Ten Principles for Good Design in Healthcare
By: Nayan Jain, Head of Engineering at Leo
We have all heard the statistics…
- Healthcare costs represent 17.5% of the GDP or about one in every five dollars spent in the United States.
- Slow growth in premiums = higher deductibles. Today, a trip to the doctor will cost more out-of-pocket.
- There are 12,011,710 practitioners, technical, and support workers in our nation’s healthcare system. Every single one of them have complaints about the software they use.
- Now EpiPens cost $600 before rebates.
And pets have better health records than people. #healthcareisbroken
In the 1970s, Dieter Rams challenged himself to think critically about the quality of his work by writing ten principles for good design. He was disillusioned by the sustainability and shortsightedness of products entering the market. With healthcare.gov behind us and the recent Theranos fallout, it is only fitting to reflect on information and software design in digital health.
How did we get here?
Before mass adoption of electronic systems, health care was transacted on charts stored in filing cabinets and vaults. Fortunately, in 2009 policymakers came up with the well-intentioned Health Information Technology for Economic and Clinical Health (HITECH) Act starting the nation down the path of moving from paper to PC by creating a system of incentives and penalties. Even though many of our modern medical institutions are not fully paperless, it is safe to say that the industry has turned a corner.
Another piece of policy that has influenced the tech landscape is the Health Insurance Portability and Accountability Act, more commonly known, as HIPAA. This act consists of two parts: restrictions imposed on health plans to protect coverage and to prevent fraud, limit medical liability, and introduce a standards-based approach to secure electronic transactions. The technical, physical safeguards, and documentation requirements have challenged even the best innovators. Move fast, but stay compliant.
Apart from the regulatory overhead, add in limited access to information and we have the perfect recipe for a system that is far from interoperable. The same data that was once collected in charts is now being written to onsite databases behind lock and key, or worse yet, being stored in large enterprise systems that have prevented access to information altogether. Silos.
A framework for tackling the hard problems
The only way to bridge the air gap between the data and the end-user in healthcare is to partner with each data provider (tedious) or specialize with a narrow dataset (restrictive). The best have tried to redirect efforts to wellness and wearables with the hopes of preempting chronic conditions and lifestyle diseases. The jury is still out on that approach.
At Leo, we believe that it is important to get involved even earlier in the process. Like Oscar, OneMedical, and others before us we are rewriting a part of the system by taking a full stack approach. We have the unique opportunity to design software that can help expecting families make the transition into parenthood by making the process of managing their family’s health easier. By doing this we will create a platform that is flexible to support the information access requirements that parents have and to assist front-line staff to provide the highest quality care and service.
Looking back at what we have built over the last two years I have summarized our learnings into the following ten key principles:
Good design in health care…
- is accurate — when clinical information is recorded it should be valid and pass stringent checks before it is stored and passed to the end-user. Care providers and support staff depend on this information being correct to make clinical decisions. As a safeguard, there should be a record of authorship for each value/section of a patient record in the event there are any questions with its validity. Patients should have the option to comment on, correct, or even question the information that is presented. Now that the patient is a part of the conversation, it is vital that the lines of communication are clear and the burden of data integrity is shared.
- is immediately available — not all decisions made in healthcare are time-sensitive, but the real-time data availability will address a wide array of use cases. If we can shorten the time between data collection and information sharing, we stand a better chance at avoiding costly medical errors and waste with repeat tests.
- is personalized — data-driven decision making is big business, there are industries built around recommendation systems and predictive models, why should healthcare be any different? This can come in the form of mining physician notes, expert guidance, or presenting supporting clinical cases that are tailored to the patient. An honest attempt at personalization will help patients better understand their unique situation. Not to mention, with the age of artificial intelligence upon us, we can expect that the growing number of datasets will assist clinicians triage previously untreatable conditions using techniques in precision medicine.
- is comprehensive — it is important that the patient has a complete view of their medical information. Outside of the standardized sections of the health record, much of the data is unstructured and cannot be fully utilized by a single company. Opening up this information to others we can build new experiences and interpretations on top of that data to better serve patients and their families.
- is secure and private — like financial, legal, and other sensitive data, health information should be kept secure. Individuals can be empowered to share (or not share) their information as they see fit simply by changing the ownership structure around data flow with the patient at the center. Data security should be developed in service of this use case. Help users securely share information and make it clear what they are making available.
- is clear for the end-user — sharing raw clinical information is not always the best option. Patients may need some translation of terms and values. The uninterpreted data in its original form should be made available, but any additional support we can add to make it consumer-friendly, the easier it is for patients to act on the data that they have.
- is historical — every piece of clinical information is vital, overwriting it with the latest and greatest is not an option. By keeping a full history with associated timestamps we can track care over time and provide clinicians (and machines) with the information they need to have a longitudinal view of a patient’s health.
- is uncluttered — file exports, long scrolling PDFs, and data dumps are not the right way to provide access to clinical information. Health data should be organized by time and relevance optimizing for readability. Software should be designed with the user and their needs in mind. Only present the relevant controls and fields for the task at hand and then get out of the way. Designing for every use case in a single screen will end up being overwhelming and headache-inducing. Unfortunately, most of enterprise software in healthcare is designed in this way. Our adaptive interfaces are only one solution to a systemic problem.
- is collaborative — open API access and structured data sharing will expand applications in population health. De-identified records and data donation will allow researchers and matching services for clinical trials to be more effective. We are starting by moving individual records securely between members of the patient’s care team to help get everyone on the same page.
- is portable — interoperability may be healthcare’s biggest challenge, but it is already a reality in other industries. Once the data has been made available to the engineering community (via APIs) it can flow between systems with little friction. The exposure of health information is distributed and limited to the types of information that each provider is allowed access to by the patient. This also means that stale data needs to be refreshed and kept up-to-date.
As with all things in software these guidelines are a work in progress. Feel free to share, comment, and contribute.
Join us — leohealth.com/careers
Sources:
