixed support for a decline in preterm birth associated with the lockdowns.2–5 In numerous reports, samples were small and difficult to generalize given limited geographic coverage. The sudden change in preterm birth associated with the lockdown described in this literature suggests changes around delivery as opposed to changes in prenatal conditions that existed before the pandemic. In this analysis, we use the census of births in the United States from 2010 to 2020 to characterize monthly changes in preterm birth by method of delivery adjusted for seasonality and trend.
Methods
We use data on preterm birth and maternal characteristics among singleton births from the National Center for Health Statistics’ (NCHS) public use national natality files from 2010 to 2020. The files contain 100% of all registered births in the United States, including home births. Our sample is limited to singleton births (n = 41 394 390). Six percent of singletons from 2010 to 2020 are missing information on inductions or cesarean deliveries and are dropped from our final sample (n = 38 891 271). However, we drop only 0.12% of births between 2016 and 2020 because of improved reporting. The primary outcomes are monthly percentage of US births that are preterm and are delivered by cesarean or induced delivery, and monthly percentage of births that are preterm and are delivered without a cesarean or induced delivery. We use the obstetric estimate of gestational age (GA) to characterize birth as preterm (<37 weeks). The NCHS has used the obstetric estimate of GA as its standard measure since 2014.6 We fit a linear regression to the monthly series of preterm births from January 2010 to February 2020. We include an indicator variable for each month of the year to adjust for seasonality and a linear and quadratic trend term to capture the time-series pattern in our measures. We assume March 2020 is the first month of the national lockdown. President Trump declared a national emergency on March 13, 2020, and 43 states issued stay-at-home orders of varying restrictiveness in the same month.7 We use the model to forecast the path of preterm birth from March to December of 2020 and compare forecasted to actual outcomes. We estimate 95% confidence intervals (CI) as 2 times the SD of the prediction error above or below our forecast. Actual values outside of the CIs indicate a change that is unlikely to occur by chance.
Results
There is a clear quadratic trend in the deseasonalized monthly percentage of preterm births involving cesarean or induced delivery (Fig 1A). The blue solid line is the fitted values up to February 2020. The blue dotted line shows the forecasted values thereafter. Before March 2020, monthly percentage of preterm births involving cesarean or induced delivery never varies >0.15 percentage points from the fitted line in either direction. A close-up from January 2019 to December 2020 shows that the actual series of preterm births (red points) is well below the CI of the forecasted series (Fig 1B) and drops immediately in March 2020 by 0.4 percentage points from the fitted line. After February 2020, the percentage of preterm births involving cesarean or induced delivery is always below the forecasted line and is 0.35 percentage points, on average, from the projected values. The decline is >6 SD of the prediction error. As a sensitivity analysis, we replicate this forecasting exercise using the fitted model from January 2010 to February 2019 and project outcomes from March to December of 2019. If our model fits the data well, there should be little deviation between the actual and predicted rate of preterm birth in the year before the pandemic. All actual rates of preterm birth fall within or near the CI of the forecasted months (Supplemental Fig 3). Given concerns about the measurement of GA, we replicate the analysis in Fig 1 using the rate of low birth weight (<2500 g) involving cesarean or induced delivery.5 The actual rate of low birth weight is consistently below the projected rate beginning in March of 2020 (Supplemental Fig 4). A figure that uses the date of the last normal menses and birth date to measure GA shows a similar pattern in preterm birth to the obstetric estimate of gestation (Supplemental Fig 5). Overall preterm birth fell by 0.53 percentage points (Supplemental Fig 6) with patterns similar to those involving cesarean or induced delivery.
Preterm birth not involving a cesarean or induced delivery falls slightly relative to the forecasted series after February 2020 (Fig 2A and B). The decline is less evident than with preterm birth involving a cesarean or induced delivery. A similar analysis using low birth weight not involving cesarean or induced delivery shows little evidence of a decline after February 2020 (Supplemental Fig 7).
Discussion
The US data suggest a decline in preterm birth after the lockdown in March of 2020. The decline is driven primarily by deliveries that were by cesarean or were induced. We consider this measure an approximate characterization of iatrogenic births, and their absence an approximate characterization of spontaneous preterm birth (acknowledging some mischaracterization because of factors such as premature rupture of membranes in the former). This decline in c-sections and induced deliveries is limited to preterm births as the trend in cesarean and induced deliveries was stable or even increased slightly through 2020 (Supplemental Fig 8).
It could be that less time seeing an obstetrician in person during the lockdown resulted in fewer diagnoses of indications for delivery. As a suggestion of behavioral changes during the pandemic, home births rose from 1.03% of all births in 2019 to 1.26% in 2020.8 Preterm home births, however, were only 0.05% of all births in 2019 and 0.06% in 2020. We interpret the minor decline in preterm births not involving a cesarean or induced delivery as suggestive of the lockdown’s effect, but the evidence is less convincing than the clear decrease in preterm birth involving a cesarean or induced delivery. This slight decline in what we consider a proxy for spontaneous preterm birth may be related to compositional change in cohorts of women who delivered in 2020. The NCHS reports that births declined 4% overall from 2019 but 8% among teens.9 Definitive evidence on how compositional changes affected spontaneous preterm birth awaits data on induced terminations in 2020.
Drs Dench and Joyce conceptualized and designed the study, drafted the initial manuscript, collected the data, and analyzed the data; Dr Minkoff conceptualized and designed the study; and all authors reviewed and revised the manuscript, approved the final manuscript as submitted, and agree to be accountable for all aspects of the work.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
