With the passage of the HITECH Act in 2009, the Department of Health and Human Services successfully pushed the medical community to adopt electronic health records. Leading that effort was the Office of the National Coordinator for Health Information Technology (ONC).
After 10 years of use and billions of dollars of investment, electronic health records (EHRs) have not only failed to live up to their potential but have helped create a crisis in medicine.
To be sure, developing computer software to cover modern medical care is a daunting task. But what has been virtually ignored in the blame game is how designs mandated by ONC have virtually assured that electronic health records will be poorly designed and excessively complex.
Current electronic health records have a number of usability issues. A 2018 survey of physicians across the U.S. by Stanford Medicine and the Harris Poll found that 59% of physicians users felt that their EHR needed a “complete overhaul.” Users estimated that 62% of their time was spent entering data into the EHR, leaving only 38% of their time to be spent directly with patients. That is just one of many studies linking the use of electronic health records to physician burnout. Critics complain of complex rigid user interfaces, interminable cascading check boxes, lack of clinical content, and a focus on business and regulatory issues instead of patient care.
Through a series of expert committees, ONC developed certification and testing criteria that electronic health records had to meet to become certified in 2011. By that time, my company, Codonix, had been providing clinical EHR systems for 15 years, first to hospital emergency departments and later to physician-owned clinics. It was clear to me from the beginning of the new ONC testing protocols that almost no clinical oversight had gone into their development.
For example, the protocols required electronic health record companies to create test users who were then instructed by the testing proctor to place an order for two different medications and then go back and change the order without leaving any trace that the order had been modified — something that goes against best practices. An order should never be erased once they are entered. The clinical portion was rife with other errors: test protocols that included drugs that had been withdrawn from the U.S. market, incorrect laboratory values, and the like.
Another major cause of user frustration is ONC’s mandate that electronic health records force users to document to certain coding schemes. For example, providers are required to document a patient’s smoking history using the Systematized Nomenclature of Medicine — what everyone calls SNOMED — codes. There are nine different ways to “smoke” (or not), and these overlap in confusing ways.
A “never smoker” is an individual who has smoked fewer than 100 cigarettes in his or her lifetime. After making that determination, however, providers have to pick from eight different ways to smoke: every day smoker, some day smoker, heavy tobacco smoker, light tobacco smoker, and the like.
This kind of complexity and reliance on rigid coding systems permeated the ONC requirements. It affects how clinicians document a patient’s spoken language, allergies, ethnicity, diagnosis assigned, drugs administered, and so on. The list covers virtually every aspect of EHR documentation.
To add to the complexity, there was no uniformity in the design of the back-end data. For example, one set of diagnosis codes (ICD-10) was required for reimbursement, while another more complex set of diagnosis codes (SNOMED) was required to document patient care. The net effect is that a patient’s clinical issues are stored in two incompatible formats. This duplication and inconsistency affects all aspects of documentation, from medications dispensed to procedures performed.
These design criteria forced EHR developers to require clinical users to follow complex coding rules, virtually guaranteeing user dissatisfaction.
ONC also required electronic health record developers to support several complex and incompatible protocols used to communicate with different EHRs, with the patient, with state immunization records, and with the Department of Health and Human Services itself to report quality measures. A total of five different incompatible communication protocols are required. Because each of these require different data elements, EHR designers must force users to document encounters to meet the requirements of these various protocols. Submitting an influenza immunization message to a state registry, for example, requires users to enter the patient’s next of kin and mother’s maiden name. This level of complexity and inconsistency is not user friendly.
The mission of ONC has transformed over time from one of encouraging the adoption of electronic health records to one that purports to improve health care quality through the use of ONC-mandated systems.
Under the Merit-based Incentive Payment System (MIPS), users must report quality measures using a system that is so burdensome and confusing that a mini industry has sprouted up to coach and guide users on how to aggregate and submit data — if not game the system. These reporting requirements are mandated despite the overwhelming clinical consensus that this type of data submission is time consuming, probably counterproductive, and is becoming cost prohibitive.
Although studies have shown that electronic health records can lead to better reporting (by which I mean box-checking), there is almost no evidence that they lead to better outcomes for patients. In Stanford’s poll, 92% of respondents felt that their EHR had little to no value for clinical decision support. In a survey conducted by the Medical Group Management Association, 76% of respondents felt that MIPS reporting played no role in improving patients’ clinical outcomes.
The Department of Health and Human Services has a poor track record designing software — just think Healthcare.gov — and EHR software is several orders of magnitude more complex. As currently implemented and designed by committee, the existing ONC requirements are best described as spaghetti code.
Instead of simplifying and standardizing back-end data collection — which would be transparent to the doctor, nurse, or other user and would allow EHR designers to focus on clinical usability — ONC has instead opted to use electronic health records as a reporting tool instead of as clinical a clinical documentation tool, putting the design paradigm backwards.
Electronic health records should not be expected to support 28 different coding schemes and multiple communication protocols, each with their own limitations, which subtly or blatantly force users to follow certain rules. Instead, these potentially transformative systems should be designed with the provider and the patient at the center.
Rather than designing clinical software, ONC should focus on analyzing, simplifying, and standardizing the huge amount of information stored in the background. Rigid, outdated, and incompatible coding and communication protocols need to be simplified and rationalized.
Physicians have long been trained to document patient encounters in organized and structured ways. A medical chart should be designed so clinicians can quickly scan and abstract all of the necessary information about a patient in a short period of time. At the same time, EHRs can transparently generate a huge amount of discrete data ranging from drug codes to diagnosis codes to procedure codes. In short, normal clinical records generated by a typical electronic health record system are a treasure trove of organized and coded information; there is no need to force them into cumbersome reporting structures.
Google can abstract an enormous amount of information from unstructured text. Imagine the power of similar algorithms when used to analyze health data that is not only already organized but also has a huge amount of attached discrete data. In this regard, the ONC and electronic health record makers can learn a lot from Google.
Andrew Muchmore, M.D., is the founder, chairman, and chief technology officer of Codonix.