Physicians’ opinion of patients’ CPR decision: Secondary analysis from a pilot RCT

Highlights

• •Physicians tend to agree with the code status choices that their patients make.
• •Physicians disagree with patients’ full code decision more frequently than no code.
• •GO-FAR and CCI scores were not associated with physician code status judgment.

Abstract

Background

While patients in the United States generally have final say in their code status, discussion with their physician plays an important role in decision-making. However, physicians do not discuss code status with every patient, and do not consistently mention patients’ prognosis following cardiopulmonary resuscitation (CPR). Understanding how physicians perceive patients’ CPR decisions is prerequisite to improving code status discussions.

Methods

We report a planned secondary analysis from a prospective randomized controlled trial of 102 English-speaking adults aged ≥65 evaluating whether “Allow Natural Death” was preferred to “Do Not Resuscitate” as the “no code” option in code status discussions. We measured physician agreement/disagreement with patient code status decisions and the correlation with objective outcome measures. Two clinically validated instruments—measuring likelihood of surviving resuscitation (Good Outcomes Following Attempted Resuscitations (GO-FAR)) and morbidity level/1- and 10-year mortality (Charlson Comorbidity Index (CCI))—were calculated for each participant.

Results

Physicians agreed with patients’ code status decisions 88.3% of the time. Physician agreement with code status was not correlated with GO-FAR or CCI scores. GO-FAR and CCI scores do not always align, indicating that illness severity and CPR outcome are not directly linked.

Conclusions

This study highlights that while physicians tend to agree with patient’s code status, their decisions do not align with data from clinically validated predictors of coding success or illness severity/mortality prediction. Further research is required as to how physicians perceive whether attempting CPR is appropriate or not.

Introduction

Patients are frequently unsatisfied when it comes to their end of life care decision-making.1, 2 Advanced care planning, including code status discussions, is often lacking.3, 4 Patients—who in the United States choose their own code status based on their values⁵—often lack understanding about resuscitation, and overestimate prognosis.6, 7 Physicians, expected to routinely discuss cardiopulmonary resuscitation (CPR) in every hospitalization,⁵ may address CPR prognostication from as little as 5% to as high as 65% of the time.8, 9 Reasons for this range from feeling unprepared to communicate prognostic information to discomfort in communicating a recommendation.8, 9, 10

Decision-making is also complicated by physicians’ tendency to selectively initiate code status discussions.¹¹ Even in critically ill patients “able to communicate verbally about their medical care” for whom physicians discuss care more frequently, nearly 70% had no code status discussion; this defaults decision-making to full code, even if not reflecting the patient’s values.⁸ Potential reasons include lack of time and fear of distressing patients.¹² Even when a comprehensive advanced care planning discussion is conducted, variability persists in provider prognostication.13, 14

Clinically validated tools, such as the 2013 Good Outcome Following Attempted Resuscitation (GO-FAR) score¹⁵ and the 1987 Charlson Comorbidity Index (CCI),¹⁶ can provide objective measures to reference during code discussions. These scores are not typically used, however, and no study has examined their correlation with code status decisions.¹⁷

Patients’ code status decisions may differ from their physicians’ opinion. Understanding factors contributing to such disagreements—including the role of objective measures like GO-FAR and CCI—is critical. This study examines the relationship between patients’ decisions, physicians’ agreement with those decisions, and objective measures on code attempts/patients’ illness severity, with a goal of enhancing shared decision-making and patient autonomy.

Methods

Study design

A pre-planned secondary analysis from a prospective, randomized, controlled trial in a single academic institution evaluated the impact of how word choice (“allow natural death” vs. “do not resuscitate” for “no code” phrasings) affected code status decisions (see Supplemental section).¹⁸ English-speaking adults age ≥65 years admitted to the Internal Medicine teaching service from August 2021 to August 2022 were recruited. Patients were block-randomized into “Do Not Resuscitate” versus “Allow Natural Death” arms, with “allow CPR” as the full code phrasing in both. After being read a standardized CPR statement using the designated “no code” decision, patients discussed their code status with a physician. All physicians were trained (via writing and video) to have standardized code status discussions to keep unstructured discussions as close to uniform as possible. Code status decisions were reported to the attending of record, who were asked if they agreed with it.

