Surrogate and Patient Discrepancy Regarding Consent for Critical Care Research

Abstract

Objective

Critically ill patients frequently display impaired decision making capacity due to their underlying illness and the use of sedating medications. Health care providers often rely on surrogates to make decisions for medical care and participation in clinical research. However, the accuracy of surrogate decisions for a variety of critical care research studies is poorly understood.

Design

Cross-sectional observational study

Setting

Academic medical center

Patients

Medical ICU patients and their designated surrogates

Intervention

Patients were asked if they would consent to participate in hypothetical research studies of increasing complexity, and surrogates independently indicated whether they would consent to enroll the patient in the same scenarios.

Results

Overall, 69 MICU patients were enrolled into the study. The majority of surrogates were either the spouse (58%) or parent (22%) of the patient. The percentage of patients that would agree to participate in a research study and the percentage of surrogates that would agree to have the patient enrolled into a research study both declined as the risk of the study increased (P < 0.001 for both analyses). In addition, the overall discrepancy, the false negative rates, and the false positive rates between patient and surrogates were greater as the risk of the study increased (P <0.001, P < 0.001, and P = 0.049 respectively). Kappa values for all seven scenarios demonstrated less than moderate agreement (range 0.03–0.41).

Conclusions

There are significant discrepancies in the willingness to participate in various types of clinical research proposals between critically ill patients and their surrogate decision makers. The results of this study raise concerns about the use of surrogate consent for inclusion of critically ill patients into research protocols.

INTRODUCTION

Making clinical research available to critically ill patients has been defined as a priority for our healthcare system ¹. However, most critically ill patients have diminished decision making capacity related to cognitive impairment from their acute illness or the use of sedative agents ². Therefore, critically ill patients are usually unable to participate in the informed consent process and a legal representative or surrogate is identified on their behalf to make decisions ^3–5.

Previous studies examining the accuracy of surrogate consent for enrollment into research protocols have demonstrated that their decisions are often inaccurate ⁶. Two studies evaluating concordance between patients and surrogates for participation in hypothetical randomized clinical critical care trials reported discrepancy rates as high as 46% ^7;8. However, randomized clinical trials represent only a portion of the research opportunities available to critically ill patients. Observational epidemiological and translational studies are also conducted in intensive care units (ICUs). The results of many observational studies have resulted in significant advances in the care of our critically ill patients ^9–11. However, the accuracy of decisions made by surrogates for critical care research other than randomized clinical trials, and impact of the risk associated with the research study on the discrepancy rates between the patient and surrogate have not been studied. Therefore, we independently surveyed patients who required admission to an ICU and their surrogate to determine the accuracy of surrogate decision making for a variety of hypothetical research protocols. We hypothesized that the discrepancy between surrogate and patient decisions would increase with the risk of the research study.

METHODS

We performed a cross-sectional observational study of patients admitted to a medical ICU and their designated surrogates in a single academic university hospital. Inclusion criteria were patient age greater than 18 years and admission to the medical ICU. Patients were excluded if they did not regain decisional capacity, they were unable to read or speak English, or no surrogate was identified during the hospitalization. Patients and their surrogates provided written informed consent prior to data collection. The proposal was approved by the Colorado Multiple Institutional Review Board. This research has previously been presented in abstract form at the 2011 International American Thoracic Society meeting held in Denver, CO.

The hypothetical scenarios were constructed based on common critical care research protocols and were vetted with a group of critical care physicians at our institution. All components of each hypothetical research study were standardized and the survey was pilot tested among critical care physicians to assess for clarity, completeness, and realism of the scenarios. The description of each hypothetical research scenario was adjusted based on the results of our pilot testing. The final hypothetical research scenarios were 1) demographic and clinical data collection, 2) urine collection, 3) blood sample collection, 4) performance of CT scan, 5) skeletal muscle biopsy, 6) bronchoalveolar lavage through an existing endotracheal tube, and 7) participation in a randomized placebo controlled trial of a medication for stress gastritis. For each scenario, the survey also included a short description of the potential risks and chance of their occurrence. Based on their potential risks, the seven scenarios were listed on the survey in the same order as presented above. No potential benefit of participation was included on the survey for any of the seven scenarios. The seven hypothetical research scenarios are included in the Appendix.

