For CPR or Not‐For‐CPR? Demographic Factors Predict Resuscitation Order Documentation and CPR Status in General Surgery Patients

ABSTRACT

Background

To improve healthcare equity, this study aimed to determine which demographic factors are associated with resuscitation order documentation and not‐for‐cardiopulmonary resuscitation (CPR) status in general surgery patients.

Methods

A retrospective cohort study was conducted including patients admitted under general surgical services of two hospitals in South Australia over 2 years. Logistic regression evaluated associations between demographic factors, and resuscitation order documentation and not‐for‐CPR status.

Results

12 846 patients were included, with 1853 (14.4%) having documented resuscitation orders. Of those with resuscitation orders, 964 (52.0%) were for CPR in cardiac arrest. Increased age (OR 1.05, 95% CI 1.04–1.05, p < 0.001), increased Charlson comorbidity index (OR 1.15, 95% CI 1.13–1.17, p < 0.001) and lower socioeconomic status (OR 0.997, 95% CI 0.995–0.999, p = 0.008) were associated with a greater likelihood of having a resuscitation order documented. Female sex (OR 0.76, 95% CI 0.60–0.95, p = 0.016), increased age (OR 0.92, 95% CI 0.91–0.93, p < 0.001), and Charlson comorbidity index (OR 0.89, 95% CI 0.86–0.92, p < 0.001) were significantly associated with being not‐for‐CPR. Having a resuscitation order documented (OR 18.0, 95% CI 10.9–29.7, p < 0.001) and being not‐for‐CPR (OR 20.7, 95% CI 10.0–42.8, p < 0.001) were significantly associated with increased in‐hospital mortality. Having a specified religion was associated with an OR of 1.29 for being for CPR (95% CI 1.02–1.62, p = 0.032).

Conclusion

This study demonstrated that multiple demographic factors predict resuscitation order documentation and content in general surgery patients, including which patients are not‐for‐CPR. These findings may improve equity in care for general surgery patients in clinical deterioration and end‐of‐life situations.

Multivariable predictors of resuscitation order documentation and association with mortality.

1. Introduction

Clarification of patient wishes through resuscitation order documentation is crucial for effective and ethical end‐of‐life care. Resuscitation orders outline the preferred care in the event of critical deterioration, including the desire for cardiopulmonary resuscitation (CPR), and are developed from shared decision making between clinicians and patients or their next of kin [1]. They can prevent inappropriate institution of traumatic and unhelpful interventions, including inappropriate CPR and other forms of resuscitation, and help guide effective and ethical care. It has been demonstrated that additional considerations must be undertaken for formulating resuscitation orders in the perioperative setting [2]. Emergency general surgery in elderly patients with not‐for‐CPR orders is associated with a poor prognosis, although this is largely reflective of poor preoperative baseline health [3]. Resuscitation order content has also been shown to impact on general surgery staff decision‐making [4, 5]. However, limited data exist regarding resuscitation orders in non‐emergency patients admitted under general surgical services. Further, no study has investigated the demographic factors that may result in disparate resuscitation order documentation for both non‐emergency and emergency general surgery patient populations. To improve health equity in end‐of‐life care and situations of clinical deterioration for general surgery patients, this study aimed to determine which demographic factors are associated with resuscitation order documentation and not‐for‐cardiopulmonary resuscitation (CPR) status in patients admitted under general surgical services.

2. Methods

2.1. Study Design, Setting and Population

This study followed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines [6]. A retrospective cohort study was conducted including consecutive elective and emergency adult patients admitted under general surgery services (colorectal, upper gastrointestinal, surgical trauma, hepatobiliary and acute surgical units) at the Royal Adelaide Hospital and the Queen Elizabeth Hospital in South Australia over a two‐year period from 1 April 2020 to 31 March 2022. This timeframe was selected as this was the timeframe from which the relevant data were reliably available from the institutional electronic medical record.

