An informatician walks into a surgery…

I wish there was a catchy punchline…

All kidding aside, my recent experience shadowing a surgery, from pre-op and surgical team prep to the surgery itself, was an eye-opening experience as an informatician. Understanding the healthcare delivery process and the lived stories of patients humanized the data within electronic health records.

As a data scientist, my day-job consists of calculating risk and predicting responses using patient data within electronic health records to mitigate adverse outcomes, such as post-heart transplant graft failure and doxorubicin-induced heart failure. In the past, I’ve done my best to prevent myself from knowing the data too intimately, to prevent myself from letting biases affect my analyses. At least, that’s what I thought would happen; weren’t my reasons analogous to the reasons a physician would not treat their own family members? Not exactly.

By witnessing the patient experience behind the data with which I work, I came to better appreciate a patient’s journey through the healthcare system, and its importance to my own work.

A patient’s journey

The morning sun was barely sitting on top of the buildings outside the windows of the pre-operation floor. Henrietta, a pseudonym for a real patient, was lying in her hospital bed being prepared for brain surgery. If you knew her before that day, you wouldn’t think there was any issue with her brain. She was comfortably within her retirement years and enjoying the last days of summer. A few months beforehand she noticed a small amount of fluid coming out of her ears, which she shrugged off as some harmless discharge that would go away with time. However, as time passed, she noticed more of this fluid draining, so much so that she kept a handkerchief close by to catch it. It was her daughter noticing fluid dripping from her mother’s ear that precipitated a much-needed visit to the doctor. Sure enough, the primary care doctor was concerned by the discharge, and took a sample to perform a lab test to determine exactly what was draining from her ears. The test revealed the culprit to be cerebrospinal fluid, the fluid that helps buoy and protect the nervous system. If left untreated, the leakage of this fluid could lead to intolerable headaches or meningitis1. She was quickly scheduled for a surgical consultation that recommended surgery to repair a tear allowing the leak.

So there Henrietta was, showing everyone her dry humor before she was rolled into surgery. Before starring in her surgical theatre performance, a team of nurses, physicians, technicians, surgeons, and anesthesiologists followed a standard procedure as they would for all surgeries that day and every day. The placement of the bed, monitors, and surgical tools are meticulously placed about the room and sanitized to support the needs of the particular surgery. The team even has a “huddle” where the patient’s name and surgery are announced to make sure they do the correct surgery on the right patient. For me, witnessing the surgery itself - seeing a half-opened skull with surgeons working away and patient vitals plastered on multiple monitors around the room - put patient-generated data into perspective. How is data stored before, during, and after surgery? Which healthcare worker is inputting, or which machine is recording, the data? What can the data capture, including what might have led to the CSF leak, how the surgery is going, and if there are any adverse events or outcomes?

Respecting the patient journey in informatics

Patient electronic health record (EHR) data is precious

Many factors affect a single person’s healthcare experience, but only a fraction of these will appear in the medical record.

While the operating procedures are standardized, Henrietta’s patient story was unique because her history and experiences as a patient were unique. Henrietta received a life-saving surgery, but other patients may be admitted for treatment of a chronic condition or after visiting an outpatient clinic for a routine medical checkup. The diverse reasons and distinct interactions with the healthcare system result in variable information within a patient’s electronic health record. Even outside the patient’s experience, the amount of information entered into the system may vary, such as unrecorded demographic information or the absence of a specific diagnosis code. Nevertheless, each patient has an experience to learn from that informaticians ought to capture. In other industries such as e-commerce and advertising, information is ubiquitous, disposable, and is used without the need for motivations or the hidden story. In healthcare, each patient’s journey has value. Each patient, and their information, must be treated with respect. This means the informatician ought to use methods and approaches that include as many patients as possible.

Missing EHR data ≠ bad EHR data

Electronic health information may not contain disease-specific diagnosis codes. Procedure codes stored in the EHR are required for insurance reimbursement rather than fully describing the procedure and its purpose. Also, there may have been missed family history or demographic information during the rush of the encounter and preparation. The absence of recorded family history, for instance, does not mean that the patient does not have a family history. This missing data may actually be useful information from which data scientists can learn. In health informatics, patients should not be omitted because of the absence of recorded information. Including patients with experiences that deviate from the norm, whether recorded or unrecorded, could assist in informing life-or-death situations. All data can be useful.

The limits of EHR data

At the end of the day, Henrietta’s surgery costs money. Recorded information in patient’s EHRs are, at a minimum, used to compensate the healthcare system. While there is tremendous promise to leverage patient data to improve healthcare delivery and medicine, electronic health record information primarily serves a billing purpose. Moreover, those with more wealth and resources may have greater access to services that potentially affect the outcomes and care utilization stored within the electronic health records of most patients. The potentially unequal experiences of patients, especially from marginalized populations, results in biased information in EHR data that, at face-value, does not accurately represent who is healthy, who is sick, nor the most effective or safe interventions for a given patient. Additionally, previous lessons learned from patient experiences may not extend to new or existing patients who may access different services or come from different backgrounds. It is the duty of data scientists to respect patients’ journeys by acknowledging the limits of the information available, identifying the opportunity presented by their EHR data, and evaluating the generalizable lessons to apply towards others.

Conclusion

Electronic health records represent the experiences of millions of patients, like Henrietta’s, through the healthcare system. It is the job of a data scientist to accurately capture and learn from EHR data to improve healthcare delivery and the practice of medicine. This necessitates basic principles of respect for the patient journey, the limits of electronic health information, and the inequities that EHR data may contain. It is our responsibility to learn only what can be learned for the good of patients and the best delivery of evidence-based medicine.

Edited by Vijendra Ramlall