Abstract
A questionnaire was mailed to 300 Iowa family physicians to determine the influence of a prior psychiatric history on decision making. The response rate was 77%. Respondents were less likely to believe that a patient had serious illness when presenting with a severe headache or abdominal pain if the patient had a prior history of depression ( P < .05) or prior history of somatic complaints ( P < .05), compared with a patient with no past history. Respondents were less likely to report that they would order testing for a patient with headache or abdominal pain if the patient had a history of depression ( P < .05, P = .08, respectively) or somatic complaints ( P < .01). Differences in likelihood of ordering tests were not significant after adjusting for differences in estimated probability of disease. We conclude that physicians respond differently to patients with psychiatric illness because of their estimation of pretest probability of disease rather than bias. We conclude that past psychiatric history influences physicians' estimation of disease presence and willingness to order tests.
Keywords: bias, medical decision making, depression, somatization
To estimate the probability of disease, physicians rely on heuristics (cognitive shortcuts or rules of thumb).1,2 One frequently used heuristic is the representative heuristic.3 Estimations using the representative heuristic are based on the “representativeness” of a clinical finding or how closely the finding resembles the essential features of a population with known disease. However, by overemphasizing the resemblance factor and underemphasizing the prevalence of disease, representativeness can lead to errors. For example, a healthy 19-year-old is not likely to have cardiac disease despite having classic cardiac symptoms.
The representative heuristic allows physicians to make use of a patient's past medical history. This history might have great relevance, such as a history of myocardial infarction in a patient with chest pain. However, the past history can also be used inappropriately, with weight being given to features that are not diagnostic of the condition in question.4
This study addresses three questions. First, do physicians take psychiatric history into account when deciding whether or not a patient has a serious illness? Second, are physicians less likely to order further testing for patients with a history of psychiatric disease? Third, if physicians are less likely to order further testing, is this a function of the estimation of disease probability or does some other factor (e.g., bias) play a role?
METHODS
A questionnaire was mailed to 300 systematically chosen family physicians practicing in Iowa. Every third name was selected to reach the predetermined number of participants. Demographic information was obtained from a database including physician gender, age, and year of graduation from a family medicine residency (if applicable). Participants were asked to read 2 scenarios about patients with new, potentially serious complaints (severe headache and acute abdominal pain) that were suggestive of potentially fatal illnesses (subarachnoid hemorrhage and aortic aneurysm). Participants were randomized to 3 experimental groups. Group 1 was given no additional past medical history, group 2 was told about a past history of depression, and group 3 was told about a long history of somatic complaints without obvious organic etiology. Physicians were asked to estimate the probability of serious disease and the likelihood that they would recommend further investigations (Table 1). The information was presented in the format of history, past medical history, and physical examination that is generally used to collect clinical information. This protocol was approved by the University of Iowa Human Subjects Committee.
Table 1.
Case Scenarios and Variations
| Case Description | Experimental Conditions | Questions |
|---|---|---|
| Case 1:43-year-old female who presents tothe office with a history of a severe headachewith a sudden onset 4 days ago. She relatesno prior history of headaches and notesassociated photophobia and nausea. She alsocomplains of a stiff neck. On exam, the vitalsigns are normal and there are no focalneurologic findings. Funduscopic exam innormal. | 1. No past history included. 2. Past medical history: History of major depressive disorder. Social history: The patient reports a number of social stressors in the past several months. | 1. What is the probability that you would order further immediate investigations for this patient?2. What is the probability that this patient has a subarachnoid hemorrhage? |
| Case 2:62-year-old male presents to your officecomplaining of abdominal pain. The pain issomewhat ill defined, but on further historyyou find that the pain had a sudden onsetin the mid-abdomen and is colicky in nature. There are no other associated gastrointestinalor genitourinary symptoms. Vital signs arenormal except for a pulse rate of 100 beatsper minute. | 3. Past medical history: Multiple visits for somatic complaints for which no organic basis has been found. | 1. What is the probability that you would order further immediate investigations for this patient?2. What is the probability that this patient has a serious problem? |
Responses were summarized using means and standard deviations. An analysis of covariance (ANCOVA) model was used to test for difference among groups, adjusting for gender and year of residency completion. However, because the responses were not normally distributed, the findings were confirmed using an ordinal logistic regression model, which does not assume normality. All analyses were performed in Stata 4.0 (StataCorp, College Station, Tex, 1995).
RESULTS
Of the 300 questionnaires, 232 were returned (77% response rate).
As shown in Table 2, respondents were less likely to believe that a patient had a serious illness when presenting with a severe headache or severe abdominal pain if the patient had a prior history of depression or prior history of somatic complaints compared to a patient with no past history provided. Respondents also were less likely to report that they would order additional testing for a patient with a severe headache or abdominal pain if the patient had a history of depression (P < .05 for headache; borderline P value of .08 for abdominal pain) or a history of somatic complaints (P < .01 for either symptom) compared to a patient with no past psychiatric history.
Table 2.
Estimated Probability of Disease and Likelihood of Ordering Tests By Presenting Symptoms andType of Psychiatric History in the Case Scenarios
| Means of Physician-Reported Probabilities, % | P Values* | ||||||
|---|---|---|---|---|---|---|---|
| CaseScenario | Question | Group 1, No Past History | Group 2, Depression | Group 3, Somatic Complaints | Group 2 vs Group 1 | Group 3 vs Group 1 | Overall |
| Headache | Order more tests? | 94 (17) | 87 (22) | 80 (30) | .04 | .008 | .03 |
| Subarachnoidhemorrhage? | 39 (29) | 30 (29) | 21 (24) | .02 | <.001 | .002 | |
| Abdominal pain | Order more tests? | 90 (19) | 81 (29) | 72 (34) | .08 | .003 | .01 |
| Serious problem? | 46 (24) | 36 (23) | 30 (24) | .03 | <.001 | .002 | |
P values are adjusted for gender and year of residency completion.
