Translating evidence-based information into effective risk communication: Current challenges and opportunities

REVIEW ARTICLE Translating evidence-based information into effective risk communication: Current challenges and opportunities AMIT KUMAR GHOSH and KARTHIK GHOSH ROCHESTER, MINNESOTA

Recent medical advances and the easy availability of evidence-based information at the point of care are believed to provide physicians with improved tools for risk communication. However, evidence indicates that physicians still display marked variability in ordering tests. Factors that determine a physician’s test-ordering tendencies vary by specialization, practice, geographical location, defensive practice, and tolerance of uncertainty and are also modified by patient requests. Understanding of statistical terms on the part of both physicians and patients remains limited. Physicians may display limited ability to assess pretest and posttest probabilities, especially in low- and intermediate-risk patients, even after attending short courses in epidemiology, or may find the process impractical. Presentation of diagnostic-test results in a natural-frequency format might improve understanding. Both physicians and patients have difficulty grasping the term “number needed to treat” compared with “relative risk reduction” when comparing therapeutic options. Other patient-related factors that limit understanding include low literacy, individual risk tolerance, and framing patterns of the problem (potential gains vs losses). Despite numerous available modalities (quantitative and qualitative) of risk communication, consensus over the advantage of any single modality in translating evidence into risk communication is limited. It is essential that physicians remain patient-centered, generate trust, and build a partnership with the patient to achieve consensus for medical decision-making. Future studies are indicated to assess the effectiveness of novel risk-communication modalities based on patients’ and physicians’ characteristics and identify appropriate modality of translating evidence (quantitative or qualitative information). (J Lab Clin Med 2005;145:171– 80) Abbreviations: AR ⫽ absolute risk; ARR ⫽ absolute risk reduction; EBM ⫽ evidence-based medicine; GP ⫽ general practitioner; LR ⫽ likelihood ratio; NNT ⫽ number needed to treat; PPV ⫽ positive predictive value; ROC ⫽ receiver-operator characteristic; RR ⫽ relative risk; RRR ⫽ relative risk reduction

From the Division of General Internal Medicine, Mayo Clinic College of Medicine. Submitted for publication October 19, 2004; accepted for publication February 14, 2005. Portions of this work were presented at the Midwest Society of General Internal Medicine (SGIM) Regional Meeting, Chicago, Ill, in September 2002; at the Second International Conference of Evidence-Based Health-Care Teachers and Developers, Palermo, Italy, September 2003; and at the Annual Society of General Internal Medicine Meeting, Chicago, May 2004.

I

n an effort to provide the most appropriate health information to patients, physicians are faced with a dilemma. Given the time constraints, physicians must focus on critical information while avoiding

Reprint requests: Amit Kumar Ghosh, MD, FACP, West 17, GIM, Mayo Clinic, 200 First Street SW, Rochester, MN 55905; e-mail: [email protected]. 0022-2143/$ – see front matter © 2005 Mosby, Inc. All rights reserved. doi:10.1016/j.lab.2005.02.006

171

172

Ghosh and Ghosh

irrelevant information during the physician-patient encounter.1 Physicians often find themselves in the role of risk communicators and must determine and describe the content of their message in a manner that will be understood by their patients. The final message should incorporate the patient’s values and opinions and should be tailored by the best existing evidence. The 2001 Institute of Medicine report emphasizes patient-centeredness as an essential component of patient-physician relationship.2 There is an increasing emphasis on providing patients with quantitative information about the risks and benefits of diagnostic tests and treatment that they can use to make the best decisions for themselves. Not transparent in this assumption are the factors that can affect the patient’s and physician’s understanding of the tests and their role in the decision-making process. The current emphasis on a curriculum in evidencebased medicine is based on the assumption that increased ability to appraise current evidence and awareness of best evidence transforms medical students and practitioners into expert risk communicators.3 Commonly stressed statistical terms include such diagnostic terms as “pretest probability,” “sensitivity,” “specificity,” and “likelihood ratio,” as well as terms used to express therapeutic efficacy, including “absolute risk reduction,” “relative risk reduction,” and “number needed to treat.” This last term has been stressed as the most important indicator of clinical significance of a therapeutic intervention.4 The authors of a recent article expressed great alarm at the infrequency of reporting of NNT in articles published in prestigious journals (only 8 of 356 eligible pieces).5 The translation of current evidence into clinical practice presents challenges, both in cases in which the diagnosis is known with certainty and in those in which it remains uncertain.6,7 It has been presumed that physicians with adequate training develop the ability to assimilate current information and thereby decrease practice variability. The final assumption is that patients are able to understand quantitative information reported in the medical literature and assume equal partnership in their health care. In this review, we will present an account of current evidence regarding physician variability in test-ordering characteristics; physicians’ understanding of probability statistics and terms describing therapeutic efficacy, including NNT, ARR, and RRR; and factors that may impair a patient’s understanding of evidence-based information. We will also suggest strategies for effective risk communication.

J Lab Clin Med April 2005

Fig 1. A, Linear process of medical decision-making; B, actual steps of the decision-making process.

