A typical medical record may begin:
The patient is a 55-year-old male who presented with retrosternal chest pain of acute onset with radiation to the left shoulder. Over the past few months he has noted increasing dyspnea on exertion. He reported a family history of ischemic heart disease. On examination, he was diaphoretic, with a rapid, thready pulse.
In reality, unless the patient was a cardiologist, it would be very unlikely for him to use terms like retrosternal chest pain, acute onset, or dyspnea. For that matter, whatever his appearance, a layperson would be unlikely to describe it as diaphoretic, or his pulse as thready. All these terms are, in fact, a standard form of medical terminology, the stuff of which medical dictionaries are made. While some terms such as dyspnea, pharyngitis, pneumonia, and cardiomyopathy are little more than Latin translations of English terms, literally breathing problem, inflammation of the nasal passage, water in the lung, and damage to the heart muscle, respectively, others have the appearance of normal English but have special status. A moon-shaped face, a butterfly rash, crushing chest pain, and chest rales are almost synonymous with Cushing’s syndrome, system lupus erythematosus, myocardial infarction, and pneumonia, and every medical student rapidly learns these associations.
It could be argued that this specialized terminology is simply designed (or evolved) to add precision to the meaning of the words, a precision that is essential for communication among health care professionals. If so, then the special language is no more dangerous than the byte, random-access memory, or jpeg of a computer specialist or the lintel, stud, and rafter of a carpenter. But the studies in this article show that these specialized terms, which we call medicalese because they are not consistently English or Latin, but rather are a language all their own, are more than just a precise communication device.
We believe this is a relatively new area of investigation. Others have had laypersons and clinicians rate whether individual conditions (hypertension or tuberculosis, for example) are a disease,1 and estimate the prevalence of various diseases.2 In a previous study from this laboratory, Eva et al.3 had residents make diagnoses in a series of ambivalent cases in which features of one differential diagnosis were stated in medicalese and the other in lay language. They showed that the differential diagnoses were biased toward the medicalese version, but it was not possible to determine whether this reflected a privileged status of the medicalese nomenclature, or simply that medicalese is more similar to the syntax used when the participants learned the diagnostic rules (i.e., transfer-appropriate processing4).
In the first study reported here, we show that both medical students and laypersons accord medicalese terms special status, one might even say prestige, that is not accorded a synonym expressed in ordinary English. In the second study, we show that, when symptoms are described in medicalese, they are more heavily weighted in diagnoses and that the cause of this weighting also appears to be attributed to prestige rather than to transfer-appropriate processing.
Study 1: Influence of Medical Terminology on Estimates of Seriousness, Prevalence, and Disease
The goal of this study was to examine the impact of medicalese versus lay terminology on layperson and medical student judgments of seriousness, prevalence, and “disease-like status.” Forty-seven undergraduate psychology students and 25 medical students completed a questionnaire. The surveys were conducted after a class. No payment or other incentive was used.
The questionnaire consisted of 22 disease names. Four were fillers and were not manipulated. The remaining diseases were presented in lay terminology (e.g., mad cow disease, high blood pressure, heart attack, chronic fatigue syndrome) or medicalese (bovine spongiform encephalitis, hypertension, myocardial infarction, myalgic encephalitis). Two forms were created; in each, medicalese and lay versions of disease names were interspersed and counterbalanced across form A and form B. The second page of each form was a dictionary of terms in which the equivalence of the two terms was explicitly described.
To elicit prevalence ratings, participants were asked “How many of your classmates do you think have this problem?” They responded by inserting a number, either of 100 (percent) or of 1,000, dependent on the instructions provided. The results revealed no difference between the two reporting procedures, so this variable was collapsed in all reported analyses. Seriousness was rated on a ten-point scale from 1 = not at all serious to 10 = very serious. Disease was rated on a four-point scale in which 1 = definitely not a disease; 2 = probably not a disease; 3 = probably a disease; and 4 = definitely a disease.
Analyses were conducted using repeated measures analyses of variance with one between-subject factor (psychology/medical student), one repeated covariate (lay terminology/medicalese), and 18 repeated measures (disease). Lay terminology/medicalese was treated as a covariate to deal with the counterbalanced forms.
The overall results are shown in Table 1. Overall, medical students regarded the individual disease descriptions as more serious, less prevalent, and more disease-like than did the psychology students. The difference between lay terminology and medicalese was highly significant for prevalence, with medicalese versions seen as less common. There was a highly significant interaction with disease for seriousness, prevalence, and disease-like status, implying that these differences did not occur for some diseases. Accordingly, Table 2 lists all the diseases, indicating the impact of lay terminology/medicalese on judgments.
With only a few exceptions, the medicalese version was viewed as more serious, less prevalent, and more disease-like, sometimes quite dramatically. For example, pharyngitis was 1 point more serious than cold, senile dementia was 1.5 points more serious than Alzheimer’s disease, myalgic encephalitis was 2 points more serious and 1 point more disease-like than was chronic fatigue syndrome. There were two exceptions: Lou Gehrig’s disease was viewed as more serious than ALS and flesh-eating disease as more serious than necrotizing fasciitis. Perhaps the lay descriptions of the latter two diseases may invoke highly vivid and memorable images.
Similarly, the lay versions were viewed as more prevalent; flu was twice as prevalent as influenza and senile dementia was seven times as common as Alzheimer’s disease (while the former may be a more inclusive category, actual differences are far smaller than indicated by these data). Prevalence figures for rare diseases were dramatically overestimated, particularly for psychology students. For example, they said that 6% of their classmates had a heart attack, 10% had ALS, 4% had Alzheimer’s disease, and 3% had mad cow disease (with fewer than 1,000 reported cases worldwide).
