How much risk are emergency department patients willing to accept to avoid diagnostic testing.

OBJECTIVES: There is a paucity of research evaluating the risk tolerance of Emergency Department (ED) patients. We hypothesized that a significant percentage of ED patients surveyed would be comfortable with ≥5% risk of adverse outcome if they avoided testing in several hypothetical scenarios.
MATERIALS AND METHODS: This was a cross-sectional study of a convenience sample of stable inner-city ED patients. Patients completed a written survey and were asked four closed-answer questions regarding risk tolerance/willingness to refuse a test/procedure, including the following scenarios: #1: LP following CT head; #2 overnight serial troponins for rule out myocardial infarction; #3 CT for possible appendicitis, #4 parent deciding whether child should undergo head CT for low risk head injury. Risk preferences were stratified to ≥5% (high) and <5% (low). Multivariate logistic regressions performed for each scenario to control for confounding factors.
RESULTS: There were 217 patients in the study group; mean age 42 ± 15 years, 48% female, 66% Hispanic, 87% income < $40,000 income group. A substantial percentage of patients rated high risk tolerance for each scenario, including 31% for #1 (avoid LP), 25% for #2 (avoid cardiac rule-out admission), 27% for #3 (avoid CT for appendicitis), and 19% for #4 (avoid head CT for child). DISCUSSION: For 3 of 4 scenarios, 25% or more of our patients would accept a high risk tolerance of adverse outcome to avoid further testing.
CONCLUSION: Our findings contribute further evidence to the growing body of literature supporting patient interest in shared decision making in the ED.

Introduction

Despite higher spending per capita, research suggests that the US does not provide its citizens with objectively better quality and access to health care. Kanzaria et al note that unnecessary testing and unwarranted hospital admissions are a reflection of physician risk intolerance that contributes to rising health care costs. Further, they contend that physicians are significantly influenced by fear of litigation when they order extensive testing for patients.

While there are a significant number of studies that focus on the risk tolerance of physicians, there is a paucity of information regarding the risk tolerance of patients, particularly with respect to care they receive in the emergency department (ED) setting.2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 From the standpoint of autonomy, beneficence, and cost, it is important to understand whether physicians are ordering tests to reduce risk well beyond what patients might otherwise tolerate. We conducted a cross-sectional study to test the hypothesis that a significant percentage of patients surveyed in our ED would be comfortable with ≥ 5% risk of adverse outcome if they avoided testing within several hypothetical scenarios.

Materials and methods

Study design

This was a prospective, cross-sectional study of emergency department patients designed to assess their risk tolerance for adverse events, including death, when presented with 4 different common clinical scenarios in the ED.

Setting

The study was conducted in the Emergency Department at Christus Spohn Memorial Hospital (Corpus Christi, Texas, USA). The facility is a teaching affiliate of the Texas A&M Health Science Center, a level-two trauma center, and serves an inner-city population. The annual Emergency Department census is 45,000 patients. The study was reviewed and provided with exempt status by the Christus Spohn Institutional Review Board prior to initiation of patient enrollment.

Population

All consenting, medically stable, oriented, adult patients age >18 years were eligible for inclusion into the study group. Patients were excluded for any of the following reasons: cognitively impaired, medically unstable, or if they refused to provide verbal consent. Our ED census is represents a predominantly Hispanic, indigent, inner-city, population.

Study protocol

Consenting patients were consecutively enrolled at times when trained research associates were available to assist with the study protocol (convenience sample). Such hours were variable but did include normal business type hours, evenings, overnights, and weekends. Demographic/historical features were recorded on a structured form (Fig. 1). Among the variables collected, we asked patients to provide insurance information. In the U.S., while many patients have private insurance, government sponsored insurance is made available to the poor (Medicaid) and to the elderly (Medicare). We provided a written survey in English to the patients that included 4 common clinical scenarios of ED patients who were undergoing evaluation for chief complaints in which the differential diagnosis included conditions with high rates of morbidity/mortality for missed diagnosis. Patients filled out their survey responses in writing. The survey instrument was developed by the senior investigator who has extensive experience in the design of emergency medicine-based survey instruments.

