- Email: [email protected]
Admission Decisions in Emergency Department Chest Pain Patients at Low Risk for Myocardial Infarction: Patient Versus Physician Preferences
GENERAL CLINICAL INVESTIGATION/ORIGINAL CONTRIBUTION
AdmissionDecisions in Emergency Department Chest Pain Patients at Low Risk for Myocardial Infarction: Patient Versus Physician Preferences From the UCLA Emergency Medicine Center* and the Departments of Medicine": and Pediatrics~, UCLA 5chooI of Medicine, Los Angeles, California. Receivedfor publication January 10, I996. Revision received June 7, i996. Acceptedfor publication Augctst i2, 1996. Presented at the Societyfor Academic Emergency MedicineAnnual Meeting, San Antonio, Texas, May I995. Copyright © by the American College of Emergency Physicians.
Mark A Davis, MD, MS** Amanda Keerbs* Jerome R Hoffman, MA, MD, ** Larry J Baraff, MD *§
See related editorial, p 702. Study objective: Patient involvement in medical decisionmaking is acceptedas an ethical and a legal imperative. Medical decisions are based in part on individuals' knowledge and acceptance of risk of adverse consequences.It is unclearwhether actionstaken to protect against low risk of poor outcome reflect patient or physician preferences. We sought to test the hypothesis that emergency department chest pain patients presented with a hypothetical situation involving a low risk of myocardial infarction are more willing than ED physicians to accept the risk associated with discharge from the hospital. Methods: We prospectively surveyed 89 ED patients with chest pain and a cohort of physicians in the ED who had been presented a hypothetical case in which the risk of AMI was quoted as 5% and the risk of death or disability if the patient was discharged was 1% and .2% if the patient was admitted. All the patients had presented to the ED with a chief complaint of chest pain; the 31 physicians, all residents, were approached at a teaching conference separate from their clinical duties. Results: Twenty-eight patients (31%), compared with 2 physicians (6%), chose discharge for the hypothetical patient with chest pain (25% difference; 95% confidence interval [Cl], 6% to 41%). Forty-four patients (49%), compared with 30 physicians (97%), correctly identified the risks associated with admission and discharge(46% difference; 95% CI, 29% to 63%). Of the subjects who correctly identified the risks, 19 patients (43%} preferred discharge, compared with 1 physician (3%)(40% difference; 95% CI, 18% to 60%). Conclusion: ED patients with chest pain appear to be more likely than physicians to accept a small risk of poor outcome in a hypothetical circumstance. Many patients cannot identify the risks associated with their decision.
606
ANNALS OF EMERGENCY MEDICINE
28:6
DECEMBER1996
CHEST PAIN Davis et al
[Davis MA, Keerbs A, Hoffman JR, Baraff LJ: Admission decisions in emergency department chest pain patients at low risk for myocardial infarction: Patient versus physician preferences. Ann EmergMed [:)ecember 1996;28:606-611.]
INTRODUCTION Ethical and legal considerations have recently prompted a departure from the traditional paternalistic model of medicine in favor of patient involvement in health care decisions, z Shared decisionmaking is a complex process in which the patient must recognize that a choice is needed, be aware of alternatives~ have the ability and the freedom to assess consequences, and be able to implement choices. 2 Patients vary in their desire to participate in medical decisionmaking3-6, but "informed choice "7 is impossible if information on treatment alternatives is unavailabie or is not effectively communicated to the patient. Protocols i~ve been derived that can characterize a population of patients with chest pain for whom the risk of acute myocardial infarction (AMI) is less than 7%, with a rate of emergency complication of .35% when protocol and physician agree that CCU admission is not required, s Additional clinical and electrocardiographic criteria available at the time of presentation may identify patients at even lower risk of AMI and emergency complication. 9q2 This has led to non-ICU admissions for canny patients deemed to be at low risk for AM113-1s and, more recently, to the development of shortstay chest pain centers. ~9 In the United States a minority of patients admitted to acute care hospitals because of chest pain prove to have AMI. 2° This reflects the fact that it is impossible to rule out AMI with 100% certainty in any patient at the time of presentation. It ts not known whether this practice is based on a lack of physician knowledge of risks involved or due to the preferences of physicians or patients. We devised this study to test the hypothesis that--when presented with a hypothetical circumstance in which the risk of AMI is low (5%), with a risk of death or serious disability of .2% with hospitalization and 1% with discharge-patients would be more likely than physicians to prefer discharge with outpatient follow-up. We further hypothesized that this would be true whether or not patients could identify the quantitative likelihood of adverse outcomes presented in the scenario. Finally, we hypothesized that patients and physicians would cite different factors as most important with regard to hospitalization or outpatient care, with risk of bad ,outcome being the most important factor to physicians but not to patients.
