Application of COVIDTracer to Inform Public Health Workforce Planning in Alaska During the COVID-19 Pandemic

Abstract

The COVID-19 pandemic has caused social and economic disruption worldwide and spurred numerous mitigation strategies, including state investments in training a large contact tracing and case investigation workforce. A team at the University of Alaska Anchorage evaluated implementation of the COVID-19 contact tracing and case investigation program of the State of Alaska Department of Health and Social Services, Division of Public Health, Section of Public Health Nursing. As part of that evaluation, the team used COVIDTracer, a spreadsheet modeling tool. COVIDTracer generated projections of COVID-19 case counts that informed estimates of workforce needs and case prioritization strategies. Case count projections approximated the reported epidemiologic curve with a median 7% difference in the first month. The accuracy of case count predictions declined after 1 month with a median difference of 80% in the second month. COVIDTracer inputs included previous case counts, the average length of time for telephone calls to cases and outreach to identified contacts, and the average number of contacts per case. As each variable increased, so too did estimated workforce needs. Decreasing the average time from exposure to outreach from 10 to 5 days reduced case counts estimated by COVIDTracer by approximately 93% during a 5-month period. COVIDTracer estimates informed Alaska’s workforce planning and decisions about prioritizing case investigation during the pandemic. Lessons learned included the importance of being able to rapidly scale up and scale down workforce to adjust to a dynamic crisis and the limitations of prediction modeling (eg, that COVIDTracer was accurate for only about 1 month into the future). These findings may be useful for future pandemic preparedness planning and other public health emergency response activities.

The COVID-19 pandemic has caused social and economic disruption throughout the globe and spurred the application of traditional measures of infectious disease control, such as policies that support social distancing, face mask wearing, and the creation and implementation of large-scale contact tracing and case investigation programs. From the beginning of the pandemic, case investigation and contact tracing were recommended interventions by the Centers for Disease Control and Prevention (CDC) ¹ and the World Health Organization. ² Rapid identification and isolation of cases, followed by swift quarantine of exposed contacts, are long-standing effective methods for reducing transmission of infectious diseases such as SARS-CoV-2. ³ As part of the pandemic response in Alaska, a team at the University of Alaska Anchorage (UAA) partnered with the State of Alaska Department of Health and Social Services, Division of Public Health, Section of Public Health Nursing (SOPHN) to support and evaluate implementation of the state’s COVID-19 case investigation and contact tracing program. The team used COVIDTracer, a spreadsheet-based modeling tool developed by CDC and publicly available for download, ⁴ to estimate projected case counts and workforce needs for case investigation and contact tracing.

Purpose

This case study describes the application of COVIDTracer to estimate case counts, explore case investigation prioritization strategies, and estimate anticipated workforce needs for SOPHN. We describe how the information was used, including its application for workforce planning purposes. ⁵ Our experience has implications for infectious disease modelers, national and state emergency preparedness planning efforts, and public health program directors responsible for ensuring an adequate and competent public health workforce.

Background

In early 2020, public health entities in the United States scrambled to develop electronic case investigation and contact tracing management systems. SOPHN was primarily responsible for case investigation and contact tracing in Alaska. A primary goal of case investigation and contact tracing is to quickly notify those who have been exposed to prevent further transmission. Reducing time from exposure to isolation or quarantine averts predicted cases of COVID-19.^6,7

The Alaska case investigation and contact tracing workforce had a scope of work that included calling Alaskan residents who had received a positive test result for COVID-19, informing them to isolate, providing self-care and monitoring instructions, asking about their close contacts, and then calling those close contacts to tell them to quarantine. To accomplish this scope of work efficiently, it was essential to estimate the workforce that would be needed. Regularly and systematically estimating needs for the contact tracer workforce during the pandemic was a priority of state leadership.

SOPHN developed and implemented a team-based management structure to coordinate the COVID-19 case investigation and contact tracing workload in Alaska. Based on actual and projected case counts from COVIDTracer, additional teams were added or removed. Each team had 8 to 12 investigators and was led by a team captain. SOPHN used the Association of State and Territorial Health Officials’ recommendations for a tier-structured workforce to recruit, train, and integrate new contact tracers. ⁸ Most teams comprised Tier 2 or Tier 3 staff who had responsibility for full case investigation and outreach to the close contacts of cases. A separate team of Tier 1 staff completed the monitoring of cases and contacts until their release from isolation or quarantine.

SOPHN initially provided case investigation and contact tracing using permanent employees in the State of Alaska Department of Health and Social Services, Section of Epidemiology (hereinafter, Section of Epidemiology) and locally placed public health nurses. As the pandemic grew and additional contact tracing workforce was needed, SOPHN hired nonpermanent state employees; awarded contracts to governmental, nongovernmental, and tribal entities; partnered with the National Guard; and collaborated with tribal health organizations and the US military.