Fig. 1

Survey instrument.

The scenarios then revealed detailed descriptions of each potential step in management, including tests and procedures (see Appendix 1). The risks and benefits of those were related as well. Finally, each scenario detailed potential adverse outcomes if tests/procedures were refused. Scenario #1 was in the setting of the patient with headache and possible SAH with indeterminate head CT necessitating lumbar puncture Scenario #2 involved chest pain, use of serial cardiac enzyme assays, and subsequent stress testing with hospital admission or observation Scenario #3 involved undifferentiated abdominal pain (posed as potential appendicitis) and recommended CT imaging for further diagnostic assurance The optional fourth scenario involved obtaining a head CT for a child after minor head trauma. The last scenario (#4) was listed as optional, because it has been our practical experience that patients in our setting are more likely to not participate if survey material is more lengthy.

Statistical analysis

Data was entered into STATA (College Station, TX) for analysis. Variables were coded in binary fashion (with dependent variables being high or low risk aversion). Risk preferences were stratified to ≥5% (high) and <5% (low) for each scenario. Multivariate logistic regression was utilized to control for confounding factors. We calculated odds ratios and 95% confidence intervals. The primary outcome parameter of the study was to test the null hypothesis that patients would not accept a high risk preference (≥5%) for each hypothetical clinical scenario.

Results

A total of 217 patients were in the final study group. Our ED population is relatively young, poor, and predominantly Hispanic. Their characteristics are summarized in Table 1.

Table 1

Study group characteristics (N = 217).

Mean age (years)	42 ± 15
% Female	48%
% Hispanic	66%
% Income < $40,000/annual	87%

Within the study group, 31% of patients rated high risk tolerance to avoid LP after negative CT in the evaluation for subarachnoid hemorrhage (scenario #1). For scenario #1, the results of multivariate logistic regression are summarized in Table 2. Individuals whose annual income was > $40,000/yr (OR 4.1; 95 CI 1.5–11.1) and those with Medicaid (OR 3.4; 95% CI 1.2–9.3) were more risk tolerant.

Table 2

Scenario #1 Multivariate Regression (Refusal of lumbar puncture after negative CT).

Characteristic	Odds ratio	95% Confidence interval
% Female	0.51	0.27–0.97
Age ≥ 65 years	1.26	0.37–4.2
Hispanic race	1.74	0.84–3.6
Income > $40,000	4.1	1.51–11.1
Education > high school	0.60	0.30–1.21
Insurance: Clinic Card	1.55	0.52–4.6
Insurance: Medicaid	3.39	1.2–9.3
Insurance: Medicare	1.59	0.29–8.3
Insurance: self-pay	1.62	0.63–4.2

Table 3 summarizes the results of multivariate logistic regression for scenario #2. We observed that 25% of patients rated high risk tolerance to avoid cardiac rule-out admission. Patients with Medicaid or who were self-pay were more likely to be risk tolerant (OR 4.2; 95% CI 1.4–12.4 and OR 3.3; 95% CI 1.2–9.0).

Table 3

Scenario #2 Multivariate Regression (Refusal of cardiac “rule-out” testing, admission).

Characteristic	Odds ratio	95% Confidence interval
% Female	0.76	0.41–1.4
Age ≥ 65 years	1.26	0.39–4.0
Hispanic race	1.35	0.64–2.85
Income > $40,000	2.04	0.76–5.44
Education > high school	0.76	0.39–1.5
Insurance: Clinic Card	1.80	0.55–5.9
Insurance: Medicaid	4.2	1.4–12.5
Insurance: Medicare	1.2	0.18–8.5
Insurance: self-pay	3.3	1.2–9.0

For scenario #3 the results of multivariate logistic regression are summarized in Table 4. We found that 27% of patients rated high risk tolerance to avoid CT for appendicitis. When analyzing patient characteristics and their relationship to risk high risk tolerance, Hispanics were more likely to accept such risk (OR 2.9; 95% CI 1.2–6.5).

