Health Care Utilization Before and After the “Muslim Ban” Executive Order Among People Born in Muslim-Majority Countries and Living in the US

Key Points Question Was the 2017 “Muslim ban” executive order associated with changes in health care utilization by people born in Muslim-majority countries living in Minneapolis-St. Paul, Minnesota? Findings This cohort study of 252 594 patients found that after the executive order was issued, there was an increase in missed primary care appointments and increased emergency department visits among people from Muslim-majority countries living in Minneapolis-St. Paul. Meaning Changes in health care utilization among people from Muslim-majority countries after the Muslim ban may reflect changes in population health influenced by federal immigration policy.


ICD-10 codes used in analysis of stress-responsive diagnoses
Prior to collecting or observing any data, research team members chose the following ICD-10 codes to include in analysis of stress-responsive diagnoses.

Analysis of pre-Order trends
In order to interpret the difference in differences presented in the manuscript as evidence of effects of the Muslim Ban, we would need to assume that in the absence of the executive order, the change in utilization and diagnoses would have been identical across groups. Although we cannot verify this assumption, we can observe whether or not differential trends appear before the executive order is issued. As the visualization in Figure 1 suggests, individuals from Muslim Ban targeted nations began to increase clinic utilization, relative to non-Latinx U.S.-born individuals, before the issuance of the executive order. Increases in stress-responsive diagnoses in the clinic are also observed before the executive order is issued. After restricting the comparison group through matching on patient demographics, these pre-Order trend differences are still present.

Difference in differences estimates for all outcomes
Although we do not ascribe a causal interpretation to all difference in differences estimates, the full set of analyses originally planned are presented in Table S7. This is an expanded version of p < 0.05*, p < 0.01** Note: As in the main text, difference in differences estimates are additional increases in each outcome (per 1000 people per 30-day time period) from the year before to the year after the Muslim Ban was issued among individuals from targeted nations, beyond the increases observed in a reference group. Robust standard errors are included in parentheses for difference in difference estimates, clustered at the individual level, with and without demographic matching. Parametric bootstrap standard errors are included in parentheses for generalized synthetical control model estimates. The estimated sampling distributions do not necessarily follow an approximately normal distribution centered at the point estimate.

Robustness checks
To assess the robustness of the difference in differences analysis presented in the manuscript, we conducted two additional checks. First, we report estimates from a model which allows for a linear time trend as an alternative to period fixed effects. Second, we report regression coefficients and sample characteristics for the difference-in-differences models comparing the individuals from targeted nations to U.S.-born non-Latinx individuals with similar age, sex, race, and insurance to those observed for individuals from targeted nations. This reference group was selected using exact matching on all available demographics. Points indicate weekly average counts per 1,000 people in each group after demographic matching for A) clinic visits, B) missed clinic appointments, C) clinic stress-responsive diagnoses, D) ED visits, and E) ED stress-responsive diagnoses. A LOESS regression line summarizing the time trend is included for each re-weighted group, based on daily average counts. For all clinic outcomes, panels A, B, and C, nonbusiness days are excluded from the analysis. The solid line marks the Order issuance and the dotted line marks the 2016 election, for reference.

Differences following the Muslim Ban among individuals from Muslim-majority nations not targeted in the Order
Because we observed utilization and diagnosis trends only for 1,254 individuals from other Muslimmajority nations, we are not able to robustly compare the experiences of these individuals from Muslimmajority nations that were and were not named in the executive order. Nonetheless, we note that a difference-in-differences analysis comparing individuals born in a Muslim majority nation not named in the Muslim Ban and U.S.-born, non-Latinx individuals reveals qualitatively similar trends to those observed for individuals from targeted nations.