Strategies for Designing Preparedness Plans for HCIDs

CDC guidelines suggest sharing of resources stratified by 3 levels of assessment and care: frontline health care facilities, Ebola assessment hospitals, and Ebola treatment centers. ¹ Estimating how to distribute limited resources among facilities while ensuring adequate population coverage presents a spatial optimization problem that must consider the locations of where and how people are likely to access treatment. We show that the patterns of health care access and utilization are likely to vary across rural and urban areas and present a tradeoff between geographic access and availability. Areas with increased access are shown to actually have less acute care beds per population than areas with fewer facilities, because the population demands are higher in such urban areas (Figure 1).

Figure 1 Hospital acute care bed density. Beds available per 100k population. This shows areas in which beds available for acute care are limited in an influx of patients, such as an outbreak.

Rural-urban disparity in access to and utilization of medical care is a commonly explored problem.Reference Braden and Beauregard ⁴ Identifying gaps in residents’ access to care without the stress of an Ebola outbreak can help providers be better prepared. For instance, rural residents are characteristically older and poorer—attributes that may affect health care access and utilization.Reference Braden and Beauregard ^{4 ,} Reference Larson and Fleishman ⁵ They tend to have a usual place of health care, which makes a resident more likely to seek care when ill compared with urban dwellers.Reference Goodman, Fisher, Stukel and Chang ⁶ However, they may also have to travel longer distances for care,Reference Braden and Beauregard ^{4 ,} Reference Larson and Fleishman ⁵ resulting in greater geographic access disparities compared with urban areas.Reference Goodman, Fisher, Stukel and Chang ⁶ However, due to the lack of a robust, publically available baseline data set on how populations access health care resources, it is difficult to identify which hospitals have disparities or need resources to treat an influx of Ebola cases. Further, while CDC guidelines provide guidance on the types of resources that are needed to ensure adequate preparedness at each hospital, the number of resources needed to handle all possible Ebola patients in a given geographic region remains unknown. Understanding this, the CDC recommends that government stockpiles of Ebola-related health resources, such as personal protective equipment, be distributed within health service region. ⁷ We argue that the subsequent distribution of such resources to hospitals within health service regions must be guided by population demand for services—which tends to be uneven across geographic space.

It is assumed that people will use their closest health service, based on the Dartmouth Hospital Referral Regions. ⁸ To approximate such service regions, Voronoi partitions are constructed around each hospital in Texas (Figure 1). For each facility, a ratio of the number of acute care beds to the total population was used as a measure for available resources. These include intensive care units but not pediatric beds, which are subject to a separate CDC plan. As noted in the inset maps (Figure 1), the smaller Voronoi-based service regions in the 2 major metropolitan areas in Texas suggest better spatial access to available health care facilities. However, those areas also show significant local variations in the number of acute care beds that are available to the population. Existing disparities in access due to transportation, education, and income further complicate the assessment of resources required at the various clinics.Reference Larson and Fleishman ⁵

The demand for services may exhibit a dynamic, evolving behavior, as Ebola is known to spread differently in populations with different demographic and socio-economic characteristics. ^{3 ,} Reference Comber, Brunsdon and Radburn ⁹ Potentially rapid increases in demand for services can place a significant strain on the availability of existing resources, thereby resulting in a breakdown of the mitigation effort. Estimating the dynamic behavior of Ebola through the construction of mathematical models based on prior outbreaks would be one manner response plans could be evaluated. Such models incorporate local and regional population characteristics along with disease-specific parameters to simulate the dynamics of the disease as it spreads with varying levels of intensity across a given population.Reference Kadri, Chaabane and Tahon ^{10 -} Reference Jimenez, Mikler and Tiwari ¹² One approach to estimate the dynamic behavior of Ebola is through the construction of mathematical models based on prior outbreaks.Reference Kadri, Chaabane and Tahon ^{10 ,} Reference Rachah and Torres ^{13 -} Reference Do and Lee ¹⁶ Such models incorporate local and regional population characteristics along with disease-specific parameters to simulate the dynamics of the disease as it spreads with varying levels of intensity across a given population.

The SEIR (Susceptible, Exposed, Infected, Removed) model can be used to simulate the spread of a disease among a given population with counts that are determined using US Census data along with other disease parameters obtained from published CDC reports.Reference Rachah and Torres ^{13 -} Reference Do and Lee ¹⁶ In the SEIR model, individuals can either be susceptible to the disease (S), exposed, meaning the person has made contact with an infected individual but is yet asymptomatic (E), infected (I), or, eventually, considered recovered or removed (R). The total population at time t is to be denoted by N where

$$N\left( t \right)\,{\equals}\,S\left( t \right){\plus}E\left( t \right){\plus}I\left( t \right){\plus}R\left( t \right).$$

The equations driving the SEIR model are as follows:

$${\rm d}S/{\rm d}t\,{\equals}\,\,{\minus}\,\beta \,SI\,N$$

$${\rm d}E/{\rm d}t\,{\equals}\,\beta \,SI\,N{\minus}\delta E$$

$${\rm d}I/{\rm d}t\,{\equals}\,\delta E{\minus}\gamma I$$

$${\rm d}R/{\rm d}t\,{\equals}\,\gamma I$$

The rate of incubation β, latent period δ, and infectious period γ are used to estimate the number of Ebola cases in a population with a given set of characteristics. The estimation of appropriate values for these parameters can be challenging given the lack of data that is currently available for Ebola and uncertainties in how different strains of the disease may spread differently among populations. Further, uncertainties in how the onset of the disease is determined and its incubation period (ie, the time between exposure and visible symptoms of Ebola) can lead to large variances in the simulation results. Currently, the World Health Organization estimates the incubation period for Ebola to be between 2 and 21 days. ¹⁷ However, recent mathematical models based on recent outbreaks use an average of 7-12.7 days.Reference Legrand, Grais, Boelle, Valleron and Flahault ¹⁴ The latent period (ie, the period during which no symptoms are evident) is essential to planning mitigation efforts as it provides information on how long a potentially exposed individual must be kept under observation for Ebola symptoms. In several models, this period is usually assumed to be between 10 and 21 days. Previous studies have determined the infectious period to be between 6.5 and 21 days.Reference Rachah and Torres ^{13 - 17} Uncertainties in the parameters used in the SEIR model require the development of computational tools that allow end-users to easily manipulate these simulation parameters and measure their impacts on how the disease spreads through a population. The results of such simulation models can also be used to develop strategies to proportionally re-allocate existing resources from one area to another or to develop plans for offloading patient demand from one geographic region to another.