Table 4

Scenario #3 Multivariate Regression (CT to evaluate for possible appendicitis).

Characteristic	Odds ratio	95% Confidence interval
% Female	0.84	0.45–1.6
Age ≥ 65 years	1.1	0.30–4.2
Hispanic race	2.9	1.3–6.5
Income > $40,000	1.3	0.47–3.7
Education > high school	0.74	0.38–1.4
Insurance: Clinic Card	1.4	0.49–4.1
Insurance: Medicaid	1.3	0.50–3.5
Insurance: Medicare	0.26	0.026–2.7
Insurance: self-pay	0.90	0.35–2.3

Finally, for the optional scenario #4, which was answered by a portion of our overall study group (n = 159), 19% of patients rated high risk tolerance to have their child avoid head CT. For scenario #4 the results of multivariate logistic regression are summarized in Table 5. There were no significant associations with this risk tolerance and the independent variables.

Table 5

Scenario #4 Multivariate Regression (Parent refusal of CT head to evaluate minor head trauma for child).

Characteristic	Odds ratio	95% Confidence interval
% Female	0.99	0.46–2.1
Age ≥ 65 years	1.1	0.17–6.6
Hispanic race	1.3	0.48–3.4
Income > $40,000	1.2	0.32–4.5
Education > high school	1.6	0.68–3.8
Insurance: Clinic Card	2.4	0.60–10
Insurance: Medicaid	3.3	0.90–12.4
Insurance: Medicare	0.57	0.05–6.3
Insurance: self-pay	2.0	0.5 8–7.2

Discussion

With advancing medical research and the widespread availability of highly sensitive imaging modalities, physicians are increasingly able to reduce the chance of missed diagnosis for serious conditions. While physicians employ more of these tests to reduce the risk of adverse events, these strategies are contributing to higher health care costs. There have been numerous studies that have examined the risk related behavior of physicians.2, 3, 4, 5, 6, 7, 8

The majority of investigations specific to emergency physician decision making have confirmed the influence of malpractice and fear of uncertainty on test ordering. Wong et al surveyed 246 Michigan-based board eligible/certified emergency physicians and found that in one of two hypothetical clinical scenarios there was an association between physician fear of malpractice and increased ordering of a head CT for Pediatric minor trauma. Katz et al studied the influence of emergency physician fear of malpractice on their management of patients with possible cardiac ischemic in two academic settings. They reported that malpractice fear was associated with a higher admission rate as well as increased use of diagnostic tests such as Troponin. In a similar fashion, Kanzaria et al found that 97% of the 435 emergency physicians they surveyed routinely ordered medically unnecessary imaging for fear of missing a low-probability diagnosis or fear of litigation.

By way of contrast, Pines et al explored more deeply into the risk related behavior of emergency physicians by breaking the concept into several components, including malpractice fear, risk taking, and stress from uncertainty as measured by distinct scales. Their study involved the evaluation of 2871 ED patients with chest pain that were evaluated by any one of 31 emergency physicians. The authors found that the decision by physicians to utilize coronary CT angiography, order cardiac markers and admit patients was influenced most strongly by risk aversion rather than fear of malpractice and stress from uncertainty.

Several studies have examined a patient's tolerance for risk related to medical treatment,9, 10, 11, 12, 13 but less so as it relates to treatment in the emergency setting. In a pilot study, Brown et al evaluated the risk tolerance for adverse events for 68 patients with chest pain. The median acceptable threshold for adverse event was 6.5% This threshold is relatively consistent with our definition of 5% as the cut-off for high risk in the current study. Subsequently, Chen et al conducted a cross-sectional study of 213 patients assess emergency department (ED) patients' risk thresholds for preferring admission versus discharge when presenting with chest pain. The authors found that approximately one-quarter of patients reported their risk tolerance threshold inconsistently depending on whether the provider gave a numeric assessment of risk versus showing the patient a visual aid to define the risk.

