Finally, A Tool to Re-Engineer Health Care: The Workflow Engine
The Challenge of Health Care
There is a paradox of productivity in health care. In every other industry, the adoption of more technology, such as the assembly line or the microprocessor, has led to significantly lower costs and increased productivity. Over the last 25 years, American health care has incorporated a spectacular array of technologies. While these technologies have allowed us to peer deep into the cellular processes of the human body, they have not reaped lower costs or greater operational efficiency. Why?
The Complexity of Health Care
Health care's technological advances require increases in labor in order to manage the technology. This results in a paradox of limited productivity in spite of increasing technology. Fundamentally, health care is costly because it is labor intensive. Many industry problems currently under review, such as medication safety, nurse burn-out, and medical errors resulting in malpractice litigation are, at their root, indicative of labor-intensive processes: when so many hands touch each patient, error is inevitable.
Dissecting Health Care Processes Today
Visit any major inpatient hospital nursing unit, ICU, OR, or ER: the number of staff involved is impressive. Nurses, doctors, unit clerks, nurses' aides, orderlies, transporters, respiratory therapy technicians, and social workers are routinely present. A plethora of workers follow pre-assigned roles and routines, and coordination of care is haphazard. When the staff acts within pre-defined and self-contained roles such as diagnosing physician or nurse, they tend to be efficient. But care is, in reality, interactive. For instance, an X-ray may entail many participants: the physician gives the X-ray order, the nurse prepares the patient, transport moves the patient, the radiology tech generates the image, and the radiologist interprets the study.
What makes health care complex is that each patient is different in seemingly unpredictable ways; hence, the provision of care has to be individually customized. Patient variation induces process variance. Health care providers typically have only one tool to adjust for this process variance — their brains.
In other industries, great technical advances have been achieved by repeating the same process over and over. As Henry Ford's famous comment regarding Model T manufacturing standardization illustrated, "the customer can have any color he wants, so long as it's black."
Many health policy experts identify the need to reduce variance as the central challenge in health care. We know variation is pervasive. In fact, Dr. John Wennberg of Dartmouth has compiled a well-known atlas mapping variation in medical and surgical practices by geography throughout the United States.
Almost every large medical enterprise has tried to reduce variance. Throughout the country, many clinicians have volunteered (or been volunteered!) to serve on multi-disciplinary guideline, care plan, TQM, or QI, committees. Typically, the output of these team efforts has been a paper flowchart. The flowchart may show a process for pneumonia, CHF, atrial fibrillation, or some common disease. The new process is then "sold" to the nurses, doctors, and other staff with meetings and reminders. Often the initial enthusiasm and adoption wane over time as clinicians forget about the flowchart amidst their current concerns and patients. The individual doctor or nurse has to make an effort to remember to use the re-engineered process, which may be hard to do if weeks or months pass until the doctor or nurse sees the next patient who specifically qualifies for the flowcharted workflow. Without a way to automatically incorporate re-engineered processes, clinicians fall back on their standard workflows.
While patients are rarely identical, their care is often similar. Therefore, much of the work of health care providers is repetitive. The tasks of drawing blood, administering a medication, taking an EKG, or providing a special diet have broad similarities from patient to patient. The variance in patient care is often around the timing of each of these processes. The question is typically not whether to do the CT scan or perform the surgery, but when.
Clinicians have to manually coordinate each action in a patient's care. If one could eavesdrop on the many phone calls and pages in the hospital, one would often hear conversations that involve the effort of coordinating and collaborating. For example, "When will I get the meds? When will transport come? When will the MRI be ready?" Much of the clinician's paperwork in the modern hospital is about arranging and scheduling care rather than about the actual care itself. Nurses and doctors, especially, perceive the resulting forms and phone calls to be a necessary evil in diagnosing and treating the hospitalized patient. Interns and residents have even given this low-value-added work that fills most of their days a pejorative name, "SCUT."
We can see in this description of the processes of health care that automation in health care likely will have to address both the coordination of care as well as intrinsic patient variation.
The Tool
While it seems that these problems are unique to hospitals and clinical practices, it turns out that other industries have struggled with complex and unpredictable workflows. A bank about to make a loan, or an insurer about to issue a policy, must also deal with complexity surrounding differing risk profiles of potential borrowers or the insured. In response, these two industries, like hospitals, created "paper factories" to manage their processes.
After a similar search for ways to automate, many banks and insurance firms now actively manage and automate their underwriting with the help of workflow engines.
