Estimating Ebola Treatment Needs, United States

To the Editor: By December 31, 2014, the Ebola epidemic in West Africa had resulted in treatment of 10 Ebola case-patients in the United States; a maximum of 4 patients received treatment at any one time (1). Four of these 10 persons became clinically ill in the United States (2 infected outside the United States and 2 infected in the United States), and 6 were clinically ill persons medically evacuated from West Africa (Technical Appendix 1 Table 6). 
 
To plan for possible future cases in the United States, policy makers requested we produce a tool to estimate future numbers of Ebola case-patients needing treatment at any one time in the United States. Gomes et al. previously estimated the potential size of outbreaks in the United States and other countries for 2 different dates in September 2014 (2). Another study considered the overall risk for exportation of Ebola from West Africa but did not estimate the number of potential cases in the United States at any one time (3). 
 
We provide for practicing public health officials a spreadsheet-based tool, Beds for Ebola Disease (BED) (Technical Appendix 2) that can be used to estimate the number of Ebola patients expected to be treated simultaneously in the United States at any point in time. Users of BED can update estimates for changing conditions and improved quality of input data, such as incidence of disease. The BED tool extends the work of prior studies by dividing persons arriving from Liberia, Sierra Leone, and Guinea into the following 3 categories: 1) travelers who are not health care workers (HCWs), 2) HCWs, and 3) medical evacuees. This categorization helps public health officials assess the potential risk for Ebola virus infection in individual travelers and the subsequent need for post-arrival monitoring (4). 
 
We used the BED tool to calculate the estimated number of Ebola cases at any one time in the United States by multiplying the rate of new infections in the United States by length of stay (LOS) in hospital (Table). The rate of new infections is the sum of the rate of infected persons in the 3 listed categories who enter the United States from Liberia, Sierra Leone, or Guinea. For the first 2 categories of travelers, low and high estimates of Ebola-infected persons arriving in the United States are calculated by using low and high estimates of both the incidence of disease in the 3 countries and the number of arrivals per month (Table). Calculating the incidence among arriving HCWs required estimating the number of HCWs treating Ebola patients in West Africa (Technical Appendix 1, Tables 2–4). For medical evacuations of persons already ill from Ebola, we calculated low and high estimates using unpublished data of such evacuations through the end of December 2014. 
 
 
 
Table 
 
Calculated monthly rates of Ebola disease among persons arriving in the United States and additional secondary cases, 2014 
 
 
 
Although only 1 Ebola case has caused additional cases in the United States (7), we included the possibility that each Ebola case-patient who traveled into the United States would cause either 0 secondary cases (low estimate) or 2 secondary cases (high estimate) (Table). Such transmission might occur before a clinically ill traveler is hospitalized or between a patient and HCWs treating the patient (7). To account for the possibility that infected travelers may arrive in clusters, we assumed that persons requiring treatment would be distributed according to a Poisson probability distribution. Using this distribution enables us to calculate, using the BED tool, 95% CIs around the average estimate of arriving case-patients. The treatment length used in both the low and high estimate calculations was 14.8 days, calculated as a weighted average of the LOS of hospitalized case-patients treated in West Africa through September 2014 (Technical Appendix 1 Table 5) (8). We conducted a sensitivity analysis using LOS and reduced case-fatality rate of patients treated in the United States (Technical Appendix 1 Table 6). 
 
For late 2014, the low estimate of the average number of beds needed to treat patients with Ebola at any point in time was 1 (95% CI 0–3). The high estimate was 7 (95% CI 2–13). 
 
In late 2014, the United States had to plan and prepare to treat additional Ebola case-patients. By mid-January 2015, the capacity of Ebola treatment centers in the United States (49 hospitals with 71 total beds [9]) was sufficient to care for our highest estimated number of Ebola patients. Policymakers already have used the BED model to evaluate responses to the risk for arrival of Ebola virus–infected travelers, and it can be used in future infectious disease outbreaks of international origin to plan for persons requiring treatment within the United States. 
 
Technical Appendix 1: 
Data inputs and assumptions; sensitivity analysis (length of stay and case-fatality rate); comparison with other published estimates; and limitations. 
 