DECEMBER 1996
28:6
ANNALS OF EMERGENCY MEDICINE
MATERIALS AND METHODS We interviewed a convenience sample of 89 patients who presented to the UCLA ED with a chief complaint of chest pain between September 1994 and February 1995 during times when trained research assistants working on the study were available. A patient was eligible to participate if a physician or nurse believed the patient's chest pain might be of cardiac origin and required cardiac monitoring in the ED. Other inclusion criteria were age greater than 25 years, fluency in English, and willingness to participate. Patients were excluded if they demonstrated obvious trauma or physical or mental incapability or if the intensity of the medical management activities was believed by the interviewer to preclude participation. All subjects were verbally informed of the purpose of the research and that their participation was voluntary. Written informed consent was waived by the institutional review board because this research involved only voluntary interviews. The information collected was recorded in such a manner that the participants could not be identified and the disclosure of responses could not conceivably be expected to place subjects at risk of criminal or civil liability or to be damaging to their financial standing, employability, or reputation. The physician subject cohort comprised 19 senior emergency medicine and 12 senior internal medicine residents who, in the course of training, regularly evaluated and made disposition decisions for ED chest pain patients under the
Figure. Summary of a scenariofor Zow-dsk chest pain. The emergency department at UCLA is studying methods to improve health care delivery and patient satisfaction with medicat care. We are interested in determining patients' choices for in-hospital versus out-of-hospital testing for possible heart attack. The example that follows does not apply to you personally. This is a research study only, is not being used by your physicians to care for you, and should not influence any decisions. Suppose that you are in the emergency department and you are told the following: There is a 5% (I in 20) chance that you are having a heart attack. There is a ,2% (1 in 500)chance that you will die or have serious disability if you stay in the hospital. There is a 1% (1 in 100)chance that you will die or have serious disability if you go home, If you were to stay in the hospital, you would undergo testing and observation during your hospital stay. If you were to go home, you would be given an appointment to return to the hospital within the next few days to meet with a cardiologist for further testing. In the event of an emergency or recurrence of symptoms you would call 911. The hospital charges approximately $3,000 for 1 day of hospital monitoring and tests. If you were to go home, the costs for testing and seeing a cardiologist would be approximately $1,500,
607
CHEST PAIN Davis et al
supervision of attending emergency physicians. Physician subjects were presented with the hypothetical scenario and interviewed at a teaching conference separate from their clinical duties. Subjects in both groups were given a one-page hypothetical scenario involving a low-risk chest pain patient. The scenario provided treatment alternatives with associated risks and costs in a narrative and review format (Figure). This scenario was read aloud to each subject in the two groups. Patient subjects received a study questionnaire, which was read aloud to them by research assistants, who also recorded the patients' responses. Physician subjects read and completed the questionnaire on their own. Patient subjects were instructed to answer the questionnaire as if they were in the situation described in the hypothetical scenario, although they were advised that the case presented in the scenario was hypothetical and did not apply to their own current situation. Research assistants were instructed to tell any patient who asked questions regarding his or her own case after reading the scenario that such questions should be addressed to the patient's physician and to restate that the hypothetical case was neither designed nor intended to apply to the patient's case. Physician subjects were asked to assume while answering the questionnaire that the risks presented in the scenario applied to a patient under their care. Table 1.
Subject identification of risks associated with discharge and admission.
Each subject was asked to indicate whether he or she would choose admission or discharge with outpatient follow-up for the hypothetical patient. Physician subjects were also asked whether they would personally prefer admission or discharge if they faced as patients the risks presented for the hypothetical case. Every subject was asked to identify the risks associated with each course of action (the answers to which were listed on the scenario, which was still in his or her possession) as two multiple-choice questions with possible answers being .2%, 1%, 5%, 10%, and "don't know." Finally, each subject was asked to indicate the level of importance of patient preference, cost, risk of death or disability, and physician preference in their decision. The hypotheses that patients would be more likely than physicians to prefer discharge and that physicians wou]d be more likely to correctly identify risk information were each tested with the use of Fisher's exact test. The hypotheses that patients and physicians would cite different factors as most important to them in choosing hospitalization or outpatient care, with risk being most important to physicians and less so to patients, was tested with Fisher's exact test with extensions. Given the lack of prior information on the likely percentage of patients who might prefer discharge or admission in a scenario involving low-risk chest pain, we had no basis on which to perform power calculations to determine the sample size needed to find specific differences in disposition preference. Therefore no single cutoff for "significance" was chosen, beyond the arbitrary nature of statistical testing in Table 2.
Risk (%) Discharge 0 .2 1t 5 10 Don't know Admission 0 .2~ 1 5 10 Don't know Both risks correct
No. of Patients (%) [n=89]
No. of Physicians (%) [n=31]
2 (2.2) 10(11) 48 (54) 11 (12) 7 (7.9) 11 (12)
0 0 30 {97) 1 (3) O 0
7 (8) 54 (61) 5 (5.6) 6 (6.7) 1 (1.0) 16 {18) 44 (49)
0 31 (100) O 0 0 0 30 (97)
p,
Group <.0001
<.0001
<.0001
*Fisher's exact test. tDenotes correct responses as presented in the chest pain scenario. Subjects held the scenario with the correct responses as they answered interviewer questions.
608
Preference for discharge or admission for chest pain scenario patient.
All subjects No. of subjects No. preferring discharge for scenario patient (%) No. preferring admission for scenario patient Subjects correctly identifying risks of discharge and admission No. of subjects No. preferring discharge for scenario patient (%) No. preferring admission for scenario patient (%)
Patients
Physicians
89 28 (31)
31 2 (6)
61 (69)
29 (94)
44 19 (43)
30 1 (3.0)
25 (57)
29 (97)
P*
<,01
<001
*Fisher's exact test.
ANNALS OF EMERGENCY MEDICINE
28:6
DECEMBER 1996
CHEST PAIN Davis et al
general. 21 We chose to present absolute P values and 95% confidence intervals (CIs) for multiple-comparison data. RESULTS
The 89 subject patients included 46 men and 43 women; mean and median ages were each 61 years. Subject patients came from 155 chest pain patients approached by research assistants, of whom 66 were excluded for the following reasons: 23 did not speak English, 36 refused or were incapable of participating, 1 could not complete the questionnaire because of intense medical management activities, and 6 failed to complete the questionnaire. Seventy-seven of the subject patients were themselves ultimately admitted to the hospital. Of the 73 patients for whom dischage summades were available, 63 had a discharge diagnosis that included coronary artery disease, of whom 10 had AMI ruled in. The physician subject group comprised 31 resident physicians, 19 in emergency medicine and 12 in internal medicine. Only half of the 89 patients correctly answered both multiple-choice questions regarding the risks associated with admission or discharge, compared with virtually all the physician subjects (46% difference; 95% CI, 29 to 63%; Table 1). Patients were more likely than physicians to choose discharge for the hypothetical patient presented in the case scenario (31% versus 6%; 25% difference; 95% CI, 6% to 41%). This was also true when the responses of only those subjects who correctly identified all risks were counted (Table 2). Seventy-sevelq of the patients were themselves ultimately admitted to the hospital. Of the 73 patients for whom discharge summaries were available, 63 had a discharge diagnosis including coronary artery disease. Of the 63, 10 were found to have sustained an AMI. When physicians were asked whether they would prefer to be admitted if they faced the risks in the hypothetical
scenario, three altered their responses. These three had chosen to admit the hypothetical patient but said they would choose discharge for themselves. Patients attached greater importance to patient preference and cost than did physicians. All patients and physicians said they believed physician preference was very important or somewhat important, and we detected no significant difference between the two groups with regard to this factor. Physicians identified risk as more important in decisionmaking than did patients (Table 3). DISCUSSION
Our subject patients and physicians differed in their disposition preferences for a hypothetical low-risk patient presenting to an ED with chest pain. Physicians were significantly less likely to prefer discharge with outpatient follow-up. Physicians were better able to identify correctly the risks presented in the case scenario, but the greater willingness of patients to choose a riskier disposition (discharge) remained present even among only that subgroup of subjects (patients and physicians) who correctly identified those risks. This finding suggests that the difference in preference between the two groups related not just to differential comprehension of risk but to a difference in underlying comfort with risk-taking, as well as other factors. Patients and physicians differed in the importance they attached to several factors related to their decisions. It is not clear why physicians might be less willing than patients to take the risks associated with discharge than were the patients themselves. We can only speculate that such factors as physician experience with bad outcomes in other (real) patients, training biases resulting in avoidance of error rather than analysis of net benefit22-24, and fear of legal consequences25 play roles. Physicians also placed less importance than did patients on cost as an important factor in their choices. Although
Table 3.