Patient-surrogate dyads were identified and enrolled following resolution of the patient’s critical illness and prior to ICU discharge. Demographic data about the patient and surrogate, and information regarding whether the dyad had previously discussed end-of-life decisions and participation in research studies were collected. Surveys for both participants contained identical hypothetical research scenarios. Patients were asked whether they would agree to participate in each of the seven hypothetical scenarios (yes or no) while surrogates were asked if they would agree to enroll the patient in the seven hypothetical scenarios (yes or no). The hypothetical nature of the scenarios was reinforced prior to completion of survey. Dyads completed their individual surveys simultaneously and independently, and study personnel monitored completion to ensure independence. In the event of questions on the part of either the patient or the surrogate, the answer was provided in the presence of both participants prior to completion of the surveys. The information from each survey was then paired, de-identified, and entered into a secure database.

Statistical Analysis

Continuous variables that were normally distributed are reported as means with standard deviation while those that are not normally distributed are reported as medians with inter-quartile range (IQR). We assigned a numerical complexity scale for each scenario: scenario 1 as the least complex and invasive and 7 as the most. Each patient/proxy pair provided an answer for the seven scenarios. Because the data was correlated a repeated measure logistic regression was used to model the outcome of consent based on the seven specific scenarios. A compound symmetric correlation structure was assumed for the estimated covariance matrix. The independent variable was the seven scenarios where patients and surrogates were asked if they would give consent for the patient. As there was a perceived linear relationship between the levels of the scenario and the log odds of the outcome, the ‘scenario’ variable was modeled as a continuous variable. This assumption of linearity was verified using orthogonal polynomial contrasts, along with the associated tests. Five outcome variables were analyzed: patient decision to participate in the research study, the surrogate decision to enroll the patient into the research study, overall discrepancy between the patient and surrogate, the false positive response when the surrogate would enroll a patient who would not choose to participate in research, and the false negative response when the surrogate would decline to enroll a patient who would choose to participate in research. The analysis for overall discrepancy was repeated adding the following covariates: age of patient, relationship of the surrogate (spouse/other), the number of years the patient and surrogate had known each other, and whether they had previously discussed involvement in a research study (yes/no). To find the degree of discrepancy between the patient and surrogate for each hypothetical study, we applied the Kappa statistic as a global test of agreement, considering a Kappa value of 0–0.2 to be poor agreement, 0.21–0.4 to be fair agreement, 0.41–0.6 to be moderate agreement, 0.61–0.8 to be substantial agreement, and 0.81–1.0 to be almost perfect agreement. SAS (SAS Institute Inc, Cary, NC) was used to perform the analysis. For all other statistical tests, statistical significance was defined by a p-value ≤ 0.05.

RESULTS

A total of 69 patient and surrogate dyads consented and were enrolled in the study. Their primary admission diagnoses for the patients were: acute respiratory failure/sepsis 25% (n=17), endocrine or metabolic disturbances 15% (n=10), heart failure 10% (n =7), gastrointestinal issues 9% (n=6), renal failure 7% (n=5), neurological reasons 4% (n=3), and a variety of other diagnoses 30% (n=21). The majority of the surrogates were spouses (58%), parents (22%), partners (7%), or children (7%) with a median length of their relationship of 28.5±16 years. Most patients reported having previously discussed their end-of-life issues with their surrogate (68%) whereas only 28% reported previously discussing their willingness to participate in medical research with their surrogate. None of the patient/surrogate dyads were approached for enrollment into other research studies during the ICU stay. Additional demographic information is included in Table 1.

Table 1.

