epimodel/epimodelhiv

Network Models of HIV Transmission Dynamics among MSM and Heterosexuals

EpiModelHIV

Modules for simulating HIV/STI transmission dynamics among men who have sex with men and heterosexual populations, developed as an extension to our general network-based epidemic modeling platform, EpiModel.

EpiModel and EpiModelHIV use the statistical framework of temporal exponential-family random graph models to fit and simulate models of dynamic networks. These statistical methods have been developed and implemented as open-source software, building on the extensive efforts of the Statnet research group to build software tools for the representation, analysis, and visualization of complex network data.

These packages combine these Statnet methods with an agent-based epidemic modeling engine to simulate HIV transmission over networks, allowing for complex dependencies between the network, epidemiological, and demographic changes in the simulated populations. Readers new to these methods are recommended to consult our EpiModel resources, including our main methods paper Vignette describing the theory and implementation.

Installation

You can install EpiModelHIV in R using remotes:

install.packages("EpiModel", dependencies = TRUE)
remotes::install_github("statnet/tergmLite")
remotes::install_github("statnet/EpiModelHPC")
remotes::install_github("statnet/EpiModelHIV")

Documentation on using this software package is forthcoming, although limited function documentation is provided within the package and available with the help(package = "EpiModelHIV") command.

epimodelhiv's Issues

VL module rewrite

Rewrite VL updates for treatment naive in more readable form than this:

  new.vl <- (inf.time.bp.tn <= vlard) * (vlap * inf.time.bp.tn / vlard) +
            (inf.time.bp.tn > vlard) * (inf.time.bp.tn <= vlard + vlafd) *
            ((vlsp - vlap) * (inf.time.bp.tn - vlard) / vlafd + vlap) +
            (inf.time.bp.tn > vlard + vlafd) * (inf.time.bp.tn <= vldo) * (vlsp) +
            (inf.time.bp.tn > vldo) * (vlsp + (inf.time.bp.tn - vldo) * vlds)

Instead, use conditional logic for each line to subset and update for each group

Represent persistent partnership ID in disclosure and partnership matrices

Instead of retaining two columns in these matrices with uid[id1], uid[id2], form the unique partnership id and store as

uid[id1] * 1e7 + uid[id2]

and save in one column. This should also cut down on the computational time in creating these ids within each module.

Check that time.unit control/param works

Optimize the riskhist module

Currently, this module is inefficient as it tracks the risk history of men over time because: 1) it calculates behavioral indication metrics for PrEP scenarios that it does not use; and 2) it stores this as a n by t matrix where t is the number of time steps looking backward over which risk is assessed. This can be addressed by limited the calculation to only those indications of interest; and storing not the history of indications but the last time of indication as a vector of length n.

Clean up of STI initialization submodule

These are mainly stylistic edits to make in the init_status_sti_msm function:

• Several "defined but not used" variables that may be removed from the function. In Rstudio, under tools --> global options --> code --> diagnostics, check all of the check boxes in this tab. Relatedly, make sure that the two check boxes under the code --> saving, make sure these two checkboxes are checked. If you had planned to use these but just not currently, then keep them.
• Are you actually using the stage.time.syph variable for progression, as it appears like you are basing progression based on time.since.inf.syph instead. If it's not being used outside this submodule, you shouldn't need to store it on dat$attr
• The time.since.inf variables (e.g., on ln 844) are not named usefully. Instead should be related to time since stage started.
• Insert section markers (code --> insert section) for each of the three diseases in this function
• Lines 754-805 set a large number of attributes to NA, and then they are never operated upon during the function, just written to dat$attr at the end. For clarity, just write each of these directly as dat$attr$X <- rep(NA, num) at the end of the function.
• I think we should write everything out to dat$attr at the end of the function to increase clarity. That includes existing code from me for GC and CT. You can set that as a separate section header to further create the distinction.

Clean up probabilities and beta coefficients for condom use

These are a mix of probs and logit coefficients, some for P(condom use) and others for P(UAI). Clean these up!