Importance of var/ousfactors in decisionmakingby patients and physicians. Very Important (%) Factors
Patient
Risk 73 Patient preference 67 Cost 27 Physician recommendation 84 *Not correctedfor multiplecomparisons,
Somewhat Important (%)
Small Importance (%)
Not Important (%)
Physician
Patient
Physician
Patient
Physician
Patient
Physician
Uncorrected P*
81 35 0 74
8 13 17 16
19 39 10 26
10 I1 17 0
0 26 71 0
9 8 39 0
0 0 19 0
.025 .0007 <,0001 .28
DECEMBER 1996 28:6 ANNALS OF EMERGENCY MEDICINE
6 O9
CHEST PAIN Davis et aI
IIIIII
physicians have recently been exposed to the increasing importance of cost-efficient care in a societal sense, it is perhaps not surprising that they are uncomfortable factoring in economic considerations when making potentially lifeand-death decisions about individual patients. Physicians also placed less significance on patient preference than did patients, although most physicians did say they believed it was at least somewhat important. Patients and physicians attached great significance to the opinion of the physician. Our study has several clear limitations. Although patients were queried during an actual episode of chest pain, for which most were hospitalized and in several cases proved to be the result of AMI, they were made aware that the case scenario was hypothetical and unrelated to their own situation. It is possible that patients would be more cautious about accepting any increase in risk (even if quite small) if and when such risk actually applied to them. It is our experience, however, that patients are often willing to take such risks in actual practice and frequently stress other personal priorities when discussing possible hospitalization. Similarly, physicians were removed from actual patient care when they participated in the study, and they too might have chosen differently when confronted with an actual patient. We believe it most likely, however, that they would be even more cautious if dealing with a real patient. The scenario and questions provided to subjects were intended to highlight risks associated with disposition decisions, with no mention of probability of favorable outcome, and are therefore subject to the effects of framing bias. 26-29 Had we emphasized the expectation of positive results (99% if discharged, 99.8% if admitted), it is likely that more subjects would have chosen the higher risk alternative (discharge), although it is unclear whether patients and physicians would have done so at the same rate. In addition, no formal analysis of patients' psychologic state 3°, rationality 31, or identification of 'terrors in thinking" that did not maximize expected utility 3~ was performed. More sophisticated educational techniques might have yielded greater patient understanding 3>34, although the question of whether ED patients can be educated to truly fulfill the requirements for informed consent requires further clarification. 35,36 There are also limitations to the external validity of our study. We investigated patient preferences with regard to a low-risk scenario only; therefore our findings should not be generalized to situations of higher risk. Our physician subjects were residents. Most had less than 4 years of clinical experience as physicians, they were generally younger than the patients for whom they cared, and all were actively involved in formal training at the time of the study. It is
6 10
possible that physicians with greater experience and those in other practice environments would approach the choices presented in a different fashion. Finally, our study assumed a situation in which risks involved in the hypothetical scenario could be accurately defined for each of the alternative dispositions. Such definition is possible m only a few areas of medicine. Even with the extensive information available on chest pain, for example, such precision is generally not possible for any individual patient. Given the current interest in evidencebased medicine, however, and ongoing efforts to better define short-term risks for many conditions, this may pose somewhat less of a problem in the future. Nevertheless, we do not mean to suggest that patients who present with chest pain consistent with AMI be read a set of numbers regarding risk associated with various choices and then simply be asked to decide whether to be admitted to the hospital. Information on which to base individual probability estimates is insufficient, as is the validation of methods to ensure that patients can be made to truly understand the various options. Perhaps most important, patients, physicians, and society at large must determine the acceptable level of risk and the cost, recognizing that risk-free alternatives do not exist. 37 A reasonable paradigm for medical decisionmaking would be a process whereby its successful implementation leads to a course of action that would be chosen by a patient if he or she had the knowledge and experience of a caring and informed physician. Medical decisions would be limited only by predetermined societal guidelines regarding such issues as cost, "reasonable risk," and patient competence. This approach suggests patient-centered decisionmaking that truly reflects patient wishes rather than patient "autonomy," which as usually practiced is subject to bias and incomplete patient understanding. This paradigm allows for differences in patient circumstance and variable desire on the part of patients to participate directly in decisionmaking. In some cases full patient education should be feasible. When this is not possible, a decisionmaking procedure that involves indirect measures for a patient's preferences could be used. Further study is needed to identify and validate a practical framework on which such an approach might rest. In summary, chest pain patients in the ED appear to be more likely than physicians to be willing to assume the risk of poor outcome associated with discharge from the hospital. However, many patients cannot identify the risks associated with their disposition preferences.