This study was approved by the Rutgers IRB, registered on ClinicalTrials.gov (NCT04896411), and is in accordance with the Declaration of Helsinki.

Experimental measures

Information collected on each patient included: demographics (age, gender, race, education, insurance status, income, and religious affiliation) and code status decision.

We used two objective measures—GO-FAR and CCI—that might correlate with physician judgment of CPR appropriateness. The GO-FAR predicts survival to discharge with minimal to no neurologic deficits (Cerebral Perfusion Category score of one) after CPR for in-hospital cardiac arrest.¹⁵ GO-FAR scores range from −15 to 76, with lower scores indicating better prognosis. The CCI estimates 1- and 10-year survival for patients, with scores ranging from 0 to 37. Higher scores indicate greater medical illness/lower 1- and 10-year survival.¹⁶ Medical history, admission diagnosis, and laboratory results were abstracted from the participant’s electronic health record to calculate the GO-FAR and CCI scores. GO-FAR was stratified into 3 prognostic grades: above average (>15%, GO-FAR score −15 to −6), average (3–15%, GO-FAR score −5 to 23), and below average (<3%, GO-FAR score > 14). CCI was also categorized into 3 grades: mild (1 to 2), moderate (3 to 4), and severe (≥5).

Statistical analysis

Patients unsure of their code status decision were presumed full code during analysis, while physicians unsure about code status agreement were presumed to agree/defer to the patient.

Data were analyzed using R statistical software version 3.4.3. The Shapiro-Wilk test was used to determine distribution of GO-FAR and CCI scores. The Mann-Whitney U and Fisher Exact tests were used to compare GO-FAR and CCI scores by participant’s code status decisions and attending agreement. P < 0.05 was considered significant.

Results

Of 923 eligible patients age ≥65 years, 394 were approached and 108 enrolled (see Supplementary Material). 102 participants completed the study and were analyzed. The average age was 72 (IQR 67.25–78), with 53% males (n = 54). Patients had the same illness level as other comparable New Jersey centers (44/102 were rate as ‘severe’ via CCI score).¹⁹ The effect of “no code terminology” on code status decision was not significant, and reported in detail elsewhere.¹⁸

17 (16.7%) patients chose no code and 84 (82.4%) chose full code; one participant was undecided. Physicians agreed with participant’s code status 88.3% (90/102), disagreed 5.8% (6/102), and were unsure 5.8% (6/102) of the time. Physician disagreement tended against patient selecting full-code (5/6, 83.3%).

By GO-FAR score, 42 (41.2%) participants were estimated to have above average, 56 (54.9%) to have average, and 4 (3.9%) to have below average CPR survival (Table 1.). The median GO-FAR scores for full code and no code patients were −4 and −2 (p = 0.10), respectively. For full code and no code patients, there was no significant difference in score if a physician agreed versus if a physician disagreed with the code decision (p = 0.94 and 0.13, respectively; Fig. 1); both groups had median GO-FAR score of −4. There was no significant difference in rate of physician disagreement with code status decision by GO-FAR score category (7.1% (3/42), 5.4% (3/56), 0% (0/4) from above average to below average respectively; p = 1).

Table 1.

Physician agreement with code status.

			Physician agree		Physician disagree		Total
			n	%	n	%	n	%
GO-FAR*	Full code	Above average	35	43.8	2	40.0	37	36.3
		Average	44	55	3	60.0	47	46.1
		Below average	1	1.3	0	0.0	1	1.0
	No code	Above average	4	25.0	1	100.0	5	4.9
		Average	9	56.3	0	0.0	9	8.8
		Below average	3	18.8	0	0.0	3	2.9
CCI†	Full code	Mild	9	11.3	1	20.0	10	9.8
		Moderate	35	43.8	0	0.0	35	34.3
		Severe	36	45.0	4	80.0	40	39.2
	No code	Mild	2	12.5	0	0.0	2	2.0
		Moderate	6	37.5	1	100.0	7	6.9
		Severe	8	50.0	0	0.0	8	7.8
Total			96	94.1	6	5.9	102

There was no statistically significant difference between categories.

^*Good Outcome Following Attempted Resuscitation (GO-FAR).

^†Charlson Comorbidity Index (CCI).

Fig. 1.