DETERMINANTS OF PHYSICIANS’ TEST-ORDERING CHARACTERISTICS

At first glance, ordering a test appears a rather straightforward exercise in clinical decision-making (Fig 1, A). One must take a thorough history and perform a meticulous examination of the patient to determine the pretest probability of the differential diagnosis, choose a test with favorable characteristics (high sensitivity and specificity), and thereby refine the posttest probability of the likely cause of patient’s problem while relegating the other causes to insignificance. In reality, the process of choosing a test is more complicated and frequently non-Bayesian. After other variables are controlled, factors that affect physicians’ test-ordering characteristics vary with their geographic location (American neurologists order more tests than do neurologists in the United Kingdom),8 specialization (specialists order more tests than do primary-care physicians),9 practice settings (solo practitioners order more tests than do physicians in group practices),10 financial incentives (physicians who own imaging facilities order more radiographs than do those who do not own such facilities),11 defensive practice (malpractice fears),12 and perception of official guidelines.13 Recent studies have also revealed that physicians’ personal intolerance of uncertainty may account for 17% of excessive cost in medical management.14 In fact, in 1 study, most physicians did not use a Bayesian approach in their practices because they found it impractical or too unfamiliar.15 A physician’s test-ordering behavior is often guided by patient-related factors. It has been shown that physicians are inclined to order more tests if a patient demands them,16 if a patient has private insurance,17 if a patient is of higher socioeconomic status,18 if a pa-

J Lab Clin Med Volume 145, Number 4

tient has a family history of the disease in question,19 or if the patient does not speak English.20 PHYSICIANS’ UNDERSTANDING OF PROBABILITY STATISTICS

Effective risk communication with regard to diagnostic tests warrants thorough understanding of the characteristics of diagnostic tests. It is assumed that physicians with adequate training possess the ability to rapidly understand the nuances behind positive and negative test results. However, physicians vary tremendously in their ability to ascertain numeric estimates of probability to the same verbal description.21,22 We sought to carry out a systematic review of physicians’ ability to correctly describe and understand the terms used in diagnostic tests. Nine articles (6 involving case scenarios, 2 involving questionnaires, and 1 involving a telephone survey) met the inclusion criteria (Table I).23–30 The most common physician error was overestimation of a test’s PPV.28 Medical students could not rule out diseases in low- and intermediateprobability case scenarios by applying Bayesian estimates. In 1 study from Australia, 13 of 50 physicians (26%) stated that they could describe PPV, although on direct interviewing only 1 physician could actually illustrate it with an example.27 In another, the percentages of physicians using Bayesian calculations, ROCs, and LRs were 3%, 1%, and 0.66%, respectively.15 However, presentation of the data in natural-frequency format increases the accuracy with which physicians determine PPV to 46%.25 Inattention to pretest probability and inability to accurately assess PPV may therefore result in increased anxiety in patients because unnecessary tests and consultations are generated. Recent studies have shown that medical students remain unable to correctly assess pretest probabilities in lowand intermediate-risk scenarios even after receiving adequate EBM instruction.30 Alternative forms of expression of diagnostic tests in the natural-frequency format might improve understanding of diagnostic tests. Hoffrage et al31 used the same clinical problem—a cancer-screening test—and gave 24 physicians information in a probability format (percentages for population base rate, test sensitivity, and false-positive rates provided) and in a naturalfrequency format: “Thirty of every 10,000 people have colorectal cancer. Of these, 15 will have positive hemeoccult test results. Of the remaining 9970 patients without colorectal cancer, 300 will still test positive. How many of those who test positive actually have colorectal cancer?” Only 1 of the 24 physicians accurately solved the problem when it was posed in terms of probability, whereas 16 of the 24 answered correctly when the problem was presented in terms of natural frequency.

Ghosh and Ghosh

173

Physician innumeracy, therefore, remains an impediment to popularizing EBM. Noguchi and colleagues30 have stressed that medical students should develop a “well-balanced clinical sense” instead of focusing solely on “ruling in” a diagnosis. PHYSICIANS’ UNDERSTANDING OF NNT

NNT has been described as an essential paradigm in the understanding of the clinical significance of a new therapy. It often presumed that physicians understand NNT, but few studies have been conducted to assess the ability of physicians to understand and apply it.32 Recent studies from the United States, Europe, and Australia have consistently shown that both medical students and physicians prefer results presented as RRR to those presented as NNT.27,29,33–39 In a survey of 50 Australian physicians, only 8 (16%) could understand and explain NNT to others.27 In 1 study, 77% of physicians deciding against treatment considered the term “NNT-1” (1 patient was helped by the treatment) “wasted effort” of a treatment.39 Despite design limitations and the absence of large numbers of subjects in most of the studies, these results are generalizable because they have been documented in different practice situations in various countries with near identical findings, i.e. poor understanding of NNT (Table II). A less frequently used expression is number of lives per thousand patients. In the APPROVe (Adenomatous Polyp Prevention on Vioxx) study, designed as a means of evaluating the efficacy of rofecoxib 25 mg in preventing the recurrence of colorectal polyps in 2,600 patients with a history of colorectal adenoma, increased cardiovascular risk appeared after 18 months of treatment with rofecoxib and persisted.40 The cumulative incidence of cardiovascular events after 3 years was 7.5 per 1000 patients receiving placebo, compared with 15 per 1000 receiving rofecoxib. This translates to a relative risk of 2 and a number needed to harm of 133 for cardiovascular event at 3 years. Extrapolating these results to the population taking rofecoxib; for every 1 million people taking the medication there would be 75,000 additional deaths per million. Hence, despite the theoretical ease of explaining therapeutic efficacy as NNT, one must account for the current limited understanding of NNT and include RRR and ARR, lives per 1000 saved or harmed, to maximize understanding of treatment efficacy. FACTORS IMPAIRING PATIENTS’ UNDERSTANDING OF EVIDENCE-BASED INFORMATION

Physicians often presume that their patients grasp completely the subtle nuances of the risks and benefits of a test or treatment. Numerous patient-related factors—namely, difficulty understanding percentages, in-