ANNALS OF EMERGENCY MEDICINE
28:6
DECEMBER 1996
CHEST PAIN Davis et al
REFERENCES 1. Redelmeier RA, Rozin P, Kahneman O: Understanding patients' decisions: Cognitive and emotional perspectives. JAMA 1993;270:72-76. 2. Coulton CJ: Decision making in support: Patient perceptions and preferences. J Clin Epidemiol 1990;43 (suppl):51 54. 3. Degner LF, Sloan JA: Decision making during serious illness: What role do patients really want to play? J Clin Epidem/el1992;45:941-950. 4. Strull WM, Lo B, Charles G: Do patients want to participate in medical decision making? JAMA 1984;252:2990-2994. 5. EndeJ, Kazis L, Ash A, at ah Measuring patients' desire for autonomy: Decision making and ]nformatiomseeking preferences among medical patients. J Gen InternMad 1989;4:23-30. 6. CassiIeth BR, Zupkis RV, Sutton-Smith K, et ai: Information and participation preferences among cancer patients. Ann intern Med 1980;92:832-836. 7. Spaeth GL: Informed choice, not informed consent [editorial]. OphthalmSurg1992;23:648-649. 8. Goldman L, CookEF, Brand DA, et ah A computer protocol to predict myocardial infarction in emergencydepartment patients with chest pain. N EnglJ Mad 1B88;318:797-803. 9. Brush JE, Brand DA, Acampora D, eta[: Use of the initial electrocardiogram to predict in-hospital complications of acute myocardial infarction. NEnfllJMed1985;312:1137-1141. 10. Slater DK, Hlatky MA, Mark DB, at ak Outcome in suspected acute myocardial infarction with normal or minimally abnormal admission electrocardiographic findings. Am J Cardiel 1987;60:766-770. 11. FesmireFM, Percy RF, Wears RL, at ah Risk stratification according to the initial electrocardiogram in patients with suspected acute myocardial infarction. Arch InternMad 1989;149:12941297. 12. Stark ME, Vacek JL: The initial electrocardiogram during admission for myocardial infarction, Arch InternMad 1987;147:843-846. 13. Fineberg HV, Scadden D, Goldman L: Care of patients with a low probability of acute myocardial infarction: Cost effectivaness of alternatives to coronary-care-unit admission. N EnglJ Mad 1984;310:130i-1387. 14. FuchsR, Scheidt S: Improvedcriteria for admission to cardiac care units. JAMA 1981;246:2037-2041. 15. Jonsbu J, Aase O, Rollag K, et ak Prospective evaluation of EDB-baseddiagnostic program to be used in patients admitted te hospital with acute chest pain. EurHeartJ 1993;14:441-446. 16, Aase O, Jonsbu J, Liestrol K, et ak Decision support by computer analysis of selected case history variables in the emergencyroom among patients with acute chest pain. EurHeartJ 1993;14:433-440.
27. Malenka DJ, Baron JA, Johansen S, at al: The framing affect of relative and absolute risk. J Pen InternMad 1993;8:543-548. 28. Mazur DJ, Merz JF: How the manner of presentation of data influences older patients in determining their treatment preferences.JAm GeriatrSac 1993;41:223-228. 29. Schneider SL: Framing and conflict: Aspiration level contingency, the status qua, and current theories of risky choice. J ExpPsycho/1992;18:1040-1057. 30. SenskyT, Catalan J: Asking patients about their treatment: Why their answers should not always be taken at face varua. BMJ1992;305:1109-1110. 31. Brock DW, Wartman SA: When competent patients make irrational choices, N £nglJ/i/led 1990;322:1595-1599. 32. Nease RF, KneelandT, O'ConnorG~, et ah Variation in patient utilities for outcomes of the management of chronic stable angina: Impfications for clinical practice guidelines. JAMA 1995;273:1185-1190. 33. Sox H, B[at M, Higgins M, et ak Medical DecisionMaking. Boston: Butterworths, 1988. 34. Randall T: Producersof videodisc programs strive to expand patient's role in medical decision-making process. JAMA 1993;270:160q62. 35. Finkebtein D, Smith MK, Faden F: Informed consent and medical ethics. Arch Ophthalmol 1993;111:324-325. 36. Kassirer JP: Adding insult to injury: Usurping patients' prerogatives. N EnglJ Mad 1983;308:8%-901. 37. Keeney RL: Decisions about fife-threatening risks. NEnglJMed1994;331:193-196. The authors thank William Mower, MD, for his assistance with experiment design and statistical analysis and Geoffrey Rutledge, MD, PhD, for assistance with statistical analysis.
Reprint no. 47/1/77994 Address for reprints: Marl< A Davis, MD Division of Emergency Medicine Beth Israel Hospital/Harvard Medical School 330 Brookline Avenue Boston, Massachusetts 02215
17. PozenMW, Agostino RB, Selker HP, at al: A predictive instrument to improve coronary-careunit admission practices in acute ischemic heart disease. N EnglJ Mad 1984;310:1273-1278. 18. Goldman L, Weinberg M, Weisberg W, et ah A computer-derived protocol to aid in the diagnosis of emergencyroom patients with acute chest pain. N EnglJ Marl 1982;307:588-596. 19. @iblerWB, RunyonJP, Levy RC, et ah A rapid diagnostic and treatment center for patients with chest pain in the emergencydepartment. Ann EmergMad 1995;25:1-8. 20. Lee TH, Cook EF, Weisberg M, et ah Acute chest pain in the emergency room: Identification and examination of low-risk patients. Arch Intern Mad 1985;145:65-69. 21. Sackett SL, Haynes RB, (]uyatt GH, et al: ClinicalEpidemiology,ed 2. Boston: Little, Brown, 1991. 22. Cohen BJ, PaukerSG: Hew do physicians weigh iatrogenic comgricat]ons?J Gen Intern Mad 1994;9:20-23. 23. Elstein AS, Holzman GB, Ravltch MM, et ah Comparisonof physicians' decisions regarding estrogen replacemant therapy for menopausal women and decisions derived from a decision analytic model Am J Mad 1!]86;80:246-258. 24. Schroth WS, Riegelman BK, Blacklow BG: Medical student's risk-taking behavior in "patient" versus "physician" decision frames: Implications for medical education. AcadMad 1994;69 (suppl):54-56. 25. Weisman CS, Morlock LL, Teitelbaum MA, et ak Practice changes in responseto the malpractice litigation climate: Results of a Maryland physician study. Mad Care1989;27:16-24. 26. Mazur DJ, Hickam OH: Interpretation of graphic data by patients in a generai medicine clinic. J GenIntern Mad 1990;5:402-405.
