In vjcitn/AnvBiocFHIR: Use pyAnVIL and SMART infrastructure to work with AnVIL FHIR in R

Introduction

This package uses pyAnVIL and SMART FHIR python infrastructure to work with FHIR resources in AnVIL.

FHIR stores

library(AnvBiocFHIR)
allst = list_stores()
length(allst)
head(allst)

Resources in a store

rgru = store_resources(tail(allst)[1])
length(rgru)
class(rgru[[1]])

A store has one or more ResearchStudy instances.

sapply(allst[1:4], function(x)length(store_resources(x)))

Methods for an instance.

names(rgru[[1]])

Patients in a FHIR store

Listing and counting

p1 = anv_fhir_test_list_patients(anvurl("phs001272-DS-CSD-MDS"))
p1$title
length(p1$pats)
names(p1$pats[[1]])

There are many available fields, but few are actually populated.

names(p1$pats[[1]]$as_json())

Multicore operations

We can dispatch to cluster nodes but must serialize all results before returning. To illustrate we define a function that counts patients (up to 1000, until pagination is straightened out) in a store:

countpats = function (x) 
{
    y = anv_fhir_test_list_patients(anvurl(x))
    ans = length(y$pats)
    names(ans) = y$title
    ans
}

No python references are returned.

We won't test this in this document, but this is a typical result:

> system.time(pp <- lapply(allst[2:5], countpats))
   user  system elapsed 
 11.474   0.802  16.643 
> library(parallel)
> options(mc.cores=4)
> system.time(pp2 <- mclapply(allst[2:5], countpats))
   user  system elapsed 
 12.353   1.393   6.106 

vjcitn/AnvBiocFHIR documentation built on May 31, 2022, 10:39 a.m.