Methods

The COVIDTracer Model and Inputs

COVIDTracer is a compartmental infectious disease transmission model developed by CDC and implemented in Microsoft Excel. It is intended to be accessible to public health practitioners without specialized knowledge in scientific programming or infectious disease dynamics. The model is described elsewhere. ⁹ Input parameters (Table) fell into 3 broad categories: (1) data on the epidemiology of the pandemic (eg, cumulative laboratory-confirmed cases to date and laboratory-confirmed cases in the past 14 days), obtained from Alaska’s COVID-19 Data Hub dashboard ¹⁰ ; (2) indicators of the effectiveness of contact tracing (eg, average number of contacts per case, percentage of people who had received a positive test result for COVID-19 who isolated and had their contacts traced and quarantined), obtained by analyzing data routinely collected in SOPHN’s contact tracing software; and (3) estimates of the time required for case investigation and contact tracing tasks (eg, average initial case interview length, percentage of time spent on overhead), obtained through monthly surveys of contact tracers.

Table.

Fields and input values for COVIDTracer-generated COVID-19 case estimates, Alaska, 2020-2021 ^a

Fields	September 1, 2020	November 1, 2020	April 1, 2021	August 1, 2021
Total population at risk (731 007 [fully vaccinated population])	731 007	731 007	549 250	412 040
No. of laboratory-confirmed cases to date (by onset date)	5438	17 536	61 002	73 197
No. of laboratory-confirmed cases in past 14 days (by onset date)	972	5487	2264	3522
Average no. of contacts per case	4.0	5.0	3.0	3.3
Average no. of times contacts were monitored during quarantine	3	1	4	4
Average no. of days after infection that laboratory-confirmed case isolated	6.95	7.00	7.20	6.25
Cases successfully isolated and contacts traced and monitored, %	26.0	20.0	33.7	26.4
R0 (basic reproduction number) b	2	2	2	5
Rt (reproduction number at time t) c	1.00	1.17	1.07	1.15
Estimated reduction in COVID-19 transmission in Alaska because of case investigation/contact tracing, %	8.3	6.4	7.9	8.5
Average length of initial case interview, min	60	60	60	60
Average length of initial contact interview, min	24	24	24	24
Average length of case follow-up telephone call, min	32	32	32	32
Average length of contact follow-up telephone call, min	12	12	12	12
Time per day spent on follow-up calls to each case and their contacts, minutes	80	92	68	72
Estimated time spent on overhead, %	18	18	26	26
Pattern of change in case counts during previous 14 d	Stable	Slowly increasing	Stable	Slowly increasing

^aEstimated case counts were made using COVIDTracer, a spreadsheet modeling tool developed by the Centers for Disease Control and Prevention and available for public use.

^bR₀ (R naught) is a general indicator of infectiousness and estimates the average number of new infections a single case causes.

^cR_t is the time-varying reproduction number, or average number of new infections a single case causes at time t.

R₀ is a general indicator of infectiousness and estimates the average number of new infections a case may cause. ¹¹ The COVIDTracer-produced models shown before August 1, 2021, assumed an unmitigated R₀ of 2.0, which was a value estimated for SARS-CoV-2 early in the pandemic. ¹² R_t (the time-varying reproductive number, or reproductive number at time t), also called R_e (the effective reproductive number), is the average number of people infected by an individual at a specific time. The Section of Epidemiology produced daily estimates of the R_t in Alaska based on a parametric model using observed case counts. We input COVIDTracer parameters so that the R_t in the model matched the R_t reported by the Section of Epidemiology at the beginning of each model period. Because our inputs were calculated to match the reported R_t, the R₀ estimate used in the model had limited effect.

The UAA evaluation team selected the COVIDTracer modeling tool to help leaders at the Division of Public Health to predict hiring needs based on the level of viral transmission in the state and nonpharmaceutical policies that may have been in place. Because of ongoing changes in the SOPHN case investigation and contact tracing management software platform, as well as changing case investigation and contact tracing protocols, the UAA team refined the methods for calculating the COVIDTracer inputs over time.

SOPHN’s case investigation and contact tracing software was not initially configured to collect the data necessary to inform COVIDTracer estimates. The available data changed over time as data field definitions were clarified and data entry reliability improved. For example, the UAA evaluation team used a field for “self-isolate start date” when calculating the number of days after infection that a case was isolated. However, the team learned that this field was hidden for most case investigators and, consequently, rarely completed. The team then identified reliably completed date fields such as exposure date, specimen collection date, and interview date to determine the number of days since infection that a case was isolated. The team used reliable data and consistent calculation methods for all inputs from January 2021 through February 2022. This project underwent administrative review by the University of Alaska Institutional Review Board, which acknowledged that it was not human subjects research.