DECEMBER 1896
28:6
ANNALS OF EMERGENCY MEDICINE
611
AdmissionDecisions in Emergency Department Chest Pain Patients at Low Risk for Myocardial Infarction: Patient Versus Physician Preferences From the UCLA Emergency Medicine Center* and the Departments of Medicine": and Pediatrics~, UCLA 5chooI of Medicine, Los Angeles, California. Receivedfor publication January 10, I996. Revision received June 7, i996. Acceptedfor publication Augctst i2, 1996. Presented at the Societyfor Academic Emergency MedicineAnnual Meeting, San Antonio, Texas, May I995. Copyright © by the American College of Emergency Physicians.
Mark A Davis, MD, MS** Amanda Keerbs* Jerome R Hoffman, MA, MD, ** Larry J Baraff, MD *§
See related editorial, p 702. Study objective: Patient involvement in medical decisionmaking is acceptedas an ethical and a legal imperative. Medical decisions are based in part on individuals' knowledge and acceptance of risk of adverse consequences.It is unclearwhether actionstaken to protect against low risk of poor outcome reflect patient or physician preferences. We sought to test the hypothesis that emergency department chest pain patients presented with a hypothetical situation involving a low risk of myocardial infarction are more willing than ED physicians to accept the risk associated with discharge from the hospital. Methods: We prospectively surveyed 89 ED patients with chest pain and a cohort of physicians in the ED who had been presented a hypothetical case in which the risk of AMI was quoted as 5% and the risk of death or disability if the patient was discharged was 1% and .2% if the patient was admitted. All the patients had presented to the ED with a chief complaint of chest pain; the 31 physicians, all residents, were approached at a teaching conference separate from their clinical duties. Results: Twenty-eight patients (31%), compared with 2 physicians (6%), chose discharge for the hypothetical patient with chest pain (25% difference; 95% confidence interval [Cl], 6% to 41%). Forty-four patients (49%), compared with 30 physicians (97%), correctly identified the risks associated with admission and discharge(46% difference; 95% CI, 29% to 63%). Of the subjects who correctly identified the risks, 19 patients (43%} preferred discharge, compared with 1 physician (3%)(40% difference; 95% CI, 18% to 60%). Conclusion: ED patients with chest pain appear to be more likely than physicians to accept a small risk of poor outcome in a hypothetical circumstance. Many patients cannot identify the risks associated with their decision.
606
ANNALS OF EMERGENCY MEDICINE
28:6
DECEMBER1996
CHEST PAIN Davis et al
[Davis MA, Keerbs A, Hoffman JR, Baraff LJ: Admission decisions in emergency department chest pain patients at low risk for myocardial infarction: Patient versus physician preferences. Ann EmergMed [:)ecember 1996;28:606-611.]
INTRODUCTION Ethical and legal considerations have recently prompted a departure from the traditional paternalistic model of medicine in favor of patient involvement in health care decisions, z Shared decisionmaking is a complex process in which the patient must recognize that a choice is needed, be aware of alternatives~ have the ability and the freedom to assess consequences, and be able to implement choices. 2 Patients vary in their desire to participate in medical decisionmaking3-6, but "informed choice "7 is impossible if information on treatment alternatives is unavailabie or is not effectively communicated to the patient. Protocols i~ve been derived that can characterize a population of patients with chest pain for whom the risk of acute myocardial infarction (AMI) is less than 7%, with a rate of emergency complication of .35% when protocol and physician agree that CCU admission is not required, s Additional clinical and electrocardiographic criteria available at the time of presentation may identify patients at even lower risk of AMI and emergency complication. 9q2 This has led to non-ICU admissions for canny patients deemed to be at low risk for AM113-1s and, more recently, to the development of shortstay chest pain centers. ~9 In the United States a minority of patients admitted to acute care hospitals because of chest pain prove to have AMI. 2° This reflects the fact that it is impossible to rule out AMI with 100% certainty in any patient at the time of presentation. It ts not known whether this practice is based on a lack of physician knowledge of risks involved or due to the preferences of physicians or patients. We devised this study to test the hypothesis that--when presented with a hypothetical circumstance in which the risk of AMI is low (5%), with a risk of death or serious disability of .2% with hospitalization and 1% with discharge-patients would be more likely than physicians to prefer discharge with outpatient follow-up. We further hypothesized that this would be true whether or not patients could identify the quantitative likelihood of adverse outcomes presented in the scenario. Finally, we hypothesized that patients and physicians would cite different factors as most important with regard to hospitalization or outpatient care, with risk of bad ,outcome being the most important factor to physicians but not to patients.
