Issues with Electronic Medical Records

Introduction
A medical record is a descriptive medical ‘image’ of a patient. Each time, on seeing a patient, the clinician captures or causes the development of information about the patient and deposits this into the medical record. In the past, we recorded this information on paper. Can you remember those records? Amazing – and not in a good sense! Some were a foot thick, containing lab reports pasted onto heavy paper and not well organized.

Often, these records contained unreadable hand-written notes and others in the patient’s circle of care had difficulty accessing them. Worse still, the record was besmirched by inappropriate comments and was often quite incomplete. Note: it was voluminous but incomplete – we can achieve the same results with electronic support if we are not thoughtful! Interestingly, it was far more difficult to “hack” large volumes of records. But, getting a single one was relatively easy, as Justice Horace Krever’s Royal Commission addressing the confidentiality of health records found in the 1970s. Scoring thousands of records, though, would have required trucks and a major B+E by a team of thieves.

Today, computer-based information systems provide electronic support for recordkeeping. Unfortunately, despite great hopes and substantial investments, the use of technology has not delivered on the dreams of those of us who promoted it. Many years ago, electronic health (eHealth) experts promised that information systems would dramatically reduce the burden of capturing patient information and provide many subsidiary benefits, such as more complete, correct, shareable and processable records. Unfortunately, mundane issues have stymied our fulfillment of this promise. We would like to address a few of these here.

What Could Go into the Record
Maybe the easiest way to explain the issues with record content and completeness is to begin with the realization that every patient embodies from thousands to millions of pieces of associated medical information. There are certainly thousands of possible data fields about a patient running from the person’s identity, weight, height, birth date, address, gender, etc., up to quite esoteric medical measures revealed through laboratory testing and investigations (think cardiac studies for example). The larger number, ‘millions’, recognizes that we also need to account for medical images (X-ray, CT, ultrasound, PET, MRI…as well as pathology images for microscopic examination), genetic information and so on. This means that clinicians could spend a lifetime doing nothing but data entry on their patients. Some feel they do!

What Should Go into the Record?
Which of these many information elements should the clinician record at an encounter with the patient? Surely not everything!
Logic tells us that information relevant to the status and care of the patient is the highest priority. This would at least ensure that a low resolution ‘informational photograph’ of the encounter existed. There are issues with this, however. For example, perhaps there should be a ‘minimum data set’ in the record describing at least the patient’s condition and the physician’s interventions, but more than that judged by the clinician to be essential. This likely means that the initial encounter with the patient would require capturing essential patient identification, demographic and history information. If the record is shareable, this would reduce the burden for later care providers.

Let’s assume that all the basic personal and demographic information, as well as essential past medical information about the patient is already there. Which details need to be in our ‘photograph’ of the current visit to ensure the record is adequate?

Well, there are many views about that. In addition to the directly relevant patient state and physician interventions information, there may be medical-legal requirements or regulatory protocols that demand other information. Regarding legal stuff, in some jurisdictions, patients may sue physicians for what the patients perceive as mistakes or deficiencies in care. The insurance companies, which back-up the physician financially related to litigation, may require additional information, such as any indication that the patient was intoxicated, under the influence of drugs or suffering from a problem irrelevant to the current visit but that the agency feels should be noted by the physician. In addition, medical professional organizations, like medical colleges or boards, may set minimum record standards. Clinicians are under pressure to record these additional pieces of information despite their peripheral nature to the immediate care of the patient.

Who is Served and Who Does the Work?
This means that the colleges, boards, insurance companies and others place demands on physician time (the time associated with data capture and entry) and can quite dramatically increase the duration of the medical visit but without additional benefit to the patient. This is a serious problem. In a certain sense, these external agents transform the physician into a data entry clerk serving their interests. Furthermore, getting that data into the record becomes a major logistical and financial issue. Research indicates that using existing eHealth systems to record all this information can easily double the time the physician must spend in recordkeeping. Just imagine if a home renovations contractor had to document all the leasehold improvements performed for a client as well as their visual effect. If that were the case, we would expect for the contractor to charge for that time and get far less construction work done in a day. The problem is that physician fees are typically fixed and not time dependent. The result: the physician is turned into an unpaid data clerk. Perhaps even worse is that the system steals time that could be invested in patient care and interferes in the intimacy of the encounter.

There are other arguments for more detailed recordkeeping that may be more cogent for the physician, however. If medical records are sparse, then public health surveillance can be impaired. Also, many have very convincingly argued for assessment of the effectiveness of care. A recent study of Carotid Endarterectomy surgeries (Chart Documentation Quality and its Relationship to the Validity of Administrative Data Discharge Records, L. So, et al., Health Informatics Journal, pp. 101-113, 2010) found that many records were inadequate to support accurate coding, a process crucial for assessing care. Effectiveness studies are rendered impossible unless physicians accurately record details of the patient’s health status, testing, interventions and resultant patient state.

Therefore, many physicians find themselves in a bind.