174

J Lab Clin Med April 2005

Ghosh and Ghosh

Table I. Studies of medical students’ and physicians’ understanding of diagnostic terms Type of study

Author (year)

Study design

Generic cases

Lyman GH (1993)

31 physicians and 19 health-care workers were presented with 2 hypothetical cases (a 30- and a 70-year-old woman with breast lumps) and asked to estimate pretest and posttest sensitivity and specificity 31 physicians and 19 nonphysicians were presented with cases in which the sensitivity, specificity, and pretest probability varied and asked to calculate the PPV

Generic case scenarios

Lyman GH (1994)

Telephone survey

Reid MC (1998)

Clinical cases

Hoffrage U, Gigerenzer G (1998)

Controlled questionnaire

Steurer J (2002)

Questionnaire and survey

Young JM (2000)

Case scenarios

Noguchi Y (2002)

Questionnaire survey with case scenarios

Heller RF et al (2004)

535 physicians asked to calculate pretest probability from case scenarios

Survey involving case scenarios

Noguchi Y et al (2004)

90 Japanese medical students were given case scenarios with low, intermediate, and high probability of coronary-artery disease and asked to estimate posttest probability before and after taking an EBM course

Survey of 300 physicians on their frequency of use of quantitative diagnostic methods, sensitivity, specificity, ROCs, LRs, and Bayesian logic 48 German physicians were asked to calculate the PPVs of 4 diagnostic tests; information was presented as probability or natural frequency 263 Swiss GPs were surveyed on the definitions of the terms “sensitivity” and “PPV” and were asked to calculate PPV; accuracy was tested with the use of a clinical vignette in which tests were presented as test only, as test plus sensitivity and specificity, and as test plus description of LR in plain language 50 Australian GPs were surveyed for their descriptions of the terms “PPV,” “sensitivity,” and “specificity,” then interviewed directly by a study author 234 Japanese medical students were given case scenarios with low, intermediate, and high probability for coronary-artery disease; estimates of pre- and posttest characteristics of stress tests were elicited from the students (intuitive estimates) and from the literature (reference estimates)

Results

Physicians and nonphysicians overestimated the PPV

Overestimation of PPV in scenarios presented with lower pretest probability; nonphysicians’ estimates of PPV in cases with negative test results were inconsistent 8 (3%) used Bayesian logic, 3 (1%) used ROCs, and 2 (0.66%) used LRs; 97% were unfamiliar with LRs and ROCs and 76% were unfamiliar with Bayesian logic Overall correct answer: Bayesian format, 10%; natural-frequency format, 46% Correct definition: sensitivity, 76%; PPV, 61%; PPV was calculated accurately by just 22% of GPs and was best estimated when the results of the LR of the test were presented in plain language 13 of the 50 said they knew about PPV, but just 1 met the criteria for knowledge of PPV Students could not rule out disease in low- and intermediate-probability situations because of errors in estimating the pretest diagnosis and applying Bayesian estimates in clinical practice; could result in the ordering of unnecessary tests Pretest probability varied from 5% to 100%, with a 56% response rate No improvement in the estimation of posttest probability in lowand intermediate-probability cases after the EBM course

Modified with permission from Ghosh AK, Ghosh K, Erwin PJ. Do medical students and physicians understand probability? QJM 2004;97:53-5. Oxford University Press.22

ability to understand NNT, low literacy, and framing of problems—may impair understanding of evidencebased information. Fuller and colleagues41 demonstrated that older pa-

tients (⬎75 years) misunderstood numeric expressions of risk and confused probabilities expressed as percentages and those expressed as fractions (ie, a probability of 20% would be 1/5). Sheridan et al42 found that

J Lab Clin Med Volume 145, Number 4

Ghosh and Ghosh

175

Table II. Studies of medical students’ and physicians’ understanding of NNT Type of study

Author (year)

Study design

Results

235 physicians were questioned on problems in hypertension and hyperlipidemia presented as relative and absolute risks Each of a random sample of 802 internists and GPs was given 1 of 2 questionnaires on the effectiveness of lipid therapy in RRR and ARR, NNT 73 GPs were given with a questionnaire on the treatment of hypertension with results presented as ARR, RRR, difference in event-free patients, and NNT 215 Australian family physicians

108 (46%) gave different responses to the same result presented differently (ie, RRR, ARR); 97 (89.8%) preferred RRR Physicians were less inclined to treat hyperlipidemia when numbers were presented as NNT and ARR

Survey of students, GPs, and faculty

Forrow et al (1992)

Random allocation of 2 questionnaires to physicians

Bucher et al (1994)

Questionnaire to GPs

Cranney and Walley (1996)

Randomized trail with case scenarios on hormone-replacemtn therapy

Nikolajevic-Sarunac J et al (1999)

Randomized crosssectional survey

Sheridan and Pignone (2002)

62 medical students at the University of North Carolina were presented data as RRR, ARR, and NNT

Questionnaire format

Nexoe J et al (2002)

Cross-sectional study, postal questionnaire

Halvorsen P et al (2003)

1500 Danish GPs were randomized to 4 groups to receive data presented as RRR, ARR, NNT, or a combination 1305 Norwegian physicians were randomized to receive different case scenarios

Questionnaire survey

Heller RF et al (2004)

