Development of healthcare software and implementing enterprise healthcare IT systems and software has a major hurdle. Often the people designing the system have little experience in using the tool they’re creating. It is rare that a healthcare program used by a physician was designed and created by a physician. Good design should make things easy to see and do, but the tendency in clinical healthcare IT is to show every piece of information regardless of how meaningful it might be; bogging the physician down with noise. This is further compounded by the fact that clinical healthcare is incredibly complicated on its own. The good news is that much of the complexity of healthcare software can be mitigated with good design practices around the physician experience.
Problems in Current Healthcare IT Design
The major issues facing clinical healthcare software design are:
- It is difficult for a physician to find the meaningful information necessary to make decisions.
- Physicians are required to do extra clerical documentation necessary for non-clinical purposes.
- Common workflows are not minimized for work-time by the physician.
- Operational requirements within the organization do not support rapid clinical documentation.
- Personalization of the system by physicians is insufficient for their needs.
- Too many alerts hide important automated clinical decision-making tools
Because of these problems, The Annals of Family Medicine in Oct 2017 found that physicians spend 6 hours of their work day during and after office hours interacting with the EHR. A 2017 time motion study of the physician day by the Society of Teachers of Family Medicine Journal found that 43% of a physician’s visit accounted for EHR interaction.
These factors led to the Mayo Clinic’s 2016 finding that clerical tasks in the EHR was one of the leading contributors to physician burnout.
Fixing the Design
All changes to physician centered design are in three categories.
- Improving user interface (UI) that increases the efficiency of operations.
- Simplifying the interface to make the right choices easier, presenting all and only the information necessary to complete the task.
- Doing everything that the computer can do on behalf of the physician.
Improving the efficiency of your UI is more of an art than a science. Common strategies include leveled functionality, identified and optimized workflows, and layered details.
Leveling your functionality, or gating, is a process to selectively show more powerful and complicated tools to super users, while showing an easier functionality to general users. Individuals learning functionality might start with a bare bones selection of tools, and slowly graduate to higher levels of use. In this way, we avoid overwhelming physicians with choice while providing maximum complexity to the users who can appreciate it.
Optimizing workflows translated directly from paper systems, or the previous antiquated EHR, can be a complicated process. There is no substitute for shadowing a physician and watching, measuring, and documenting every step they take both in and out of the system. Use that shadowing time to determine what the most common workflows for your users are; and then pull out any step that unnecessarily slows the process down. A click to a different page that needs to load, another manual login, or even picking the patient off a list; every step removed adds significant value to a physician’s week. Each change is multiplied by the number of patients accessed by the physician. In high volume practices, this can be up to 80 patient charts per day!
Layering details, or selectively showing commonly used information and hiding or abstracting the less used components in an accessible way is another technique that is used to improve efficiency. This technique also assists in simplifying your UI to be more easily used. Mechanisms include; collapsible details, tooltips to allow the user to hover to see more details and using decision trees to determine which components to show. Note: it is important to show in some fashion that there are more details to be seen – otherwise it is easy for a user to miss this functionality.
Looking at the Mayo Clinic study, one large area of physician burnout is caused by physicians having to do documentation that was once the province of data entry specialists and front desk staff. These are the clerical tasks of managing an electronic medical record. Many of these tasks are not due to computer use, but rather the compliance office, state, and federal requirements such as Meaningful Use. While some of these guidelines can’t be avoided, such as marking that a physician has reviewed allergies each clinic visit, many of these actions can be abstracted or combined. For example, instead of one click each for allergies, medications, and history, a designer of software can combine these actions. Instead of manually signing their verbally given orders, these actions can be signed along with the open chart and notes. Clinician notes themselves can be further streamlined using templates and automated links to various pieces entered by interfaces to other systems, nurses, MAs, and front desk staff.
In the spirit of letting computers do the computer work, many elements of medical care are protocol driven. Some examples of protocol driven care are: determining appropriate tests that can be administered, refilling prescriptions, delivering chemotherapy to treat cancer, or many treatments in orthopedics. Using decision trees, and physician designed protocols, many of these treatments can easily be handled by the computer without day-to-day physician interaction. Where decisions need to be made – creating a set of nursing review protocols can create care that is delivered in a less costly way that removes the burden from physicians.
In situations where protocol can drive care, but patients need to be given information or a clinical design needs to be made; the patient chart can be used to alert a physician. For example, “Your patient hasn’t had an Hg A1C recently for their diabetes chronic disease management” while adding a button that orders the lab with a single click. Many of these alerts can even be handled by nursing staff by protocol to ease the burden on the physician. With a verbal message that is automatically signed with the chart, a physician is alerted with the minimal amount of interaction.
While letting computers do the computer work, it should be possible then to provide all patient information necessary to do a single workflow on one screen. Avoid bouncing between different tools that each has its own loading time. These One-Stop Shops allow physicians to focus on one task a time and helps a clinician to make informed choices without having to hunt down elements to the task.
The Role of Artificial Intelligence in Healthcare Software
In the spirit of having computers do the computer work, in today’s world we can also have the computers do some of the physician work as well. While this subject is deep and involved, we can graze the surface here.
There are many types of artificial intelligence that can play in the healthcare arena. AIs we are most familiar with are voice recognition in transcription (Dragon, M-Modal Dictation, etc.) and many forms of text searches through a patient chart.
Popular AI in the news today is in the use of neural networks to do supervised or unsupervised learning on healthcare data sets can speed the physician workflow. For example, by learning the most commonly used sets orders by a physician, we can recommend these orders and prepopulate many of the order fields on their behalf.
Neural networks can also be used to conduct raw research by finding relationships between unexpected elements of data. It could then be used to form the framework of the research paper, leading to a faster communication of findings and peer review.
Within the population health management model, neural networks can help determine who is managing their chronic diseases well and learn to leverage the information to suggest changes in care. This model would learn which treatments are more effective given a set of comorbidities, more narrow lab result values, or other collections of drugs. Then, these findings can be turned into another research paper to share what it has learned with other healthcare institutions.
Artificial intelligence and decision trees could be also combined to replace many individuals in health triage – helping patients achieve faster results at a lower cost to the health network while involving the physician only when necessary. Given the popularity of Siri, Google Assistant, and Alexa – these in-home AIs could bet trained to conduct the nurse triage without physician involvement. Given access to the patient charts through a portal, this system could also create appointments on behalf of the patient given the triage results.
The hardest part of building an AI for healthcare is getting access to the data necessary for training. This information is generally walled off due to privacy and HIPAA concerns.
Benefits of Physician Centered Design
Employing these design principles both in the production and setup of your electronic medical record, or clinical focused IT systems could save 10 minutes of the typical 45 minutes spent on a typical primary care appointment.
This time could be used in two ways:
- Make the physician’s day shorter – improving physician happiness.
- Improve patient throughput by up to 30%.
Focusing the clinician’s time on patient care and less on clerical activities allows the physician to do what they love, patient care, which also helps with reducing the rate of physician burnout.
Also, making the physician’s job easier; rapidly and easily providing them with information needed to make decisions, and assisting with the decision-making process will also help improve overall outcomes for the patient.
Employing these principles at your organization while creating or implementing healthcare software will speed the delivery of care, improve physician and patient satisfaction, and improve outcomes for patients.
For details and methodology on how to implement these principles or artificial intelligence at your organization, contact me through email or my LinkedIn page.