In unmnn/uhregmisc: Miscellaneous Helper Functions For The Univ. Regensburg Tinnitus Data

knitr::opts_chunk$set(echo = TRUE)

uhregmisc

Miscellaneous Helper Functions For The Univ. Regensburg Tinnitus Data.

Important: For most of the functions to work, you would need to put into the data-raw/ folder the original dataset (200622_uhreg.xlsx) which cannot be uploaded to GitHub due to copyright reasons.

Overview

• data-raw/uhreg.R: preprocessing script for the raw UHREG data
• datasets uhreg and data_dict (+ json export at inst/extdata/data_dict.json)
• function discretize_score() for discretizing a numeric questionnaire score according to predefined cutoff values from literature.
• Shiny app to explore missingness / available data.

Installation

remotes::install_github("unmnn/uhregmisc")

library(uhregmisc)

Shiny app on data missingness

knitr::include_graphics(file.path("app-ume", "app-ume.png"))

Provides an overview of missing data patterns. You can specify a data subset based on:

• visit type: screening, baseline, interim visit, final visit and/or followup
• treatment (10 most frequent treatment types are selectable)
• gender
• age range

Additionally, you can choose to order questionnaires (y-axis) either by mutual missingness (determined by a hierarchical clustering) or just alphabetically by their name.

How to run app

# When running the first time, create the necessary pre-processed dataset with:
source("app-ume/ume_prep-data.R")
# Afterwards, the folder `app-ume` should contain the file `data.rds`.
# Then, just run the app via:
shiny::runApp("app-ume")

Data preprocessing steps

The script data-raw/uhreg.R preprocesses the UHREG data.

• part of column name representing the questionnaire is written in caps
• establish consistent name convention: prefix of variable name (part until first underscore) represents the name of the corresponding questionnaire
• example: TQ_q01 → TQ questionnaire
• meta variables, e.g. patient ID or treatment type, have the prefix .META
• some variables are renamed for consistency of chronology, e.g. TSCHQ_q36_1_otologic_additional instead of TSCHQ_q361_otologic_additional
• abbreviations in names: q instead of question, lvs instead of last_validation_status
• example: TBF12_q10 instead of TBF12_question10
• in column names, zero-padding is added to allow for alphabetical ordering, e.g. TBF12_q01 instead of TBF12_q1 (there is also a TBF12_q10)
• all BDI columns are removed due to complete missingness
• variable types are converted where appropriate:
• TSCHQ_q02_sex to factor
• .META_patient_id to character
• .META_visit_day, TSCHQ_q05_begin_tinnitus to date
• -1 values are replaced with NA_integer_ where appropriate
• some (sub-)scales are created:
• TFI subscales: intrusive, sense of control, cognitive, sleep, auditory, relaxation, quality of life, emotional
• TQ subscales: emotional distress, cognitive distress, intrusiveness, auditory perceptual difficulties, sleep disturbances, somatic complaints
• MDI: depression categorization

Open questions

• How to deal with the huge variety of treatment pathways?

109 unique treatment pathways

                               S                         S-B-FV-FU 
                              869                                38 
                             FV-B                            S-B-FV 
                               32                                24 
                             B-FV                               S-S 
                               22                                17 
                     B-FV-I-I-I-S                     B-FV-FU-FU-FU 
                               16                                15

• What do the columns with "last_validation_status" mean? What do their values (-1, 0, 1) mean?
• Documentation for "v_exp_tschq_q36_otologic" and "v_exp_tschq_q361_otologic_additional"

TODOS

• Improve rudimentary documentation for uhreg and data_dict
• Add function that dummifies categorical variables

data_dict

library(dplyr)
library(purrr)
data_dict

var <- "TSCHQ_q09_perception"
data_dict %>% filter(variable == var) %>% pull(description)
data_dict %>% filter(variable == var) %>% pluck("value", 1)

uhreg overview

skimr::skim_without_charts(uhreg)

unmnn/uhregmisc documentation built on Nov. 13, 2020, 4:05 a.m.