535 British physicians were presented data as NNT and RRR

despite a self-proclamation of numeracy by 70% of their patients, only 2% of all patients could correctly answer all 3 numeracy questions posed to them. Lipkus et al43 showed that even highly educated people had difficulty with simple numeracy problems. Surprisingly, 16% to 20% of patients incorrectly answered such basic risk questions as “What constitutes a higher risk: 1%, 5%, or 10%?” Hence innumeracy in patients, even those provided accurately presented data, can result in erroneous understanding leading to misinformed medical decision-making. Different techniques have been used to present the therapeutic efficacy of drugs to patients. Qualitative statements about a drug’s efficacy were common practice until the EBM movement required proof of the clinical significance of a drug. The effect of a drug is often presented to the patient as RRR, ARR, and NNT. Studies have consistently shown that patients prefer to receive information as RRR.42,44,45 The concept of

75% of GPs had problems understanding statistics (RRR ⬎ ARR, etc)

Physicians were less likely to prescribe hormone-replacement therapy when data were presented as NNT than when they were presented as RRR 61% accurately interpreted the data, but correct interpretation was lower in the NNT format (NNT, 25% vs RRR, 75% vs ARR, 75%) 91% prescribed treatment when data were presented as RRR, compared with 63% when data were presented as NNT ⫹ ARR 77% of physicians did not recommend treatment as a result of concern over “wasted effort” (NNT-1) GPs preferred RRR to NNT (57% vs 38%)

NNT may not be readily understood by patients. In 1 randomized study, 80% of patients opted for hypothetical treatments presented as NNT ranging from 10 to 400, as long as the side effects were not severe.45 In another study, only 21% of patients were able to calculate the effect of treatment when presented with a case expressed in RRR format, compared with 6% of those presented with information in NNT format.42 NNT was considered akin to a lottery, with the benefit to patients from treatment being 1/NNT.44 Patients also found the concept of varying NNT, changing with different time points, confusing. Recent work has revealed that the reading level required to comprehend the information written on informational postcards often exceeds patients’ literacy skills.46,47 It is estimated that around 90 million Americans have literacy skills insufficient to help them comprehend current medical instructions.46 Sheridan et al42 reported that the readability of the format of risk re-

176

J Lab Clin Med April 2005

Ghosh and Ghosh

Table III. Quantitative studies of risk communication Modality

RR and AR

Author (year)

Study design

Result

Weissler AM (1989) Farrow L (1992) Malenka (1993) Naylor CD (1992)

Literature review Physician survey Outpatient survey Medical student, physicians

Bucher H (1994)

Physician survey

Hux JE (1995)

Outpatient survey

Nexoe JD (2002)

Physician survey

Pictorial display

Mazur DJ (1993) Mazur DJ (1993) Fuller R (2001)

Patients, physicians, and students Patients Patient ⬎ 75 yr

Decision aids

O’Connor AM (1998)

Randomized clinical trial in postmenopausal women (pamphlet vs pamphlet plus tape and booklet) Breast-cancer patients, lumpectomy plus radiotherapy vs mastectomy Internet, computer, various media ⬎65 yr vs ⬍65 yr

RR, AR, NNT

Whelan T (1999)

Survival curves

Deyo RA (2001) Mazure DJ (1993) Weeks JC (1998)

Calman chart

Calman K (1996)

Written instructions

Bogaert MG (1994) Bauman A (1997) Garrud P (2001)

Probabilities

Kong A (1986)

Acceptable regret Natural-sampling frequency

Gurm HS (2000) Djulbegovic B (1999) Wilkness SM (2000)

Cohort stages III and IV non-small–cell lung cancer (NSCC) Patient risk vs risk of daily activities Risk communication, drug-package insert, evaluation in outpatients Pamphlets for asthma

RR preferred over AR RR preferred over AR RR preferred over AR RR was preferred to AR, which was preferred to NNT RR was preferred to AR, which was preferred to NNT RR was preferred to AR, which was preferred to NNT RR was preferred to AR, which was preferred to NNT Ability to interpret graphic data is variable Ability to interpret graphic data is variable Limited ability to interpret graphic data; preferred modality over numerical percentages Increased understanding of risk and benefits

Decision aid improved decision capability of patients Improved understanding Patients younger than 65 yr preferred treatment with better short-term survival Patients overestimated 6-mo survival Easy to understand; disadvantage was numericals Inadequate patient-physician discussion of risk Written for higher reading levels, resulting in a disadvantage Decreased anxiety on the part of the patient

Pamphlets explaining laparoscopic technique Quantitative estimation of quantitative terms among physicians Outpatients; 1:100 risk vs 99% safe Patients with cancer; uncertain outcomes

Framing effect Routine use difficult

Genetic counseling for cancer risk

Risk perception better with frequency format

duction of a treatment varied quite extensively in their study: Flesch-Kincade grade levels of 5.8 (RRR), 8.3 (ARR), and 11.5 (NNT), respectively. It is therefore recommended that written instructions be kept at FleschKincade grade level 6. This may also account for the difficulty encountered by patients in understanding NNT. At the time of this writing, the Flesch-Kincade grade-level system is available in Microsoft Word. Widely used to assess the readability of text, it has high validity. The formula depends on average sentence length (number of words⫼number of sentences) and the average number of syllables per word (number of syllables⫼number of words). Use of this formula could lead to underestimation of the actual reading level in

Wide variation in probability estimation

complex text and short but unfamiliar medical terms, as well as the level at which many medical documents are written. It is therefore possible that an adult reader, even a well-educated one, could have difficulty reading a complex medical term and would be erroneously labeled as reading at a lower grade level. The Flesch Reading Scale, which measures reading material on a scale of 100 (easy) to 0 (difficult) could be more useful in an adult population. A Flesch score of 65 is usually identified as the “plain-English score.” In a recent study, patients with diabetes and low literacy levels who were subject to intensive management, including 1-on-1 educational sessions, telephone reminders, and help in addressing problems with insurance and com-

