Permalink
Browse files

minor updates for a revised submission.

  • Loading branch information...
nickreich committed Sep 3, 2017
1 parent c6def17 commit 4fbc4620157234ee66d81498769c5831a5e25e6b
@@ -1,8 +1,8 @@
---
title: 'Methodological supplement for: Quantifying the Risk and Cost of Active Monitoring
for Infectious Diseases'
author: ""
date: "February, 2017"
author: "Nicholas G Reich*, Justin Lessler, Jay K Varma, Neil M Vora"
date: "September 2017"
header-includes:
- \usepackage{setspace}\doublespacing
- \usepackage{lineno}\linenumbers
@@ -18,7 +18,7 @@ output:
fig_height: 5
fig_width: 5
md_extensions: +implicit_figures+grid_tables
csl: elsevier-vancouver.csl
csl: emerging-infectious-diseases.csl
bibliography: active-monitoring.bib
---
Binary file not shown.
@@ -1,7 +1,7 @@
---
title: "Quantifying the Risk and Cost of Active Monitoring for Infectious Diseases"
author: "Nicholas G Reich, Justin Lessler, Jay K Varma, Neil M Vora"
date: ""
author: "Nicholas G Reich*, Justin Lessler, Jay K Varma, Neil M Vora"
date: "September 2017"
output:
word_document:
fig_caption: yes
@@ -15,9 +15,7 @@ output:
fig_width: 5
keep_tex: yes
md_extensions: +implicit_figures
header-includes:
- \usepackage{setspace}\doublespacing
- \usepackage{lineno}\linenumbers
header-includes: \usepackage{setspace}\doublespacing
csl: emerging-infectious-diseases.csl
bibliography: active-monitoring.bib
---
@@ -58,26 +56,14 @@ Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA (J. Le
New York City Department of Health and Mental Hygiene, USA (J.K. Varma, N.M. Vora);
Centers for Disease Control and Prevention, Atlanta, Georgia, USA (J.K. Varma, N.M. Vora)
__Corresponding author:__
Nicholas G Reich
Department of Biostatistics and Epidemiology
University of Massachusetts-Amherst
School of Public Health and Health Sciences
Amherst, MA 01003
Email: nick@schoolph.umass.edu
Telephone: 413.545.4534
\
<!-- \endgroup-->
<!--\clearpage-->
# Abstract
During outbreaks of deadly emerging pathogens (e.g., Ebola, MERS-CoV) and bioterror threats (e.g., smallpox), actively monitoring potentially infected individuals aims to limit disease transmission and morbidity.
Guidance issued by CDC on active monitoring was a cornerstone of its response to the West Africa Ebola outbreak.
@@ -99,7 +85,7 @@ CDC’s guidance on active monitoring was a cornerstone of its response to Ebola
Recommendations for active monitoring were discontinued in February 2016.[@CentersforDiseaseControlandPrevention:lyxYF_Hp]
Over 20% of all individuals actively monitored for Ebola in the United States were monitored in New York City (NYC), more than any other jurisdiction.[@StehlingAriza:2015vd]
Active monitoring may help prevent and contain outbreaks of rapidly spreading emerging pathogens that pose a grave threat to public health. Such outbreaks may occur naturally, via a bioterrorist attack, or via unintended release from a laboratory. The decision to implement active monitoring depends on an assessment of the risk posed by a pathogen and the ability of active monitoring to reduce that risk. Key considerations include the transmissibility and pathogenicity of the pathogen, the potential size of an outbreak, and the relationship between the time of symptom onset and infectiousness.[@Fraser:2004ci] Ebola, Middle East Respiratory Syndrome Coronavirus (MERS-CoV), and smallpox are examples of viral illnesses for which active monitoring could play a pivotal role in preventing a large-scale outbreak.[@CentersforDiseaseControlandPrevention:2015tk]
Active monitoring may help prevent and contain outbreaks of rapidly spreading emerging pathogens that pose a grave threat to public health. Such outbreaks may occur naturally, via a bioterrorist attack, or via unintended release from a laboratory. The decision to implement active monitoring, or another non-pharmaceutical intervention, depends on an assessment of the risk posed by a pathogen and the ability of the chosen intervention to reduce that risk.[@peak2017comparing] Key considerations include the transmissibility and pathogenicity of the pathogen, the potential size of an outbreak, and the relationship between the time of symptom onset and infectiousness.[@Fraser:2004ci] Ebola, Middle East Respiratory Syndrome Coronavirus (MERS-CoV), and smallpox are examples of viral illnesses for which active monitoring could play a pivotal role in preventing a large-scale outbreak.[@CentersforDiseaseControlandPrevention:2015tk]
Active monitoring programs must balance conflicting priorities. The central goals of these programs are to identify, isolate, and treat infected individuals quickly. Setting an active monitoring period many times longer than any known incubation period of the pathogen of interest could virtually guarantee that all infected individuals would exhibit symptoms while being monitored. However, such a program would be unreasonably expensive, inconvenience monitored individuals, and incur many financial and social costs through frequent responses to false positive cases. Evidence-based monitoring periods and appropriate tailoring of the monitoring intensity to disease risk should therefore be used to balance costs with biosecurity risks.
