Search the KHIT Blog

Monday, August 31, 2015

Health IT: process mining and analytics for healthcare QI

So, I saw this tweet by UberGeek Chuck Webster, whom I've known for a while (mostly online, though we met at HIMSS one year)...

Went and looked it up on both Springer and Amazon.

The Amazon blurb copy:
What are the possibilities for process mining in hospitals?  In this book the authors provide an answer to this question by presenting a healthcare reference model that outlines all the different classes of data that are potentially available for process mining in healthcare and the relationships between them. Subsequently, based on this reference model, they explain the application opportunities for process mining in this domain and discuss the various kinds of analyses that can be performed.

They focus on organizational healthcare processes rather than medical treatment processes. The combination of event data and process mining techniques allows them to analyze the operational processes within a hospital based on facts, thus providing a solid basis for managing and improving processes within hospitals. To this end, they also explicitly elaborate on data quality issues that are relevant for the data aspects of the healthcare reference model.

This book mainly targets advanced professionals involved in areas related to business process management, business intelligence, data mining, and business process redesign for healthcare systems as well as graduate students specializing in healthcare information systems and process analysis.
Being as yet too cheapskate to pay $52.24 for the Kindle version, I fired up Dragon and read in the following from the "Look Inside" preview.
Chapter 1

Abstract: Health care costs have increased dramatically and the demand for high quality care will only grow in our aging society. At the same time, more event data are being collected about care processes. Healthcare information systems (HIS) have hundreds of tables with patient related event data. Therefore, it is quite natural to exploit these data to improve care processes while reducing costs. Data science techniques will play a crucial role in this endeavor. Process mining can be used to improve compliance and performance while reducing costs.

Process mining has been applied successfully in a variety of domains, e.g., banking, insurance, logistics, production, and e-government, customer relationship management, remote monitoring, and smart diagnostics. Through process mining one can relate the actual behavior of people, machines, and organizations with modeled behavior. This often leads to surprising insights showing that reality is very different from the perceptions, opinions, and beliefs stakeholders have. This is particularly relevant for healthcare processes. These processes are often only partly structured with many exceptional behaviors and different stakeholders. Healthcare requires flexibility and ad hoc decision-making. These characteristics make it impossible to apply rigorous business process management (BPM), workflow management (WFM), and business process reengineering (BPR) techniques. Clearly, a hospital is not a factory and patients cannot be cured using a conveyor belt system. However, the abundance of data collected in today’s hospitals can be used to improve care processes dramatically. Unlike many other domains, there is still room for dramatic improvements in healthcare processes. Process mining can be used to improve compliance and performance while reducing costs…

One approach... is to focus on the many complex time-consuming and nontrivial processes that are undertaken within these organizations.  In order to give suggestions for improving and redesigning these processes they need to be analyzed. Such an analysis is typically done by conducting interviews. Unfortunately, this is time-consuming and costly.  furthermore, typically the subjective view is provided on how a process is executed. That is, people involved in the performance of these healthcare processes (e.g., physicians, managers) tend to have an ideal scenario in mind, which in reality is only one of the many scenarios possible. Moreover, in many hospitals “political battles” take place due to organizational issues. Different stakeholders may have different views, e.g., some parties may not be interested in reducing the overall costs and improving transparency. Therefore in order to give objective suggestions for improving and redesigning processes one needs to exploit the event data readily available. Such an analysis is possible using process mining.

1.2 process mining: data science in action

Although our capabilities to store and process data have been increasing exponentially since the 1960s, suddenly many organizations realized that survival is not possible without exploiting available data intelligently. This of course  also holds for healthcare organizations. Society, organizations, and people are “always on”. Data are collected about anything, at any time, and at any place. Gartner uses the phrase “the nexus of forces” to refer to the convergence and mutual reinforcement of four interdependent trends: social, mobile, cloud, and information. The term  “big data” is often used to refer to the incredible growth of data in recent years. For hospitals of course the goal is not to collect more data, but to exploit data to realize more efficient and effective care processes.

Obviously the term “big data” has been hyped in recent years. However there is a growing demand for data scientists that can turn data into value. Just like computer science emerged as a new discipline from mathematics when computers became abundantly available, we now seem worth of data science as a new discipline driven by the huge amounts of data available today. Data science aims to use the different data sources to answer questions that can be grouped into the following four categories:

Reporting: what happened?
Diagnosis: why did it happen?
Prediction: what will happen?
Recommendation: what is the best that can happen?

Figure 1.2 deliberately emphasizes the process aspect. The goal is not to analyze data, but to improve care processes. Process mining aims to discover, monitor and improve real processes by extracting knowledge from event logs readily available in today’s information systems. Starting point for process mining is an event log. Each event in such a log refers to an activity (i.e., a well-defined step in some process) and is related to a particular case (i.e., a process instance)...

Yeah. That's enough. I get the drift. Better late than never, I guess. My experience in this area of "operations analytics" antedates my Health IT involvement by quite some time. As I posted a year ago:
@BobbyGvegas says...

