kassonlab / covid19-epi Goto Github PK

Reproducing query from earlier issue:
"/* Modified factors to account for difference between population in Southeast Asia from Ferguson Model and Swedish population. This adjustment maintains 1/3 transmission in community, 1/3 transmission in household, and 1/3 transmission in workplace.
Overall population is 8.4X smaller in Sweden, attributed as a 8.4X increase in community transmission factor.
Household size is approximately 2X smaller in Sweden, 2 vs. 4, attributed as a 2X increase in household transmission factor.
R0 Scaling factor used to set approximate doubling time. R0_scale=2.2 attributes to a doubling time of approximately 3 days while R0_scale=1.8 attributes to a doubling time of approximately 5 days.
*/"

Roger--I'm not sure this was properly answered earlier, and I'd like to make sure this is fully addressed.
Am I correct in thinking that there may be unresolved questions in your mind?

Intentionally or?

There is probably existing data that has been used to calculate the percentages you use to put individuals in different classes, by doing so you could get vastly different starting states.
From all working to none working...

Why not a simple calculation without randomness?
(count number in group) place in sub-groups using percentages
Then the randomness would move only to infection transfer

This is a list of things that need to be looked at, it is by no means complete, so if anyone thinks of anything else just add a comment to this issue.

One of the goals is to be able to run with the full population of Sweden for at least 500 days.
And to run multiple scenarios with varying parameters.

• Strategies for parallelizing, both OpenMP and MPI if possible.
• Memory footprint need to be reduced, this might very well include restructuring heavily. This needs to be done carefully so as to not disrupt Jasmines development.
• Make the code more modular, needs to be done slowly and carefully and in constant contact with me or Jasmine so she can continue to develop the interactions.

Ic, Ih, Iw is never initialized will be null causing core dump

valgrind ./build/covid19
==12714== Memcheck, a memory error detector
==12714== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==12714== Using Valgrind-3.15.0 and LibVEX; rerun with -h for copyright info
==12714== Command: ./build/covid19
==12714==
Using seed 14755718885144
==12714== Conditional jump or move depends on uninitialised value(s)
==12714== at 0x40AA80: household_lat_long (in /home/roger/Documents/2_code/covid19-epi/build/covid19)
==12714== by 0x4032AE: main (in /home/roger/Documents/2_code/covid19-epi/build/covid19)
==12714==
max_HH_size = 7
All infections initialized
Starting density kernel calculations
Done with density kernel calculations in 65220.19
Total initialization time 67010.45
Starting simulation
Timestep 0.00

HH_count is limited, but this accesses outside the limit...

(gdb) list 0x40aa80
0x40aa80 is in household_lat_long (covid19.c:139).
134 tmp_lat = lat_locale[HH_count];
135 tmp_lon = lon_locale[HH_count];
136 / Scale population density */
137 pop_density_init_num[HH_count] = ceil(pop_density_init_num[HH_count] * population / land_pop_total_density);
138 tot_pop_density += pop_density_init_num[HH_count];
139 if (tot_pop_density > population || ceil(pop_density_init_num[HH_count+1] * population / land_pop_total_density) < 0.5) {
140 *num_locale = HH_count + 1;
141 }
142
143 // Determine city of each population square. Use city data to determine which schools students attend. Workplaces are placed by county. //

HH_count is limited to number, but here one more is added...

Builds correctly but then crashes when run on a stock Amazon Ubuntu 18.04 Deep Learning image AMI as of 2020-05-13T1412.

A machine instantiated from this image is pre-installed with cmake 13.3.3, CUDA 10.0, and nvcc. It has the GNU compiler, not the Intel compiler.

I'm not sure if this is intended to be a supported configuration, but the README suggests it is, and it compiles without errors, so I thought I would report this.

I tried this on the following exactly AWS AMI ID and instance type:

aws_ami_id=ami-09517d3f1a9a1cd9a
aws_instance_type=g3s.xlarge

Mortality rate in ICU should depend on whether ICU resources are overwhelmed. Just testing the code a little bit with 100k population yields scenarios with up to 43 ICU patients per 100k, which is about 7 times the number that Sweden usually maintains.

I recall seeing an article a couple of weeks ago estimating the change in CFR between a non-overwhelmed healthcare system and an overwhelmed one, but I cannot locate it anymore.

An alternative, if mortality rate cannot be quantified accurately, is to not report deaths in this simulation. Number of hospital and ICU beds required is perhaps more useful information in any case.

Usually 'i' loops over population, here the loop is missing => 'i' reuses old value
most likely from last loop, i.e i=population
This is likely to result in an access out of range 0..population-1

This first loop is missing a comment, something along the line of
/* Place one household with one adult at each location */
(first loop HH_count is thus both counting households and locations, making code like

/* Modified factors to account for difference between population in Southeast Asia from Ferguson Model and Swedish population. This adjustment maintains 1/3 transmission in community, 1/3 transmission in household, and 1/3 transmission in workplace.
Overall population is 8.4X smaller in Sweden, attributed as a 8.4X increase in community transmission factor.
Household size is approximately 2X smaller in Sweden, 2 vs. 4, attributed as a 2X increase in household transmission factor.
R0 Scaling factor used to set approximate doubling time. R0_scale=2.2 attributes to a doubling time of approximately 3 days while R0_scale=1.8 attributes to a doubling time of approximately 5 days.
*/

used as

double betac=0.103 ; // Scaled from betac=0.075 in influenza pandemic with R0=1.6, COVID-19 R0=2.2 (Ferguson 2020)

fd=1/(1+pow((d/4), 3)); //kernel density function as parameterized for GB.
return (betac_scale*betac*kappa*fd*(1+severe*(omega-1)));

shouldn't betac_scale rather be 1/8.4 if anything? "8.4X smaller"

Do Ferguson really scale this too in his calculations depending on population size?
Same for the 2x factor for household size (can't imagine why you need an extra factor if simulation of houshold tranmissions are correct)