Editorial Lost for words: using verbal terms to express ...

3 downloads 0 Views 155KB Size Report
express probabilities in oral radiology ... What is wrong with words as a way to express ... August issue of this journal' highlights the problem. .... Am J Med. 1983; 74: 1061-5. 4. Kong A, Barnett GO, Mosteller F, Youtz C. How medical.
Editorial Lost for words: using verbal terms to express probabilities in oral radiology D.R.M. Timmermans and P.A. Mileman * Medical Decision Making Unit, Faculty of Medicine, University of Leiden, Leiden and • The Department of Oral Radiology, Academic Centre for Dentistry in Amsterdam, Free University, Amsterdam, The Netherlands

Dentomaxillofac. Radiol., 1993, Vol. 22, 171-172, November

What is wrong with words as a way to express uncertainty? What does a radiologist mean when he says of the radiographic appearance of an ossifying fibroma that 'the borders of this lesion are usually well defined'? The recent article by Stheeman et al. in the August issue of this journal' highlights the problem. They studied the meaning of verbal probability terms, such as 'not uncommon', used in a textbook of oral radiology and found that they had a variety of interpretations. The range of meaning, for example, of the expression 'not uncommon' was between a probability of 32% and 65%. This variation in interpretation confirms what was already known in rnedicine/r' and should be a cause of some concern. Two pertinent problems lie behind this variation: that of ambiguity of interpretation and of imprecision of the meaning of verbal terms. Ambiguity in the interpretation of verbal probability terms affects communication between clinicians and patients, students and teachers, specialists and referring clinicians, and parties in medicolegal proceedings. This ambiguity of verbal probability terms has lead researchers to suggest that, if these phrases are used, there should be a clear consensus about their meaning. In particular, phrases that are more ambiguous than others should be avoided. According to Stheeman et al. I these are not only phrases with double negatives but also phrases with meanings in the middle of the probability range (30-80%). Unfortunately, the latter are also the phrases which are used most frequently'. In addition to being ambiguous, verbal probability terms are also imprecise. Verbal terms have a range of meaning and do not represent a specific probability. The objective nature of radiology as a diagnostic science assumes that the relationships between signs and diseases are established by clinical epidemiological research. How can knowledge of the strength of relationships between radiological features and the underlying disease be updated by new research if probabilities cannot be recorded precisely? In the current situation, writers of radiographic reports and

© 1993 Butterworth-Heinemann Ltd 0250-832X/93/040171-02

textbooks may easily become more involved in trying to improve their style by using a variety of probability phrases, rather than with the transmission of their precise meaning.

An alternative approach to the use of verbal probability terms To improve the clarity of probability information the use of verbal phrases for expressing uncertainty should be avoided. Clinicians should use percentages (e.g. 10%) and a range representing the uncertainty of the probability (e.g. 5-15%). However, there are some drawbacks to this solution and the proposal could meet with some resistance. Another solution to increase the precision and decrease the ambiguity of verbal terms has been suggested. This is to use a limited number of categories of phrases (five to seven) with specific numerical probability boundaries 1.4.5. This proposal is based upon the idea that the human mind has difficulty in coping with more than seven categories at a time when ordering phenomena'':", Kong et .at." and Hamm'' have also recommended the use of an ordered hierarchical list of phrases with specific boundaries from which people could choose a term to express a probability. This lexicon of verbal terms should cover the full range of probabilities.

Limitations of using an ordered list of phrases The meaning of verbal probability terms may change with the context. Mapes" has demonstrated that the term 'rare' is interpreted as less likely when used to describe the chance of the side-effects of beta-blockers than when used to describe the side-effects of antihistamines. He suggested that this discrepancy arises because the side-effects of an antihistamine are mild when compared with those of a beta-blocker. The seriousness of the consequences influences the interpretation of the verbal term used to express the probability of occurrence.