Physician agreement with patient code status decision by GO-FAR (A) and CCI score (B). There was no significant difference between physician agreement or disagreement with patient code status decision, both by Good Outcome-Following Attempted Resuscitation (GO-FAR) for full code (p = 0.94) and no code patients (p = 0.13); and by Charlson Comorbidity Index (CCI) (p = 0.46 and p = 0.30, respectively).

By CCI, 12 (11.8%) participants had mild grade, 42 (41.2%) had moderate grade, and 48 (47.1%) had severe grade comorbidities (Table 1.). Full code and no code patients had the same median CCI score of 4 (p = 0.52). Full code and no code patients had no significant difference in CCI score whether physicians agreed or disagreed with the code decision (p = 0.46 and 0.30, respectively; Fig. 1.).

GO-FAR and CCI score categories were discrepant. A quarter (12/48) of patients who in the “severe” category for CCI had an above average GO-FAR score. The Cohen’s Kappa between GO-FAR and CCI was 0.002 (p-value = 0.97), indicating none to slight agreement. For the five cases where physicians disagreed with the patient’s full code status, 2 had above average estimate and 3 had average estimates of good outcome resuscitation by GO-FAR score. By CCI score, 1 was in the mild, and 4 were in the severe category.

Discussion

This novel study examines whether the physician of record agrees with their patient’s code status decision, at an 88.3% rate. Physicians typically disagreed with full-code status decision (5/6 times). Discordance between predicted code status outcomes (GO-FAR) and estimated measure of chronic illness (CCI) may be impairing physicians’ decision-making in some patients.

Prior outpatient studies found physician’s code status recommendation aligned with the patient’s decision at 84.2% and 84.6% of the time20, 21; however, in those studies, physicians did not know patients’ CPR preference. Although attending physicians in our study were biased by knowing the patients’ CPR preference, because patients decide on code status in our setting, we felt it useful to understand if physicians felt a need to intervene in discussions (noted by disagreeing with the patient). Physicians agreeing with patients seven in eight times helps identify a baseline rate for more in-depth studies. While some physicians dismiss routine code status discussion as leading to “perfunctory and superficial discussions,”²² the rate of in-hospital cardiac arrest has been increasing.²³ Patients without confirmed code status pick no resuscitation as frequently as 20% when asked,23, 24 and communication interventions—which improve patient knowledge—lead to lower preference for CPR.³ As code status discussions take less than five minutes on average,¹⁸ it is clear that the added time burden from screening all patients is small for potentially significant benefits.

Two objective measures of mortality and CPR success (CCI and GO-FAR) that may aid code status prognostication diverge in patient categorization, and neither closely aligns with physician agreement on code status appropriateness (Fig. 1.). While patients decide code status based on their values,⁵ physicians may attempt to persuade patients otherwise; sometimes, addressing the futility of CPR changes code status decision from full code to DNR.²⁵ It would be beneficial if physician persuasion was based on objective clinical criteria, and that physicians provide accurate, individualized assessments to patients.²⁶ However, the use of pre-arrest clinical calculators is lacking.¹⁷ While objective measures may not change patient-specific preferences or code status on their own, they may improve code status decisions overall by anchoring the discussion on the patient’s own circumstances.

There are several limitations of this study, including the small sample size, single-center design, and exclusion of patients who were non-English speaking. Study physicians were also not blinded to a patient’s study arm.

Conclusions

While physicians typically agree with patient’s code status decisions, there is no correlation between either of two clinically validated calculators and physician agreement with code status decision.

Disclosures

Partial results were presented at the American Geriatrics Society National Meeting Abstract competition and the American College of Physicians Poster Finalist competition.

Sponsor’s role

There was no external funding for the study.

Disclaimer

The contents of this work are solely the responsibility of the authors and do not necessarily reflect the official views of Rutgers University.

CRediT authorship contribution statement

Stephanie Ji: Writing – original draft, Formal analysis, Data curation. Alexa Pisciotti: Writing – review & editing. Mitsu Patel: Writing – review & editing. Catherine Chen: Writing – review & editing, Supervision, Resources, Methodology, Investigation, Conceptualization. Michael B. Steinberg: Writing – review & editing, Supervision, Methodology. Karthik J. Kota: Writing – review & editing, Validation, Supervision, Resources, Project administration, Methodology, Investigation, Data curation, Conceptualization.

Funding

There was no external funding for the study.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

The following are the Supplementary material to this article: