It’s a familiar situation to many of us as patients. We met with our doctor, potentially about a severe problem, and during our consultation, they seemed distracted by their computer. Of course, in the past, doctors would write prescriptions on paper and write down notes to remember what we tell them, what they see, and what they hear from us during exams. But something seems different in this era of Electronic Health Record (EHR) systems. What is going on with those computers? Why do doctors sometimes seem frustrated with them? Why do they have to pay so much attention to them? Was it supposed to end up like this?
As patients, we are not alone. Doctors can also be frustrated with what EHR systems have imposed on them. Even before COVID-19, physicians were struggling with burnout, and increased computerization of practice was cited as one of the major causes, often due to increased documentation demands. But why is that?
EHRs: Easy on management, but pushing doctors to burnout
One of the major reasons EHR systems are so challenging for physicians to use is that hospitals originally started to adopt them for accounting/financial and operation management purposes. Clinical treatment, or helping patients and clinical personnel, were all secondary considerations at best. A major feature of how they support hospital administration is by managing the processes they need to complete to be correctly reimbursed by insurance providers. This involves the enormously tricky problem of medical coding.
Medical coding involves alphanumerically encoding each diagnosis, treatment prescription, medical action taken, and equipment used. The codes used are standardized and universal. It represents a continuous record of each patient's medical journey and is an essential reference for insurance providers. Medical coding is how providers can determine what diseases a given insured person has, how necessary treatment is, how complex it will be, what factors will affect treatment outcomes and more.
Advancements in technology and knowledge, treatment techniques, disease classifications, and EHRs have also led to increasing complex medical coding systems. For example, there are now more than 68,000 diagnosis codes and 90,000 treatment codes that physicians need to consider when selecting the correct ones for a patient’s record. These codes are not something they need to know to treat patients, and memorizing the entire system would be a tremendous waste of time that they could spend on becoming a better doctor.
For this reason high-quality medical coding can only be achieved by combining the experience and wisdom of doctors with that of coding experts that understand the insurance system. This involves the physician recording the diagnosis and treatment of the patient in their medical record. At the same time, coding experts are responsible for assigning corresponding codes to the diagnoses and treatments, using anatomy, physiology, treatment details, and healthcare reimbursement guidelines to ensure all the codes are correct from the perspective of the insurance system.
So, rather than making life easier, EHRs have created new challenges for physicians. They need to work with medical coding experts and combine their expertise to complete the new task of creating sophisticated documentation that insurance companies can understand. This process can be exhausting. When a physician misses a diagnosis or inputs the wrong one, the coding team will “query” them to see if they can fix the patient record. Each query takes around 20 minutes to complete, in a context where physicians are seeing 10-20 patients a day, and the average number of diagnoses is over 50 per patient.
Hospital coding improvements: Aiming to treat the root cause, but end up treating only the symptoms
Let's look at the example of hospitalization due to a traffic accident. Within 72 hours after discharge, the doctor had to type in the diagnosis, treatment, and everything else from the patient's treatment process – only then could the all-important discharge record be generated. Based on their clinical experience and personal habits, the doctor listed the discharge diagnosis as "left lower leg crushing injury with necrosis."
However, after entering the final diagnosis, this doctor faced a predicament: The system popped a warning stating, "No corresponding standard disease code found"! The doctor did a little more searching, showing 418 options for necrosis and 231 options for crushing injuries. The system also provided two seemingly identical options: "Excision of left lower leg subcutaneous tissue" and "Extraction of left lower leg subcutaneous tissue."
While pondering all this, the doctor was also presented with a reminder from the hospital: "Maintain medical record quality! Maintain the balance between reimbursement and revenues!" And that's all just one case. As a medical professional, navigating the complex process of reviewing and coding medical records can be challenging while focusing on both clinical rationality and reimbursement.
With a large volume of notes to review, doctors can quickly become overwhelmed and potentially lose sight of these necessary considerations. Doctors need to find balance, and a way to effectively manage this process to provide the best care for their patients while ensuring they are properly compensated for their services.
Inevitably gaps between coding and the reality of the patient’s condition and treatments occur. Despite all this hard work, insurance billing problems, and documentation that can even affect healthcare quality downstream are still commonplace.
Studies show that over 21% of US medical bills have diagnosis coding errors! CMS, the federal agency that runs the Medicare, Medicaid, and Children's Health Insurance Programs, also issued an alert in 2020 that federal healthcare insurance payments were wrong by as much as $900 million in the case of Medicare Part D, and a whopping $86.5 billion with Medicaid. Challenges physicians have with medical coding are a huge part of this. In the end, it results in devastating waste, abuse, and fraud in healthcare that causes a lose-lose situation for patients, hospitals, physicians, and taxpayers.
Helping doctors focus on what matters
To help doctors focus more on clinical practice, we at AESOP have developed DxPrime: a knowledge base constructed from 4.5 billion pieces of medical data, using machine learning to analyze inpatient and outpatient data from Taiwan and the US. Refining the combined wisdom of doctors and classification experts, we developed a proprietary model that utilizes AI multi-analysis to help doctors summarize discharge diagnoses.
The system integrates directly into EHR systems and provides physicians with suggestions for discharge diagnosis packages by combining structural data from admission diagnoses, lists of multi-specialty questions, drug use, tests, exams, treatments, and surgeries. The doctor only needs to assess the diagnoses DxPrime presents to them to ensure clinical rationality, and choose one. DxPrime then automatically converts it into one of the built-in Diagnosis Related Groups (DRGs) that CMS and other insurance companies use to decide on the reimbursement amount.
It drastically improves medical coding quality and costs, enhancing the reimbursement process and outcomes. This gives physicians more time to spend on what really matters, their patients. Medical AI needs to play a facilitator role in helping make up for human and systemic limitations. It will help doctors get insights from big data to create value in medical decision-making to achieve real data-driven clinical decision support.