J Lab Clin Med Volume 145, Number 4

Ghosh and Ghosh

177

Table IV. Qualitative studies of risk communication Modality

Audiotape videotape Observation

Author (year)

Resident physicians and inpatients

Risks discussed 14% of the time

Makoul GP (1995)

Video/audio/questionnaire telephone interview/chart review; GPs

Browner CH (1996)

Pregnant women watched videotape of prenatal screening Audiotape, outpatient interview Audiotape of physicians with malpractice claims

Physicians overestimated the amount of information given and the patient’s ability to understand Better retention of knowledge

Edwards A (1998)

Focus group of nurses GPs

Pho K (2000)

Telephone survey of patients ⬍ 60 yr with colorectal adenoma

Krouse HJ (2001)

Video-modeling studies, evaluations of preprocedure video were reviewed Retrospective audiotape review of oncology consultation

Leighl N (2001)

Schapira MM (2001)

Pearson SD (1994) Scheitel SM (1996)

Closed-circuit TV

Result

Wu WC (1988)

Braddock (1997) Levinson (1997)

Questionnaire

Study design

Worobey JL (1985)

Focus group of women on breast-cancer risk, graphical display of data vs probability format Risk attitudes of emergency physicians, patients with chest pain Questionnaire filled out by patient and physician after examination Patient education

munication achieved better glycemic control than did similar patients who received standard care.48 The manner in which information is framed may also dissuade patients from seeking treatment. Gurm and Litaker49 demonstrated that when the information of a procedure was framed as 99% safe, versus one with a 1-in-100 risk of complications, patients overwhelmingly favored the first procedure, even though there was no real difference in the data! Information about decisions involving risk can be framed with emphasis either on potential gains or on potential losses.50 In risky situations such as detection behaviors (eg, mammography for breast-cancer diagnosis), loss-framed messages were thought to be more persuasive than gain-framed messages. In a study of women with no family history of breast cancer who did not adhere to the recommendation of screening mammography, a video-message system was used to communicate a gain-framed message to 1 group and a loss-framed message to another.50 The authors found that the loss-framed message resulted in 1.7 times greater use of mammography than did

Discussion on risk/benefit in 9% cases Physicians who were sued used less humor, spent less time, and often lacked a faciliatory style of talking with patients Concern about available information/ communication Poor awareness of increased risk of colorectal cancer in first-degree relatives Decreased anxiety in patients and increased understanding Anxious patients were less satisfied; patients receiving less informa-tion were more satisfied. Uncertainty of risk decreases trust

Low risk takers admitted more patients Patients failed to report 68% of all health problems, summary letter suggested Patient population is heterogeneous, should modify to suit needs

gain-framed messaging and that this increase was not associated with an alteration of perceived risk. Patients’ potential innumeracy, limited literacy, and tendency to be influenced by the framing of data should all be considered during the medical decision-making process. TRANSLATING EVIDENCE INTO EFFECTIVE COMMUNICATION

It has been shown that communication of risks of disease and treatment modalities remains central to the physician-patient relationship and that it is an essential determinant of patient satisfaction.51 Understanding of risk remains germane to patients’ desire to effect change through behavior modification and ensure treatment compliance. We conducted a systematic review to identify the modalities by which risk is communicated to patients and to determine its effectiveness,52 studying 22 articles involving quantitative methods33,34,38,41,44,49,53– 69 and 13 articles in which qualitative methods70 – 82 were used to describe risks (Tables III and IV). Among the quantitative methods terms used to ex-

178

J Lab Clin Med April 2005

Ghosh and Ghosh

plain risks were RRR and absolute risk, RRR along with ARR and NNT, pictorial display of risks , decision aids, survival curves, Calman Chart Risk, written instructions, probabilities, the acceptable-regret technique, and natural sampling frequency (Table III). Among the qualitative studies used to study risk communication were focus-group interviews (involving audiotape and videotape techniques in 10), questionnaires (n ⫽ 2), and closed-circuit TV (n ⫽ 1). The qualitative and quantitative studies (n ⫽ 4) included the study of focus groups (audiotape) with graphical display (n ⫽ 2) and percentages (n ⫽ 2). Most of the studies were conducted to study a single disease model. Patients expressed greater desire to follow therapy when reduction of risk was presented as RRR rather than as ARR and NNT. Additionally, risktaking behavior on the parts of physicians and patients determined their communication style and choice of therapy, respectively.80 Education level, patient age, cultural differences, errors in risk estimation, and technique of expression (eg, quantitative vs qualitative) were all related to overall perception of risk. Despite the large body of evidence there seems to be a lack of consensus regarding the most appropriate method with which to communicate medical risks.83 Appropriate techniques for the presentation of accurate information about actual risks in several disease models can be challenging. At this time, a combination of quantitative and qualitative techniques, along with pictorial display of data, seems to be helpful in explaining risks to patients. Understanding of the complexity of the current medical decision-making process remains important in the management of a patient’s problems (Fig 1, B). Recent literature indicates that evidence available in full-text form online may be cited more often just because it is more visible.84 Physicians should be able to critically appraise evidence to determine the credibility of information and not just accept evidence because it is easily available. Epstein et al51 recommended 5 communication techniques with which to convey evidence to patients: (1) understand the patient’s and his or her family’s experiences and expectations; (2) build a partnership; (3) provide the medical evidence, including a discussion of uncertainties; (4) present recommendations based on consensus; and (5) check for understanding and agreement. One of the many strategies for communicating risk is known as CARE: Cite basic risk in general terms, add estimated probabilities for positive and negative outcomes to descriptive terms such as “low-risk” reinforce effectiveness by using visual aids, and express encouragement and hope to the patient.85 It is possible that in the setting of an established physician-patient relation-

