Estimated COVID-19 Cases and Hospitalizations Averted by Case Investigation and Contact Tracing in the US

Key Points Question What are the estimated numbers of COVID-19 cases and hospitalizations averted by case investigation and contact tracing (CICT) programs in the US? Findings This decision analytical model study used CICT program data from 23 jurisdictions and estimated that CICT programs averted 1.11 to 1.36 million cases and 27 231 to 33 527 hospitalizations over 60 days during the 2020 to 2021 winter peak of the pandemic. The upper estimate assumes that all interviewed cases and monitored contacts complied with isolation and quarantine guidelines, whereas the lower estimate assumes that fractions of interviewed cases and monitored or notified contacts did so. Meaning These findings suggest that CICT programs likely played a critical role in curtailing the pandemic.


eAppendix 1. Case Investigation and Contact Tracing Effectiveness
The effectiveness of case investigation and contact tracing (CICT) is determined by the proportion of cases and their infected contacts that are effectively isolated and quarantined, preventing further transmission in the susceptible population. The duration of quarantine and isolation is described in Centers for Disease Control and Prevention (CDC)'s guidance. 9 We assumed that confirmed cases are effectively isolated following case interviews. We further assumed that contacts are quarantined upon either contact notification or through active monitoring.
We calculated the average proportion of cases and contacts isolated and quarantined by CICT for each location as follows: Step 1: We first calculated the proportion of cases that effectively isolated: * # ℎ Step 2: We then calculated the proportion of infected contacts that effectively Step 3: We took the weighted average between the results of steps 1 and 2 (Terms A and B) by weighting quarantined contacts by R0, since undetected infected contacts will infect R0 additional individuals on average (or 2.5 new infections per infected contact). This resulted in the final equation: Average proportion of cases and contacts (which become cases) isolated by CICT = [% Cases interviewed*Compliance + (R0 * % Contacts identified * (% Contacts monitored *Compliance + % Contacts notified but not monitored*Compliance))] / (1+R0) By populating this equation with the assumed compliance to isolation/quarantine guidance (described in Table 1), we assessed the following three scenarios.

Equation 1: Baseline Low Estimate
80% of interviewed cases and monitored contacts, and 30% of notified contacts (who are not monitored), isolate or quarantine: Average proportion of cases and contacts (which become cases) isolated = [% Cases interviewed*0.8 + (R0 * % Contacts identified * (% Contacts monitored *0.8 + % Contacts notified but not monitored*0.3))] / (1+R0) where R0 is the assumed number of new infections per case without any interventions and when the population is entirely susceptible to infection (Table A2).
In addition, reducing the time from case identification to effective isolation is critical for case investigation and contact tracing to succeed. The longer the cases and contacts interact with the susceptible population, the greater the opportunity for onward transmission. In practice, cases with no known exposure are predominantly identified and isolated after symptom onset 1 , and cases with known exposures (i.e., contacts that eventually become infected cases) can begin quarantine upon contact notification (even potentially prior to symptom onset). We assumed asymptomatic cases can only be identified and isolated if they are notified through case investigation and contact tracing.
For the purposes of our study, we assumed the proportions of cases with no known exposure and cases with known exposures were equal (i.e., 50/50 breakdown) because we did not have data on what prompted case identification in each location. Therefore, for each location the days to effective case isolation was determined by taking the average of the days to effective isolation between case groups with known and no known exposures. The time to effective case isolation for each of the two case groups was determined as follows: For symptomatic cases with no known exposures (i.e., symptoms prompt identification): We assumed that cases experience a 5-day pre-symptomatic period (See Table A2), get tested the day after symptom onset (i.e., 6 days would have transpired since infection at the time of testing). We then obtained the number of days from testing to result notification by adding the reported "Median days from specimen collection to case reporting to the health department (HD)". We also assumed that confirmed cases begin isolation the day after their result notification (i.e., we added 1 to the total obtained above). Our assumptions regarding the "next-day" timing of testing and entry into isolation are based on symptoms and notifications beginning or occurring throughout the day, with a sizeable portion occurring sufficiently late enough in the day to prevent testing and entry into isolation the same evening. This assumption takes into account practical considerations such as time needed to find a testing site and arrange an appointment, and for notified individuals to prepare to isolate (e.g., purchasing food or medications, setting up childcare, handling work or other commitments).
