It's all about workflow

One of the words that I heard repeatedly at HIMSS10 this month was workflow. As I set out to write this post, I realized that I was not sure I really knew what workflow was. I visited the Wikipedia webpage concerning workflow. It had a lot of interesting information that helped me realize that I probably did not understand all the possible applications of workflow. I highly recommend this page to others. Interestingly, the shortest and most apropos definition described workflow, when used in the computing domain, as the interaction of human and machine. The tight association of workflow and quality topics was an unexpected enlightenment. I have tried to look at how workflow issues affect clinicians who use electronic health records in the paragraphs below. I will provide examples of how workflow is impacted.

Augmented productivity, improved safety or quality of medical care. 1) Order sets. Clinicians learn to write orders from day one in their clinical training. Ordering is a major process used to provide medical care. The use of order sets has many advantages over handwritten orders. First, teams of clinicians come together to design the order set and try to incorporate best practices from their own experiences and recent medical literature. For the clinician at the point of care, it is possible to "write" lengthy orders with just one or several clicks of the mouse. For recipients of orders, they are concise and legible, and standardized. For patients, order sets may assure that they are receiving the best care possible. 2) Electronic prescribing (eRx). A number of published studies have shown the efficacy of eRx in increasing the efficiency of care, minimizing medical errors caused by poor legibility of handwritten prescriptions, similarity of certain drug names, and incorrect dosing decisions. The ambulatory office workflows with eRx, especially for prescription renewals, can greatly increase clinician efficiency. As information exchanges grow, formulary checking and optimization present the potential of time and money savings for all. A thorny issue for the workflow of electronic prescribers has been the regulatory prohibition against the use of eRx for controlled substances. The FDA should be publishing an interim final rule (IFR) on this topic in the Federal Register today (preliminary document can be found here.) This long-awaited rule should help eliminate one of the major barriers to more wide-spread adoption of eRx. 3) Simple Decision Support. Alerts and reminders, when thoughtfully implemented, have been shown to improve the safety and quality of care, especially with respect to the medication ordering process. Already, innumerable errors have been prevented where drug-drug, drug-allergy, or drug-disease checking are performed automatically. These alerts fit perfectly into the clinical workflow because they appear while the drug ordering process is underway. I will discuss the more complex forms of clinical decision support below.

Examples where the workflow effects have primarily been detrimental in the clinical environment. 1) When the clinician moves quickly from one geographical point of service to another. Provider access to computer workstations and other hardware remains problematic. Clinicians continue to complain about the barriers created by security and privacy policies and procedures. Usernames and frequently changing passwords are a source of frustration. Use of single sign-on methods help but still are not a panacea. Furthermore, there is a trend toward true two factor authentication (see requirements in electronic prescribing of controlled substances rule above) in health care systems that may exacerbate the problem. Smart cards, proximity cards and biometric techniques may help mitigate clinician frustration. Somehow, I believe a technical solution to this conundrum will be developed in the future. 2) Infrequent tasks. It is human nature to forget how perform tasks that are done infrequently. It can be more difficult to create electronic orders for treatments that don't fit neatly into order sets. Good initial and recurrent training, superuser availability, and access to effective help assets-help desks, help menus, online-on-demand training, can all serve to limit clinician dissatisfaction.

Examples with variable, questionable or unknown effects. 1) Advanced Clinical Decision Support. This flavor of CDS usually addresses the more complex scenarios involved in acute and chronic disease management. CDS here is usually implemented through the use of structured order sets. The workflow challenge is to automatically deliver needed information to the clinician when and where needed (or at least make manual access to detailed information simple and imbedded in the same application.) Another difficulty is integration of CDS systems and EHRs. The correct data need to be collected and stored in the EHR to enable maximum CDS utility. Strategically placed info buttons allow clinicians to drill deeper for detailed information and to examine the original sources for evidence-based guidelines. Another challenge is the use of diagnostic decision support. When the diagnosis of a patient's condition is not known, then use of diagnostic CDS is an obvious course of action. Here the problem is making sure that enough of the correct data from history, physical, laboratory studies, radiology, and other studies is available to the CDS system so that it can provide a list of diagnoses that has a high probability of containing the one that is correct. Another problem for diagnostic CDS is that it cannot be helpful unless the clinician activates it. If a clinician decides to treat a patient based on an incorrect diagnosis, then the patient is not likely to get better. The difficulty for the clinician is knowing what you don't know. Here is a good example. A colleague of mine treated a patient with a large blister on the inner side of his foot with antibiotics and surgery. He thought it was an infection. Later, he asked me to consult because I had special training in foot and ankle surgery. I reviewed the patient's chart and I noted there was a history of regional enteritis and in fact the patient had a small fistula from the bowel to his abdominal wall. The photograph of the initial skin lesion looked like a bullous to me. There are relatively few conditions that manifest with bullous-like skin lesions. A quick check on the Internet yielded the correct diagnosis-pyoderma gangrenosum-a skin condition associated with inflammatory bowel disease that requires treatment with high dose steroids. This is a medication we would not use to treat infections. As a clinician there is always the fear-how do you know what you don't know? 2) Clinical Documentation. Many clinicians find documentation to be more difficult and slower with EHR technology than with traditional paper-based record keeping methods. Often, clinicians use keyboard entry of data but a large proportion finds this method to be slow, awkward, and intrusive. Template driven documentation can be efficient in some settings but is difficult to use for more nuanced requirements of treating a patient with multiple chronic diseases, complex histories of the present illness, and customized treatment plans. On the other hand, dictation can be incorporated into all or parts of the EHR record but the resulting narrative is usually unsatisfactory for meeting the ultimate goal of having structured, computable documentation needed for such diverse uses as CDS, Quality Measure computations, clinical research, and population and public health reporting. 3) Quality Reporting. The IFR for Meaningful Use focuses a lot of attention on quality reporting. Whether individual EHRs will best be able to provide all the quality measure outputs or whether use of an intermediary dedicated to extracting the data from the EHR and then calculating the measures will be the direction to go remains to be determined. Finding ways to perform the requirements automatically, in the background, with accurate results is what clinicians need. 4) Reconciliation. Clinicians must reconcile numerous types of information-problem lists, medication lists, allergies, and outside medical records to name just a few. Many of these tasks are amenable to automation to help reduce clinician workload. I haven't heard much about the reconciliation challenge at conferences but I expect it to move up in prominence during the next few years. Fitting reconciliation tasks into clinician workflows is not easy. 5) Health Information Exchange (HIE.) One of the thrusts of the IFR for Meaningful Use is to increase the private and secure exchange of health care information among the various stakeholders. Relatively few health information exchanges are up and running in the U.S. Clinicians will have to learn how to incorporate health information exchange into their workflows, in the least disruptive manner. Some of these exchanges may take place synchronously but most will probably be asynchronous. We also don't yet know what flavor of HIE will work best. Point-to-point seems most simple but requires knowledge of the recipient's contact information. Exchange of information within a network has satisfactory standards for guidance and a number of successful demonstration sites exist. The network of networks concept of early versions of the NHIN seems to be evolving. The effect on workflow is difficult to discern because the technology is so immature at this point.

Workflow effects of information technology in health care are important. I am sure we will be hearing more and more about this topic in the coming years.