ship, the physician can have enough understanding of a patient’s risk preferences to be able to present the risks in a more comprehensible form for the specific patient and be able to guess accurately what the patient likely desires to know. Though this could pose an ethical challenge to the physician trying to uphold the trust of the patient and engage him or her in a well-informed decision-making process. Having carefully evaluated a patient’s condition and being aware of the efficacy of the best available evidence, the clinician must also incorporate the patient’s values, needs, and preferences into the process of identifying the best treatment plan. Finally, the physician needs to be able to support the patient when diagnosis and treatment options remain uncertain despite exhaustive evaluation.7,86 In summary, physicians’ test-ordering tendencies may be based on Bayesian and several non-Bayesian factors. Despite recent increased focus on EBM, physicians demonstrate widely varying understanding of probability terms and NNT. Patients similarly vary in their ability to grasp information about risk presented as numbers and percentages. It is unclear whether the variability of medical management could to some extent be contributed to by variability in the understanding of current evidence by physicians and patients. At this time, both qualitative and quantitative approaches to risk communication seem appropriate. Further research is indicated to identify the most appropriate “specific-risk” communication technique tailored to answer a patient’s questions. REFERENCES

1. Morgan MG, Fischhoff B, Bostrom A, et al, editors. Risk communication. A mental model approach. Cambridge, UK: Cambridge University Press; 2002. 2. Envisioning the National Health Care Quality Report. Washington, DC: Institute of Medicine; 2001. 3. Barzansky, B, Etzel SI. Educational programs in US medical school, 2002–2003. JAMA 2003;290:1190 – 6. 4. Cook RJ, Sackett DL, Roberts RS. An assessment of clinically useful measures of treatment effect. BMJ 1995;310:452– 4. 5. Nuovo J, Melnikow J, Chang D. Reporting number needed to treat and absolute risk reduction in randomized controlled trials. JAMA 2002;287:2813– 4. 6. Ghosh AK, Ghosh K. Misinterpretation of numbers: a potential source of physician’s error. Lancet. 2004;364:1108 –9. 7. Ghosh AK. On the challenges of using evidence-based information: the role of clinical uncertainty. J Lab Clin Med 2004;144: 60 – 4. 8. Vickrey BG, Gifford DR, Belin TR, et al. Practice styles of US compared to UK neurologists. Neurology 1998;50:1661– 8. 9. Cohen MC, Stafford RS, Misra B. Stress testing: national patterns and predictors of test ordering. Am Heart J 1999;138: 1019 –24. 10. Landon BE, Blumenthal D, Reschovsky J, Reed M, Landon BE. Personal, organizational, and market level influences on physicians’ practice patterns: results of a national survey of primary care physicians. Med Care 2001;39:889 –905.

J Lab Clin Med Volume 145, Number 4

11. Carey TS, Garrett J. Patterns of ordering diagnostic tests for patients with acute low back pain. The North Carolina Back Pain Project. Ann Intern Med. 1996;125:807–14. 12. Van Boven K, Dijksterhuis P, Lamberts H. Defensive testing in Dutch family practice. Is the grass greener on the other side of the ocean? J Fam Pract 1997;44:468 –72. 13. Rello J, Lorente C, Bodí M, Diaz E, Ricart M, Kollef MH. Why do physicians not follow evidence-based guidelines for preventing ventilator-associated pneumonia? A survey based on the opinions of an international panel of intensivists. Chest 2002; 122:656 – 61. 14. Allison JJ, Kiefe CI, Cook EF, Gerrity MS, Orav EJ, Centor R. The association of physician attitudes about uncertainty and risk taking with resource use in a Medicare HMO. Med Decis Making 1998;18:320 –29. 15. Reid MC, Lane DA, Feinstein. Academic calculations versus clinical judgements: practicing physicians use of quantitative measures of test accuracy. Am J Med 1998;104:374 – 80. 16. Pazirandeh M. Does patient partnership in continuing medical education (CME) improve the outcome in osteoporosis management? J Contin Educ Health Prof 2002;22;142–51. 17. van Merode GG, Stroink AE, Maarse JA, Goldschmidt HM. Impact of insurance coverage type on laboratory test ordering behavior of general practitioners. World Hosp Health Serv 2000; 36:7–12, 36 –7. 18. Franks P, Fiscella K. Effect of patient socioeconomic status on physician profiles for prevention, disease management, and diagnostic testing costs. Med Care 2002;40:717–24. 19. Nakar S, Vinker S, Neuman S, Kitai E, Yaphe J. Baseline tests or screening: what tests do family physicians order routinely on their healthy patients? J Med Screening 2002;9:133– 4. 20. Waxman MA, Levitt MA. Are diagnostic testing and admission rates higher in non–English-speaking versus English-speaking patients in the emergency department? Ann Emerg Med 2000; 36:456 – 61. 21. Bryant GD, Norman GR. Expressions of probability: words and numbers. N Engl J Med 1980;302:411. 22. Ghosh AK, Ghosh K, Erwin PJ. Do medical students and physicians understand probability? QJM 2004;97:53–5. 23. Lyman G, Balducci L. Overestimation of test effects in clinical judgement. J Cancer Educ 1993;8:297–307. 24. Lyman GH, Balducci L. The effect of changing disease risk on clinical reasoning. J Gen Intern Med 1994;9:488 –94. 25. Hoffrage U, Gigenzenger G. Using natural frequencies to improve diagnostic inferences. Acad Med 1998;73:538 – 40. 26. Steurer J, Fischer JE, Bachman LM, et al. Communicating accuracy of tests to general practioners: a controlled study. BMJ 2002;324:824 – 6. 27. Young JM, Glasziou P, Ward JE. General practitioners’ self rating of skills in evidence based medicine: a validation study. BMJ 2002;324:950 –1. 28. Noguchi Y, Matsui K, Imura H, et al. Quantitative evaluation of the diagnostic thinking process in medical students. J Gen Intern Med 2002;17:839 – 44. 29. Heller RF, Sandars JE, Patterson L, McElduff P. GPs’ and physicians’ interpretation of risks, benefits and diagnostic test results. Fam Pract 2004;21:155–9. 30. Noguchi Y, Matsui K, Imura H, Kiyota M, Fukui T. A traditionally administered short course failed to improve medical students performance: a quantitative evaluation of diagnostic thinking. J Gen Intern Med 2004;19:427–32. 31. Hoffrage U, Lindsey S, Hertwig R, Gigerenzer G. Medicine. Communicating statistical information. Science 2000;290:2261–2.