DECEMBER 1996
28:6
ANNALS OF EMERGENCY MEDICINE
MATERIALS AND METHODS We interviewed a convenience sample of 89 patients who presented to the UCLA ED with a chief complaint of chest pain between September 1994 and February 1995 during times when trained research assistants working on the study were available. A patient was eligible to participate if a physician or nurse believed the patient's chest pain might be of cardiac origin and required cardiac monitoring in the ED. Other inclusion criteria were age greater than 25 years, fluency in English, and willingness to participate. Patients were excluded if they demonstrated obvious trauma or physical or mental incapability or if the intensity of the medical management activities was believed by the interviewer to preclude participation. All subjects were verbally informed of the purpose of the research and that their participation was voluntary. Written informed consent was waived by the institutional review board because this research involved only voluntary interviews. The information collected was recorded in such a manner that the participants could not be identified and the disclosure of responses could not conceivably be expected to place subjects at risk of criminal or civil liability or to be damaging to their financial standing, employability, or reputation. The physician subject cohort comprised 19 senior emergency medicine and 12 senior internal medicine residents who, in the course of training, regularly evaluated and made disposition decisions for ED chest pain patients under the
Figure. Summary of a scenariofor Zow-dsk chest pain. The emergency department at UCLA is studying methods to improve health care delivery and patient satisfaction with medicat care. We are interested in determining patients' choices for in-hospital versus out-of-hospital testing for possible heart attack. The example that follows does not apply to you personally. This is a research study only, is not being used by your physicians to care for you, and should not influence any decisions. Suppose that you are in the emergency department and you are told the following: There is a 5% (I in 20) chance that you are having a heart attack. There is a ,2% (1 in 500)chance that you will die or have serious disability if you stay in the hospital. There is a 1% (1 in 100)chance that you will die or have serious disability if you go home, If you were to stay in the hospital, you would undergo testing and observation during your hospital stay. If you were to go home, you would be given an appointment to return to the hospital within the next few days to meet with a cardiologist for further testing. In the event of an emergency or recurrence of symptoms you would call 911. The hospital charges approximately $3,000 for 1 day of hospital monitoring and tests. If you were to go home, the costs for testing and seeing a cardiologist would be approximately $1,500,
607
CHEST PAIN Davis et al
supervision of attending emergency physicians. Physician subjects were presented with the hypothetical scenario and interviewed at a teaching conference separate from their clinical duties. Subjects in both groups were given a one-page hypothetical scenario involving a low-risk chest pain patient. The scenario provided treatment alternatives with associated risks and costs in a narrative and review format (Figure). This scenario was read aloud to each subject in the two groups. Patient subjects received a study questionnaire, which was read aloud to them by research assistants, who also recorded the patients' responses. Physician subjects read and completed the questionnaire on their own. Patient subjects were instructed to answer the questionnaire as if they were in the situation described in the hypothetical scenario, although they were advised that the case presented in the scenario was hypothetical and did not apply to their own current situation. Research assistants were instructed to tell any patient who asked questions regarding his or her own case after reading the scenario that such questions should be addressed to the patient's physician and to restate that the hypothetical case was neither designed nor intended to apply to the patient's case. Physician subjects were asked to assume while answering the questionnaire that the risks presented in the scenario applied to a patient under their care. Table 1.
Subject identification of risks associated with discharge and admission.
Each subject was asked to indicate whether he or she would choose admission or discharge with outpatient follow-up for the hypothetical patient. Physician subjects were also asked whether they would personally prefer admission or discharge if they faced as patients the risks presented for the hypothetical case. Every subject was asked to identify the risks associated with each course of action (the answers to which were listed on the scenario, which was still in his or her possession) as two multiple-choice questions with possible answers being .2%, 1%, 5%, 10%, and "don't know." Finally, each subject was asked to indicate the level of importance of patient preference, cost, risk of death or disability, and physician preference in their decision. The hypotheses that patients would be more likely than physicians to prefer discharge and that physicians wou]d be more likely to correctly identify risk information were each tested with the use of Fisher's exact test. The hypotheses that patients and physicians would cite different factors as most important to them in choosing hospitalization or outpatient care, with risk being most important to physicians and less so to patients, was tested with Fisher's exact test with extensions. Given the lack of prior information on the likely percentage of patients who might prefer discharge or admission in a scenario involving low-risk chest pain, we had no basis on which to perform power calculations to determine the sample size needed to find specific differences in disposition preference. Therefore no single cutoff for "significance" was chosen, beyond the arbitrary nature of statistical testing in Table 2.
Risk (%) Discharge 0 .2 1t 5 10 Don't know Admission 0 .2~ 1 5 10 Don't know Both risks correct
No. of Patients (%) [n=89]
No. of Physicians (%) [n=31]
2 (2.2) 10(11) 48 (54) 11 (12) 7 (7.9) 11 (12)
0 0 30 {97) 1 (3) O 0
7 (8) 54 (61) 5 (5.6) 6 (6.7) 1 (1.0) 16 {18) 44 (49)
0 31 (100) O 0 0 0 30 (97)
p,
Group <.0001
<.0001
<.0001
*Fisher's exact test. tDenotes correct responses as presented in the chest pain scenario. Subjects held the scenario with the correct responses as they answered interviewer questions.
608
Preference for discharge or admission for chest pain scenario patient.
All subjects No. of subjects No. preferring discharge for scenario patient (%) No. preferring admission for scenario patient Subjects correctly identifying risks of discharge and admission No. of subjects No. preferring discharge for scenario patient (%) No. preferring admission for scenario patient (%)
Patients
Physicians
89 28 (31)
31 2 (6)
61 (69)
29 (94)
44 19 (43)
30 1 (3.0)
25 (57)
29 (97)
P*
<,01
<001
*Fisher's exact test.
ANNALS OF EMERGENCY MEDICINE
28:6
DECEMBER 1996
CHEST PAIN Davis et al
general. 21 We chose to present absolute P values and 95% confidence intervals (CIs) for multiple-comparison data. RESULTS
The 89 subject patients included 46 men and 43 women; mean and median ages were each 61 years. Subject patients came from 155 chest pain patients approached by research assistants, of whom 66 were excluded for the following reasons: 23 did not speak English, 36 refused or were incapable of participating, 1 could not complete the questionnaire because of intense medical management activities, and 6 failed to complete the questionnaire. Seventy-seven of the subject patients were themselves ultimately admitted to the hospital. Of the 73 patients for whom dischage summades were available, 63 had a discharge diagnosis that included coronary artery disease, of whom 10 had AMI ruled in. The physician subject group comprised 31 resident physicians, 19 in emergency medicine and 12 in internal medicine. Only half of the 89 patients correctly answered both multiple-choice questions regarding the risks associated with admission or discharge, compared with virtually all the physician subjects (46% difference; 95% CI, 29 to 63%; Table 1). Patients were more likely than physicians to choose discharge for the hypothetical patient presented in the case scenario (31% versus 6%; 25% difference; 95% CI, 6% to 41%). This was also true when the responses of only those subjects who correctly identified all risks were counted (Table 2). Seventy-sevelq of the patients were themselves ultimately admitted to the hospital. Of the 73 patients for whom discharge summaries were available, 63 had a discharge diagnosis including coronary artery disease. Of the 63, 10 were found to have sustained an AMI. When physicians were asked whether they would prefer to be admitted if they faced the risks in the hypothetical
scenario, three altered their responses. These three had chosen to admit the hypothetical patient but said they would choose discharge for themselves. Patients attached greater importance to patient preference and cost than did physicians. All patients and physicians said they believed physician preference was very important or somewhat important, and we detected no significant difference between the two groups with regard to this factor. Physicians identified risk as more important in decisionmaking than did patients (Table 3). DISCUSSION
Our subject patients and physicians differed in their disposition preferences for a hypothetical low-risk patient presenting to an ED with chest pain. Physicians were significantly less likely to prefer discharge with outpatient follow-up. Physicians were better able to identify correctly the risks presented in the case scenario, but the greater willingness of patients to choose a riskier disposition (discharge) remained present even among only that subgroup of subjects (patients and physicians) who correctly identified those risks. This finding suggests that the difference in preference between the two groups related not just to differential comprehension of risk but to a difference in underlying comfort with risk-taking, as well as other factors. Patients and physicians differed in the importance they attached to several factors related to their decisions. It is not clear why physicians might be less willing than patients to take the risks associated with discharge than were the patients themselves. We can only speculate that such factors as physician experience with bad outcomes in other (real) patients, training biases resulting in avoidance of error rather than analysis of net benefit22-24, and fear of legal consequences25 play roles. Physicians also placed less importance than did patients on cost as an important factor in their choices. Although
Table 3.
