extras/notes.md

In OHDSI/Iris: What the study does (short line)

This file contains various notes for Iris development

changes

In May 2016 - most original Iris measures were incorporated into Achilles Iris is being used for new features and data quality

update: June 2016 some of the experimental features were moved to Achilles Derived measures. This mechanism can also be used to exchagne data density and other data quality parameters about a CDM dataset. Running latest version of Achilles Heel (where derived measures (=analyses) are generated is highly recommended)

running Iris just after you executed Achilles or just Achilles Heel

#remove older version of Iris
remove.packages('Iris')
#install Iris (assumes you have  devtools library installed)
devtools::install_github("OHDSI/Iris")

#re-use your connection object and point to the same "results" database but add where the results schema is 
connectionDetails$target_database_schema='results'

#also make sure you original set up of connectionDetails included a parameter called schema. 
#If not add it like this (otherwise the final execute call will fail) (think myCdm)
connectionDetails$schema='cdm5_inst'

#execute experimental parts of Iris by changing the part parameter (e.g., 3) 
#(see the SQL Iris folder to see all experimental parts)
source('c:/d/z_connect.R')
iPart<-Iris:::executePart(part=3,connectionDetails,cdmVersion = 5)
iPart<-Iris:::executePart(part=6,connectionDetails,cdmVersion = 5)

☺#results are in working R folder ( see it via commnad getwd()  )
#review .csv files generated or the iPart variable to see the outputs

Executing Iris on multiple datasets (sample code)

This sample code allows to execute a set of Iris parts on multiple datasets

#use your previous connectionDetails object with username and psw for database

# use this for single dataset Iris ZIP file generation
 dataLinks=c('mdcr_v5');resultsLinks=c('nih')


#use this for multiple datasets ZIP file process (modify the strings)
 dataLinks=c('ccae_v5','mdcr_v5','mdcd_v5')
 resultsLinks=c('ccae_v5_results','mdcr_v5_results','mdcd_v5_results')
 #ignore this line resultsLinks=c('nih','nih','nih')



library(Achilles)
for (i in seq_along(dataLinks)){
 print(dataLinks[i])

 cdmDatabaseSchema=dataLinks[i];resultsDatabaseSchema=resultsLinks[i]
 connectionDetails$schema=cdmDatabaseSchema
 connectionDetails$target_database_schema=resultsDatabaseSchema

 iPart<-Iris:::executePart(part=2,connectionDetails,cdmVersion = 5)
 iPart<-Iris:::executePart(part=3,connectionDetails,cdmVersion = 5)
 iPart<-Iris:::executePart(part=6,connectionDetails,cdmVersion = 5)

 #execute early implementation of Achilles Share
 shareRes<-Iris:::achillesShare(connectionDetails,cdmDatabaseSchema=cdmDatabaseSchema,resultsDatabaseSchema=resultsDatabaseSchema)
 #optionaly include that in export
 write.csv(shareRes,paste0(cdmDatabaseSchema,'-iris_part-',1,'.csv'),na='',row.names=F)


  #there are some new rules implemented in Achilles (from May 6th) 
  heelRes<-Achilles:::fetchAchillesHeelResults(connectionDetails,resultsDatabaseSchema)

  heelDerivedMeasuresTable<-Iris::fetchResultsTable(connectionDetails,resultsDatabaseSchema,'achilles_results_derived')
  #optionaly include Heel Derived measures output
  #write.csv(heelDerivedMeasuresTable,paste0(connectionDetails$schema,'-iris_part-',99,'.csv'),na='',row.names=F)


  #optionaly include Heel output
  write.csv(heelRes,paste0(connectionDetails$schema,'-iris_part-',0,'.csv'),na='',row.names=F)


}

zip('iris-export.zip',files='*iris_part*.csv')
#inspect the zip file to see what is being exported

More SQL files

To pilot new ideas quickly (and not let people wait for full Iris results), the SQL code is now split into more parts. (see SQL folder). To execute just part of Iris, alternative execute functions are used. (internal, so use triple semicolon to get to it)

Connection data

Iris plans to use ConnectionDetails method to specify server and login data to various functions. The beta execute functions already use this method.

Connection data can be in external .R file (and that file can be outside the project (so you don't upload it to public site accidentaly, for example). See example below

source('c:/d/z_connect.R')
iPart<-Iris:::executePart(part=3,connectionDetails,cdmVersion = 5)

Connecting using Redshift driver
iris_parameterized_3.sql
Executing Iris Part 3 on ccae_v5 ...
Execution time: 3.260972 secs
  ANALYSIS_ID                                                   ACHILLES_HEEL_WARNING RULE_ID COUNT_VALUE
1         499 WARNING: percentage of unmapped rows exceeds threshold (concept_0 rows)      27   0.0648880
2         699 WARNING: percentage of unmapped rows exceeds threshold (concept_0 rows)      27   0.1503383
3         799 WARNING: percentage of unmapped rows exceeds threshold (concept_0 rows)      27   2.2974252

We do encourage you to email feedback to vojtech huser at nih dot gov. (even if you just sucesfully ran the tool)

Rule notes

Heel should allow for some "US centric measures" (even though they may not apply world wide). CDMv5 has some "US centric features".

Yes vote 27,iris3," % of unmapped rows,"looks for concept 0 mapped rows in multiple tables and generates output if over a threshold" (not as percentage)

         28,iris3,"percentage of deceased patients","for general population this should be over certain threshold"
     32,iris3,"data from < 3 states"  (US centric)
no vote  38, data from les than <2 countries  
? vote   33,i3,"W:data by 3 digit zip are not recorded in PERSON table  (not global)
No vote  34(should rename),i3,"w:measuremtn rows with no time data are over threshold
Yes vote 35,TODO, w:new measure added to Achilles,ALL VALUES ARE NON NUMERICAL (strange), indicates "mapping based measurement data"
36 count   ration providers/patients is too low
34, percentage of unmapped row (in rule overview huser_dev)

Other notes

#achillesShare (experimental part, for later integration into Achilles)
library(Achilles)
source('c:/d/z_connect.R')
cdmDatabaseSchema       ='ccae_v5'  #modify this for your context e.g.,XYZdata
resultsDatabaseSchema   ='nih'      #modify this for your context e.g.,XYZresults

Iris:::achillesShare(connectionDetails,cdmDatabaseSchema=cdmDatabaseSchema,resultsDatabaseSchema=resultsDatabaseSchema)

OHDSI/Iris documentation built on May 7, 2019, 8:26 p.m.