New Note - R_registerRoutines and R_useDynamicSymbols

New note during build and reload:

File 'EpiModelHIV/libs/x64/EpiModelHIV.dll':
  Found no calls to: 'R_registerRoutines', 'R_useDynamicSymbols'
It is good practice to register native routines and to disable symbol search

@smjenness - Wasn't sure about the best course/location of fixing it right now, so creating an issue rather than just trying to address it right now. Looks like a line or so might need to be added to any Cpp files

Suggested that a single call to this R file (https://raw.githubusercontent.com/stan-dev/rstantools/master/R/init_cpp.R) might address it

(RcppCore/Rcpp#636)
(RcppCore/Rcpp#651)
(https://stackoverflow.com/questions/42313373/r-cmd-check-note-found-no-calls-to-r-registerroutines-r-usedynamicsymbols)
(https://github.com/kevinushey/sourcetools/blob/master/R/register.R)

Access to .rda files

Hello, when I try running the msm-test-script.R file, it tries to read 3 .rda files that I can't find in the GitHub:

netstats <- readRDS(file.path(scr.dir, "data/artnet.NetStats.Atlanta.rda"))
epistats <- readRDS(file.path(scr.dir, "data/artnet.EpiStats.Atlanta.rda"))
est <- readRDS(file.path(scr.dir, "data/artnet.NetEst.Atlanta.rda"))

Would I be able to get access to these .rda files so I can run this script? Thank you!

Hard return of roxygen code

Preferred margin for roxygen parameter definitions is 80 spaces. You can add a light margin line with Tools --> General options --> code --> display --> ✔️ Show margin, Margin column = 80.

Then add a hard return, especially in the params.R file.

Standardize order of new.edges matrix in same way as partnership list matrix

The partnership list matrix orders the new edgelist like so:

    highlow <- el[which(el[, 1] > el[, 2]), , drop = FALSE]
    lowhigh <- el[which(el[, 1] < el[, 2]), , drop = FALSE]
    part.el <- rbind(highlow[, 2:1], lowhigh)

That approach would make comparing within the disclose and condom modules simpler.

Initial distribution of syphilis by stages moved to initial conditions

Set ups for stage.syph.B.prob and stage.syph.W.prob should go into init_msm, and be drawn from there in the initialization module

Change HIV testing interval parameter to 2x the parameter when using deterministic interval

Don't index on the active attribute

We use dat$attr$active now only as a placeholder mainly to count the population size. But through some historical coding practices, related to the fact that we allow inactive (aka dead) people to stay on the network in the broader EpiModel framework, we use the variable for indexing vectors. That can and should be removed from any indexes. Make sure to pre/post test this to ensure no negative consequences.

Allow any infection to be initialized at 0

This would be for debugging and single epidemic models

Tag MSM vs HET modules with roxygen keywords

And also do a close read of the docs

absdiffby summary statistic

Hi Sam, Hi all,

I want to use absdiffby than I need to compute the target.

I can't find any fully explicit help file of absdiffby term. (here is the actual help file right? https://github.com/statnet/EpiModelHIV/blob/master/R/ErgmTerms.R )

Could you please tell me what this summary stat is exactly?

Which is the ref category? (e.g. by arg is sex and coded 0/1)
What the arg assym stands for?
Is the term code is https://github.com/statnet/EpiModelHIV/blob/master/src/changestats.users.c ?

Thanks,
Damien

Add unit tests for the simulation workflow in EpiModelHIV

Can use the presaved estimation and network statistics data. We can discuss further details...

HIV progress module

Hello,

I am wondering if the new values for stage should be updated only at the end of the function?
https://github.com/statnet/EpiModelHIV/blob/CombPrev/R/mod.hivprogress.R

I could not test this specific function, but a similar version.
For example:
when getting the index for stage == 2
https://github.com/statnet/EpiModelHIV/blob/26ea27a0a89f5b78a2b0123fc74875b0f5b26b44/R/mod.hivprogress.R#L64

then in line 68, it could potentially be selected again. Although, it might not be selected again based on time.since.inf
https://github.com/statnet/EpiModelHIV/blob/26ea27a0a89f5b78a2b0123fc74875b0f5b26b44/R/mod.hivprogress.R#L68

Cheers,
Fabricia.

Track number treated for STIs due to intervention

E.g. syph.tx.ept or syph.tx.test

het-test-script.R error: $ operator is invalid for atomic vectors

Hello,

I get this error when running the het-test-script.R that occurs in line 21 and after:

param <- param_het()
Error: $ operator is invalid for atomic vectors
init <- init_het(i.prev.male = 0.25, i.prev.feml = 0.25)
control <- control_het(nsteps = 2600)

sim <- netsim(est, param, init, control)
Error in UseMethod("as.control.list") :
no applicable method for 'as.control.list' applied to an object of class "control.net"

Would this be due to version problem? My sessionInfo() is the following:

sessionInfo()
R version 4.1.2 (2021-11-01)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Monterey 12.2.1

