Analyzing screening strategies for the COVID19 disease
1. Analyzing screening strategies for the COVID19 disease
Andréa Duhamel
andrea.cynthia-duhamel@univ-lehavre.fr
Full Professor, LITIS-RI2C
Research director of the ISEL
Vice-President of the digital transformation at the ULHN
LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Seine
Maritime
> 1.2 M
Inhabitants
Source:
INSEE
35.054
COVID19
cases in
Seine-
Maritime
COVID19 up-to-date situation
2
> 67.5M
COVID19
cases in
the world
> 1.5 M
of deaths
> 2.2 M
COVID19
cases in
France
55.521
of deaths
From 14th to 19th of December a large scale screening tests will be
done in the Havre Seine Metropole
576
of deaths
2. LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Context and Motivation
3
• The COVID19 caused a strong impact on the society, economy, health
system, etc (mundialization effects).
• Several weakness in terms of logistics:
• Manage the protection material (masks, blouses, glasses,…)
• Lack of strategy to test the population
• Level of stock not well managed
• Level of drugs not well managed
• The strategy test, isolate and track do not work
• ….
• Communication problems
• Lack of strategy (resilience) over the time
LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Moving to the action
4
• What could we do ? In which way a logistician could help ? I have been
interested on:
• The strategies to screening the population
• The location to perform the test
• I have looked to answer some questions such as:
• Are there some interesting strategy to test the population using virological ou
serological tests ?
• If so, under which criteria ?
• Which are the best strategy according to evolution dynamics of the COVID
epidemic over the time ?
• Massif vs. target screening tests
Participation de Guillaume Dageons – I4
Participation de Florent Lore – I5
3. LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Problem definition
5
• Lets us consider two type of tests: virological and serological
• Given a limited number of tests (capacity), a time period and an epidemic
dynamics, we look for maximizing the number of tested patients
• Scheduling the test for patients in a time horizon
• 300 days are considered, i.e 10 months
• Decision: (i) test a patient with a serological test; (ii) test a patient with a
virological one; (iii) not test a patient
• Time to test: (i) immediately; (ii) as soon as possible; (iii) after a time
• Key points:
• The data
• The strategies
• The evaluation criteria
LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Data
6
• Public database to check the epidemic evolution (Cascoronavirus, statista,
banque des territoires, gouvernement et géodes)
• Demand status:
• Low : demand between 50% and 66 % of the global tests’ capacity
• Medium: demand between 66% and 90% of the global tests’ capacity
• High: demand between 90% and 110% of the global tests’ capacity
• Critical: demand between 110% and 140% of the global tests’ capacity
• Epidemic evolution:
• Increasing: if the demand increases over the time
• Stable: if the demand does not change much over the time
• Decreasing : if the demand decreases over the time
• Demand vs. epidemic evolution generates 12 scenario combinations
• For each combination, we randomly generated 10 instances
4. LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Data
7
• Four patients’ profile
• Symptomatic patients not in a cluster
• Symptomatic patients in a cluster
• Asymptomatic patients not in a cluster
• Asymptomatic patients in a cluster
• Test capacity is inspired on the first time period of the epidemic in France
• 25 000 virological tests
• 2 500 serological tests
• Considering 300 days : 5 700 000 virological tests and 570 000 serological tests
LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Strategies
8
• Hypothesis: no symptomatic patient is assigned to a serological test
• Simulation: uses the rule First In First Out, but
• Strategy 1: adopted by the French government at the first COVID wave
• Symptomatic patients, belonging to a cluster or not are assigned to a virological
test using the arrival order
• No priority order
• This strategy uses the stocks of virological tests for the symptomatic patients
• Strategy 2:
• Priority for symptomatic patients and patients in a cluster
• This strategy is interesting for the tracking whenever the epidemic is under
control
5. LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Strategies
9
• Strategy 3: Two levels of priorities
• The first level: the symptomatic patients belonging to a cluster or not
• The second level: the asymptomatic patients belonging to a cluster
patient symptomatique hors cluster ; patient asymptomatique hors cluster ;
patient symptomatique d’un cluster ; patient asymptomatique d’un cluster ;
Ordre d’arrivée des
patients
Liste
d’attente
Jour J
Liste
d’attente
Jour J+1Jour J Jour J+1
Tête de la liste Tête de la liste
Queue de la liste
Queue de la liste
LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Strategies
10
• Strategy 4:
• Two queues: one for the symptomatic patients and another for the
asymptomatic patients
patient symptomatique hors cluster ; patient asymptomatique hors cluster ;
patient symptomatique d’un cluster ; patient asymptomatique d’un cluster ;
Ordre d’arrivée des
patients
Listes
d’attente
Jour J
Listes
d’attente
Jour J+1Jour J Jour J+1
Tête des listes Tête de la liste
Queue des listes
Queue de la liste
6. LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Strategies
11
• Strategy 5:
• Three queues: one for the symptomatic patients, another for the asymptomatic
patients belonging to a cluster and the last for asymptomatic patients not in a
cluster
patient symptomatique hors cluster ; patient asymptomatique hors cluster ;
patient symptomatique d’un cluster ; patient asymptomatique d’un cluster ;
Ordre d’arrivée des
patients
Listes
d’attente
Jour J
Listes
d’attente
Jour J+1Jour J Jour J+1
Tête des listes Tête de la liste
Queue des listes
Queue de la liste
LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Criteria evaluation
12
• The percentage of patients not tested
• The percentage of symptomatic (in a cluster or not) patients not tested;
• The percentage of patients belonging to a cluster not tested;
• The percentage of stock break in terms of number of days
• For each criteria, the smallest is the percentage, the best the strategy is.
• The goal is then to minimize the number of ….
7. LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Examples of results
13
LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Examples of results
14
8. LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Remarks
15
• The strategy 1 adopted by the French government is the worst in three
criteria, except for the stock break.
• The strategies 4 and 5 performed globally better over the four criteria, and it
would be interesting in a massif screening tests.
• The strategy 2, one of the simplest, lost only in a criteria (the number of
untested patients). It seems interesting for a target screening tests
campaign.
LITIS-RI2C 2020, Le Havre, Andréa Cynthia Santos http://pagesperso.litislab.fr/acduhamel/
Main contributions and perspectives
16
• Raise the importance of logistics in a health crises like the one of the
COVID19
• Formalization of the problem of strategies to test the population
• Two possible Forthcoming studies:
• Probabilistic methods
• A use case will be done in the S8 with students under the direction of Sonia
Froufe and myself, using existing simulating software
• Congratulations to Guillaume Dageons that works on this study in a training
time, financed by SFLog
• Last but not least : an indexed publication
• Dageons G., Santos A. C. Analyses de stratégies de dépistage dans la lutte contre le
covid-19. Cahiers de la logistique, CL2020 – N°01, p. 1-12, 2020.