2.2. Resuscitation Orders at Participating Institutions

Both participating centres within the study followed the same network‐wide protocol regarding the documentation of resuscitation orders, which should ideally be in place for all patients at significant risk of clinical deterioration. At these centres, resuscitation orders were documented in the electronic medical record, normally by the admitting team, and after discussion with the patient and their next of kin if required. The process of discussing and documenting a resuscitation order is initiated by a medical officer (i.e., level of seniority above medical intern), normally of the admitting team. The section within the electronic medical record where discrete resuscitation orders were entered has been in place since the inception of the electronic medical record. Resuscitation orders include the capacity to specify that a patient is for “full resuscitation”, including consideration for CPR, aggressive organ support, or medical emergency response (MER) calls for deterioration. Resuscitation orders are enduring; however, they are expected to be re‐assessed upon each admission and can be updated or modified as clinical circumstances change.

2.3. Data Collection

Data were prospectively entered within the electronic medical record by clinical staff throughout each patient admission, and then were retrospectively collected for the present study. Documented resuscitation orders, demographics (including age, sex, primary language, religion), Charlson comorbidity index [7], time of admission, and in‐hospital mortality were collected from the electronic medical record. Patient postcode was used in conjunction with Australian Bureau of Statistics data to provide an estimate of socioeconomic status as a percentile relative to the rest of the state [8]. Exact percentile was used to ensure optimal transparency regarding this datapoint. If an individual was admitted on more than one occasion during the study period, only the first admission was included. If an individual had more than one resuscitation order documented, only the first documented resuscitation order was included in the analysis.

2.4. Statistical Analysis

Descriptive statistics were used to describe the cohort characteristics. Multivariable logistic regression analysis was employed to evaluate for factors that were independently associated with having a resuscitation order documented, and having the resuscitation order describe that the individual is not‐for‐CPR. In the logistic regression analyses, explanatory variables included age, sex, primary language (English vs. non‐English), religion (specified religion vs. not specified), Charlson comorbidity index, socioeconomic percentile, and admission time (08:00–20:00 vs. 20:00–08:00). Marital status was not included in the logistic regression analysis. When evaluating for association between documentation of a resuscitation order, or content of a resuscitation order, and in‐hospital mortality, the presence or absence of resuscitation order documentation was included as an explanatory variable. Variance inflation factor was used to evaluate for multicollinearity, and no explanatory variables had a score greater than 10 that would have necessitated removal from the logistic regression analyses. Median imputation was used to replace missing input data (Charlson comorbidity index: 6774 patients; socioeconomic status: 3526 patients). Logistic regression results are presented as odds ratios (OR) with 95% confidence intervals (95% CI) unless otherwise specified. Statistical analysis was conducted using Python and R (version 1.1.456).

2.5. Ethics Statement

Institutional ethical approval was obtained from the Central Adelaide Local Health Network Human Research Ethics Committee (reference number: 16 409), with a waiver of individual consent.

3. Results

3.1. Patient Characteristics

The study included 12 846 patients. The mean age was 54.9 years (standard deviation (SD) 20.5); 47.9% were female. 90.8% spoke English as their primary language, 31.7% specified a religion, and the mean socioeconomic percentile was 59.0 (SD 28.5). The median Charlson comorbidity index of the cohort was 6 (interquartile range (IQR) 4–8). The overall rate of in‐hospital mortality was 1.1%. Characteristics of the study population are presented in Table 1.

TABLE 1.

Cohort characteristics.

Variable	Individuals without a resuscitation order	Individuals with a resuscitation order
n	10 993	1853
Age years; mean (SD)	52.0 (19.5)	71.6 (18.5)
Socioeconomic percentile; mean (SD)	58.7 (28.6)	59.0 (28.0)
Female sex; number	5763 (52.4%)	926 (50.0%)
Primary language English; number	10 078 (91.7%)	1590 (85.8%)
Religion specified; number	7698 (70.0%)	1077 (58.1%)
Admitted overnight (20:00–08:00); number	5318 (48.4%)	930 (50.2%)
Charlson comorbidity index; median (IQR)	6 (2–8)	8.0 (6–10)
Survival to discharge; number	10 971 (99.8%)	1730 (93.4%)

Abbreviations: IQR, interquartile range; SD, standard deviation.