Demographics

Characteristics	Number (n=69)
Patient Age (in years)	51 IQR* [38–63]
Sex of Patients (% male)	57%
Race (% Caucasian)	81%
ICU Length of Stay (in days)	3 IQR* [2–5]
Previously Mechanically Ventilated	25%
Proxy Relationship (% spouse)	59%
Length of Relationship (in years)	29 IQR* [16–39]
Discussed Medical Research	28%
Discussed End-of-Life Care	68%

^*IQR = interquartile range

Patients were significantly less likely to participate in research as the risk associated with the study increased (odds ratio = 0.67, 95% CI = 0.60–0.74, p < 0.001) (Table 2). Surrogates were also less likely to agree to enroll the patient into a research protocol as the risk associated with the study increased (odds ratio = 0.54, 95% CI = 0.49–0.60, p < 0.001) (Table 2). Discrepancy between patients and surrogates was common for the seven hypothetical studies, and their overall percentage of discrepancy increased as the risk associated with the study rose (odds ratio = 1.35, 95% CI = 1.20–1.52, p < 0.001) (Table 3). The differences in the rates of discrepancy for the seven scenarios were significant for both the false positive situation when the surrogate would have agreed and the patient would have declined (odds ratio = 1.14, 95% CI = 1.01–1.29, p = 0.049) and the false negative situation when the surrogate would have declined and the patient would have agreed (odds ratio = 1.42, 95% CI = 1.21–1.68, p < 0.001) (Table 3). The kappa values for each of the scenarios were categorized as poor. Using the decision of the patient as the “gold standard” and decision of the surrogate as the diagnostic test, the positive and negative predictive values for each of the seven scenarios are displayed in Table 4.

Table 2.

Patient and Surrogate Decisions regarding Critical Care Research with 95% Confidence Intervals

	Patient would agree to consent for self	Surrogate would agree to consent for patient
1: Collection of confidential medical information	94% (88–100%)	99% (97–100%)
2: Collection of urine	81% (72–90%)	96% (91–100%)
3: Collection of blood	75% (65–85%)	90% (83–97%)
4: CAT scan	58% (46–70%)	62% (51–73%)
5: Muscle biopsy	49% (37–61%)	52% (40–64%)
6: Bronchoalveolar lavage while intubated	39% (27–51%)	42% (30–54%)
7: Randomized clinical trial	46% (34–58%)	41% (29–53%)

Table 3.

Patient and Surrogate Discrepancy with 95% Confidence Intervals

	Surrogate: No Patient: Yes	Surrogate: Yes Patient: No	Patient/Surrogate Discrepancy	Kappa value
1. Collection of medical information	0%	4% (0–9%)	4% (0–9%)	0.39
2: Collection of urine	4% (0–9%)	18% (9–27%)	22% (12–32%)	0.04
3: Collection of blood	7% (1–13%)	22% (12–32%)	29% (20–38%)	0.03
4: CAT scan	13% (5–21%)	17% (8–26%)	30% (19–41%)	0.41
5: Muscle biopsy	17% (8–26%)	20% (11–29%)	37% (26–48%)	0.10
6:Bronchoalveolar lavage with intubated	20% (11–29%)	23% (13–33%)	43% (31–55%)	0.24
7. Randomized Clinical trial	26% (16–36%)	20% (11–29%)	46% (34–58%)	0.06

Table 4.

Predictive Values of Surrogate Decision Making with 95% Confidence Intervals

	Positive Predictive Value of Surrogate	Negative Predictive Value of Surrogate
1. Collection of medical information	96% (91–100%)	100%
2: Collection of urine	80% (70–90%)	0%
3: Collection of blood	76% (65–87%)	29% (18–40%)
4: CAT scan	72% (61–83%)	65% (54–76%)
5: Muscle biopsy	45% (33–57%)	65% (54–76%)
6:Bronchoalveolar lavage with intubated	61% (50–72%)	64% (53–75%
7. Clinical trial	50% (38–62%)	56% (44–68%)

There were no associations between the discrepancy for each scenario and any of the demographic factors collected including: race, education level, age of the patient, previous discussions regarding end-of-life issues or participation in medical research, surrogate relationship nature, or the length of their relationship. Additionally, we observed no associations between patient willingness to participate in each research scenario and their length of stay in the ICU, age, race, or gender.

