Code status orders involve decisions about using or forgoing cardiopulmonary resuscitation (CPR) in the event of cardiopulmonary arrest. In discussions about code status, it is important that doctors and patients communicate effectively to provide appropriate care that respects patients’ values and preferences. However, such communication can be difficult,1 and studies suggest that doctors often misunderstand patients’ code status preferences,2 even in the setting of serious illness.3 There is a need to improve the quality of code status discussions so that patient preferences can be ascertained and implemented.4

One potential barrier to appropriate code status decisions may be the absence of shared understandings of patients’ goals of care. Goals provide basic orientation for clinical and ethical decision making,5 and code status discussions may be compromised if doctors and patients have differing goals of care in mind.6 7 Accordingly, clinicians have been encouraged to use goals to frame end-of-life decision making,1 8^–11 an objective supported by a recent review that identifies six common goals of care for this purpose.12 Furthermore, some authors believe that patients’ goals are more important than their treatment preferences documented in advance directives.7 13^–20 Beyond goals, an additional barrier in code status discussions may be patients’ inadequate understanding of the outcome probabilities of resuscitative treatments.21^–24

To investigate these barriers, we interviewed hospitalised adults to:

• assess the concordance between patients’ code status preferences and their code status orders;

• assess patients’ understanding of outcome probabilities after inhospital resuscitation;

• describe patients’ goals of care during acute hospitalisation;

• investigate associations between patients’ code status preferences and their goals of care.

METHODS

Participants

After approval by the University of Iowa institutional review board, we interviewed adult inpatients within 48 h of admission to the general medical services at a large Midwestern academic medical centre from June to August 2007. Prior to contacting potential participants, attending doctors responsible for patients on the general medical services were notified by email about the study. Potential participants were identified by reviewing administrative census lists, and interviews were conducted from Mondays to Fridays. Patients were excluded if they were only available for interview within the 48-h time frame on a Saturday or Sunday.

Structured interviews

Two medical students enrolled patients, obtained consent and conducted the interviews; they were trained by the principal investigator, and each medical student performed their first three interviews under the observation of one of the doctor-investigators to ensure that interview procedures were followed appropriately.

The interviewers established a daily list of potential participants, assigned a random order to the list and approached as many patients as possible on a given day. Potential participants were invited to participate as soon after admission to the hospital as was feasible. The informed consent document was read by the patient or by the interviewer. Ability to consent was assessed after the informed consent document was read. Patients were deemed capable of providing consent if they were:

• alert and able to communicate;

• able to explain what would be expected of them during the interview;

• able to explain what they would do if they no longer wanted to participate in the study;

• able to explain what they would do if they experienced distress or discomfort during the interview.

Capable patients then signed the informed consent document and the interview was conducted at the bedside. Family members were welcome to be present.

Survey instrument

Interviewers used a structured survey instrument, an earlier version of which had been tested in a pilot sample of 34 patients and then revised and amplified. The instrument provided a step-by-step, verbatim guide for the entire interview and consisted of closed and open-ended questions. After the initial interviews, the instrument underwent minor revisions for increased clarity. Patient responses were recorded onto a copy of the instrument during the interviews, which lasted approximately 30 min.

The instrument included questions regarding knowledge about CPR and its outcome probabilities, code status discussions and preferences, goals of care, advance directives, self-reported health status,25 a two-question depression screen,26 activities of daily living27 and demographic variables. CPR was presented as having three components: defibrillation, chest compressions and intubation with mechanical ventilation (see table 2 in the Results section for the exact wording). We queried six common goals of care based on existing literature1 10: be cured; live longer; improve current health; maintain current health; accomplish something particular in life; and be comfortable. Patients were asked which of these goals (including a seventh option of “other”) pertained to them, and for each goal affirmed patients were asked if they were expecting or hoping for that goal; among the goals they affirmed, patients were also asked to indicate which single goal was most important to them.

Regarding CPR outcome probabilities, the interviewer asked:

“If someone has a cardiac arrest in the hospital and receives CPR, what do you think are the chances, on a scale from 0 to 100%, of their surviving long enough to leave the hospital?”

To assess how information may affect preferences, the interviewer then said:

“Research has shown that if someone has a cardiac arrest in the hospital and is treated with CPR, their chance of surviving long enough to leave the hospital is approximately 15% (15 out of 100)”

and added:

“Does this information make you want to change your preferences about receiving CPR while you are in the hospital?”

The interviewer then said:

“Research has also shown that if someone has a cardiac arrest in the hospital and is treated with CPR, their chance of leaving the hospital with good brain function is approximately 7% (7 out of 100)”

and added:

“Does this information make you want to change your preferences about receiving CPR while you are in the hospital?”

Statements regarding outcome statistics were based on existing data.28

Medical record review, communication with doctors and prognostic estimates

After the interviews, the patients’ medical records were reviewed for demographic, medical and code status information. Based on institutional practice, a patient’s code status was assumed to be “full” (do resuscitate) if no code status documentation was found. In cases of discrepancy between a patient’s code status order in the medical record and their treatment preferences expressed during the interview, the principal investigator notified the patient’s attending doctor. If a patient answered positively to one or both depression screening questions, the interviewer notified one of the patient’s doctors. Attending or resident doctors were queried about patients’ estimated prognoses (<6 months, 6–24 months, >24 months).

