CreateStrategusAnalysisSpecification.R

################################################################################
# INSTRUCTIONS: Make sure you have downloaded your cohorts using 
# DownloadCohorts.R and that those cohorts are stored in the "inst" folder
# of the project. This script is written to use the sample study cohorts
# located in "inst/sampleStudy/Eunomia" so you will need to modify this in the code 
# below. 
# 
# See the Create analysis specifications section
# of the UsingThisTemplate.md for more details.
# 
# More information about Strategus HADES modules can be found at:
# https://ohdsi.github.io/Strategus/reference/index.html#omop-cdm-hades-modules.
# This help page also contains links to the corresponding HADES package that
# further details.
# ##############################################################################
library(dplyr)
library(Strategus)

# Time-at-risks (TARs) for the outcomes of interest in your study
timeAtRisks <- tibble(
  label = c("30d"),
  riskWindowStart  = c(1),
  startAnchor = c("cohort start"),
  riskWindowEnd  = c(30),
  endAnchor = c("cohort start")
)

# If you are not restricting your study to a specific time window, 
# please make these strings empty
studyStartDate <- '20100101' #start year
studyEndDate <- ''   #present
# Some of the settings require study dates with hyphens
studyStartDateWithHyphens <- gsub("(\\d{4})(\\d{2})(\\d{2})", "\\1-\\2-\\3", studyStartDate)
studyEndDateWithHyphens <- gsub("(\\d{4})(\\d{2})(\\d{2})", "\\1-\\2-\\3", studyEndDate)

# Consider these settings for estimation  ----------------------------------------
useCleanWindowForPriorOutcomeLookback <- FALSE # If FALSE, lookback window is all time prior, i.e., including only first events

# Shared Resources -------------------------------------------------------------
# Get the list of cohorts - NOTE: you should modify this for your
# study to retrieve the cohorts you downloaded as part of
# DownloadCohorts.R
cohortDefinitionSet <- CohortGenerator::getCohortDefinitionSet(
  settingsFileName = "inst/Cohorts.csv",
  jsonFolder = "inst/cohorts",
  sqlFolder = "inst/sql/sql_server"
)

# Subset Operators
subsetOperators <- list()
subsetOperators[[length(subsetOperators) + 1]] <- 
    CohortGenerator::createLimitSubset(
    name = 'prepandemic',
    calendarStartDate = '2010-01-01',
    calendarEndDate = '2019-12-31'
  )
subsetOperators[[length(subsetOperators) + 1]] <- 
    CohortGenerator::createLimitSubset(
    name = 'pandemic',
    calendarStartDate = '2020-01-01',
    calendarEndDate = '2022-12-31'
  )
subsetOperators[[length(subsetOperators) + 1]] <- 
    CohortGenerator::createLimitSubset(
    name = 'postpandemic',
    #priorTime = 0,
    #followUpTime = 0,
    #limitTo = "all",
    calendarStartDate = '2023-01-01',
    calendarEndDate = '2025-06-06'
  )


for (i in 1:length(subsetOperators)) {
subsetDefs <- 
  CohortGenerator::createCohortSubsetDefinition(
    name = "",
    definitionId = i,
    subsetOperators = subsetOperators[i]
  )

cohortDefinitionSet <- cohortDefinitionSet |>
  CohortGenerator::addCohortSubsetDefinition(cohortSubsetDefintion = subsetDefs, targetCohortIds = c(1, 2))
}

knitr::kable(cohortDefinitionSet[, names(cohortDefinitionSet)[which(!names(cohortDefinitionSet) %in% c("json", "sql"))]])

if (any(duplicated(cohortDefinitionSet$cohortId))) {
  stop("*** Error: duplicate cohort IDs found ***")
}

# Create some data frames to hold the cohorts we'll use in each analysis ---------------
# Outcomes: The outcome for this study is cohort_id >= 3 
oList <- cohortDefinitionSet %>%
  filter(.data$cohortId > 2 & .data$cohortId <= 22) %>%
  mutate(outcomeCohortId = cohortId, outcomeCohortName = cohortName) %>%
  select(outcomeCohortId, outcomeCohortName) %>%
  mutate(cleanWindow = 0)

# Df for TP
cohorts <- cohortDefinitionSet %>%
  select (cohortId, cohortName)

cohorts$type <- ifelse(cohorts$cohortId %in% oList$outcomeCohortId, 'event', 'target')

# temp remove 1 and 2 to try to overcome stuck

cohorts = cohorts%>%filter(cohortId>2)


# CohortGeneratorModule --------------------------------------------------------
cgModuleSettingsCreator <- CohortGeneratorModule$new()
cohortDefinitionShared <- cgModuleSettingsCreator$createCohortSharedResourceSpecifications(cohortDefinitionSet)
cohortGeneratorModuleSpecifications <- cgModuleSettingsCreator$createModuleSpecifications(
  generateStats = TRUE
)

# CharacterizationModule Settings ---------------------------------------------
cModuleSettingsCreator <- CharacterizationModule$new()
characterizationModuleSpecifications <- cModuleSettingsCreator$createModuleSpecifications(
  targetIds = cohortDefinitionSet$cohortId, # NOTE: This is all T/C/I/O
  outcomeIds = oList$outcomeCohortId,
  outcomeWashoutDays = oList$cleanWindow , 
  minPriorObservation = 0,
  dechallengeStopInterval = 30,
  dechallengeEvaluationWindow = 30,
  riskWindowStart = timeAtRisks$riskWindowStart, 
  startAnchor = timeAtRisks$startAnchor, 
  riskWindowEnd = timeAtRisks$riskWindowEnd, 
  endAnchor = timeAtRisks$endAnchor,
  minCharacterizationMean = .01
)

# treatmentPatternsModule Settings ---------------------------------------------
tModuleSettingsCreator <- TreatmentPatternsModule$new()
treatmentPatternsModuleSpecifications <- tModuleSettingsCreator$createModuleSpecifications(
  cohorts,
  includeTreatments = "startDate",
  indexDateOffset = 0,
  minEraDuration = 1,
  splitEventCohorts = NULL,
  splitTime = NULL,
  eraCollapseSize = 7,
  combinationWindow = 1,
  minPostCombinationDuration = 1,
  filterTreatments = "All",
  maxPathLength = 7,
  ageWindow = 10,
  minCellCount = 5,
  censorType = "minCellCount"
)

# Create the analysis specifications ------------------------------------------
analysisSpecifications <- Strategus::createEmptyAnalysisSpecificiations() |>
  Strategus::addSharedResources(cohortDefinitionShared) |>
  Strategus::addModuleSpecifications(cohortGeneratorModuleSpecifications) |>
  Strategus::addModuleSpecifications(characterizationModuleSpecifications) |>
  Strategus::addModuleSpecifications(treatmentPatternsModuleSpecifications) 

ParallelLogger::saveSettingsToJson(
  analysisSpecifications, 
  file.path("inst", "CAPAnalysisSpecification.json")
)