Click here to view.(228K, pdf) 
 
 
Technical Appendix 2: 
Beds for Ebola Disease (BED) model. 
 
Click here to view.(161K, xlsx)


HCWs
We defined an HCW as a person who has worked in >1 of the 3 West African countries in a capacity related to providing care to Ebola patients. The monthly rate of new HCW infections (main text, Table: Input 3) in West Africa was calculated by dividing the monthly number of reported Ebola cases in HCWs (at different time points in the epidemic) by estimates of the total HCW population exposed as a result of staffing Ebola Treatment Units (ETUs) (Technical Appendix 1 Table 4) (1). A lower estimate of the rate of infected HCWs in West Africa was calculated by using the 3-month average number of cases reported among HCWs at the midpoint of the outbreak (36/month, calculated July 2014) (Technical Appendix 1 Table 3) and the highest estimate of HCWs in the 3 West African countries (4,172 workers/1000 ETU beds) (Technical Appendix 1 Table 4). This number of HCWs assumes that all HCWs, regardless of their type of employment, are at higher risk than the general population for exposure to Ebola (Technical Appendix 1 Table 2). A higher estimate of the rate of infections among HCWs was calculated by using the maximum 3-month average infections among HCWs to date (129/month, calculated in October 2014) (Technical Appendix 1 Table 3), and the lowest HCW population at risk (2,677 workers/1,000 ETU beds) (Technical Appendix Table 2). This number of HCWs assumes that a smaller subset of staff, based on their position (i.e., those more likely to have patient encounters), are at higher risk for Ebola infection.
The arrival rate of HCWs to the United States was based on 1) the number of travelers who identified themselves as having worked in a health care facility during the previous 21 days and 2) the risk category ("high," "some," or "low") assigned to them during enhanced entry screening at their airport of entry to the United States during November 5-December 1, 2014 (2,3). The low estimate value of arrivals of HCWs (30 arrivals/month) was approximately the lowest rate of "high-" and "some-risk" HCWs entering the United States (main text Table: Input 3) during the timeframe examined. The high estimate value (60 arrivals/month) was approximately the highest rate of high-, some-, and low-risk HCWs entering the United States.

Medical Evacuees
This category comprises persons who already have symptomatic Ebola-related illness and who are consequently flown to the United States for treatment in special aircraft with a special containment apparatus. Patients in this category may include HCWs who are already clinically ill with Ebola. Patients in this category do not include persons who have had a "high-risk" exposure in an affected country who enter the United States without clinical symptoms: Such persons do not require an ETU bed upon arrival but they may be admitted if they receive investigational therapies, such as postexposure prophylaxis. Based on the 3-month experience during the outbreak during August-October 2014, the number of medical evacuees to the United States was assumed to be either 3 or 1 per month (main text Table: Input 3). The high estimate (3 persons) was chosen to match the observed monthly average of the number of evacuees from West Africa to all other countries in the world (including the United States).

Secondary Transmission
Secondary transmission may occur during the period in which a traveler is clinically ill but before he or she is placed in an isolated hospital bed. Some secondary transmission may also occur between the ill patient and the US-based HCWs treating the patient (4). The number of secondary transmissions per each HCW and non-HCW case imported to the United States was assumed to be either 0 (low estimate) or 2 (high estimate) (main text Table: Input

Sensitivity Analysis: Length of Stay and Case-Fatality Rate
A sensitivity analysis of LOS was also conducted in which LOS were based on casepatients treated in the United States through November 2014. Although few in number (n = 10), case-patients treated in the United States could have longer average LOS of 22.4 days and improved survival of 80% (i.e., CFR 20%). Case-patients treated in West Africa had an average LOS of 14.8 days and CFR 40% cases treated in Africa (Technical Appendix 1 Table 6).
When data on LOS and survival were used from case-patients treated in the United States (in the sensitivity analysis) the low estimate was still 1, but the 95% CI widened slightly (95% CI 0-4). The high estimate increased from 7 cases to 12 cases (95% CI 5-19).

Comparison with Other Published Estimates
Our estimates are within the range of other published estimates (6,7). Using a similar, incidence-based risk calculation (based on incidence in September 2014), Bogoch et al.