for the IADMFR

Dentomaxillofac. Radiol., 1993, Vol. 22, November

171

Editorial The interpretation of probability terms is also influenced by the perceived base rate of events or the prior probability'). The probability expressed by, for example, 'it is likely that you will get influenza in the coming year' is interpreted as higher than 'it is likely that you will get pneumonia in the coming year'. The prevalence of influenza is higher than that of pneumonia, which affects the meaning of the verbal term. In addition, verbal terms are often used to describe the relative frequency of events rather than their base rates. This can lead to ambiguity in their meaning, for example: osteosarcoma is usually described as the most 'common' primary malignant neoplasm of bone. However in epidemiological terms these are rare tumours 10. Simply using an ordered list of phrases will not take these influences on the meaning of a specific phrase into account. It could be argued that the context-dependent meaning of verbal probability terms implies that the significance of differences between clinicians within a speciality has been overestimated. This might be even more true for those with extensive experience. However, Timmermans II has shown that experience in a specific domain (i.e. specializing in acute appendicitis) does not lead to higher agreement about the meaning of verbal terms. In her study, experienced surgeons were not found to have a higher level of agreement than specialists in internal medicine (who are less familiar with acute appendicitis), or junior surgical staff (who have less experience). It would therefore seem incorrect to assume that, when verbal terms are used, the concept of probability in the written reports of radiologists is clearly communicated even to experienced clinicians.

Subjective barriers to adopting numerical probabilities Using numerical probabilities, the obvious solution, may however meet with resistance. Verbal probability terms may be used for a specific reason. Percentages suggest a preciseness that the user may not want to communicate. The vagueness in the communication is intended for a variety of reasons. Expressing uncertainty verbally rather than numericaJly may allow clinicians leeway to incorporate an additional meaning. For a clinician, the word 'possible' may imply that he or she has a logical reason for suspecting a disease, such as the presence of a specific symptom. Clinicians may want to avoid having the quality of their diagnostic prediction judged ('I only said it was a rare occurrence'I'F. Resistance to the use of numerical probabilities may also arise where there is a lack of recognized epidemiological data. IdeaJly, probabilities should be derived from clinical epidemiological studies; however, in the interim subjective probabilities based on clinicians' expert knowledge can be used. Clinicians should learn to express in numerical terms their expert opinion about the proba-

172

Dentomaxillofac. Radiol., 1993, Vol. 22, November

bility of the presence of a disease given the presence of a radiographic feature'. This could increase the clarity of communication.

Conclusions Promoting and aiding the use of numerical probabilities in a sensitive and informed way may, despite the difficulties, turn out to be the most efficient way to increase the accuracy of the process of communication. The apparent elegance of a compact verbal expression with ample opportunity for misconceptions must be traded off against a more precise, but perhaps cumbersome numerical expression. Textbook editors and writers of radiological reports should adopt a numerical approach to expressing probabilities in order to increase precision and reduce ambiguity. However, they may have to introduce it in phases. The use of a limited number of categories of phrases with specific numerical borders is one approach I. The adoption of such a system would be explicit and as such would take into account how much people differ in their interpretation of verbal terms without avoiding their use entirely.

References 1. Stheeman SE, Mileman PA, Hoff MA van 't, Stelt PF van der. Blind chance? An investigation into the perceived probabilities of phrases used in oral radiology for expressing chance. Dentomaxillofac Radio/I993; 22: 135-9. 2. Bryant GO, Norman GR. Expressions of probability: words and numbers. N Eng/ J Med 1980; 302: 411. 3. Nakao MA, Axelrod S. Numbers are better than words: verbal specifications of frequency have no place in medicine. Am J Med 1983; 74: 1061-5. 4. Kong A, Barnett GO, Mosteller F, Youtz C. How medical professionals evaluate expressions of probability. N Eng/ J Med 1986; 305: 740-4. 5. Hamm RM. Selection of verbal probabilities: a solution for some problems of verbal probability expressions. Organ Behav Hum Decis Processes 1991; 48: 193-223. 6. Miller GA. The magical number seven, plus or minus two. Psych Rev 1956; 63: 81-97. 7. KUipers B, Moskowitz AJ. Kassirer JP. Critical decisions under uncertainty: representation and structure. Cognitive Sci 1988; 12: 177-210. 8. Mapes REA. Verbal and numerical estimates of probability in therapeutic contexts. Soc Sci Med 1979; ]3: 277-82. 9. Wallsten TS, Fillenbaum S, Cox JA. Base rate effects on the int~tation of probability and frequency expressions. J Memory Language 1986; 25: 571-87. 10. Stephens RG, Kogan SL, Speechley MR, Dunn WJ. A critical view of the rationale for routine, initial and periodic radiographic surveys. J Canad Dent Assoc 1992; 58: 825-37. It. Timmermans D. Numerical versus verbal probability statements in medical decision making. Theoret Surg 1992; 7: 162. 12. Beyth-Marom R. How probable is probable? A numerical translation of verbal probability expressions. J Forecasting 1982; I: 257-69. 13. Fryback DG. Decision Maker, quantify thyself. Med Decis Making 1985; 5: 51-60.