By Robert James Campbell, EdD, CPHIMS, CPEHR
Imagine having a friend who can identify the name of a song after hearing one or two notes, or an acquaintance who can identify a film from a single "still" frame taken from the movie. If this sounds familiar, maybe it's because of the way it reminds you of how a physician thinks about their patients. In the same way that a song or film can be categorized by style—jazz, blues, pop, rap, or classical—and movies by genre—romance, science fiction, or film noir—physicians learn to use "clinical judgment" to diagnose or classify a patient's illness into categories—such as infectious, vascular, genetic, and neoplastic—and even further into sub-categories like pneumonia, hypertension, Down syndrome, and colon cancer.1 By exploring the process physicians undergo to develop knowledge structures that allow them to diagnose and treat their patients, one can determine how the electronic health record can be used to facilitate and improve a physician's cognitive thought process.
How Physician Thinking Develops
During medical school a student will acquire knowledge in the fields of biology, biochemistry, pathophysiology, clinical epidemiology, and pharmacology. However, this knowledge remains "inert" in the sense that students have no problem recalling the information on examinations and tests yet struggle when applying it to a live patient. A shift occurs when an individual graduates from medical school, becomes a resident, and later as a physician begins to see patients. It is through this process of seeing patients and attempting to diagnose and treat them that physicians develop "clinical experience" about how different diseases manifest themselves in individual patients.2 Through this differentiation process, physicians also begin to develop "illness scripts."
An illness script is a mental representation containing encapsulated pathophysiological knowledge about a disease in the form of a diagnostic label or description.3 More importantly an illness script contains information, termed enabling conditions, defining the contexts that make it possible for disease to appear in the human body. Authors H.G. Schmidt and R.M. Rikers illustrate in a Medical Education article how an illness script is formed over time as an individual is exposed to multiple patients having the same problem. When given a detailed description of a drug user being examined in an emergency room, a fourth-year medical student asked to diagnose the problem will describe how contaminated needles have led to a bacterial infection, and with the release of antibodies into the bloodstream, the patient experiences high temperatures, shaking, chills, shortness of breath, and the feeling of exhaustion. When the same description is given to an internist, their diagnosis is "This drug user has developed sepsis as a result of using contaminated needles."4 For the internist, all the relevant pathophysiological causal information has been encapsulated into the concept of sepsis. Furthermore, the knowledge that the patient is a drug user enables the internist to choose and substantiate from a list of differential diagnoses the illness script for sepsis. As mentioned above, an illness script contains information on the enabling conditions for a disease along with three other constructs: fault, consequences, and attributes.
Enabling conditions can be described as personal features—age, gender, addiction to controlled substances, family history, previous health conditions, previous diseases, current and past medications—and provoking factors that include maxillofacial, genitourinary, present and past sexual behavior, overexposure to direct sunlight and tanning beds, and the use of tainted needles and syringes.5 A fault can be "a description of the malfunction," in the case above: sepsis.6 Or it can be a causative agent, such as rotavirus, norovirus, campylobacter, treponema pallidum, and enterobacteriaceae. Consequences are signs and symptoms of how a disease reveals itself in the human body. Consequences can vary with disease and can be classified into categories such as:
- Onset: sudden or acute
- Descriptive: temperature, sweating, nausea, indigestion
- Irregularities in the Organ Systems: shortness of breath, stenosis, inspiratory rales, enlarged spleen, kidney failure, ocular hypertension
- Test Results and Vital Signs: blood pressure, pulse, erythrocyte sedimentation rate, white blood cell count, HbA1c, and granulocyte count
Diagnostic Errors Related to Physician Decision Making
Any discussion related to physician decision making must touch on the subject of diagnostic errors. Although current error rates are unknown, research estimates that the percentages for physician diagnostic errors fall somewhere between 10-15 percent.11 Of the two system processes discussed thus far, System 1 processes are more prone to generating diagnostic errors because they rely primarily on intuition and the use of mental heuristics like illness scripts. Also of grave concern is the fact that System 1 processes are subject to bias, emotion, prejudice, and passion, which increases the likelihood of a diagnostic error being committed.12
Illness scripts contain sets of attributes whose values can be used to determine the enabling conditions, fault, and consequences. The values for the attributes are used by the physician to determine what is normal in a healthy human being and what is abnormal or an indication that something is wrong with the patient. Attributes can also be used to search, select, and verify an illness script when diagnosing a patient. For example, the sepsis illness script used above will have attributes containing the values drug user, fever, sweating, and chills. In most instances, if these values are not present in the patient, the sepsis illness script will not be activated by the physician.
