Comparison of Estimated Rates of Coronavirus Disease 2019 (COVID-19) in Border Counties in Iowa Without a Stay-at-Home Order and Border Counties in Illinois With a Stay-at-Home Order.

Key Points Question Was the stay-at-home order in Illinois associated with different rates of coronavirus disease 2019 (COVID-19) compared with Iowa, which did not issue a stay-at-home order? Findings This cross-sectional study of border counties in Iowa and Illinois used difference-in-differences design and found an increase in estimated rates of COVID-19 cases per 10 000 residents in the border counties in Iowa compared with the border counties in Illinois after a stay-at-home order was implemented in Illinois but not in Iowa. Meaning The results of this study suggest that issuing a stay-at-home order in Iowa may have helped limit the spread of COVID-19 cases in that state.


Introduction
Iowa is 1 of 5 states that have not issued stay-at-home orders for the coronavirus disease 2019 (COVID-19) pandemic. The state has issued a series of orders, including banning large gatherings and closing bars and restaurant dining on March 17, 2020; closing some nonessential businesses (eg, dental offices, clothing stores, barbershops, massage therapy, medical spas) on March 26, 2020; closing all primary and secondary schools (recommended on March 15, 2020) on April 2, 2020; and closing additional businesses (eg, malls, nongrocery stores, museums, libraries, social clubs) on April 6, 2020. 1 Multiple constituents have petitioned Iowa's governor to issue a stay-at-home order (including the Iowa Board of Medicine). 2 However, the governor has indicated that the existing restrictions are essentially equivalent to stay-at-home orders in other states. 3 There is no empirical evidence on whether issuing a stay-at-home order in Iowa could have been associated with a reduced rate of COVID-19 infections in the state. Illinois, which borders Iowa, issued a stay-at-home order effective on the evening of March 21, 2020. 4 The order also closed all nonessential businesses in that state.
Using a cross-sectional study with difference-in-differences design, we compared the cases of COVID-19 in border counties in Iowa and Illinois.

Methods
This cross-sectional study used a difference-in-differences design to compare daily changes in COVID-19 cases per 10 000 residents in 8 Iowa counties bordering Illinois (ie, Clinton, Des Moines, Dubuque, Jackson, Lee, Louisa, Muscatine, and Scott) with those in 7 Illinois counties bordering Iowa (ie, Carroll, Hancock, Henderson, Jo Daviess, Mercer, Rock Island, and Whiteside) before and after Illinois issued a stay-at-home order on March 21, 2020. County-level COVID-19 data come from a repository of state and local health agency reports. 5 The difference-in-differences design assumes similarity of COVID-19 trends across the border if Iowa and Illinois had issued similar orders. We evaluated this assumption by comparing COVID-19 trends before enacting the order. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies. Per the Common Rule, institutional review board review was not required for this study, which used deidentified, publicly available data.

Statistical Analyses
We estimated the following difference-in-differences regression using daily county-level COVID-19 data: Cases ct = α + β 1 StayHome c × Post t + θ c + ω t + ⑀ ct , in which β 1 is the difference-in-differences estimate, representing the pre-post stay-at-home order differential change in COVID-19 cases per 10 000 residents (Cases) between Iowa and Illinois, c indicates a specific county, and t indicates a specific day. The pre-period is from March 15 (when the first case was reported in those counties) to March 21. We performed a t test to test whether the β 1 equaled 0 at a significance level of P < .05. We estimated this model separately for 6 post-periods (each 5 days) from March 22 (the first full day with a stay-at-home order in Illinois) until April 20. County-specific fixed effects (θ c ) captured county time-invariant differences, such as population density and demographic and socioeconomic We also performed 2 sensitivity analyses. The first sensitivity analysis accounted for differences in the timing of closing schools and nonessential businesses between the 2 states. We reestimated the regressions adding 2 state-level time-varying indicators as covariates, as follows: (1) whether the state had issued an order for closing primary and secondary schools by that day (issued on April 2 in Iowa and March 17 in Illinois) and (2) whether the state had issued an order to close nonessential businesses (the first order in Iowa was on March 26; in Illinois, closure of nonessential businesses was part of the stay-at-home order, so it was assigned the same date). The second sensitivity analysis examined whether there were differential trends in COVID-19 cases by county population density and poverty rates, reported by census data, 6 that may have confounded the difference-indifferences estimates. For this analysis, we added an interaction between the county population density and the post-period indicator (post) and another interaction between the county poverty rate and post to each regression.