For cases with known exposures (i.e., those who were notified they were a contact and eventually became a case): We first calculated the days from index case testing to their exposed contacts' notification by summing jurisdictions' reported "Median days from specimen collection to case report to the HD", "Median days from case report to the HD to the case interview completion", and "Median days from case interview completion to contacts notification". We assumed that contacts begin quarantine the day after receiving exposure notification from their health department (i.e., we added 1 to the sum above).
The "next-day" timing of entry into quarantine is based on the same practical reasoning as cases needing time to prepare to isolate once notified (described above).
We then used the resultant sum from the procedure above to estimate the time (in days) from exposure to quarantine for contacts. Because we did not have information on when exposures actually occurred for contacts, we assumed that these individuals' exposures occurred at the midpoint of their potential exposure window (in days). We identified the earliest date in this window as the first day of infectiousness among cases to which contacts were exposed. Based on our assumed 5-day presymptomatic period for symptomatic cases (described above), this was two days prior to the symptom onset date in cases exposing the contact. We identified the latest possible exposure as the date the cases exposing them were interviewed by the health department (because they began isolation the next day). See both "Contacts" rows in Figure A2 for a visual depiction of this timeline. Notes: In this hypothetical scenario, we assume a jurisdiction needed 1 day from specimen collection (testing) to result notification and 2 days from specimen collection to contact notification. The index case (symptomatic case with no known exposure) began showing symptoms on day 6 post-infection, got tested on day 7 and was notified of test result on day 8. The case's contacts (cases with known exposure) were exposed sometime between days 4 to 8 and notified of their exposure on day 9. Therefore, the index case began isolation on day 9 and contacts went into quarantine on day 10 (based on our assumptions above). To calculate the days from contacts' exposure to their quarantine, we took the average of the maximum days a contact was infected (6 days in this example based on the earliest possible exposure) and the fewest days the contact could be infected (2 days in this example, based on the latest possible exposure), and weighted each day span by the case's infectiousness on each of possible exposure days. The result is 3.9 days in this example, meaning the contact had been exposed for 3.9 days upon initiating quarantine. We then took the average between 8 days (index case) and 3.9 days (contacts) as the number of days from exposure to isolation (for both cases and contacts). This is 6 days in this example.
The days between cases with known exposures becoming infected and their exposure notification can vary from what we assumed. For example, cases may take longer to become symptomatic, or get tested the same day that they become symptomatic or begin their isolation on the same day as their results notification. Similarly, contacts who become cases may be exposed earlier or later than we assumed and may make up a larger or smaller share of the case pool. Readers interested in more detail of the influence of varying our assumed time to case isolation may wish to see Table A6 in the CDC's Epidemiology and Laboratory Capacity (ELC)-funded jurisdictions also reported the Number of contacts that were notified within 1 day of case interview, the Number of contacts that were notified between 1-3 days after case interview, and the Number of contacts that were notified within 3 or more days after case interview. We used these additional data elements as a quality check ( Figure A3) of the reliability of jurisdictions' reported median values regarding notification timing (described above). We did this by calculating the lower limit of the average number of days from case interview to contact notification as follows: This metric assumes that all contacts were notified within 3 days of the case interview. We used this metric to exclude jurisdictions from the analysis (i.e., deemed reported data unreliable) when the lower limit of the average time to contact notification was greater than our calculated time to contact notification using reported median days AND the proportion of contacts that were