Public data and replication of the paper: Automated vs. Manual Case Investigation and Contact Tracing for Pandemic Surveillance: Evidence from a Stepped Wedge Cluster Randomized Trial
panel_data.csv
: Aggregate data from the study used made available for partial replicationFigure_Replication.ipynb
: A Jupyter notebook replicating Figures 2, 3, 4 (difference-in-differences time series plots)Regression_Replication.ipynb
: Partial replication of the regression analysis presented in Table 3 using panel data instead of individual level dataStepped Wedge CICT Protocol.pdf
: The study protocol
The data is organized as a ZIP Code x Week panel. For each week/ZIP Code the treatment condition (manual/automated), is recorded along with the number of participants assigned to automated and manual CICT and observed outcomes for participants who recieved each treatment.
Under perfect compliance a ZIP Code in the manual condition would only have individuals assigned to manual CICT, and a ZIP Code in the automated condition would only have individuals assigned to automated CICT; however, due to capacity constraints we observed a large number of overflow cases where individuals in the manual condition were given automated CICT.
As a result, for each outcome we provide the result for participants in each condition (i.e., manual outcome and automated), regardless of what the randomization protocol says. This allows for replicating the ITT analysis, IV analysis and Figures 2,3,4 which visualize the overflow cases as well. The below table shows the data structure:
Column | week | Zip | crossover | svi_label | condition | n_assigned_automated | n_assigned_manual | mean_perc_filled_assigned_automated | mean_perc_filled_assigned_manual |
---|---|---|---|---|---|---|---|---|---|
Type | <int,[2,3,4,5]> | <str, "[50,80) SVI" or "[80,100] SVI"> | <str, "Manual CICT" or "Automated CICT"> | <float, [0,1]> | <float, [0,1]> | ||||
Description | Reference week | Reference ZIP Code | Refers to the ZIP Code's belonging to the early (2), medium (3), late (4) or never (5) step down cluster | That ZIP Code's social vulnerability bucket. | The condition that ZIP Code is receiving on that week based on the randomization protocol. In other words, what form of CICT should participants be receiving in that ZIP Code. | Number of participants assigned to automated CICT. | Number of participants assigned to manual CICT | Overall completion rate of data fields for participants in the ZIP Code who were assigned automated CICT | Overall completion rate of data fields for participants in the ZIP Code who were assigned manual CICT |
Example | 2021-12-13 | 94040 | 4 | [50,80) SVI | Manual CICT | 1 | 10 | 0.8181818181818182 | 0.6545454545454545 |
There are also additional columns showing the individual data field completion rate.
week
: Reference weekZip
: Reference ZIP Codecrossover
: Refers to the ZIP Code's belonging to the early (2), medium (3), late (4) or never (5) step down clustersvi_label
: Whether the ZIP Code is high SVI (80-100) or low SVI (50-80)condition
: Whether the ZIP Code is in the manual CICT condition or the automated CICT conditionn_assigned_automated
: Number of participants in that ZIP Code assigned to automated CICTn_assigned_manual
: Number of participants in that ZIP Code assigned to manual CICTmean_perc_filled_assigned_automated
: Primary outcome, the overall completion rate for the 11 data fields for the participants who were given automated CICTmean_perc_filled_assigned_manual
: Primary outcome, the overall completion rate for the 11 data fields for the participants who were given manual CICTmean_has_race_ethnicity_assigned_automated
: Proportion of individuals who recieved automated CICT that provided their race/ethnicitymean_has_race_ethnicity_assigned_manual
: Proportion of individuals who recieved manual CICT that answered their race/ethnicitymean_has_employer_assigned_automated
: Proportion of individuals who recieved automated CICT that provided their employermean_has_employer_assigned_manual
: Proportion of individuals who recieved manual CICT that provided their employermean_has_gender_assigned_automated
: Proportion of individuals who recieved automated CICT that provided their gendermean_has_gender_assigned_manual
: Proportion of individuals who recieved manual CICT that provided their gendermean_has_sexual_orientation_assigned_automated
: Proportion of individuals who recieved automated CICT that provided their sexual orientationmean_has_sexual_orientation_assigned_manual
: Proportion of individuals who recieved manual CICT that provided their sexual orientationmean_has_language_assigned_automated
: Proportion of individuals who recieved automated CICT that provided their preferred languagemean_has_language_assigned_manual
: Proportion of individuals who recieved manual CICT that provided their preferred languagemean_has_signs_or_symptons_assigned_automated
: Proportion of individuals who recieved automated CICT that provided their COVID-19 symptomsmean_has_signs_or_symptons_assigned_manual
: Proportion of individuals who recieved manual CICT that provided their COVID-19 symptomsmean_has_travel_assigned_automated
: Proportion of individuals who recieved automated CICT that provided whether they had travelled recentlymean_has_travel_assigned_manual
: Proportion of individuals who recieved manual CICT that provided whether they had travelled recentlymean_has_attended_gathering_assigned_automated
: Proportion of individuals who recieved automated CICT that provided whether they had attened a large gathering recentlymean_has_attended_gathering_assigned_manual
: Proportion of individuals who recieved manual CICT that provided whether they had attened a large gathering recentlymean_has_able_to_self_isolate_assigned_automated
: Proportion of individuals who recieved automated CICT that provided whether they were able to self isolatemean_has_able_to_self_isolate_assigned_manual
: Proportion of individuals who recieved manual CICT that provided whether they were able to self isolatemean_has_num_contacts_generated_assigned_automated
: Proportion of individuals who recieved automated CICT that provided whether they had any recent contactsmean_has_num_contacts_generated_assigned_manual
: Proportion of individuals who recieved manual CICT that provided whether they had any recent contactsmean_has_congregate_setting_assigned_automated
: Proportion of individuals who recieved automated CICT that provided whether they had any exposure to a congregate settingmean_has_congregate_setting_assigned_manual
: Proportion of individuals who recieved manual CICT that provided whether they had any exposure to a congregate setting
This code and data were prepared by Cameron Raymond on September 13, 2022. If you have questions, reach out to Derek Ouyang at [email protected].