JAMA Network Open | Public Health
To address testing rate differences between the counties in Iowa vs Illinois, we examined daily state level testing data from the COVID Tracking Project to determine total tests per 10 000 residents in Iowa and Illinois. 7 Data analyses were conducted with Stata/SE version 16.0 (StataCorp).
Statistical significance was set at P < .05, and all tests were 2-tailed.

Results
The  Table 1). Within 10, 20, and 30 days after the enactment of the stay-at  In the first sensitivity analysis accounting for differences in the timing of closing schools and nonessential businesses between the 2 states, the difference-in-differences estimates for the stay-  Table 2). In the second sensitivity analysis examining whether there were differential trends in COVID-19 cases by county population density and poverty rates that may have confounded the difference-in-differences estimates, the difference-in-differences estimates were robust to adding these covariates (eg, difference in COVID-19 cases per 10 000 residents in Illinois at b This indicates the estimated difference-in-differences association of a stay-at-home order with COVID-19 cases in a given period relative to March 15 to March 21 (ie, the period before the stay-at-home order in Illinois was enacted).
c Heteroskedasticity robust SEs were estimated because homoscedasticity is rejected for all post-period regressions.  Table 2).
Illinois had a greater increase in tests per 10 000 residents following the stay-at-home order.
Both states started with a similar testing rate, but the testing rate in Illinois increased after the stayat-home order at a faster rate than in Iowa (10 days, 28 vs 23 tests per 10 000 residents; 20 days, 69 vs 51 tests per 10 000 residents; 30 days, 117 vs 82 tests per 10 000 residents) (Figure 2).

Discussion
This cross-sectional study with difference-in-differences design found an increase in estimated rates of COVID-19 cases per 10 000 residents in the border counties in Iowa compared with the border counties in Illinois following a stay-at-home order that was implemented in Illinois but not in Iowa.
The estimates indicated as many as 217 excess cases in Iowa after 1 month without a stay-at-home order. Our findings are also consistent with recent evidence from California, 10 where a statewide stay-at-home order was associated with reduced COVID-19 cases by 125.5 to 219.7 per 100 000 residents by April 20. Our findings suggest that issuing a stay-at-home order in Iowa while daily cases continued to increase may have helped limit the number of cases.
In the last 2 days of the study period (ie, April 19-20), Iowa began announcing an increase in cases from surveillance testing following outbreaks in meat-processing facilities, including 1 facility in Louisa County. 11 That outbreak appears to have been first reported on April 6. 12 As described earlier, there was a trend of more cases in Iowa before April 6 and the trend increased during most of the study period. Therefore, the surveillance testing reported toward the end of the study period does not appear to explain the whole trend. Meat-processing plants would not be closed during a stay-athome order because they are considered essential businesses. Whether stay-at-home orders could modify the likelihood of such outbreaks (through initial cases) and their subsequent effects on community spread is unknown and is an important question for future research, given the outbreaks reported in multiple meat-processing facilities in Iowa and other states.

Limitations
This study has limitations. Findings should be interpreted cautiously considering that possible differences in COVID-19 testing across the border may confound the difference-in-differences estimates. For example, Illinois had a greater increase in tests per 10 000 residents after the stay-athome order ( Figure 2). If state-level testing rates are representative of these counties, this would suggest that Iowa's testing rate was not higher during this period, and therefore the difference in testing would not explain the observed difference in rates per capita after the stay-at-home order.
However, county-level testing trends may be different from state trends and may still confound the