notified 3 or more days after case interview was less than 10% of total contacts (i.e., too few to exert enough influence on the average lower limit for it to plausibly exceed the median-based value  Advanced, we used the following definitions and assumptions. A "case" was defined as a person who has been exposed, infected and subsequently becomes infectious, regardless of the presence of clinical symptoms. We assumed that for the first 3 days after infection, cases do not infect others. During days 4-5 post-infection, cases are presymptomatic but shed virus in amounts that may infect others. [2][3][4][5] During days 6-14, the infected person can be symptomatic and shedding virus, albeit during days 11-14 the risk of onward transmission is relatively low but non-zero (the complete infectivity distribution is given in Table A1). We assumed that approximately 40% of cases are asymptomatic during days 6-14 yet have a risk of onward transmission equal to 75% of symptomatic cases (Table A2)    a Single large city or county in these states were separate Epidemiology and Laboratory Capacity (ELC) jurisdictions and not included in this analysis. b Low CICT impact scenario assumes only actively monitored contacts (who later became cases) effectively quarantined/isolated. High CICT impact scenario assumes notification prompted contacts (who later became cases) to quarantine effectively. In both scenarios we assumed interviewed cases fully adhered to isolation guidelines. c Percent reduction in the number of new infections per case (Rt) due to a combination of vaccination and all other nonpharmaceutical interventions (NPIs; e.g., masks use, social distancing, school/restaurant closures, etc. Calculated as the percent difference in R0 and Rt after implementation of vaccine and other NPIs. d Percent reduction in the number of new infections per case (Rt) due to CICT after the implementation of other NPIs. Calculated as the percent difference between Rt after implementation of other NPIs and Rt after implementation of both other NPIs and CICT. e After accounting for the impacts from vaccination and all other NPIs. f Cases or hospitalizations averted by CICT out of the estimated cases or hospitalizations remaining after the implementation of vaccination and other NPIs.

eAppendix 3. Instructions for Using COVIDTracer Advanced Special Edition to Estimate the Number of COVID-19 Cases and Hospitalizations Averted by CICT
These instructions will guide a user how to use the Special Edition version of COVIDTracer Advanced tool to repeat the analysis described in this manuscript to estimate COVID-19 cases averted by case investigation and contact tracing activities. The Special Edition version of COVIDTracer Advanced is a modification of the publicly available tool on CDC's website that enables users to assess the impact of CICT before vaccine was widely available. Additional modifications would be required if you intend to explicitly account for vaccinated individuals (e.g., decreasing susceptible population over time, decreased risk of hospitalization among vaccinated individuals, etc).
Readers seeking basic information about the model, data elements, and definitions should refer the COVIDTracer Advanced User Manual. However, some statements in the web manual are not applicable to the Special Edition version used in this analysis. https://www.cdc.gov/coronavirus/2019-ncov/php/contacttracing/COVIDTracerTools.html COVIDTracer Advanced uses the Windows operating system (Microsoft Windows 2010 or higher) and Excel (Microsoft Office 2013 or higher).
Before starting, complete the following: 1) Determine your 60-day study period. The first day of your study period is your "model start date." This "model start date" will be referenced later in these instructions. For example, if you are interested in estimating cases and hospitalizations averted by CICT during the 60-day period from January 1-March 1, 2021, your "model start date" is January 1, 2021.
2) Obtain these data for the jurisdiction of interest: a. Total population b. Total cases as of the day before the model start date (In the example study period above, this is the total cases reported as of December 31, 2020.) c. Cases reported during the past 14 days (In the example study period above, this is the sum of cases reported from December 18 to 31, 2020.) d. The case trend during the past 14 days (e.g., increasing, plateaued, decreasing) e. Daily (i.e., incident) case counts for the 60-day study period f. The following case investigation and contact tracing program metrics.
