R for Health Data Science

Ewen Harrison, Riinu Pius

• Publisher: Chapman and Hall/CRC
• Year: 2020
• ISBN:
  9781000226164
• Edition: 1
• Topic: Computer Science
• Subtopic: Computer Science General
• Index: Computer Science

In this age of information, the manipulation, analysis, and interpretation of data have become a fundamental part of professional life; nowhere more so than in the delivery of healthcare. From the understanding of disease and the development of new treatments, to the diagnosis and management of individual patients, the use of data and technology is now an integral part of the business of healthcare. Those working in healthcare interact daily with data, often without realising it. The conversion of this avalanche of information to useful knowledge is essential for high-quality patient care.

R for Health Data Science includes everything a healthcare professional needs to go from R novice to R guru. By the end of this book, you will be taking a sophisticated approach to health data science with beautiful visualisations, elegant tables, and nuanced analyses.

Features

• 
  
  

• Provides an introduction to the fundamentals of R for healthcare professionals

• 
  
  

• Highlights the most popular statistical approaches to health data science

• 
  
  

• Written to be as accessible as possible with minimal mathematics

• 
  
  

• Emphasises the importance of truly understanding the underlying data through the use of plots

• 
  
  

• Includes numerous examples that can be adapted for your own data

• 
  
  

• Helps you create publishable documents and collaborate across teams

With this book, you are in safe hands – Prof. Harrison is a clinician and Dr. Pius is a data scientist, bringing 25 years' combined experience of using R at the coal face. This content has been taught to hundreds of individuals from a variety of backgrounds, from rank beginners to experts moving to R from other platforms.

Part II

Data analysis

In the second part of this book, we focus specifically on the business of data analysis, that is, formulating clear questions and seeking to answer them using available datasets.

Again, we emphasise the importance of understanding the underlying data through visualisation, rather than relying on statistical tests or, heaven forbid, the p-value alone.

There are five chapters. Testing for continuous outcome variables (6) leads naturally into linear regression (7). We would expect the majority of actual analysis done by readers to be using the methods in chapter 7 rather than 6. Similarly, testing for categorical outcome variables (8) leads naturally to logistic regression (9), where we would expect the majority of work to focus.

Chapters 6 and 8 however do provide helpful reminders of how to prepare data for these analyses and shouldn’t be skipped. time-to-event data (10) introduces survival analysis and includes sections on the manipulation of dates.

6

Working with continuous outcome variables

Continuous data can be measured.

Categorical data can be counted.

6.1 Continuous data

Continuous data is everywhere in healthcare. From physiological measures in patients such as systolic blood pressure or pulmonary function tests, through to population measures like life expectancy or disease incidence, the analysis of continuous outcome measures is common and important.

Our goal in most health data questions, is to draw a conclusion on a comparison between groups. For instance, understanding differences in life expectancy between the year 2002 and 2007 is more useful than simply describing the average life expectancy across all of time.

The basis for comparisons between continuous measures is the distribution of the data. That word, as many which have a statistical flavour, brings on the sweats in many people. It needn’t. By distribution, we are simply referring to the shape of the data.

6.2 The Question

The examples in this chapter all use the data introduced previously from the amazing Gapminder project.¹ We will start by looking at the life expectancy of populations over time and in different geographical regions.

6.3 Get and check the data

# Load packages

library(tidyverse)

library(finalfit)

library(gapminder)

# Create object gapdata from object gapminder

gapdata <- gapminder

It is vital that datasets be carefully inspected when first read (for help reading data into R see 2.1). The three functions below provide a clear summary, allowing errors or miscoding to be quickly identified. It is particularly important to ensure that any missing data is identified (see Chapter 11). If you don’t do this you will regret it! There are many times when an analysis has got to a relatively advanced stage before the researcher was hit by the realisation that the dataset was far from complete.

glimpse(gapdata) # each variable as line, variable type, first values

## Rows: 1,704

## Columns: 6

## $ country <fct> Afghanistan, Afghanistan, Afghanistan, Afghanistan, Afgha...

## $ continent <fct> Asia, Asia, Asia, Asia, Asia, Asia, Asia, Asia, Asia, Asi...

## $ year <int> 1952, 1957, 1962, 1967, 1972, 1977, 1982, 1987, 1992, 199...

## $ lifeExp <dbl> 28.801, 30.332, 31.997, 34.020, 36.088, 38.438, 39.854, 4...

## $ pop <int> 8425333, 9240934, 10267083, 11537966, 13079460, 14880372,...

## $ gdpPercap <dbl> 779.4453, 820.8530, 853.1007, 836.1971, 739.9811, 786.113...

missing_glimpse(gapdata) # missin...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Dedication
• Preface
• About the Authors
• I Data wrangling and visualisation
• II Data analysis
• III Workflow
• Appendix
• Bibliography
• Index