We subsequently performed a multivariable analysis to determine if the difference between the various hypothetical research studies was independently associated with the overall rate of discrepancy between the patient and their surrogate. The effect of the scenario of the rates of discrepancy remained when adjusting for the age of the patient, the surrogate relationship, the number of years the surrogate and patient had known each other, and whether the dyad had previously discussed involvement in research protocols (odds ratio = 1.35, 95% CI = 1.20–1.52, p < 0.001).

DISCUSSION

Surrogates are commonly approached to make medical decisions for critically ill patients who have impaired decisional capacity. This study demonstrates that surrogates often make inaccurate decisions regarding the enrollment of patients in a variety of hypothetical critical care research studies. We also identified that the discrepancy between the patient and surrogate increased in association with the perceived risk of the research study. Our study builds on previous reports demonstrating similar findings for randomized clinical trials, and is the first to evaluate the accuracy of the surrogate consent process for various types of observational clinical research studies ^7;8.

There are two important and dichotomous implications of inaccurate surrogate decisions for participation in critical care research. False-negative enrollments, occurring when surrogates decline to enroll patients who would choose to participate in research infringe upon our ability to further critical care research and raises concerns about generalizability of research results. False-positive enrollments occur when the surrogate enrolls a patient who would not choose to participate in research and threatens the ethical principle of autonomy for the patient. Both of these scenarios have prompted significant controversy and continue to fuel debate and regulation on the use of surrogate consent ^2–4. Therefore, identifying methods to improve the accuracy of surrogate decisions and reduce violations of vulnerable patients’ rights are warranted.

Obtaining informed consent, particularly using surrogates, is complex and involves many factors that make it difficult to develop simple or straightforward improvements in the process. In this study, for example, if surrogates were relied upon to enroll patients in studies involving a muscle biopsy, the patient’s wishes would have been violated nearly one quarter of the time. The contributing factors and values that influence surrogate consent decisions remain poorly understood. In a recent study examining surrogate consent for critical care research, the most commonly cited reason for declining consent was the surrogate feeling too worried to consider the research study¹². This finding highlights concerns that surrogates may make decisions based on their own emotional state rather than the accepted hierarchical standards of known wishes, substituted judgments, and best interests^13;14. Furthermore, there are no established standards regarding what constitutes an acceptable degree of discordance to justify use of surrogate consent in incapacitated patient populations.

Patients who are enrolled into a research study by a surrogate should be re-consented when they regain decisional capacity. There may be important differences in the consent and re-consent process. We recently demonstrated that the majority of patients who were enrolled by surrogates into two recent ARDS network studies provided re-consent once they regained decisional capacity¹⁵. Further investigation is needed to determine if variations in the discrepancy rates are due to the fundamental differences between the consent and re-consent process.

Based on uncertainties regarding the surrogate consent process, it is important for researchers to identify ways to improve the accuracy of surrogate decisions to reduce violations of vulnerable patients’ rights. Due to ongoing concerns about the validity of surrogate consent for research participation, governmental agencies and healthcare organizations have implemented various mechanisms to minimize the potential violation of vulnerable patient populations. Restrictions on the use of surrogate consent for research have developed for many vulnerable patient populations, such as the elderly, patients with dementia and other psychiatric illness, as well as the critically ill ^1;16–19. Despite these concerns, few states have regulated surrogate consent for research involving critically ill patients. As a result, much variability exists in Institutional Review Board surrogate consent practices and the limits on risks in studies involving incapacitated adults ²⁰.

There are several limitations to our study. First, this study was only conducted in a single academic hospital and non-English speaking patients were not enrolled. Therefore, the results may not be generalizable to other critical care clinical settings. Our study was limited to patients who survived their critical illness and who had maintained or regained cognitive function after their ICU stay. The experience of surviving an ICU stay may bias the decisions made by the patient or their surrogate. After surviving an ICU stay, patients may also frequently experience some form of cognitive impairment that impacts their level of functioning ^2;20–22. We did not determine whether the decisions of the patient or surrogate changed over time after they were discharged from the hospital. In addition, the study protocols presented were hypothetical, potentially influencing the responses of the patient and surrogate. However, it is unclear whether patients and surrogates are more or less likely to consent to hypothetical studies than actual studies. Additionally, it is unclear whether surrogate or patient consent for current or future research studies is influenced by participation in previous clinical research. In this study we only examined the effect of increasing risk of the study as designated by the research protocol on the patient and surrogate’s willingness to participate. We did not formally test how perceived benefit affects the decision made by the participants ^23–26.