Importance of var/ousfactors in decisionmakingby patients and physicians. Very Important (%) Factors
Patient
Risk 73 Patient preference 67 Cost 27 Physician recommendation 84 *Not correctedfor multiplecomparisons,
Somewhat Important (%)
Small Importance (%)
Not Important (%)
Physician
Patient
Physician
Patient
Physician
Patient
Physician
Uncorrected P*
81 35 0 74
8 13 17 16
19 39 10 26
10 I1 17 0
0 26 71 0
9 8 39 0
0 0 19 0
.025 .0007 <,0001 .28
DECEMBER 1996 28:6 ANNALS OF EMERGENCY MEDICINE
6 O9
CHEST PAIN Davis et aI
IIIIII
physicians have recently been exposed to the increasing importance of cost-efficient care in a societal sense, it is perhaps not surprising that they are uncomfortable factoring in economic considerations when making potentially lifeand-death decisions about individual patients. Physicians also placed less significance on patient preference than did patients, although most physicians did say they believed it was at least somewhat important. Patients and physicians attached great significance to the opinion of the physician. Our study has several clear limitations. Although patients were queried during an actual episode of chest pain, for which most were hospitalized and in several cases proved to be the result of AMI, they were made aware that the case scenario was hypothetical and unrelated to their own situation. It is possible that patients would be more cautious about accepting any increase in risk (even if quite small) if and when such risk actually applied to them. It is our experience, however, that patients are often willing to take such risks in actual practice and frequently stress other personal priorities when discussing possible hospitalization. Similarly, physicians were removed from actual patient care when they participated in the study, and they too might have chosen differently when confronted with an actual patient. We believe it most likely, however, that they would be even more cautious if dealing with a real patient. The scenario and questions provided to subjects were intended to highlight risks associated with disposition decisions, with no mention of probability of favorable outcome, and are therefore subject to the effects of framing bias. 26-29 Had we emphasized the expectation of positive results (99% if discharged, 99.8% if admitted), it is likely that more subjects would have chosen the higher risk alternative (discharge), although it is unclear whether patients and physicians would have done so at the same rate. In addition, no formal analysis of patients' psychologic state 3°, rationality 31, or identification of 'terrors in thinking" that did not maximize expected utility 3~ was performed. More sophisticated educational techniques might have yielded greater patient understanding 3>34, although the question of whether ED patients can be educated to truly fulfill the requirements for informed consent requires further clarification. 35,36 There are also limitations to the external validity of our study. We investigated patient preferences with regard to a low-risk scenario only; therefore our findings should not be generalized to situations of higher risk. Our physician subjects were residents. Most had less than 4 years of clinical experience as physicians, they were generally younger than the patients for whom they cared, and all were actively involved in formal training at the time of the study. It is
6 10
possible that physicians with greater experience and those in other practice environments would approach the choices presented in a different fashion. Finally, our study assumed a situation in which risks involved in the hypothetical scenario could be accurately defined for each of the alternative dispositions. Such definition is possible m only a few areas of medicine. Even with the extensive information available on chest pain, for example, such precision is generally not possible for any individual patient. Given the current interest in evidencebased medicine, however, and ongoing efforts to better define short-term risks for many conditions, this may pose somewhat less of a problem in the future. Nevertheless, we do not mean to suggest that patients who present with chest pain consistent with AMI be read a set of numbers regarding risk associated with various choices and then simply be asked to decide whether to be admitted to the hospital. Information on which to base individual probability estimates is insufficient, as is the validation of methods to ensure that patients can be made to truly understand the various options. Perhaps most important, patients, physicians, and society at large must determine the acceptable level of risk and the cost, recognizing that risk-free alternatives do not exist. 37 A reasonable paradigm for medical decisionmaking would be a process whereby its successful implementation leads to a course of action that would be chosen by a patient if he or she had the knowledge and experience of a caring and informed physician. Medical decisions would be limited only by predetermined societal guidelines regarding such issues as cost, "reasonable risk," and patient competence. This approach suggests patient-centered decisionmaking that truly reflects patient wishes rather than patient "autonomy," which as usually practiced is subject to bias and incomplete patient understanding. This paradigm allows for differences in patient circumstance and variable desire on the part of patients to participate directly in decisionmaking. In some cases full patient education should be feasible. When this is not possible, a decisionmaking procedure that involves indirect measures for a patient's preferences could be used. Further study is needed to identify and validate a practical framework on which such an approach might rest. In summary, chest pain patients in the ED appear to be more likely than physicians to be willing to assume the risk of poor outcome associated with discharge from the hospital. However, many patients cannot identify the risks associated with their disposition preferences.