Limitations
The findings in this report are subject to several limitations. First, this analysis does not account for the possibility of the outbreak worsening in the future. If the incidence increases among the general population or HCWs, so would the rate of importations if air travel arrival rates remained the same. If Ebola becomes established in other countries (particularly those with many travelers to the United States) the rate of importation may also increase. However, our BED tool can be used to update and reestimate the risk for imported cases of Ebola. Second, this analysis does not specifically evaluate the effect of travel restrictions, such as reductions in airline traffic and capacity, and exit screenings (which could decrease the risk for travel by symptomatic persons or persons with higher exposure risks). Imposing reductions in air travel may not have a notable impact. Gomes et al. found that reducing air travel may delay importation only by a few weeks but not prevent or reduce the rate of importation (7). Again, our BED tool can be used to explore the potential impact of a decrease or increase in the number of monthly arrivals from West Africa. Third, we assumed that secondary cases will be very limited and easy to contain, thus preventing further infections (i.e., no tertiary cases will occur). Fourth, the upper limit for the number of non-HCW travelers with Ebola was calculated by assuming that these travelers have a risk for infection equal to that of the general population in the 3 primarily affected West African countries. Because most travelers are likely to have a higher socioeconomic status than persons in the general population, and consequently, a lower risk for Ebola infection, this assumption most likely overestimates the risk for infection among travelers.
As an alternative (as noted in Appendix Data Inputs and Assumptions, General Travelers) we estimated in the lower limit calculation, the impact of assuming that travelers had a level of risk that is one third that of the general population in the 3 affected countries. This reduction in risk for infection among travelers, compared with the general population, may still overestimate the actual risk. Again, the BED tool can be used to explore the impact of assuming a different level of reduction in risk (either higher or lower than what we assumed). Finally, these results may notably underestimate or overestimate the likelihood of HCWs entering the United States from West Africa who are infected with Ebola because data on this traveler category are insufficient. For instance, the number of HCWs working in West Africa and the number of Ebola patients being treated in non-ETU settings (e.g., hospitals, clinics) is unknown. As a result, this analysis calculated the risk for exposure to Ebola for HCWs from limited data on the number of HCWs in ETUs; and assumed this risk was equal for all HCWs, irrespective of the setting in which they worked. Recent evidence, however, indicates that HCWs in ETUs constitute <5% of all Ebola infections among HCWs (8). Furthermore, even if the risk to HCWs could be reliably calculated, it cannot be determined how it applies to workers entering the United States because the data on self-declared HCWs obtained from airport screenings do not include specific data fields that capture where HCWs worked and what they did in West Africa.  Table 3.) ETU, Ebola Treatment Unit; HCW, health care worker. †The number exposed was based on the type of HCWs working in ETUs (see Technical Appendix 1 Table 2). For the low estimate calculation, we considered all HCWs as exposed (i.e., the sum low-and high-risk personnel). The high estimate calculation was based on the high-risk personnel only, under the assumption that a smaller subset of staff, based on their position (i.e., those more likely to have patient encounters) are at higher risk for Ebola virus infection.
‡The total number of ETU beds at the end of October among the primarily affected countries of Liberia, Guinea, and Sierra Leone (Centers for Disease Control and Prevention, unpub. data). §Output (rounded to the nearest whole number) = [(Input 1)/(Input 2/1000 × Input 3)] × 100.
Technical Appendix 1 *Based on the average interval from hospitalization to discharge + 1 SD; for survivors this was 11.8 d (SD 6.1), and for nonsurvivors it was 4.2 d (SD 6.4) (9). CFR, case-fatality rate; LOS, length of stay. †Survivors' LOS (during treatment at US hospitals only) (n = 8) was based on 19.4 d + 1 SD of 8.8. Nonsurvivors' LOS (during treatment in US hospitals only) (n = 2) was the maximum LOS from the observed range of 2-10 d. CFR was obtained from 2 deaths of 10 case-patients treated (see Technical Appendix 1 Table 6). ‡Weighted Average LOS = LOS for survivors × (1-CFR proportion) + LOS for nonsurvivors × CFR.