These metrics are meant to be representative of the 60-day study period. If you don't have such data for the entire study period, you may base these metrics on a shorter period (e.g., 30 days or 4-weeks) from the model start date (and assume they are representative of the full 60 days): i. Number of days from exposure to case isolation and contact quarantine ii. Percent (%) of all cases successfully isolated and contacts quarantined 3) Open the COVIDTracer Advanced_SpecialEdition tool (Supplement 2) a. When opening the spreadsheet file, click the "Enable Macros" button for full functionality of the tool. b. Enable Excel "Solver Add-In." Instructions: in Excel, click on File → Options → Add-ins → select "Analysis ToolPak" → click "Go" (not the "Ok" button) → select checkbox for "Solver Add-In" and click "Ok." The Solver button, will appear in the "Data" menu.
In worksheet, "A. Outbreak Details" Step 1: Enter the population for the jurisdiction of interest.
Step 2: Enter the model start date, the total number of COVID-19 cases in the jurisdiction until the day before the model start date, and the number of cases reported in the last 14 days within the jurisdiction.
Step 3: Set the pattern of daily cases over the past 14-day period selected in Step 3.
The default is "Daily case counts are slowly increasing." However, if daily case counts have been changing rapidly, remaining constant, or decreasing over the last 14 days, select from the pull-down menu the pattern that best matches the jurisdiction's data.
The selection of the case trend in the past 14 days determines how reported cases are distributed over the 14 days prior to the model's initiation date. Visually inspect the case trend and choose the most appropriate option. You can also run the model with different case trend patterns and pick one that yields the "best fit" (by repeating steps 3 to 6).
In worksheet, "Case Counts" Step 4: Paste the jurisdiction's daily case counts (i.e., incident cases) for the 60-day study period into the "Daily" column (column AH) In Worksheet, "B. Impact of Contact Tracing" Step 5: Using your representative CICT program data, enter values for:  Number of days after infection that case is isolated  % of all cases successfully isolated and contacts traced and monitored Step 6: Estimate the % reduction in transmission due to community interventions (shown in cell G28) by fitting COVIDTracer Advanced's simulated curve to your observed case curve. You will use the Solver Add-in to do this: The Solver Add-in finds an optimal solution for the % reduction in transmission due to community intervention by minimizing the mean squared error (a mathematical value describing the differences between both curves; shown in cell O32).
Instructions for using the Solver: From the Excel menu tab, click "Data" and the "Solver" button, then follow the instructions described here to set up the parameters in the pop-up dialogue box (see screen shot below): Set Objective: Set objective to cell "$O$32", which is the mean squared error.
By Changing Variable Cells: Enter $G$28 (This cell refers the Solver to the "Estimated % reduction in transmission due to continued community interventions.") Select a Solving Method: For simplicity, we recommend selecting "GRG Nonlinear" from the drop-down menu.
Then the Excel Solver function will automatically find the optimal value (estimated % reduction in transmission due to continued community intervention) and populate the value in cell G28. The figure below shows a fitted curve (solid line) generated by a Using the 7-day moving incidence average for fitting and up to three periods for each jurisdiction using our low CICT impact scenario (Table 1). b Low impact scenario assumes 80% of actively monitored contacts (who later became cases), 30% of notified contacts effectively quarantined/isolated and 80% interviewed cases fully adhered to isolation guidelines. c Single large city or county in these states were separate Epidemiology and Laboratory Capacity (ELC) jurisdictions and not included in this analysis. d Percent reduction in the number of new infections per case (Rt) due to a combination of vaccination and all other nonpharmaceutical interventions (NPIs; e.g., masks use, social distancing, school/restaurant closures, etc). Calculated as the percent difference in R0 and Rt after implementation of vaccine and other NPIs. e Percent reduction in the number of new infections per case (Rt) due to CICT after the implementation of other NPIs. Calculated as the percent difference between Rt after implementation of other NPIs and Rt after implementation of both other NPIs and CICT. f After accounting for the impacts from vaccination and all other NPIs. g Cases or hospitalizations averted by CICT out of the estimated cases or hospitalizations remaining after the implementation of vaccination and other NPIs. h See eFigure 5 for fitted curves and periods used for fitting.