3.2. Resuscitation Order Documentation and CPR Status

A resuscitation order was documented in the notes for 1853 (14.4%) patients. Of these, 964 (52%) were documented as being for CPR in the event of cardiac arrest. There were 61 individuals with resuscitation orders that specified “not‐for‐CPR” and “not‐for‐medical emergency response calls” (3.3%).

3.3. Demographic Factors Predicting Resuscitation Order Documentation

Several demographic factors were associated with an increased likelihood that a resuscitation order had been documented. Increased age (years) (OR 1.05, 95% CI 1.04–1.05, p < 0.001) and increased Charlson comorbidity index (OR 1.15, 95% CI 1.13–1.17, p < 0.001) were both associated with a greater likelihood of having a resuscitation order documented. Higher socioeconomic status percentile was independently associated with a lower likelihood of having a resuscitation order documented (OR 0.997, 95% CI 0.995–0.999, p = 0.008). Female sex, English as primary language, having a specified religion, and being admitted overnight were not significantly associated with documentation of a resuscitation order. Table 2 clearly presents all logistic regression findings.

TABLE 2.

Multivariable predictors of resuscitation order documentation, CPR status, and associated in‐hospital mortality in general surgery patients.

Predictor	OR (95% CI)	p
Factors Associated with having a resuscitation order (n  = 12 486)
Age (per year)	1.05 (1.04–1.05)	< 0.001
Charlson comorbidity index (per point)	1.15 (1.13–1.17)	< 0.001
Socioeconomic percentile	0.997 (0.995–0.999)	0.008
Predictors of “Not‐for‐CPR” status among patients with resuscitation orders (n = 1853)
Age (per year)	0.92 (0.91–0.93)	< 0.001
Charlson comorbidity index (per point)	0.89 (0.86–0.92)	< 0.001
Female sex (vs. male)	0.76 (0.60–0.95)	0.016
Religion specified	0.78 (0.62–0.98)	0.032
Association between resuscitation orders and in‐hospital mortality (n = 12 846)
Having a documented resuscitation order	18.0 (10.9–29.7)	< 0.001
Having a documented “Not‐for‐CPR” order	20.7 (10.0–42.8)	< 0.001

Abbreviation: OR, Odds ratio.

3.4. Demographic Factors Predicting CPR Status

Demographic factors associated with being “not‐for‐CPR” within the documented resuscitation order included female sex, increased age, increased Charlson comorbidity index, and not having a specified religion. Females were less likely to be for CPR (OR 0.76; 95% CI 0.60–0.95, p = 0.016). Older patients (OR 0.92, 95% CI 0.91–0.93, p < 0.001) and those with a high Charlson comorbidity index (OR 0.89, 95% CI 0.86–0.92, p < 0.001) were more likely to have a “not‐for‐CPR” order in place compared with younger patients and those with a low Charlson comorbidity index. Patients with a specified religion were more likely to have a documented resuscitation order that stated they were “yes‐for‐CPR” than those without a stated religion (OR 1.29, 95% CI 1.02–1.62, p = 0.032) (Table 2). There were no significant associations between English primary language, socioeconomic percentile and being admitted overnight and being not‐for‐CPR within the documented resuscitation order.

3.5. Resuscitation Orders and In‐Hospital Mortality

When controlling for other explanatory variables including age and comorbidities, there was a strong association between having a documented resuscitation order and in‐hospital mortality. Having a documented resuscitation order was associated with an increased likelihood of in‐hospital mortality (OR 18.0, 95% CI 10.9–29.7, p < 0.001). Similarly, having a documented resuscitation order that specified that an individual was not‐for‐CPR was associated with an increased likelihood of in‐hospital mortality (OR 20.7, 95% CI 10.0–42.8, p < 0.001) (Table 2).

