ABSTRACT
Background. Mask mandates have been a widely used public health tool during the COVID-19 pandemic, but how to optimize their impact in the setting of concurrent but spontaneous population-level behavior changes due to rising case counts is not known. This study aimed to examine how earlier or later mask mandate implementation in the context of spontaneous behavior change would have affected transmission of SARS-CoV-2 and severe COVID-19 outcomes in the St. Louis, Missouri area. Methods. Our model utilized aggregated hospitalization and death data for St. Louis city and county residents admitted to nearly all hospitals in the metropolitan area. We first fit a real-life model to estimate changes in transmission after the July 3, 2020 mask mandate, and then created counterfactual scenarios in which 1) 10%, 25%, and 50% of the changes were attributed to the mandate (as opposed to spontaneous behavior change) and 2) the mandate was implemented 3 or 7 days earlier, or 7 or 14 days later. We used an SEIR (Susceptible-Exposed-Infectious-Recovered) model framework and fit models in R. Results. Assuming that 50% of increased masking was due to the mandate, implementing a mandate 7 days earlier was associated with a reduction from 12,685 (IQR: 10,463-16,560) to 12,294 (10,296-15,205) cumulative hospitalizations by September 30, while a 2-week delay was associated with an increase to 13,277 (10,808-17,908) hospitalizations. Trends were similar, but with reduced magnitude, when assuming that 10% or 25% of increased masking was due to the mandate (Figure). Depending on whether 10%, 25%, or 50% of increased masking was due to the mandate, implementing the mandate 1 week early was associated with a return to baseline (June 26) hospital census 1-7 days earlier, while delaying the mandate by 2 weeks led to a 2-12 day delay in return to baseline. Hospital census and cumulative deaths in the real-life (baseline) model and under 12 counterfactual scenarios which vary mask mandate timing (3 or 7 days earlier, or 7 or 14 days delayed) and percentage of increase in masking that is attributed to the mask mandate (Panels A-B: 10%, Panels C-D: 25%, and Panels E-F: 50%). As more of the increase in masking is attributed to the mandate, the costs of delaying the mandate and the benefits of earlier implementation increase. While differences in hospital census are most apparent several weeks after the mandate, differences in deaths gradually become more apparent over time. Conclusion. Impact of a mask mandate depends on both timing and percent of increased masking that is attributed to the mandate. Implementing a mandate even a few days earlier is associated with fewer cumulative hospitalizations and earlier return to baseline, but the overall duration of implementation is slightly longer. Given wide variations in public behavior, locally-tailored models are essential for estimating the impact of interventions and informing the local public health response.
ABSTRACT
Background: Despite longstanding guidelines endorsing isoniazid preventive therapy (IPT) for persons with HIV, uptake is low across sub-Saharan Africa. Mid-level health managers oversee IPT programs nationally;interventions aimed at this group have not been tested. Methods: We conducted a cluster randomized trial in Uganda among district-level health managers from 2017-2021. The unit of randomization was groups of 4-7 managers. Our intervention convened managers into mini-collaboratives facilitated by Ugandan TB/HIV experts and provided business leadership/management training, SMS platform access, and data feedback. The primary outcome was IPT initiation rates among adults with HIV in health facilities overseen by participants over 2 years (2019-2021). We compared incidence rates using cluster-level targeted minimum loss-based estimation. We conducted pre-specified analyses that excluded Q3-2019 to understand intervention effects independent of a national "100-day push" of IPT tied to a financial contingency during Q3-2019. Qualitative interviews were analyzed to ascertain mechanisms of intervention action. Results: Managers from 82/82 eligible districts (61% of Uganda's 135 districts) were enrolled and randomized: 43 districts to intervention, 39 to control. After one year, in 5-point-Likert quantitative surveys, intervention-group managers demonstrated greater increases in familiarity with IPT (by +0.47 points (95%CI:0.14-0.80)) and knowledge of IPT efficacy (+0.59 points (95%CI:0.06-1.12)) as compared to control. Intervention-group managers reported improved within-district communication and inter-district collaboration and feeling empowered to better manage frontline providers, in contrast to control, in qualitative interviews. Over two years, the IPT initiation rate was 0.74 vs. 0.65 starts/person-year in intervention vs control: incidence rate ratio (IRR)=1.14 (95%CI:0.88-1.46;p=0.16). Excluding Q3-2019, IPT initiation was higher in intervention vs control: 0.32 vs. 0.25 starts/person-year (IRR=1.27, 95%CI:1.00-1.61, p=0.03;Figure). Conclusion: Though overall IPT initiation rates were not significantly higher with the mid-level manager intervention in this cluster randomized trial, rates were significantly higher compared to control when excluding the massive MoH-led "100-day IPT push" in both arms. The higher rates were sustained during the COVID-19 pandemic, suggesting benefits of targeted leadership and management training for mid-level health managers.