Workflow engines are software programs that view the world as a series of tasks or processes to be coordinated. Each process or sub-process can be represented as a series of actions that need to be routed, monitored, and supported. The actions are routed to agents (either human or machine) that have the resources to complete that action. Every step is monitored — workflow engines keep track of when something is to be done and when an alert needs to be issued or the work routed elsewhere. Workflow engines typically have graphical user interfaces to design the series of tasks and associated routing and monitoring flows. Because the workflow engine is often central to the enterprise, it may be linked to dozens of other software programs and databases that provide resources to perform the sequence of tasks.
The Role of the Workflow Engine in Health Care
The re-engineering teams that came together in many hospitals to provide care plans, guidelines, and TQM initiatives were not that far from the mark in process redesign. These multi-disciplinary teams realized that their processes could be well represented as a flowchart. All that was missing was automating that flowchart and connecting it to the actual care rendered.
Providing automation to care is what the workflow engine can do. Every step where someone interacts with a networked electronic device, whether that is a personal computer, PDA, CT scanner, lab system, telephone, automated IV pump, or wireless ID tracking tag, is an opportunity to link to the workflow engine. Physicians entering orders, nurses bar-coding medications, clerks registering patients, and surgeons scheduling surgery all have the opportunity to be linked and coordinated automatically as they perform their own specialty-specific activities. Teamwork can be automated with the workflow engine.
Workflow engines can sequence, monitor, track, alert, and reroute any step in each of these patient care processes. Therefore, the personal, moral, and legal responsibility for timely care no longer rests solely with the doctor or nurse.
The workflow engine has a graphical authoring environment. The standard workflows for drawing blood, administering medicine, or performing an imaging study are now source code that can be inspected by all and transmitted everywhere in the enterprise. Workflows can start automatically as soon as the patient is identified or registered. They do not need to be "jump-started" by human memory.
With a workflow engine, every step in a workflow can be designed and redesigned. Once the design cycle is done, the workflow steps can immediately be tested. Each step can automatically be logged and tracked. What works can be kept and distributed for use by all. What does not work can be immediately re-engineered. The timing of the many steps of diagnosis and treatment of a sick patient can be determined as precisely as human variation will allow. Steps that were serially sequenced by hand can be initiated in parallel by the workflow engine. For example, bed turnaround is often poorly managed in busy hospitals. Patients can spend 24 hours lying on a stretcher in the emergency department hallway, waiting for a hospital bed. Meanwhile, the physician order to discharge a current patient, freeing up the hospital bed, travels from the doctor to the unit clerk to the nurse. The nurse has to arrange for discharge medications, transport, and social work. Following those steps, the final call to housekeeping to clean the room is made. The whole discharge process can easily take 12 hours.
The sequence of steps in hospital bed turnaround and many other care processes is predictable. Parameters of acceptable variation in timing and sequencing can be specified. Many of the current human handoffs placed into someone's inbox or outbox can be automated. The participants in discharging a patient can be notified simultaneously once the decision to discharge has been made.
Implementation of the Workflow Engine
Workflow engines are used in many industries. One leading workflow engine vendor, Staffware®, has over a million stations live in more than 5,000 customer sites. What do we need to do to get this technology into health care?
First, hospital computing has to let go of the "database mentality." Hospital IT systems have typically been designed around the building block of a field in a database. Hospital computing has always been about putting something into the database or taking something out of the database. Workflow computing is really about using the task or activity to be performed as the building block for system design. The central goal of the computer is to support specific processes by routing tasks, actions, and resources. The database management system is a supporting actor, not the lead.
Second, this paradigm shift requires rethinking hospital information systems from the ground up. It is not enough to bolt a workflow engine onto a database. The network, the user interface, system interfaces, data stores, and associated design tools all need to embrace a workflow metaphor. Those who re-engineer need to understand that workflow here refers to the fundamental arrangement of tasks and sequences between all of the resources of the health system. This is absolutely different from the usual use of the term "workflow," to refer to an optimized sequence of computer screens for one individual user performing a specific task.
Summary
One hundred years after the automation of highly repetitive tasks such as steel manufacturing or automobile assembly, health care has the opportunity to automate the repetitive secretarial coordination of care that makes health care labor intensive, expensive, and dangerous.
Those of us in health care have had to wait to automate our processes until we had enough computer power to accommodate the tremendous variation between patients and their medical needs. That time has finally come. Over the next decade, we have the opportunity to harness the power of the workflow engine to provide care for the sick.