4. Discussion

This study demonstrated that multiple demographic factors predict resuscitation order documentation and content in patients admitted to general surgical services, including which patients are not‐for‐CPR. Some associations were intuitive, such as that between increased age and increased comorbidities and having “not‐for‐CPR” status, and the increased likelihood of in‐hospital mortality associated with having a resuscitation order documented. However, the associations between sex, socioeconomic status, and CPR status identified in the present study warrant further investigation to ensure healthcare and outcome disparities do not result in this patient population.

Our results can be contextualised within the existing literature on resuscitation orders in surgical patients, which in general has shown similar patterns and considerations. For instance, Kalkman et al. stressed the need for special consideration of do‐not‐resuscitate (DNR) orders in perioperative patients [2], underscoring the importance of proactive end‐of‐life discussions in surgical care. In emergency general surgery, Scarborough et al. observed significantly higher mortality in elderly patients with pre‐existing DNR orders [3], a finding echoed by Kazaure et al. in a broad surgical cohort [9]. Scarborough et al. further attributed this mortality gap to a possible ‘failure‐to‐pursue rescue’ phenomenon among DNR patients [3], suggesting that such patients may receive less aggressive intervention during complications. This concern is supported by Neufeld et al., who found in a national survey that general surgeons' management decisions were influenced by patients' code status [4]. Similarly, Scarborough and Pappas reported that DNR status was associated with elevated postoperative mortality even after emergency surgery [3] implying that resuscitation order status can impact treatment intensity and outcomes. Collectively, these studies corroborate our finding that resuscitation order documentation and content are closely intertwined with patient outcomes and clinical decision‐making in surgical populations.

Consistent with the present findings, there is increasing evidence across various specialties that females who have certain medical presentations are more likely to be not‐for‐CPR than males [10, 11]. Potential hypotheses to explain this include increased likelihood of clear communication with family prior to the illness or systematic clinician and family biases. However, a cause for the discrepancy between the sexes is unknown. Both provider and patient gender have been found to influence resuscitation orders in cancer patients, with female patients more likely to receive earlier not‐for‐CPR orders from female physicians [12]. As there is no way to reliably identify reasons for the sex discrepancy found in the resuscitation orders of the large general surgery cohort evaluated in the present study, further research is required. However, when interpreting these results, it is helpful to remember the value of an appropriate resuscitation order regardless of patient sex. The finding of higher socioeconomic status being associated with a lower likelihood of having a documented resuscitation order is a novel one within this patient population. The reasons for this are unclear, and sub‐analyses of this association could be a useful avenue for future research. It may be that the patients had better overall health and did not require a resuscitation order; however, it could also reflect more robust family or social support structures. Another consideration could be attitudes towards prolonging end‐of‐life care and the effect that costs to the individual and system have on this decision.

The association between having a documented resuscitation order and experiencing in‐hospital mortality is consistent with previous evidence identifying that general surgery patients with “do‐not‐resuscitate” orders had worse outcomes than those without such orders [9]. Accordingly, the association between resuscitation order documentation and mortality may simply reflect clinicians correctly identifying patients most at risk of deterioration or approaching end‐of‐life and, accordingly, implementing appropriate clinical care. Conversely, in other emergency presentations, it has been shown that the prediction of poor outcomes is a self‐fulfilling prophecy [13]. The content of resuscitation orders may also be entirely independent of the treating clinical team. It may be derived from the patient or their close family, and the well‐described neural mechanisms of the brain shielding itself from existential threats, particularly death [14]. This is likely underappreciated in surgical systems of care.