ANNALS OF EMERGENCY MEDICINE
28:6
DECEMBER 1996
CHEST PAIN Davis et al
REFERENCES 1. Redelmeier RA, Rozin P, Kahneman O: Understanding patients' decisions: Cognitive and emotional perspectives. JAMA 1993;270:72-76. 2. Coulton CJ: Decision making in support: Patient perceptions and preferences. J Clin Epidemiol 1990;43 (suppl):51 54. 3. Degner LF, Sloan JA: Decision making during serious illness: What role do patients really want to play? J Clin Epidem/el1992;45:941-950. 4. Strull WM, Lo B, Charles G: Do patients want to participate in medical decision making? JAMA 1984;252:2990-2994. 5. EndeJ, Kazis L, Ash A, at ah Measuring patients' desire for autonomy: Decision making and ]nformatiomseeking preferences among medical patients. J Gen InternMad 1989;4:23-30. 6. CassiIeth BR, Zupkis RV, Sutton-Smith K, et ai: Information and participation preferences among cancer patients. Ann intern Med 1980;92:832-836. 7. Spaeth GL: Informed choice, not informed consent [editorial]. OphthalmSurg1992;23:648-649. 8. Goldman L, CookEF, Brand DA, et ah A computer protocol to predict myocardial infarction in emergencydepartment patients with chest pain. N EnglJ Mad 1B88;318:797-803. 9. Brush JE, Brand DA, Acampora D, eta[: Use of the initial electrocardiogram to predict in-hospital complications of acute myocardial infarction. NEnfllJMed1985;312:1137-1141. 10. Slater DK, Hlatky MA, Mark DB, at ak Outcome in suspected acute myocardial infarction with normal or minimally abnormal admission electrocardiographic findings. Am J Cardiel 1987;60:766-770. 11. FesmireFM, Percy RF, Wears RL, at ah Risk stratification according to the initial electrocardiogram in patients with suspected acute myocardial infarction. Arch InternMad 1989;149:12941297. 12. Stark ME, Vacek JL: The initial electrocardiogram during admission for myocardial infarction, Arch InternMad 1987;147:843-846. 13. Fineberg HV, Scadden D, Goldman L: Care of patients with a low probability of acute myocardial infarction: Cost effectivaness of alternatives to coronary-care-unit admission. N EnglJ Mad 1984;310:130i-1387. 14. FuchsR, Scheidt S: Improvedcriteria for admission to cardiac care units. JAMA 1981;246:2037-2041. 15. Jonsbu J, Aase O, Rollag K, et ak Prospective evaluation of EDB-baseddiagnostic program to be used in patients admitted te hospital with acute chest pain. EurHeartJ 1993;14:441-446. 16, Aase O, Jonsbu J, Liestrol K, et ak Decision support by computer analysis of selected case history variables in the emergencyroom among patients with acute chest pain. EurHeartJ 1993;14:433-440.
27. Malenka DJ, Baron JA, Johansen S, at al: The framing affect of relative and absolute risk. J Pen InternMad 1993;8:543-548. 28. Mazur DJ, Merz JF: How the manner of presentation of data influences older patients in determining their treatment preferences.JAm GeriatrSac 1993;41:223-228. 29. Schneider SL: Framing and conflict: Aspiration level contingency, the status qua, and current theories of risky choice. J ExpPsycho/1992;18:1040-1057. 30. SenskyT, Catalan J: Asking patients about their treatment: Why their answers should not always be taken at face varua. BMJ1992;305:1109-1110. 31. Brock DW, Wartman SA: When competent patients make irrational choices, N £nglJ/i/led 1990;322:1595-1599. 32. Nease RF, KneelandT, O'ConnorG~, et ah Variation in patient utilities for outcomes of the management of chronic stable angina: Impfications for clinical practice guidelines. JAMA 1995;273:1185-1190. 33. Sox H, B[at M, Higgins M, et ak Medical DecisionMaking. Boston: Butterworths, 1988. 34. Randall T: Producersof videodisc programs strive to expand patient's role in medical decision-making process. JAMA 1993;270:160q62. 35. Finkebtein D, Smith MK, Faden F: Informed consent and medical ethics. Arch Ophthalmol 1993;111:324-325. 36. Kassirer JP: Adding insult to injury: Usurping patients' prerogatives. N EnglJ Mad 1983;308:8%-901. 37. Keeney RL: Decisions about fife-threatening risks. NEnglJMed1994;331:193-196. The authors thank William Mower, MD, for his assistance with experiment design and statistical analysis and Geoffrey Rutledge, MD, PhD, for assistance with statistical analysis.
Reprint no. 47/1/77994 Address for reprints: Marl< A Davis, MD Division of Emergency Medicine Beth Israel Hospital/Harvard Medical School 330 Brookline Avenue Boston, Massachusetts 02215
17. PozenMW, Agostino RB, Selker HP, at al: A predictive instrument to improve coronary-careunit admission practices in acute ischemic heart disease. N EnglJ Mad 1984;310:1273-1278. 18. Goldman L, Weinberg M, Weisberg W, et ah A computer-derived protocol to aid in the diagnosis of emergencyroom patients with acute chest pain. N EnglJ Marl 1982;307:588-596. 19. @iblerWB, RunyonJP, Levy RC, et ah A rapid diagnostic and treatment center for patients with chest pain in the emergencydepartment. Ann EmergMad 1995;25:1-8. 20. Lee TH, Cook EF, Weisberg M, et ah Acute chest pain in the emergency room: Identification and examination of low-risk patients. Arch Intern Mad 1985;145:65-69. 21. Sackett SL, Haynes RB, (]uyatt GH, et al: ClinicalEpidemiology,ed 2. Boston: Little, Brown, 1991. 22. Cohen BJ, PaukerSG: Hew do physicians weigh iatrogenic comgricat]ons?J Gen Intern Mad 1994;9:20-23. 23. Elstein AS, Holzman GB, Ravltch MM, et ah Comparisonof physicians' decisions regarding estrogen replacemant therapy for menopausal women and decisions derived from a decision analytic model Am J Mad 1!]86;80:246-258. 24. Schroth WS, Riegelman BK, Blacklow BG: Medical student's risk-taking behavior in "patient" versus "physician" decision frames: Implications for medical education. AcadMad 1994;69 (suppl):54-56. 25. Weisman CS, Morlock LL, Teitelbaum MA, et ak Practice changes in responseto the malpractice litigation climate: Results of a Maryland physician study. Mad Care1989;27:16-24. 26. Mazur DJ, Hickam OH: Interpretation of graphic data by patients in a generai medicine clinic. J GenIntern Mad 1990;5:402-405.
DECEMBER 1896
28:6
ANNALS OF EMERGENCY MEDICINE
611