Matrix products: default
LAPACK: /Library/Frameworks/R.framework/Versions/4.1/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats graphics grDevices utils datasets methods base

other attached packages:
[1] ARTnet_2.5.0 ARTnetData_1.1 EpiModelHIV_1.5.0
[4] tergmLite_2.5.5 EpiModelHPC_2.1.2 EpiModel_2.2.1
[7] statnet.common_4.5.0 tergm_4.0.2 ergm_4.1.2
[10] networkDynamic_0.11.0 network_1.17.1 deSolve_1.30
[13] tictoc_1.0.1 forcats_0.5.1 stringr_1.4.0
[16] dplyr_1.0.7 purrr_0.3.4 readr_2.1.2
[19] tidyr_1.2.0 tibble_3.1.6 ggplot2_3.3.5
[22] tidyverse_1.3.1 readxl_1.3.1 rlang_1.0.1
[25] assertthat_0.2.1 truncnorm_1.0-8 ensurer_1.1
[28] gtools_3.9.2

loaded via a namespace (and not attached):
[1] colorspace_2.0-2 ellipsis_0.3.2
[3] class_7.3-19 rprojroot_2.0.2
[5] fs_1.5.2 rstudioapi_0.13
[7] proxy_0.4-26 remotes_2.4.2
[9] fansi_1.0.2 mvtnorm_1.1-3
[11] lubridate_1.8.0 xml2_1.3.3
[13] codetools_0.2-18 doParallel_1.0.17
[15] cachem_1.0.6 robustbase_0.93-9
[17] pkgload_1.2.4 jsonlite_1.7.2
[19] broom_0.7.12 dbplyr_2.1.1
[21] compiler_4.1.2 httr_1.4.2
[23] backports_1.4.1 Matrix_1.3-4
[25] fastmap_1.1.0 lazyeval_0.2.2
[27] cli_3.1.1 prettyunits_1.1.1
[29] tools_4.1.2 coda_0.19-4
[31] gtable_0.3.0 glue_1.6.1
[33] Rcpp_1.0.8 rle_0.9.2
[35] cellranger_1.1.0 vctrs_0.3.8
[37] ape_5.6-1 nlme_3.1-153
[39] iterators_1.0.14 ps_1.6.0
[41] testthat_3.1.1 trust_0.1-8
[43] rvest_1.0.2 lifecycle_1.0.1
[45] devtools_2.4.3 DEoptimR_1.0-10
[47] MASS_7.3-54 scales_1.1.1
[49] hms_1.1.1 parallel_4.1.2
[51] RColorBrewer_1.1-2 memoise_2.0.1
[53] stringi_1.7.6 desc_1.4.0
[55] foreach_1.5.2 e1071_1.7-9
[57] pkgbuild_1.3.1 pkgconfig_2.0.3
[59] lpSolveAPI_5.5.2.0-17.7 lattice_0.20-45
[61] processx_3.5.2 tidyselect_1.1.1
[63] magrittr_2.0.2 R6_2.5.1
[65] generics_0.1.2 DBI_1.1.2
[67] pillar_1.7.0 haven_2.4.3
[69] withr_2.4.3 bindata_0.9-20
[71] modelr_0.1.8 crayon_1.4.2
[73] utf8_1.2.2 tzdb_0.2.0
[75] usethis_2.1.5 grid_4.1.2
[77] callr_3.7.0 reprex_2.0.1
[79] munsell_0.5.0 sessioninfo_1.2.2

Remove all refs to attr$active throughout

Generate attribute codebook for EpiModelHIV

Dual-site (or three-site?) infection to include syphilis

@smjenness, how far should we go with this? Build structure for each combo but set to the same relative risk?
Possible combos:

• rGC and rCT
• rGC and syphilis
• rCT and syphilis
• rGC, rCT, and syphilis
• uGC and uCT
• uGC and syphilis
• uCT and syphilis
• uGC, uCT, and syphilis

Over counts based on multiple infections within timesteps

In the transmission module the transmission events and type defined by:

infected <- c(disc.ip[trans.ip == 1, 2],
              disc.rp[trans.rp == 1, 1])
inf.role <- c(rep(0, sum(trans.ip)), rep(1, sum(trans.rp)))
inf.type <- c(disc.ip[trans.ip == 1, "ptype"],
              disc.rp[trans.rp == 1, "ptype"])

These vectors are not unique which is not a problem when [infected] is used as an index but it results in over counts in the Epi summary statistics:

dat$epi$incid[at] <- length(infected)

dat$epi$trans.main[at] <- sum(inf.type == 1)
dat$epi$trans.casl[at] <- sum(inf.type == 2)
dat$epi$trans.inst[at] <- sum(inf.type == 3)

Change GC to NG in the codebase

Use stochastic draw for HIV testing default instead of deterministic interval

Check that models may be run on the backfill queue

Rename functions to be HIV or STI specific

For example, tx_msm would become hiv_tx_msm and ept_msm would become sti_ept_msm

Numeric attributes for stage and tt.traj

Easier conversion to C++ code