Our ED-based study is novel in its attempt to assess risk tolerance for several different clinical scenarios. Within our study group, we found that at least 19% of patients in each of the respective 4 scenarios were in the high risk tolerant group. For 3 of the 4 scenarios, > 25% of patients were in the high risk tolerance group. These findings suggest some overall stability in risk preference for ED patients across different clinical scenarios. Recent multi-center investigation has shown that shared decision making between physicians and patients can significantly reduce the use of cardiac testing (stress test and coronary CT) when applied within the emergency department setting. Our results present opportunities to expand research on shared decision making to such areas as LP following negative CT in the work-up of subarachnoid hemorrhage, CT for patients with abdominal pain but relatively low suspicion for acute appendicitis, and CT imaging for pediatric head injury. Further study is warranted to confirm that the views of our patients would be similar to those who are actually confronted with point-of-care decision making for these symptoms.

Limitations and future directions

This study has several limitations that warrant discussion. Due to logistical constraints, we did not track potential participant refusals. Our trained research associates work variable hours and days/nights of the week such that the study group demographics are representative of the low-income, predominantly Hispanic population to which the ED serves and consistent with previous data collected at our center. While the authors took great effort in the construction of each written scenario to help patients understand the clinical implications of testing and the risk of refusal, the questions didn't necessarily correlate with the purpose of each respective patient visit to the ED that day. It may have been difficult for patients to relate to the hypothetical clinical scenario in the context of a different current symptom. Thus, we cannot be certain that patients would make the same choices if they were confronted with the similar scenarios under real circumstances.

In addition, as the survey instrument utilized here has not been previously validated, it is possible that wording of the scenarios and/or the survey questions might have also led to a different result. Along the same lines, our decision to define high risk tolerance as greater than 5% is somewhat arbitrary, though consistent with the tolerance of adverse outcome noted by Brown et al for chest pain patients (6.5%) With most emergency physicians seeing several thousand patients per year, it is hard to imagine that many would hold their “acceptable miss rate” at 50 per thousand patients seen. It is also uncertain how our results might have varied had we utilized a continuous scale for risk tolerance ranking. Patients may also have responded differently based on the degree of pain that they were respectively experiencing at the end of an encounter, but we did not provide information nor control for this possibility in our data collection/analysis. We also note that some of our demographic variables are collinear (e.g education and income level), but this is a common problem for this type of research and, nonetheless, researchers typically include these variables to control for confounding in a multivariate regression. Future investigators should also conduct larger studies to better understand the degree of significance of the independent variable associations with higher risk tolerance.

Finally, another limitation of our study is related to the educational level, which may influence their ability to understand the term “risk” and the associated short and long term complications associated with a missed diagnoses. The fact that our patients provided a level of risk tolerance does not provide us with any information regarding their comprehension. Our results may also not be generalizable to a more educated and/or wealthier population who surely would understand the scenarios and implications of risk taking in a more sophisticated fashion. Future studies should be conducted to examine this question in other diverse settings, and to examine whether physician risk tolerance would increase if they were more aware of their patients' own tolerance for risk.

Conclusions

For 3 of 4 scenarios, 25% or more of our patients would accept a high risk tolerance of adverse outcome to avoid further testing. If future studies validate our findings in other settings, there may be opportunities for physicians to engage patients in shared decision making when pre-test probability assessment is low.

Financial disclosures/conflicts

None.

Treatment/procedure	<1%	1%	2%	5%	10%	15%	Fill in blank %	Don't understand question8
Head CT

Treatment/procedure	<1%	1%	2%	5%	10%	15%	Fill in blank %	Don't understand question8
Cardiac stress Test

Treatment/procedure	<1%	1%	2%	5%	10%	15%	Fill in blank %	Don't understand question8
Abd/pelvic CT

Treatment/procedure	<1%	1%	2%	5%	10%	15%	Fill in blank %	Don't understand question8
Head CT