Preparing Estimates

The UAA team prepared monthly reports for SOPHN, which included COVID-19–related projections to inform case investigation priorities and contact tracing workforce needs. We present several estimates as they were originally delivered to SOPHN management. Four of the COVIDTracer estimates included in our study were of case counts from 4 time points: September 1, 2020; November 1, 2020; April 1, 2021; and August 1, 2021. These time points are paired with actual reported case counts for the same periods.

Because of the case investigation/contact tracing workflow, the estimated workforce hours needed relied on projected case counts, the average length of time for telephone calls to cases and contacts, and the average number of contacts per case. To demonstrate the impact of contacts per case on workforce estimates, we modeled various scenarios for SOPHN each month. These scenarios included estimated workforce needs given varying R_t values and estimated workforce needs given varying contacts per case. Different R_t values, influenced by policies such as wearing face masks, school/business closures, and restrictions of large gatherings, changed case count estimates and predicted contact tracing and case investigation workforce needs.

COVIDTracer did not explicitly account for vaccination. The UAA team received technical assistance from the CDC Health Economics and Modeling Unit, ¹³ which had developed the tool and was advised to implement an ad hoc adjustment for vaccination by subtracting the number of fully vaccinated people from the total population size. Unless otherwise noted, projections used an input of the number of unvaccinated people at each model start date and did not adjust for an increase in the number of people who may have been vaccinated against COVID-19 during the predicted time frame.

Additional COVIDTracer-Based Analyses

In addition to workforce projections, the UAA team used COVIDTracer to conduct supplemental analyses that (1) illustrated how changing the R_e and the number of contacts per case would change the projected number of contact tracing staff hours needed and (2) illustrated pandemic trajectories under various assumptions about average time from exposure to isolation of cases and quarantine of contacts. We included 1 of these analyses in the current study; using data from March 2020 through August 1, 2021, we input 3 scenarios: 5, 7, and 10 average days from COVID-19 exposure to outreach. In these scenarios, outreach was assumed to lead to isolation or quarantine. We assumed an average of 5 days from COVID-19 exposure to outreach by a contact tracer.

Results

Estimated COVID-19 Case Counts

Case counts predicted by COVIDTracer reasonably matched the actual reported case counts for about the first month after each projection and then diverged (Figure 1). The total number of cases projected for the initial 30-day period differed from the reported cases by a median of 7%. For the second 30-day period, the case counts differed by a median of 80%. Inputs for the COVIDTracer-generated estimates are also shown (Table), illustrating that these 4 time points included very different scenarios. Despite the differences in inputs, such as the number of cases in the past 14 days (ranging from 972 to 5487), the average number of contacts per case (ranging from 3.0 to 5.0), and R₀ (ranging from 2 to 5), COVIDTracer-generated estimates consistently predicted actual reported COVID-19 cases in the short term.

Figure 1.

Actual and COVIDTracer-estimated COVID-19 case counts, Alaska, July 2020–September 2021. COVIDTracer is a spreadsheet modeling tool developed by the Centers for Disease Control and Prevention that is available for public use.

Impact of Days From Exposure to Outreach

Changing the model input for average number of days from exposure to outreach had a substantial impact on COVIDTracer outputs. COVIDTracer estimated 11 428 cases in a 5-month period (Figure 2), as compared with 81 512 cases if the average number of days from exposure to outreach was 7 and 174 622 cases if the average number of days from exposure to outreach was 10. The difference in COVIDTracer-estimated case counts represents an approximately 93% decrease in estimated case counts using COVIDTracer, reducing the average number of days from exposure to outreach from 10 to 5.

Figure 2.

COVIDTracer-estimated daily COVID-19 case counts at 5, 7, and 10 average days from viral exposure to isolation, Alaska, August–December 2021. COVIDTracer is a spreadsheet modeling tool developed by the Centers for Disease Control and Prevention that is available for public use.

Lessons Learned

The COVIDTracer tool was especially useful for SOPHN to estimate case counts and case investigation and contact tracing workforce needs about 1 month into the future. However, COVIDTracer estimates could not predict the pandemic’s trajectory beyond about a month because of factors such as changing policies, shifting levels of compliance with community-wide nonpharmaceutical interventions, and the arrival of new variants with increased transmissibility or immune escape. In public health crises, entities may benefit from proactively preparing to support the rapid scale-up and scale-down of a public health emergency response workforce based on trends predicted through models such as COVIDTracer.