A comment I just made over at THCB.
@BobbyGvegas says: August 20, 2014 at 12:39 pm
There are 3 fundamental aspects of workflow in the digital era: physical tasks, IT (EHR) tasks, and cognitive tasks. Every certified EHR has to have an audit trail to comply with HIPAA, given that every time ePHI is created, viewed, updated, transmitted, or deleted the transaction must be “date-time/who/what/about whom” captured in the audit trail log. 
The ePHI audit log, to me, is a workflow record component. It can’t tell me WHY front desk Susie or Dr. Simmons took so long to get from one transaction element to the next — i.e., physical movements or cognitive efforts — but it can tell me a lot, adroitly analyzed. 
I worked for number of years as a credit risk and portfolio management analyst in a credit card bank. We had an in-house collections department that took up an entire football field sized building, housing about 1,000 call center employees. I had free run of the internal network and data warehouse. One day I just happened upon the call center database and the source code modules (written by an IT employee in FoxPro, which I already knew at an expert level). I could open up the collections call log and watch calls get completed in real time. We were doing maybe a million outbound calls a month (a small Visa/MC bank).

(My fav in the Comments field was “CH [cardholder] used fowl language,” LOL)

It was, in essence, an ongoing workflow record of collections activity.
I pulled these data over into SAS and ground them up. I could track and analyze all activity sorted by any criteria I wished, all the way down to the individual collector level. I could see what you did all day, and what we got (or didn’t) for your trouble.

I was [able to] rather quickly show upper management “Seriously? You dudes are spending $1,000 to collect $50, every day, every hour” etc. The misalignment was stunning. I started issuing a snarky monthly summary called “The Don Quixote Report” with a monthly “winner.” …Yeah, we called this hapless deadbeat 143 times this month trying to get 15 bucks out of him…

Well, it didn’t take long to squelch all that. We saved the bank 6 million dollars in Collections Department Ops costs that year via call center reforms. Didn’t exactly endear me to the VP of Collections, whose bonus was tied to his budget.

Gimme a SAS or Stata install and SQL access to the HIT audit logs, and I will tell you some pretty interesting (Wafts-of-Taylorism 2.0) workflow stories.
Again, better late than never in the healthcare space.

Also, apropos of the inextricability of "process" and digital IT, from Dr. Jerome Carter's latest:
We are in the earliest stages of determining how current software design principles, precepts, and methods apply to clinical care systems. The paper chart is the main stumbling block preventing a critical (re)assessment of clinical software design principles. A static information archive has been used as the basis for EHR systems, leaving clinical processes out in the cold. As we are learning, processes are just as important as data. Unfortunately, legacy clinical software has made workflows implicit, and current UCD processes and usability research are mainly focused on legacy EHR systems. Legacy systems are poor at process support because they are designed to offer information, not support processes. Fixing them will require revisiting all three design aspects along with the recognition of a fourth aspect – explicit process representation. Next-generation clinical software must allow explicit process representation and execution. Likewise, clinical software development practices must be able to unambiguously represent clinical processes from requirements to deployment.

Clearly, the HIT community has embraced the idea that processes are important. But, in doing so, seems to be doubling down on trying to shoehorn process support into legacy systems whose designs are based on paper charts. Bad usability and poor interfaces are frequently cited as the reasons EHR systems are so disruptive to clinicians’ workflows. However, usability and interface issues are often symptoms of deeper problems. Design teams must address each of the four aspects of clinical software, and these teams must have a shared conceptualization of what the term “design” encompasses. Explicit representation of clinical processes and workflow technology must become part of the design discussion when addressing usability concerns, care coordination features, and CDS needs. Otherwise, we will continue to bump into the “workflow” elephant in the room. Design has four aspects—all must addressed…
UX of the IT systems themselves are part of the processes to be studied. As I noted in my comment under Dr. Carter's post:
“…The AI community began by trying to model isolated human intelligence while the emerging community of human-computer interaction designers followed in Engelbart’s augmentation tradition. He had begun by designing a computer system that enhanced the capabilities of small groups of people who collaborated. Now Gruber had firmly aligned himself with the IA community. At the Stanford Knowledge Systems Laboratory, he had interviewed avionics designers and took their insights to heart. There had been an entire era of industrial design during which designers assumed that people would adapt to the machine. Designers originally believed that the machine was the center of the universe and the people who used the machines were peripheral actors. Aircraft designers had learned the hard way that until they considered the human-machine interaction as a single system, they built control systems that led to aircraft crashes. It simply wasn’t possible to account for all accidents by blaming pilot error. Aircraft cockpit design changed, however, when designers realized that the pilot was part of the system. Variables like attention span and cognitive load, which had been pioneered and popularized by psychologists, became an integral part first in avionics and, more recently, computer system design…”

Markoff, John (2015-08-25). Machines of Loving Grace: The Quest for Common Ground Between Humans and Robots (Kindle Locations 4720-4730). HarperCollins. Kindle Edition.

More to come...

No comments:

Post a Comment