Illness scripts provide instructions for constructing a mental model for a disease, a family of diseases, or even representations of patients that have been seen in the past.7 Therefore, in one respect, clinical reasoning for a physician consists of examining a patient and if there is a close match between the presenting symptoms of the current patient and those of a previous patient, then the script used for the previous patient is selected and verified. This is known as pattern recognition and is a prime example of how physicians perform their job on a daily basis.
If, on the other hand, a particular patient does not come to mind, the physician will continue to search for clues in the presenting patient that will activate a set of illness scripts. These illness scripts represent the physician's differential diagnosis. Using the information gathered in their initial workup of the patient, the physician will eliminate those illness scripts that do not match up with the patient's symptoms, continuing this iterative process, until one illness script can be selected and verified. Another way of looking at illness scripts is as a clinical narrative for a disease that physicians construct from the knowledge they gained in medical school that is compiled over time and is based on clinical experience. This experience consists of encounters with patients whose signs and symptoms provide the cold hard facts of how disease manifests itself in a body. Therefore, clinical reasoning becomes a process of matching the narrative uncovered in the presenting patient with the correct illness script that has been compiled in memory.
The use of illness scripts begets one question: What if the physician is presented with a patient that has a set of anomalous conditions, or even more troubling, what if the physician does not have an illness script for the patient or they cannot narrow down a list of differential diagnoses to one specific script? One response to this question has been the development and explication of the universal model for diagnostic reasoning.
The Universal Model for Diagnostic Reasoning
This model holds that physicians use two types of processes to diagnose their patients: System 1 and System 2. System 1 processes have been discussed in detail and can be described as being intuitive, experiential, and pattern-oriented. An illness script is a prime example of a System 1 process. When making a System 1 diagnosis physicians will respond to the ambient and contextual conditions present in the patient encounter.8 These ambient conditions go beyond the presenting characteristics of the patient to include the healthcare environment (workload), supply issues (test results, procedures, hospital beds), and comprehensive issues (ethical, medicolegal).9 System 1 processes not only speed up the decision making process but can also be identified as a method physicians use to cope with the "flesh and blood" reality of being a medical professional such as stress, sleep deprivation, overbooking of patients, lack of resources—all of which have an influence on the decision making process.10 In many cases there is not enough time in the day to think analytically about each patient encounter. System 1 processes can be used to help overcome these obstacles.
System 2 processes can be described as being slow, analytical, resource-intensive, and mentally strenuous. System 2 processes will be activated when the patient presentation does not trigger an illness script, the symptoms are not recognized by the physician, or when an apparent diagnosis is highly uncertain. Using a linear process, the physician relies on their medical training, intelligence, and logical aptitude to analyze the signs and symptoms presented by the patient to formulate a diagnosis. For example, when engaging System 2 processes, the physician will try to explain a set of patient symptoms using their pathophysiological knowledge of disease. If they identify a set of possibilities they can then use Bayesian analysis and clinical epidemiology to determine the probabilities that a certain disease is at the heart of the patient's problem. This back and forth process between patient symptoms and clinical knowledge, checking and rechecking signs and symptoms against the biomedical literature, is a time- and resource-intensive process, but in the long run allows the physician to come to a proper diagnosis.
Over time, continued use of System 2 processes can lead to the development of illness scripts and other System 1 processes. Use of System 1 and System 2 processes is not an all or nothing proposition. There may be times when a physician uses a System 1 process to diagnose a patient, but realizes that there may be something more going on with this case, at which point they engage System 2 processes. On the other hand, a physician may be engaged in System 2 processing, and because of certain constraints on their resources, resort to System 1 processes.
Analyzing Dispositions to Respond
The discussion up to this point shows that in any given circumstance, and, more importantly, based on the context, a physician will have a tendency to respond to a patient using either a System 1 or System 2 process. Furthermore, because of their susceptibility to diagnostic errors, System 1 errors produce in the physician an inclination to use one of the following response patterns: a cognitive disposition to respond (CDR), or an affective disposition to respond (ADR). A CDR can be a type of mental shortcut that is made to speed up the decision making process. For example, two of the most common CDRs are known as anchoring and search satisficing.
