programming language for statistical computing supported by the R Foundation for Statistical Computing.R4NName of medical unitR6SName of medical unitRAA or R+A+AResponsibility, authority, and accountabilityRBBBRight bundle branch blockRBCRelationship‐Based Care®RMCRecovery management checkupsRMSEARoot mean square error of approximationRNRegistered nurseSAMSASubstance Abuse and Mental Health Services AdministrationSASStatistical Analysis System is a software system for data analysisSBPSystolic blood pressureScDDoctor of scienceSCIPSurgical care improvement projectSCNSenior charge nurseSCUStep‐down unitSEMStructural equation modelSPSSStatistical Package for the Social Sciences is a software system owned by IBM (International Business Machines)SRMRStandardized root mean square residualSTSSociotechnical systems (theory)ST‐TSegment of the heart tracing in an electrocardiographSUDSubstance use disorderTIATransient ischemic attackTIPTreatment improvement protocolsTLCTriple lumen catheterTTETransthoracic echocrdiogramUTDUnable to determineVSVital signVS: SBPVital sign: systolic blood pressureVS: DBPVital sign: diastolic blood pressure
Acknowledgments
First, the three editors of this book would like to acknowledge our developmental editor, Rebecca Smith, who has made the inaccessible, complex concepts of data analytics simple to understand and exciting to contemplate.
Secondly, we would like to acknowledge all the analysts and mathematicians from other disciplines who have enthusiastically and humbly shared their knowledge of mathematics and how it is applied in science. We have been inspired by the depth and breadth of what you know and by your eagerness to learn from others. The lead editor would also like to ask the indulgence of all of the mathematicians, analysts, and scientists who will read this book, as you encounter moments in this book where brevity and simplicity have taken precedence over thorough scientific explanations. In an effort to make this book accessible to a lay audience, much of the technical talk has been truncated or eliminated.
Thirdly, we acknowledge the visionary leaders who had the courage to step out and measure what matters—behavior and context. Without your understanding that data beyond frequencies was needed, the ability to use predictive analytics to improve healthcare outcomes would still be an elusive dream.
Finally, the editors of this book acknowledge all the staff members who took part in these studies. Every one of you made each model of measurement better, and you played a vital part in producing the groundbreaking findings in this book. Without you, this book would not exist.
1 Using Predictive Analytics to Move from Reactive to Proactive Management of Outcomes
John W. Nelson
For predictive analytics to be useful in your quest to improve healthcare outcomes, models for measurement must reflect the exact context in which you seek to make improvements. Data must resonate with the staff members closest to the work, so that action plans premised on the data are specific, engaging, and instantly seen as relevant. This chapter provides 16 steps the author has used in healthcare settings to engage staff members in outcomes improvement. Models created using these steps have proven effective in improving outcomes and saving millions of dollars because the process engages the entire healthcare team to provide input into (a) the design of measurement instruments, (b) interpretation of results, and (c) application of interventions, based on the data, to improve outcomes. Analysts and staff members build models of measurement that tell the story of the organization empirically, which makes the data not only actionable but relatable.
The Art and Science of Making Data Accessible
Data can and should read like a story. The presentation of data in healthcare should be interesting and engaging because it reflects empirically what people are experiencing operationally. It is the experience of this author that when data is presented as part of the employees' story, they love it. Everybody likes to talk about their experience, and when a data analyst is able to tell them what they themselves are experiencing with the numbers to back it up, it places staff members at the edges of their seats.
The presentation of data in healthcare should be interesting and engaging because it reflects empirically what people are experiencing operationally.
With the advent of big data, machine learning, and artificial intelligence, we now too often turn one of our oldest, most cherished human traditions—storytelling—over to machines. The stories machines tell reveal patterns and relationships that staff members are familiar with, but they leave out the context, rendering their stories unrelatable. If your goal is to provide people with information they instantly recognize as accurate and relevant, your models must be specified to the people and contexts they presume to report on, and only then should they be examined empirically.
The stories machines tell leave out the context, rendering their stories unrelatable.
You are about to meet a 16‐step process for how to tell a story, using data, that is not only interesting; it is actionable operationally. No two organizations are the same, and no organization stays the same over time. Thus, it is critical to evaluate whether data presented within an organization accurately captures the context and nuance of the organization at a point in time.
Admittedly, the idea of 16 steps may initially feel prohibitively complex. As you spend time looking at the process in terms of some practical examples, however, you will find that what I have provided is simply a template for examining and sorting data which you will find not only simple to use, but ultimately quite liberating.
As you read through the steps, you are likely to intuit what role you would play and what roles you would not play, in this process. Some of the work described in the steps will be done by staff members closest to the work being analyzed, and some will be done by mathematicians, statisticians, programmers, and/or data analysts. If some of the content is unfamiliar to you or seems beyond your reach, rest assured that someone on the team will know just what to do.
If some of the content is foreign to you or seems beyond your reach, rest assured that someone on the team will know just what to do.
Step 1: Identify the Variable of Interest
What problem or improvement opportunity is of concern and/or interest to your team? This could be an issue such as patient falls or the civility of employees, managers, or patients. This problem or improvement opportunity is referred to as the variable of interest; it will also serve as the focal point of the story your team will eventually be telling with data. In some settings, the variable of interest may be referred to as the outcome of interest.
Step 2: Identify the Things That Relate to the Variable of Interest: AKA, Predictor Variables
If your team was looking to improve an outcome related to falls, for example, you would want to examine anything that could predict, precede, or contribute to a fall. Assemble members of the care team and think together about what might lead to a fall, such as (a) a wet floor, (b) staff members with stature too small to be assisting patients with walking, (c) the patient taking a heart medication a little before the fall, and so on. As the discussion of everything that relates to your variable of interest continues, designate one person to write down all the things being mentioned, so the people brainstorming what relates to falls can focus solely on describing the experience and are not distracted by writing things down (Kahneman, 2011). Do not search far and wide for possible predictor variables or even think about the evidence from the literature at