Although some of these differences are small, the experimental conditions, for which specific definitions of equivalence are provided, are such that any difference might be unexpected. Nevertheless, although the differences are intriguing, it is not clear that they would influence decision making by patients or clinicians. Further, this study investigated the use of medicalese versions of diagnoses; our opening scenario was related to terminological differences in signs and symptoms. If a similar privileged status is accorded to signs and symptoms recorded in a medicalese version, this might effect decision making. Our second study investigated this possibility.
Study 2: Influence of Terminology on Diagnostic Decisions
As we indicated earlier, a previous study found that, when presented with ambiguous cases, medical residents were more confident in a diagnosis when the indicative features were presented in medicalese compared to when the same features were presented in lay terminology. One explanation offered by Eva et al.3 is that this represents a transfer-appropriate processing phenomenon in that the influence of syntax might be driven by the terminology used when a student initially learns the features of the disease. In contrast, it is also possible that medicalese terminology simply maintains prestige that will lead it to be more influential regardless of the terminology used during learning. To examine this possibility, we conducted a study in which undergraduate psychology students were trained to diagnose four pseudopsychiatric conditions, using diagnostic criteria that were presented in either medicalese or lay terminology. Participants were then tested with a series of ambiguous cases in which the feature descriptions were systematically varied.
The study involved 27 first-year undergraduate students as part of a course credit. The study protocol was approved by the university ethics committee. Four pseudopsychiatric conditions were created: dysthymia, cyclothymia, borderline personality disorder, and histrionic personality disorder. While these are actual psychiatric conditions, we have labeled them “pseudo” to indicate that the features associated with each condition during training were only loosely based on the actual diagnostic criteria. Each condition had four supporting features that had both medicalese (e.g., dyssomnia) and lay (disturbed sleep) versions. Each participant was first taught the four categories by presentation of the disease name, the medicalese label for the four features, and a one- or two-sentence lay description of each symptom that presented the lay terminology of the symptoms.
Each participant then saw three prototypical training cases, defined as containing three of the four diagnostic features, sequentially for one of the four diagnostic categories. The three features were buried in a one-paragraph description of the patient. All three features were presented in either medicalese or lay terminology, depending on the condition to which the participant was randomly assigned. During this training phase, the participant was asked to identify the critical features to ensure that they understood the task and were learning the diagnostic framework. After working through all four diagnostic categories in this way, participants were shown an additional four prototypical cases, one from each diagnostic category. The same terminology was used the training and practice phases. In this phase, participants were asked to circle the diagnosis that they believed to be most applicable.
During the test, each participant saw a mixed series of 18 new cases, each of which had two features consistent with one diagnosis and two features consistent with a second diagnosis. Each pair of features was presented either in medicalese (e.g., dysthymia: dyssomnia, hyperphagia) or in lay terminology (e.g., cyclothymia: mood instability, unpleasant mood). The terminology used for each case was counterbalanced. Participants were asked to assign probabilities to all four diagnostic possibilities so that they equaled 100. If the transfer-appropriate processing account in the findings of Eva et al.3 is correct, then we would anticipate finding a Training x Test interaction such that the features presented will be more influential when the terminology used in testing matches that used in training. If, in contrast, medicalese simply maintains prestige regardless of training, we would expect no such interaction, but rather would anticipate a main effect of the language used in testing.
For participants trained with medicalese terminology, the probability assigned to the diagnoses at test was greater (M = 40.4) when the features were presented in medicalese compared to when the same features were presented in lay terminology (M = 35.6). Counter to the transfer-appropriate processing hypothesis, the same can be said for those participants that were trained using lay terminology (M = 43.9 and 38.6, respectively). A 2 (training) x 2 (test) analysis of variance revealed that no evidence that training descriptions interacted with test descriptions (F[1,1,039] = 0.025, p = .87). Similarly, training condition overall had no effect (F[1,1,039] = 3.22, p = .07). However, features stated in medicalese during testing were assigned a probability on average 5.1% higher than when stated in lay language (F[1,1,039] = 7.82, p < .005).
The results of these studies support the notion that standard medical terminology has an associated prestige value that has consequences for judgments about diseases. In the first study, diseases described in medicalese were viewed as more serious, less common, and more disease-like than when the same diseases were described in lay terms. In the second study, clinical features described in medicalese led to greater endorsement of the associated diagnosis regardless of the language primarily used during training.
The studies have limitations. Both studies were conducted on novices, either undergraduate students or beginning medical students. It is possible that this prestige factor may wane with increasing experience, although the studies by Eva et al.3 and Jemmott et al.2 showed similar effects with residents and physicians.
Students should be aware that the medical terminology that they put so much effort into acquiring represents more than simply a standardized representation for communication; such language acquires privileged status among both patients and health care professionals.
1.Campbell EJM, Scadding JG, Roberts RS. The concept of disease. BMJ. 1979;2:757–62.
2.Jemmott JB, Croyle RT, Ditto PH. Commonsense epidemiology: self-based judgments from laypersons and physicians. Hlth Psychol. 1988;7:55–73.
3.Eva KW, Brooks LR, Norman GR. Does ‘shortness of breath’ = ‘dyspnea’? The biasing effect of feature instantiation in medical diagnosis. Acad Med. 76;10(Suppl):S11–13.
© 2003 by the Association of American Medical Colleges
4.Morris CD, Bransford JD, Franks JJ. Levels of processing versus transfer appropriate processing. J Verb Learn Verb Behav. 1977;16:519–33.