Update MIMIC IV to version 2.2

[prockenschaub]

MIMIC IV has introduced several changes to the data structure, including new tables and columns. This has lead to several issues for new users: #15 #28 #33 #42 #44

NOTE: this PR breaks backward compatibility with prior MIMIC IV versions. Fixing access to previous versions of the data is in process.

[nbenn]

@prockenschaub This is a pure config change, so in terms of review comments, I don't have much to offer. Have you tried something like "loading all concepts from new mimic iv" to check compatibility of current concept definitions with the new data source definition?

[nbenn]

The check CI failures seem unrelated. I'm happy to sort this out post-merge.

[prockenschaub]

I ran all non-recursive concepts and the difference was a drop in 4-5% of rows across all concepts. This should be due to a reduction in patients for version 2.2. I didn't run recursive concepts because collect_dots inload_concepts.cncpt currently throws an error if I pass src as config (related to #19).

Sys.setenv(RICU_SRC_LOAD = "")
library(ricu)

# Get concepts
dict <- load_dictionary(src = "miiv")

# Get custom miiv config
cfg <- load_src_cfg(name = "miiv-1.0", cfg_dirs = "/path/to/miiv/versions")
attach_src(cfg$miiv, data_dir = "/path/to/miiv-1.0/data")

# Try to load a recursive concept
load_concepts(dict[["avpu"]], src = data$miiv)
#> • gcs
#> • egcs
#> • mgcs
#> • vgcs
#> • tgcs
#> • ett_gcs
#> Error in `assert_that()` at ricu-package/R/callback-cncpt.R:10:4:
#> ! length(dots) not identical to 1L
#> Run `rlang::last_trace()` to see where the error occurred.

I nevertheless think this is a sufficient check and this can be merged now.

[dplecko]

Solved in ricu 0.6.0.