Ghosh and Ghosh

179

32. Ghosh AK, Ghosh K. Number needed to treat (NNT): challenges in interpretation. A literature review. J Gen Intern Med 2004; 19(Suppl 1):184. 33. Forrow L, Taylor WC. Absolutely relative: how research results are summarized can effect treatment decisions. Am J Med 1992; 92:121– 4. 34. Bucher HC, Weinbacher M, Gyr K. Influence of methods of reporting study results on decision of physicians to prescribe drugs to lower cholesterol concentration. BMJ 1994;309:761– 4. 35. Cranney M, Walley T. Same information, different decisions: the influence of evidence on the management of hypertension in the elderly. Br J Gen Pract 1996;46:661–3. 36. Nikolajevic-Sarunac J, Henry DA, O’Connell DL, Robertson J. Effects of information framing on the intentions of family physicians to prescribe long-term hormone replacement therapy. J Gen Intern Med 1999;14:591– 8. 37. Sheridan SL, Pignone M. Numeracy and the medical student’s ability to interpret data. Eff Clin Pract 2002;5:35– 40. 38. Nexoe J, Oltarzewska AM, Sawicka-Powierza J, Kragstrup J, Kristiansen IS. Perception of risk information. Similarities and differences between Danish and Polish general practitioners. Scand J Prim Health Care 2002;20:183–7. 39. Halvorsen PA, Kristiansen IS, Aasland OG, Forde OH. Medical doctors’ perception of the “number needed to treat” (NNT). A survey of doctors’ recommendations for two therapies with different NNT. Scand J Prim Health Care 2003;21:162– 6. 40. Barclay L. Vioxx withdrawal prompts reevaluation of COX-2 inhibitor safety. Medscape Available at: http//www.medscape.com/viewarticle/490979. Accessed March 17, 2005. 41. Fuller R, Dudley N, Blacktop J. Risk communication and older people— understanding of probability and risk information by medical inpatients aged 75 years and older. Age Ageing 2001; 30:473–96. 42. Sheridan SL, Pignone MP, Lewis CL. A randomized comparison of patients’ understanding of number needed to treat and other common risk reduction formats. J Gen Intern Med. 2003;18: 884 –92. 43. Lipkus IM, Samsa G, Rimer BK. General performance on a numeracy scale among highly educated samples. Med Decis Making 2001;21:37– 44. 44. Hux JE, Naylor CD. Communicating the benefits of chronic preventive therapy: Does the format of efficacy data determine patients’ acceptance of treatment? Med Decis Making 1995;15: 152–7. 45. Kristiansen IS, Gyrd-Hansen D, Nexoe J, Nielsen JB. Number needed to treat: easily understood and intuitively meaningful? Theoretical considerations and a randomized trial. J Clin Epidemiol 2002;55:888 –92. 46. Rudd RE, Comings JP, Hyde JN. Leave no one behind: improving health and risk communication through attention to literacy. J Health Commun 2003;8(Suppl 1):104 –15. 47. Kirsch I, Kader B, Jensen GV, Kopher W. Adult literacy in America. Washington, DC: National Center for Educational Statistics, US Dept of Education; 1993. 48. Rothman RL, DeWalt DA, Malone R, Bryant B, Shintani A, Crigler B, et al. Influence of patient literacy on the effectiveness of a primary care– based diabetes disease management program. JAMA. 2004;292:1711– 6. 49. Gurm HS, Litaker DG. Framing procedural risks to patients: is 99% safe the same as a risk of 1 in 100? Acad Med 2000;75: 840 –2. 50. Banks SM, Salovey P, Greener S, Rothman AJ, Moyer A, Beauvais J, et al. The effects of message framing on mammography utilization. Health Psychol 1995;14:178 – 84.