To understand these two concepts imagine a patient who comes to the emergency room with an empty bottle of a controlled substance. The patient admits to taking all the pills in the bottle. The emergency room staff quickly treats the patient for a medication overdose. The patient does not respond to the treatment, and later when a third party shows up with an empty bottle of a second controlled substance, the emergency room staff learns the patient has overdosed on a second medication. In this example the staff "anchored" to the belief that the patient's sole problem was an overdose on one controlled substance and they "then called off the search" for any other drugs that the patient may have taken inappropriately, which lead to a drastic decline in the patient's health.13 A review of the literature shows that there are more than 42 different types of CDRs that can affect a physician's decision making process. Besides anchoring and search satisficing, other CDRs include aggregate bias, availability, gambler's fallacy, and ego bias.14
An ADR is more susceptible to emotion and prejudice and can include a bias on the part of the physician toward the patient based on gender, race, ethnicity, obesity, psychiatric illness, age, economic status, sexual orientation, substance abuse disorders, and chronic and complex illness. A typical scenario where an ADR may come into play is when an older African American female has an encounter with her white primary care physician. The physician dutifully listens to the patient, provides a diagnosis, and then writes out a prescription. Later the woman complains that the physician never looked up from his desk, nor did he place his hands on her to perform a physical examination. Research shows that disparities exist in the quality of care delivered to ethnic minority patients who are more susceptible to cardiovascular disease and cancer.15
Clinical Judgment Through a Clinical Narrative
In the book How Doctors Think: Clinical Judgment and the Practice of Medicine author Kathryn Montgomery says physicians practice what she terms "clinical judgment" in their day to day routines. For Montgomery, clinical judgment is "medicine as the science of the individual," where a physician uses abduction to use the effects (signs and symptoms) to interpret a cause, rather than deduction which traces cause in the form of a general rule to effect.
Using a rational and logical approach, a physician will listen to the narrative told by the patient to generate their own coherent clinical narrative, which is based on their clinical experience, medical knowledge, and understanding of clinical research, to diagnose the patient. The concept of a coherent clinical narrative is important because it will contain the thought processes used to diagnose the patient, the differential diagnoses, and how each one was eliminated until a final determination is made regarding the patient's problem. The clinical narrative will also contain research relevant to the case and how it applies or does not apply under the current circumstances. Finally, the narrative is important because it will document and illuminate how the physician interpreted the signs and symptoms presented to them by the patient, which can be used to diagnose future cases and possibly point out when they may have been under the influence of a cognitive or affective predisposition to respond inappropriately.
Capturing both the patient's narrative and the physician's clinical narrative comes under the purview of the electronic health record (EHR). More importantly, the EHR can be used as a tool to enhance a physician's cognitive processing of a case and to help eliminate diagnostic error.
EHRs Grab Physician Attention, but Need Tweaking
With the introduction of standards included in the "meaningful use" EHR Incentive Program, the EHR has taken a prominent place in the minds of new healthcare administrators and healthcare professionals. Questions still linger for both parties whether this innovation will benefit the administrative side of healthcare or the clinical side. To bolster the clinical side of the argument, one important role that HIM professionals can play—whether they are clinical application developers, clinical analysts, or a medical record supervisor—is making sure this technology enhances the way physicians practice medicine.
As discussed earlier, an important feature of clinical judgment is the coherence that exists between the physician's clinical narrative and the patient's narrative. On the patient side, current iterations of EHRs can capture patient demographic information, vital signs, current problems, and a medication list. This information will help the physician make an accurate diagnosis of the patient. Where many EHRs are lacking, however, is the capture of the story of how the patient is experiencing the illness, and more importantly, the thinking process the physician followed to generate a diagnoses. This is where many EHRs are deficient because they do not have the tools to quickly capture narrative information, code it where necessary, and then make it conveniently available within the record.
In the era of the paper health record, patient narratives and differential diagnoses were captured and recorded via transcription services. But the use of these services has diminished with the advent of the electronic record. Consequently the practice of cutting and pasting notes from one record to another record has become an unintended consequence of EHR adoption. This practice not only negates the idea of the importance of capturing the patient's unique story, but it does not fully capture the physician thought processes leading up to a diagnosis. One challenge for the HIM professional is to investigate and develop better methods for capturing narrative information.
In medical school, students learn that all patients are the same, and when they begin practicing medicine, they learn that all patients are different. This metaphor holds true for EHRs, because for this technology to be successful it must be tailored to the people who will use it on a daily basis. To enhance the physician's cognitive thought processes and eliminate diagnostic errors, better tools need to be developed and incorporated into EHR systems.