Technology Comes to the “Rescue”
Over the last six decades, researchers and companies have worked to design, develop and implement computer-based information systems for medical recordkeeping. In the United States, the Obama administration funded a tens of billions of dollars effort to acquire and implement these systems. The idea was that eHealth systems would assist physicians in producing comprehensive, usable and shareable records that could be subject to analysis and used to predict the patient’s course. What a great idea! It may have produced a far greater ability to bill for healthcare services, but the effect on medical recordkeeping was not the one desired.

Today’s computer-based systems  in effect interrogate a physician for information and thereby increase the time it takes to produce records. This is because these systems implement record content standards. Physicians may perceive these standards to be in excess of what may be appropriate for the immediate care of the patient.

Then there is the problem of the ‘data entry bottleneck’: we have only a limited set of ways to get information into a computer system. For instance, we can type it in, select options from lists (e.g., the names of drugs) by pointing or using a mouse or we can dictate the information to the computer. However, every one of these data entry methods is problematic.

Consider the options. Typing, for most people, is slow and error prone. On the other hand, selecting items from lists by pointing or using a mouse means that the clinician must scan many, very long lists of drugs, for example, choose the desired item and sometimes slip from one item to the next, resulting in errors that are easily overlooked. Furthermore, the route of drug delivery (such as by mouth, injection), the dosage and the frequency all need to be selected. Each step affords the physician of opportunities for errors. Dictation may be a better option, but existing computer-supported dictation software’s abilities to recognize speech is still imperfect and speaking out loud in the location where the clinician works may be inappropriate for privacy or other reasons.

The information relevant to the encounter, the information demanded by external agencies or for other purposes, the corrections and even the mistakes all contribute to a lot of data entry. Worse still, the clinician only has a very restrictive ‘funnel’ for getting information into the record. This causes widescale frustration with and even rejection of quite a few of the existing record capture systems.

Solutions
When it comes to figuring out solutions, we must realize that there are a lot of smart people working on the problem. They have recognized that the capture of medical records requires more robust methods.

Clearly, one partial solution is the implementation and use of dictation systems that are highly accurate, being particularly capable when it comes to medical terminology, drug names and doses. This will not help when the clinician is in public circumstances, however.

Some have suggested the use of machine learning or artificial intelligence capabilities (Can AI Fix Medical Records? C. Willyard, Sci Amer, pp S13-S16, Feb 2020). Properly implemented, these technologies might be able to locate information and extract it directly from other sources, such as laboratory and consultation reports coming from elsewhere. These tools could conceivably dissect out the specific information of relevance to the patient and place it into the record of the encounter. There is also the possibility that the system could, in the ‘background’ (processing independent of the user), analyze the existing patient record in the light of new information acquired during the encounter. It could use this to suggest specific investigations and even to propose a differential diagnosis to the clinician. One system (Saykara virtual assistant “Kara”, www.saykara.com) already ‘listens’ to conversations and turns them into notes, diagnoses and orders that are put into the record. Robert Wachter at the University of California has pointed out that, regretfully, it and similar apps are “probably not quite ready for prime time” (ibidem, pg. S15). At the very least, tools like this could make the effort of data entry easier.

Work we have previously done suggests that we think through and specify the kinds of information needed at each patient encounter and create focused data collection efforts rather than either accepting arbitrary information or requesting masses of information. This would involve determining who will be using the data, what uses will be made of it and which information is most critical in achieving the effects desired. We have called this the “User, Uses, Effects Model” (Covvey, H.D., Berry, D., et al., Formal Structure for Specifying the Content and Quality of the Electronic Health Record”, Requirements Engineering 2003, September 2003).

Short of brain-computer interfaces, which are actually being explored in research labs, one of the above suggestions is likely all we have to work with at this time. However, we need better than what we have!

The Status Quo
Truthfully, the status quo is not acceptable. Electronic health record systems have largely merely replaced paper, the medium for recording but, unlike paper, they demand information. The good news is that the captured information is at least readable and can be shared and processed to a degree. However, physicians will generally agree the effort involved in getting the data into systems largely serves others, not themselves or their patients, providing little benefit to them in the care of their patients. Recognizing this, some organizations have hired ‘scribes’ to be present at the encounter and capture information. That is an expensive workaround to address our deficient creativity.

It is also the case that, sometimes, those unwillingly pressed into serving others come up with maladaptive shortcuts. One example is the ‘copy and paste’ method of creating a record of the current encounter. This involves accessing a previous record, copying it, pasting it into the record of the current encounter and making a few modifications and extensions. The problem with this is that it creates bulk but little added value. Further, inappropriate, out-of-date and incorrect information is just as easily copied as good stuff. It is of note that the creator of ‘cut, copy and paste’ recently shuffled off the mortal coil (https://gizmodo.com/larry-tessler-modeless-computing-advocate-has-passed-1841787408).

For the foreseeable future, physicians will either enter only a few pieces of information to get the problem off their backs or will spend inordinate and intractable amounts of time completing pages of electronic forms or responses to questions presented to them on a computer screen. Finding the balance will be a challenge.

Afterward
This is an area still ripe for research and development and may be one of interest to some readers. If so, consider applying your creativity to solving this problem. Doing that will free up people who are bogged down in a kind of trench warfare with their systems while trying to address the needs of their patients.

share this article...
Share on FacebookShare on Google+Tweet about this on TwitterShare on LinkedIn