Clinically, our findings highlight the need for systematic and equitable resuscitation planning for surgical patients. We recommend that hospitals implement protocols to ensure resuscitation orders are considered for all general surgery admissions, especially those at higher risk of deterioration, to improve the currently low documentation rate (only 14.4% of patients in our cohort). In addition, clinicians should remain vigilant about potential demographic biases—for example, the tendency in our cohort for female and higher socioeconomic status patients to have lower rates of documented resuscitation orders—and ensure that end of life discussions are patient‐centred and free and unintended bias. From a policy perspective, embedding routine prompts for resuscitation order documentation into admission workflows providing staff training in culturally sensitive, unbiased communication could help standardise these discussions across all patient groups. Such measures, alongside early involvement of palliative care for high‐risk patients, may improve the consistency and fairness of end‐of‐life decision‐making in surgical populations.

While the findings of this study are worth highlighting, there are some limitations. There was no means of systematically evaluating whether individuals had pre‐existing advanced care directives prior to hospital admission, and this was outside the scope of the present study. There was a requirement for median imputation of a significant number of patients for the Charlson comorbidity index and socioeconomic percentile, and the reasons for the degree of missing data is unknown, as these were simply the rates that were present in the institutional electronic medical record from which data were extracted. For patients that survived to hospital discharge, details of discharge pathway and readmission were outside the scope of the present study. The proportion of patients who had a palliative care consult or admission during their admission was not evaluated, and the present study was conducted at two centres in the same health network in South Australia. The overall rate of a resuscitation order being documented was low and may confound results, as there was a selection bias towards documentation of a resuscitation order in the elderly and unwell. Lastly, the relationship between sex and socioeconomic class and not‐for‐CPR status was not adjusted for illness severity or healthcare interactions prior to the index episode.

5. Conclusions

This study demonstrated that multiple demographic factors predict resuscitation order documentation and content in patients admitted under general surgical services, including which patients are not‐for‐CPR. These findings suggest opportunities to further investigate and improve end‐of‐life decision making in the general surgical population for the overall improvement of healthcare equity and outcomes. Future studies would benefit from investigating subgroups within the general surgery patient population, comparisons between hospitals that follow different clinical protocols, differences between emergency and elective surgical populations, association with MER calls, and also consider other social factors such as living arrangement, marital status, and admission source.

Author Contributions

Joshua G. Kovoor: conceptualization, data curation, formal analysis, investigation, methodology, project administration, writing – original draft, writing – review and editing. Stephen Bacchi: conceptualization, data curation, formal analysis, investigation, methodology, writing – original draft, writing – review and editing. Daksh Tyagi: data curation, investigation, writing – review and editing. Aashray K. Gupta: conceptualization, investigation, methodology, writing – review and editing. Brandon Stretton: conceptualization, investigation, methodology, writing – review and editing. Nidhi Aujayeb: conceptualization, data curation, investigation, writing – review and editing. Amy Lu: conceptualization, data curation, investigation, writing – review and editing. Kayla Nathin: conceptualization, data curation, investigation, writing – review and editing. Thomas J. Hugh: investigation, supervision, writing – review and editing. Gregory Crawford: investigation, supervision, writing – review and editing. Guy J. Maddern: investigation, supervision, writing – review and editing.

Conflicts of Interest

Professor Guy Maddern is an Editorial Board member of the ANZ Journal of Surgery and a co‐author of this article. To minimize bias, they were excluded from all editorial decision‐making related to the acceptance of this article for publication. This study was presented at the 2024 Royal Australasian College of Surgeons Annual Scientific Congress (ASC), and thus an abstract has been published online.

Kovoor J. G., Bacchi S., Tyagi D., et al., “For CPR or Not‐For‐CPR? Demographic Factors Predict Resuscitation Order Documentation and CPR Status in General Surgery Patients,” ANZ Journal of Surgery 95, no. 7‐8 (2025): 1433–1438, 10.1111/ans.70233.