One common solution to diagnostic error is the use of clinical decision support tools such as DXplain, Dynamed, Isabel, VisualDx, and PEPID to help physicians generate differential diagnoses. However, research shows that if these tools are not integrated within the EHR, it is unlikely that physicians will adopt them.16
A new and untested methodology to reduce diagnostic errors on the horizon is the use of cognitive forcing strategies. These strategies can be defined as a critical element in the execution of a process to ensure that a correct procedure is followed, or to prevent an untoward event.17 One example of a cognitive forcing function is the checklist. Checklists may be developed and used in three different ways: general checklists for diagnosis, differential diagnosis checklists, and disease-specific checklists.
A general checklist contains items that remind the physician to perform the key tasks that mark a successful diagnosis. One example would be to obtain one's own complete medical history.19 This is a key consideration because it prevents the physician from committing the cognitive mistake known as framing error, where they simply accept the diagnosis of another physician and take it as fact. Other items on the general list could include performing a concentrated and purposeful physical exam, and most importantly, reflect and take a "diagnostic time out."19 Differential diagnosis checklists can be developed to make sure that a physician, who is dealing with a common complaint that provides a diagnostic challenge—for example, chest pain—considers all the possible diagnoses for that symptom.
Finally, disease-specific checklists can be used to provide a list of differential diagnoses, tests, and assessments that should be considered and common mistakes made when diagnosing a specific disease. It is important to note that as every patient is different, the way a physician practices medicine is different and they will need a set of tools that are tailored to their needs and thinking processes. It will be up to HIM professionals, whether they are working in a healthcare facility or as part of a cardiology practice, to develop tools like checklists and templates that promote higher levels of thinking among the physicians using them and, more importantly, prevent diagnostic error.
- Montgomery, K. How Doctors Think: Clinical Judgment and the Practice of Medicine. New York: Oxford University Press, 2006.
- Schmidt, H.G., G.R. Norman, H.P. Boshuizen. "A Cognitive Perspective on Medical Expertise: Theory and Implications." Academic Medicine 65, no. 10 (1990): 614.
- Schmidt, H.G., and R.M. Rikers. "How expertise develops in medicine: knowledge encapsulation and illness script formation." Medical Education 41 (2007): 1135.
- Schmidt, H.G., G.R. Norman, H.P. Boshuizen. "A Cognitive Perspective on Medical Expertise," 616.
- Schmidt, H.G., G.R. Norman, H.P. Boshuizen. "A Cognitive Perspective on Medical Expertise," 615.
- Croskerry, P. "A Universal Model of Diagnostic Reasoning." Academic Medicine 84, no. 8 (August 2009).
- Reason, J. Human Error. Cambridge, UK: Cambridge University Press, 1990.
- Berner, E.S., and M.L. Graber. "Overconfidence as a cause of diagnostic error in medicine." American Journal of Medicine 121 (2008): S2-S23.
- Ely, J.W., M.L. Graber, P. Croskerry. "Checklists to Reduce Diagnostic Errors." Academic Medicine 86, no. 3 (2011).
- Croskerry, P., and G.R. Nimmo. "Better clinical decision making and reducing diagnostic error." Journal of the Royal College of Physicians of Edinburgh 41, no. 2 (2011).
- Croskerry, P. et al. "Cognitive and affective dispositions to respond." In Patient Safety in Emergency Medicine, edited by K. Cosby, P. Croskerry, et al. Philadelphia, PA: Lippincott Williams & Wilkins, 2008, 219-227.
- Campbell, R.J., and D.A. Nolfi. "Teaching Elderly Adults to Use the Internet to Access Health Care Information: Before-After Study." Journal of Medical Internet Research 7, no. 2 (2005): 19.
- Miller, R.A. "Computer-assisted diagnostic decision support: history, challenges, and possible paths forward." Advances in Health Sciences Education 14, no. S1 (2009): 89-106.
- Ely, J.W., M.L. Graber, P. Croskerry. "Checklists to Reduce Diagnostic Errors."
Robert James Campbell (CAMPBELLR@ecu.edu) is assistant professor, health services and information management, at East Carolina University.
Campbell, Robert James.
"Step Inside the Physician's Head: EHRs Can Enhance a Physician’s Cognitive Processing and Eliminate Diagnostic Errors, With Some HIM-led Changes"
Journal of AHIMA