@@ -127,13 +113,16 @@ Additionally, it determines the uncertainty associated with these estimates due
Our model estimates the risks and costs associated with active monitoring programs for a range of active monitoring durations. To estimate the cost per person-day of monitoring, we used data on the number of individuals actively monitored by DOHMH and costs associated with the DOHMH Ebola response. Additionally, for the purposes of hypothetical cost calculations, we assumed that an individual who becomes symptomatic with the disease of interest while under active monitoring gives rise to no secondary infections, while an individual who develops symptoms after his/her active monitoring period ends could give rise to as many as 4 new Ebola infections (an upper estimate based on prior research [@Althaus:2014cw;@Chowell:2004gh]).
We developed open-source software, including a freely-available web application at [http://iddynamics.jhsph.edu/apps/shiny/activemonitr/](http://iddynamics.jhsph.edu/apps/shiny/activemonitr/). The source code for the web app, the data for the analyses, and the code to reproduce this manuscript itself are all freely available online under an open-source license at GitHub (https://github.com/reichlab/activemonitr), with a static version in an open-access digital library [@Reich:2017z]. All analyses were run in `r R.version.string`.[@rcoreteam] These tools enable others to easily implement our model and reproduce our results.
### Stratifying by exposure risk
Classifying individuals based on prior exposure risks enables targeted strategies in a range of public health response settings, including active monitoring. For example, in response to the West Africa Ebola outbreak, CDC issued recommendations on risk stratification of individuals for a potential Ebola virus exposure ("high risk", "some risk", "low (but not zero) risk", "no identifiable risk") and for how long and how intensively individuals in each of these categories should be monitored.[@CentersforDiseaseControlandPrevention:2015tc] DOHMH's active monitoring program, described previously[@Millman:2016cz], was implemented consistent with CDC recommendations. However, creating, evaluating, and modifying such classifications in practice is a difficult task and requires situational awareness and data that would vary depending on the pathogen and outbreak setting.
For the CDC Ebola risk strata, we estimated probabilities of a monitored individual developing Ebola. These estimates were based on extrapolated numbers of actively monitored individuals in the United States during 2014-2016[@StehlingAriza:2015vd] and public data on the four domestic cases of Ebola[@CentersforDiseaseControlandPrevention:VbDBWqNH] (Supplemental Text, Table 2).
### Data Availability
We developed open-source software, including a freely-available web application at [http://iddynamics.jhsph.edu/apps/shiny/activemonitr/](http://iddynamics.jhsph.edu/apps/shiny/activemonitr/). The source code for the web app, the data for the analyses, and the code to reproduce this manuscript itself are all freely available online under an open-source license at GitHub (https://github.com/reichlab/activemonitr), with a static version in an open-access digital library [@Reich:2017z]. All analyses were run in `r R.version.string`.[@rcoreteam] These tools enable others to easily implement our model and reproduce our results.
# Results
### Incubation period estimates
@@ -473,7 +462,7 @@ min_cost_days_distr <- min_sampled_costs %>% group_by(phi, phi_lab) %>%
Our model provides ranges of expected cost for active monitoring systems. It identifies an optimal duration of active monitoring, by finding the expected cost range with the lowest maximum value. We applied the model to the case-study of Ebola in NYC, based on data from DOHMH (Table 1, Figure 4). Expected costs of short periods of active monitoring (left hand side of Figure 4) are driven by the cost of a missed case and the number of expected additional secondary cases, while the rate of decline with additional days of monitoring is driven by the shape of the incubation period distribution. The costs of longer periods of active monitoring are driven by the per day cost of monitoring and costs of false positive detections (right hand side of Figure 4).
For the low-risk individuals, the model suggests that the cost is minimized with
For the low-risk individuals, the model suggests that the cost is minimized with
`r round(min_costs[which(min_costs[,"phi"]==1e-4), "min_cost_dur_days"],1)`
days of monitoring (i.e.,
`r round(min_costs[which(min_costs[,"phi"]==1e-4), "min_cost_dur_days"]/mean(pstr_gamma_params_ebola$median),1)`
@@ -516,42 +505,55 @@ The framework presented is generally applicable, but given the lack of empirical
Recent history has shown that the unexpected emergence of new disease threats has become a recurring theme in global health and preparedness. While active monitoring does not play a role in the response to every emerging infectious disease (e.g., it has not played an important role in the response to the Zika virus epidemic), it will likely be used again in the response to future threats. Our framework provides valuable information for assessing the cost-effectiveness of various active monitoring strategies in response to critical disease outbreaks. By providing an empirical basis for evaluating active monitoring programs, these tools can strengthen biosecurity and optimize active monitoring programs in response to future global disease threats.