Estimates of disease probability and likelihood of ordering further tests were highly correlated. Once the estimated disease probability was put into the ordinal logistic regression model, the differences across groups in the likelihood of ordering tests were no longer significant for either headache or abdominal pain (P = .19 andP = .22, respectively). This suggests that physicians respond differently to patients with psychiatric illness because of their estimation of the pretest probability of disease rather than because of bias.
DISCUSSION
Physicians make mistakes when analyzing data. Some of the factors that wrongly influence physicians' decision making include the order in which information is presented,5 the proportion of positive and negative tests in a series of tests,6 the race and gender of the patient,7 and the demeanor of the patient.8
This study identifies another factor that influences decision making, a patient's history of depression or somatic disease. While this study cannot determine the exact relation between psychiatric history, estimation of disease probability, and test-ordering behavior, it suggests that prior psychiatric history influences estimates of disease likelihood and this influences decisions about further testing. This order of decision making has been borne out in other studies.9
Patients with psychiatric disease have varied responses to their illness, and not all will have prominent somatic complaints.10,11 In this study, it was specified in the case of the headache that the patient had not had similar complaints in the past. Yet physicians still generalized their belief about depressed and somatic patients to our hypothetical patients.
It can be argued that heuristics should change the pretest probability in some illnesses. For example, most physicians would estimate the probability of colon cancer to be much lower in a 19-year-old with guaiac-positive stool than in a 65-year-old with guaiac-positive stool because of the known relation between age and colon cancer. The cases in this study are not analogous. In our study, the histories were suggestive of potentially fatal acute illnesses (subarachnoid hemorrhage, possible aortic aneurysm) that should be considered regardless of the presence or absence of psychiatric disease.
The major limitation of this study is that the patients were not physically present. However, patient presence does not guarantee an accurate assessment; the demeanor of the patient may be misleading.8 To avoid confounders, we tested the effect of psychiatric history in isolation from other clinical information. The second limitation is that physicians may act differently toward patients that they know. However, this study realistically approximates the situation of a physician on-call for a colleague's patients or that of a physician working in an emergency department. The third limitation is that the subjects in the study were family physicians. However, the potential for different findings in other groups does not invalidate the findings as they apply to this group. Given the similarity of education, the findings are likely to be generalizable to other primary care physicians.
In summary, physicians in this study modified their estimates of probability of disease and the likelihood they would pursue further diagnostic evaluations in patients with a history of depression or somatic complaints. Physicians need to be cognizant of the role that psychiatric history may play in clinical decision making.
Acknowledgments
This work was supported by a predoctoral grant from the Department of Health and Human Services (5DO5 PE87007-18).
REFERENCES
- 1.Peay MY, Peay ER. The evaluation of medical symptoms by patients and doctors. J Behav Med. 1998;21:57–81. doi: 10.1023/a:1018715521706. [DOI] [PubMed] [Google Scholar]
- 2.McDonald CJ. Medical heuristics: the silent adjudicators of clinical practice. Ann Intern Med. 1996;124:56–62. doi: 10.7326/0003-4819-124-1_part_1-199601010-00009. [DOI] [PubMed] [Google Scholar]
- 3. Nisbett R, Ross L. In: Nisbett R. Ross L. Human Inference: Strategies and Shortcomings of Social Judgment. Englewood Cliffs, NJ: Prentice-Hall, Inc; 1980:26–7.
- 4.Green LA, Yates JF. Influence of pseudodiagnostic information on the evaluation of ischemic heart disease. Ann Emerg Med. 1995;25(4):451–7. doi: 10.1016/s0196-0644(95)70257-1. [DOI] [PubMed] [Google Scholar]
- 5.Bergus GR, Chapman GB, Gjerde C, et al. Clinical reasoning about new symptoms despite preexisting disease: sources of error and order effects. Fam Med. 1995;27:314–20. [PubMed] [Google Scholar]
- 6.Egglin TK, Feinstein AR. Context bias: a problem in diagnostic radiology. JAMA. 1996;76:1752–5. doi: 10.1001/jama.276.21.1752. [DOI] [PubMed] [Google Scholar]
- 7.Schlman KA, Berlin JA, Harless W, et al. The effect of race and sex and physicians' recommendations for cardiac catheterization. N Engl J Med. 1999;340:618–26. doi: 10.1056/NEJM199902253400806. [DOI] [PubMed] [Google Scholar]
- 8.Birdwell BG, Herbers JE, Kroenke K. Evaluating chest pain: the patient's presentation style alters the physician's diagnostic approach. Arch Intern Med. 1993;53:1991–1995. doi: 10.1001/archinte.153.17.1991. [DOI] [PubMed] [Google Scholar]
- 9.Christensen-Szalanski JJ, Diehr PH, Bushyhead JB, et al. Two studies of good clinical judgment. Med Decis Making. 1982;2:275–283. doi: 10.1177/0272989X8200200303. [DOI] [PubMed] [Google Scholar]
- 10.Kirmayer LJ, Robbins JM. Patients who somatize in primary care: a longitudinal study of cognitive and social charateristics. Psychological Medicine. 1996;26:937–951. doi: 10.1017/s0033291700035273. [DOI] [PubMed] [Google Scholar]
- 11.Suh T, Gallo JJ. Symptom profiles of depression among general medical service users compared with specialty mental health service users. Psychological Medicine. 1997;27:1051–1063. doi: 10.1017/s0033291797005205. [DOI] [PubMed] [Google Scholar]