180

Ghosh and Ghosh

51. Epstein RM, Alper BS, Quill TE. Communicating evidence for participatory decision making. JAMA 2004;291:2359 – 66. 52. Ghosh AK, Erwin P, Ghosh K. Effective risk communication: multiple modalities, unclear consensus. A review of literature. J Invest Med 2002;50:182. 53. Malenka DJ, Baron JA, Johansen S, Wahrenberger JW, Ross JM. The framing effect of relative and absolute risk. J Gen Intern Med 1993;8:543– 8. 54. Weissler AM, Miller BI, Boudoulas H. The need for clarification of percent risk reduction data in clinical cardiovascular trial reports. J Am Coll Cardiol 1989;13:764 – 6. 55. Naylor CD, Chen E, Strauss B. Measured enthusiasm: does the method of reporting trial results alter perceptions of therapeutic effectiveness? Ann Intern Med 1992;117:916 –21. 56. Mazur DJ, Hickam DH. Patients’ and physicians’ interpretations of graphic data displays. Med Decis Making 1993;13:59 – 63. 57. Mazur DJ, Hickman DH. Patient preferences: survival vs qualityof-life considerations. J Gen Intern Med 1993;8:374 –7. 58. O’Connor AM, Tugwell P, Wells GA, Elmslie T, Jolly E, Hollingworth G, et al. Randomized trial of a portable, selfadministered decision aid for postmenopausal women considering long-term preventive hormone therapy. Med Decis Making 1998;18:295–303. 59. Whelan T, Levine M, Gafni A, Sanders K, Willan A, Mirsky D, et al. Mastectomy or lumpectomy? Helping women make informed choices. J Clin Oncol 1999;17:1727–35. 60. Deyo RA. A key decision maker: the patient. BMJ 2001;323: 466 –7. 61. Mazur DJ, Merz JF. How the manner of presentation of data influences older patients in determining their treatment preferences. J Am Geriatr Soc 1993;41:223– 8. 62. Weeks JC, Cook EF, O’Day SJ, Peterson LM, Wenger N, Reding D, et al. Relationship between cancer patients’ predictions of prognosis and their treatment preferences. JAMA 1998;279:1709 –14. 63. Calman KC. Cancer: science and society and the communication of risk. BMJ 1996;313:799 – 802. 64. Bogaert MG, Stichele RV. Drug treatment: risk perception by the public and the health professionals Eur J Clin Res 1994;6:203–7. 65. Bauman A. The comprehensibility of asthma education materials. Patient Educ Couns 1997;32(1 Suppl):S51–9. 66. Garrud P, Wood M, Stainsby L. Impact of risk information in a patient education leaflet. Patient Educ Couns 2001;43:301– 4. 67. Kong A, Barnett GO, Mosteller F, Youtz C. How medical professionals evaluate expressions of probability. N Engl J Med 1986;315:740 – 4. 68. Djulbegovic B, Hozo I, Schwartz A, McMasters KM. Acceptable regret in medical decision making. Med Hypotheses 1999;53: 253–9. 69. Wilkness SM. Communication of cancer risk information in genetic counseling: does format of presentation affect understanding? [abstract]. University of Virginia, Dissertion Abstracts International, section B: The Sciences and Engineering 2000; 61(6-B):3296.

J Lab Clin Med April 2005

70. Wu WC, Pearlman RA. Consent in medical decision making: the role of communication. J Gen Intern Med 1988;3:9 –14. 71. Makoul G, Arntson P, Schofield T. Health promotion in primary care: physician-patient communication and decision making about prescription medications. Soc Sci Med 1995;41:1241–54. 72. Browner CH, Preloran M, Press NA. The effects of ethnicity, education and an informational video on pregnant women’s knowledge and decisions about a prenatal diagnostic screening test. Patient Educ Couns 1996;27:135– 46. 73. Braddock CH III, Fihn SD, Levinson W, Jonsen AR, Pearlman RA. How doctors and patients discuss routine clinical decisions. Informed decision making in the outpatient setting. J Gen Intern Med 1997;12:339 – 45. 74. Levinson W, Roter DL, Mullooly JP, Dull VT, Frankel RM. Physician-patient communication. The relationship with malpractice claims among primary care physicians and surgeons. JAMA 1997;277:553–9. 75. Edwards A, Matthews E, Pill R, Bloor M. Communication about risk: the responses of primary care professionals to standardizing the ’language of risk’ and communication tools. Fam Pract 1998;15:301–7. 76. Pho K, Geller A, Schroy PC III. Lack of communication about familial colorectal cancer risk associated with colorectal adenomas (United States). Cancer Causes Control 2000;11:543– 6. 77. Krouse HJ. Video modelling to educate patients. J Adv Nurs 2001;33:748 –57. 78. Leighl N, Gattellari M, Butow P, Brown R, Tattersall MH. Discussing adjuvant cancer therapy. J Clin Oncol 2001;19:1768 –78. 79. Schapira MM, Nattinger AB, McHorney CA. Frequency or probability? A qualitative study of risk communication formats used in health care. Med Decis Making 2001;21:459 – 67. 80. Pearson SD, Goldman L, Garcia TB, Cook EF, Lee TH. Physician response to a prediction rule for the triage of emergency department patients with chest pain. J Gen Intern Med 1994;9: 241–7. 81. Scheitel SM, Boland BJ, Wollan PC, Silverstein MD. Patientphysician agreement about medical diagnoses and cardiovascular risk factors in the ambulatory general medical examination. Mayo Clin Proc 1996;71:1131–7. 82. Worobey JL. Using closed-circuit television as a teaching tool: implications for health communication educators. Proceedings of the 76th Annual Meeting of Eastern Communication Association; 1985 May 2–5; Providence, RI. 83. Edwards A, Unigwe S, Elwyn G, Hood K. Effects of communicating individual risks in screening programs: Cochrane systematic review. BMJ 2003;327:1–7. 84. Murali NS, Murali MR, Auethavekiat P, Erwin PJ, Manjrekar J, Ghosh AK. Impact of FUTON and NAA bias on visibility of research . Mayo Clin Proc 2004;79:1001– 6. 85. Paling J. Strategies to help patients understand risks. BMJ 2003; 327:745– 8. 86. Ghosh AK. Dealing with medical uncertainty: a physician’s perspective. Minn Med 2004;87:48 –51.