Future studies are needed to characterize how surrogates make these decisions and to identify factors that assist them in making decisions that more accurately reflect the wishes of the patient. Understanding the etiologies of the discrepancy in the surrogate consent process may enable researchers to enhance the protection of the rights of their patients. Additionally, the development of modalities to improve the surrogate consent process could mitigate some of the inaccuracies in the current process of surrogate informed consent for the critically ill.

CONCLUSION

Surrogates for critically ill patients often lack the ability to accurately predict whether patients would participate in clinical research studies, particularly as the research becomes invasive, complex, and associated with increased risk ^8;27. Future research is needed to identify the factors that influence the inaccurate decisions by surrogates. In particular, characterizing modifiable factors may assist in devising interventional studies to reduce discrepancies. Ultimately, improvements in the surrogate consent process may allow researchers to both protect the rights of our patients and simultaneously advance the field of critical care medicine.

Appendix

Research Participation Questionnaire

The following questions refer to potential research projects that may occur in the future and will provide no therapeutic benefit to the patient. However, please note that all information given here will be kept confidential and none of the answers you provide here will affect the standard of care for you or your ability to receive necessary therapeutic interventions during your current hospital stay. Thank you.

1. Would you have allowed the collection of your medical information (such as vital signs or laboratory information) while you were in the Intensive Care Unit (ICU) and kept for a later study?
- All information would be kept confidential and would not be linked to your name or information.
	YES	NO
2. Would you have allowed the collection of a sample of your urine that would have otherwise been thrown away?
- There is no excess risk to the patient and samples would be taken from an indwelling catheter or voided sample.
	YES	NO
3. Would you have allowed the collection of 5 mL (approximately one tablespoon) of extra blood be taken from the existing venous catheter for a study at a later time?
– There is a (0.01%) risk of creating a line infection with this extra sample collection.
	YES	NO
4. Would you had been willing to have been transported to the radiology department and had a computed tomography scan (“CAT scan”) be performed for review at a later time?
- There are potential risks of having catheters and tubes becoming dislodged during the move to radiology (< 1%) and unknown risks involved in the minimal radiation exposure from the scan.
	YES	NO
5. Would you have allowed a small muscle biopsy (the size of a pea) with local anesthetic to have been performed while you were under heavy sedation for review at a later time?
-The risks for this procedure include a 25% chance of mild pain at the site and a 0.5% chance of minor infection.
	YES	NO
6. Would you have allowed a bronchoalveolar lavage to have been performed while you had already had tube down your throat and were under heavy sedation, and have the results stored for study at a later time. This procedure involves have a flexible tube passed inside of the existing tube that was already in your throat. The flexible tube would be passed down into your lungs and a small amount of fluid (50 mls) would be placed into the lungs and then quickly suctioned back out. This procedure is often routinely performed to help identify infections in the lung?
- The risks for this procedure include: 0.5% chance of minor complication such as a temporary lower blood oxygen levels or lower blood pressure. There is a 25% chance of short term fever.
- The procedure has a <1% chance of detecting some other treatable illness
	YES	NO
7. Would you have allowed your participation in a clinical trial where one half of the patients are given a placebo (normal glucose) and the other half are given a dose of antacid for treatment of stress gastritis, which is a complication of serious illness?
- There is a 1% chance of getting stress gastritis by getting the sugar dose.
- There is a 50% chance of getting the antacid and side effects include: abdominal pain (1.8%), diarrhea (8%), and nausea (3.7%). There is a <0.1% risk of hip fracture with use greater than 6 months.
	YES	NO