The information used for COVIDTracer projections was based on available data at the time of each projection; however, data fields and inputs continued to be refined for increased accuracy. For example, the R_t used for each projection was taken from the State of Alaska Department of Health and Social Services Data Hub.^10,14 However, the estimated R_t for those periods is now different, potentially because more cases have been found through epidemiologic data review. Consistent calculation methods for inputs were also challenging to identify given the evolving nature of the case investigation and contact tracing software and protocols. Data fields in the software were often not well-defined or completed consistently, resulting in iterative adjustments.

The COVIDTracer model as used in this analysis includes several assumptions: (1) that the population mixes homogeneously (ie, anyone is equally likely to infect anyone else), (2) that all cases are detected, (3) that people who have been previously infected are completely protected against reinfection, (4) that only people reached by a contact tracer will isolate or quarantine, and (5) that people with COVID-19 who are interviewed are accurately able to report their total number of contacts. While a more complex model that allowed for relaxing some of these assumptions might have led to more reliable projections, that is not a foregone conclusion. Indeed, COVID-19 forecasts based on various methods have been generally unable to accurately project the course of the pandemic multiple months into the future. ¹⁵ The emergence and spread of variants has shaped the course of the pandemic and is a key factor that makes reliable long-term projections difficult to obtain.

The COVIDTracer tool provided SOPHN with insights needed to make short-term adjustments to workforce capacity. However, because the model estimates decreased in accuracy when looking further than 1 month ahead, the tool was not sufficient to make staffing plans beyond that period. This shortcoming was especially challenging in managing COVID-19 surges, as demand for contact tracers escalated more quickly than staff could be hired, trained, and operationalized. While this was a limitation when applying model results to workforce decision-making, it is not a reflection of the COVIDTracer tool itself, as longer-term predictions of the pandemic trajectory are affected by multiple variables, including changes in the public’s behavior. ¹⁵

COVIDTracer model estimates were particularly helpful for managing case prioritization during pandemic escalations and de-escalations. For example, during the 2020 winter surge in Alaska, SOPHN hired additional staff in response to estimated case counts; however, SOPHN realized it would not be able to recruit, hire, and train contact tracing staff fast enough to meet predicted case counts. To maximize impact, SOPHN worked with UAA and the Section of Epidemiology to develop a case prioritization structure to focus workforce resources on reaching out to people infected with COVID-19 closest to the date they were tested (0-4 days from specimen collection). As the surge subsided and the workforce grew, SOPHN was able to return to conducting outreach and case investigation of all people with a laboratory-confirmed positive polymerase chain reaction test result reported to the State of Alaska Department of Health and Social Services within 10 days of specimen collection.

However, the divergence of long-term COVIDTracer estimates from reported case counts created mismatches between hired workforce and needed workforce. For example, COVIDTracer projections in June 2021 indicated a small number of new cases, and the COVID-19 vaccine was widely available to the public. Consequently, SOPHN and state partners decreased staffing substantially. The Delta variant surge in Alaska at nearly the same time led to a substantial mismatch between workload demands and workforce capacity moving into fall 2021. It would have been helpful for the tool to integrate the risks associated with new variants of concern or variants of interest into the analysis to increase the predictability of variant surges.

Planning for monitoring and evaluation (which may include transmission models such as COVIDTracer) should be an integral part of the process of designing contact tracing information management systems so that key metrics of program and client services are consistently documented and assessed. Close collaboration between infectious disease experts studying contact tracing effectiveness and public health practitioners who develop and implement contact tracing systems may contribute to improved approaches for implementing and evaluating contact tracing systems.

Public health agencies can mitigate the inherent difficulty of accurately projecting long-term workforce needs in a pandemic by developing systems to react to the changing dynamics quickly and efficiently. Until such a time as modeling tools can more accurately predict long-term trends in transmission, Alaska should prepare internal state systems to become better able to adjust to rapid changes in workforce needs. Challenges in scaling up the response included difficulties in recruiting, hiring, and training staff to obtain access to the information technology systems that underpinned contact tracing and training staff on the team-based management structure. Preparing for low probability but high-impact public health events such as pandemics may require developing the capacity to scale up and train a workforce quickly and to implement systems for data collection, management, and analysis. Human resources and information technology staff and systems are an integral part of expanding and contracting a public health emergency response and, especially, data- and personnel-intensive interventions such as contact tracing. Public health agencies may wish to consider how organizational components responsible for human resources and information technology can participate in exercises to identify and remedy potential issues prior to a real-world crisis. Investment in mobile exposure notification technologies, secure data integration among laboratories, reporting entities, and investigation data systems could streamline outreach and reduce transmission of a fast-moving virus. ¹⁶