<!--
# Acknowledgments
We thank the individuals who underwent active monitoring in NYC, Ifeoma Ezeoke for preparing the DOHMH active monitoring data and Margaret Pletnikoff for preparing the DOHMH cost data. We thank Dr. Simon Cauchemez, Dr. Ousmane Faye, and Dr. Amadou Sall for providing access to the Ebola incubation period data from Guinea and giving permission for this data to be made publicly available. We also acknowledge John Maher for providing details on costs of evaluating Ebola patients at Bellevue Hospital in New York City.
The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of DOHMH or CDC.
-->
NGR was funded by a grant from the MIDAS program of the National Institutes of General Medical Sciences (R35GM119582). The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of DOHMH, CDC, NIGMS, or the National Institutes of Health.
# Author Contributions
NGR, JL, JKV, and NMV designed the study. NMV and JKV oversaw the active monitoring data collection and provided data for the study on active monitoring programs. NGR obtained incubation period data from existing publications and via personal communication with other authors. NGR ran the data analysis, prepared the figures, and drafted the manuscript. All authors participated in writing and providing substantial feedback on the manuscript.
# Competing Financial Interests
All authors claim no competing financial interests associated with this work.
# Tables
```{r print-the-table, results='asis'}
kable(out,
caption="Table 1. Total expected costs to the public health system (including active monitoring and response) per 100 monitored individuals, at two exposure-risk categories. Columns represent multiples of the median incubation period of Ebola. Rows represent the CDC exposure-risk category. Costs are inclusive of active monitoring and public health response and are based on model outputs for Ebola. Values are in $'000s.")
caption="Total expected costs to the public health system (including active monitoring and response) per 100 monitored individuals, at two exposure-risk categories. Columns represent multiples of the median incubation period of Ebola. Rows represent the CDC exposure-risk category. Costs are inclusive of active monitoring and public health response and are based on model outputs for Ebola. Values are in $'000s.")
```
\clearpage
# Figure legends
# Figures
__Figure 1__ Model schematic representing four outcome scenarios for a person under active monitoring. Estimated costs shown are based on published costs[@Yacisin:2015wm] as well as new data from DOHMH and Bellevue Hospital in NYC. Details on the model formulation are available in Supplemental Table 3. Our model uses probabilities to calculate the likelihood of each of the possible model outcomes. Additionally, we estimate the probability that an individual who does not develop the disease of interest develops symptomatic illness necessitating hospitalization to rule-out the disease of interest. (See Supplemental Text for details.)
![](static-figures/figure1-model-schema-v9.jpg)
<!-- static-figures/figure1-model-schema-v9.jpg-->
\
\clearpage
__Figure 2__ Estimates and credible regions for incubation period distributions for Ebola, MERS-CoV and smallpox. The shaded elliptical areas represent regions that contain 95% of the estimated posterior distributions for each of the three diseases. The disease-specific curves plotted on the right show the estimated distribution for the incubation period for each disease (dark line). To show some of the uncertainty associated with these estimates, a random selection of density functions sampled from the joint posterior are represented by colored transparent lines around the heavy lines. Shaded vertical bands indicate the marginal credible regions for the median and 95th percentile.
![](static-figures/figure2-estimated-distribution-dynamic-5.pdf)
<!--static-figures/figure2-estimated-distribution-dynamic-5.pdf-->
\
\clearpage
__Figure 3__ Estimated probabilities of symptoms occurring after active monitoring (AM) ends across different active monitoring period durations, shown as multiples of the median incubation period. Figures are shown for 'some or high risk' and 'low (but not zero) risk' scenarios, with probabilities of developing symptomatic infection set to 1/1,000 and 1/10,000, respectively. The vertical dashed line indicates the 21 day duration recommended for Ebola active monitoring (i.e., approximately 2.4 times the median incubation period of Ebola).
![](static-figures/figure3-scaled-prob-symptoms-1.pdf)
<!--static-figures/figure3-scaled-prob-symptoms-1.pdf-->
\
\clearpage
__Figure 4__ Estimated cost ranges of actively monitoring 100 individuals for Ebola, calculated separately for some or high risk individuals and low (but not zero) risk individuals. The dashed lines intersect at the minimum point for the upper limit of each cost range.
![](static-figures/figure4-optimized-cost-analysis-1.pdf)
<!--static-figures/figure4-optimized-cost-analysis-1.pdf-->
@@ -131,6 +131,17 @@ @article{Reich:2009jq
month = sep
}
@article{peak2017comparing,
title={Comparing nonpharmaceutical interventions for containing emerging epidemics},
author={Peak, Corey M and Childs, Lauren M and Grad, Yonatan H and Buckee, Caroline O},
journal={Proceedings of the National Academy of Sciences},
volume={114},
number={15},
pages={4023--4028},
year={2017},
publisher={National Acad Sciences}
}
@manual{rcoreteam,
title = {{R: A Language and Environment for Statistical Computing}},
author = {{{R Core Team}}},
Binary file not shown.

0 comments on commit 4fbc462

Please sign in to comment.