Statistical analysis

Data from the questionnaires were entered manually into a Microsoft Access data file and then uploaded into SAS V.8. Frequency, χ² and analysis of variance (ANOVA) analyses were performed using statistical software to describe variable frequencies and determine associations between variables.

RESULTS

Demographic data and admitting diagnoses

Figure 1 shows the inclusion process for the 135 participants and how their sex, age and race/ethnicity compared with non-participants. Table 1 provides the participants’ demographic information. Admitting diagnoses represented a broad range of disease categories: gastroenterology (n = 52, 38.5%), infectious diseases (n = 29, 21.5%), pulmonology (n = 16, 11.9%), nephrology (n = 10, 7.4%), neurology (n = 7, 5.2%), substance misuse (n = 7, 5.2%), endocrinology (n = 6, 4.4%), haematology-oncology (n = 4, 3.0%), connective tissue diseases (n = 3, 2.2%) and cardiology (n = 1, 0.7%).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Inclusion details and comparison of patients included in/excluded from the study.

Knowledge and preferences regarding code status

Table 2 shows the differences between patients’ perceived and actual knowledge about CPR. A total of 41 patients (30.4%) agreed that “During this hospitalisation, my doctor discussed with me whether I would want CPR in case my heart stopped beating or my lungs stopped breathing, which would mean that I was dying.” Older patients were more likely to have discussed their code status (>64 vs 40–64 vs <40: 36.8% vs 37.5% vs 11.1%, respectively, p = 0.014), as were patients with fair or poor self-rated health compared with those with good, very good or excellent health (42.1% vs 21.8%, p = 0.011).

Table 2 also describes patients’ preferences for the main components of CPR, and 116 (85.9%) patients preferred all three components (full code status). Older patients were less likely to prefer full code status (>64 vs 40–64 vs <40: 57.9% vs 88.8% vs 94.4%, respectively, p<0.001), as were patients with living wills compared with those without (72.7% vs 90.2%, p = 0.012). Nineteen patients preferred to forgo one or more components of CPR, and 12 did so because “once doctors start this kind of treatment they won’t know when to stop and let you die”; eight of these patients would have reconsidered their preference if they were assured that this treatment “would be stopped in a few days if it was not helping.”

Documentation and discrepancies regarding code status preferences

Advance directives were infrequently found in patients’ medical records, whether living wills (n = 6, 4.4%) or powers of attorney for healthcare (n = 7, 5.2%). Code status orders were found in 60 (44.4%) of medical records, and code status preferences were described in 90 (66.7%) of admission or progress notes. For 11 (8.1%) patients there was a discrepancy between code status preferences ascertained in the interview and code status documentation in the medical record; eight of these patients preferred less treatment, and three more treatment, than indicated in their medical records.

Patients’ goals of care

Large proportions of patients affirmed each of the six goals (with a mean of 4.9 goals affirmed per patient), and patients’ single most important goals were broadly distributed (table 3). Patients were more likely to say they were expecting goals (range 40.6–70.0%) than hoping for them (range 30.0–47.8%). In all, 53 (39.3%) patients recalled speaking with their doctors about goals of care, and those who preferred full code status were more likely to have done so (43.1% vs 15.8%, p = 0.023). χ² analyses comparing code status preferences (full code status vs not full code status) across the seven goals of care did not reveal significant associations. Two patients (1.5%) said discussing goals of care made them want to change their CPR preferences while in the hospital (both were more interested in receiving CPR).

Perceptions of CPR outcome probabilities

Patients’ mean prediction of survival to hospital discharge after cardiac arrest and CPR was 60.4% (range 2–100%). After being informed that survival to discharge after CPR is approximately 15%, 11 (8.1%) patients said this information made them want to change their code status preferences: 9 patients were less interested in receiving CPR, 1 patient was more interested and 1 patient gave no answer. After being informed that the chance of good brain function after inhospital CPR is approximately 7%, 28 (20.7%) patients said this information made them want to change their code status preferences: 25 patients were less interested in receiving CPR, 1 patient was more interested and 2 gave no answer.

Attitudes at end of interview

A total of 95 (70.4%) patients believed it had been helpful to talk about goals of care and 112 (83.0%) believed it had been helpful to talk about the chances of surviving CPR in the hospital. Twenty-five (18.5%) patients said their preferences about CPR changed because of the interview, and 21 (15.5%) patients specifically stated they were less interested in receiving CPR. Five (3.7%) patients said the interview made them feel uncomfortable and 3 (2.2%) said they felt uncomfortable talking about CPR.

DISCUSSION

To our knowledge, this is the first study to investigate code status preferences and goals of care in a population of hospitalised adults. To reflect the conditions of clinical practice in the inpatient setting, we queried patients within 48 h of hospitalisation, and our findings suggest that most patients welcome discussions about code status decisions and goals of care. The need for such discussions is evidenced by discrepancies between code status orders and patients’ code status preferences, patients’ poor understanding of CPR and its outcome probabilities, the impact of CPR outcome information on patients’ treatment attitudes, and the heterogeneity of patients’ most important goals of care.

One of the barriers to appropriate code status decisions involves knowledge. Few patients were able to identify the three main components of CPR when used by doctors in hospitals, a finding consistent with knowledge deficits in other studies.23 29 30 Patients also greatly overestimated the probability of surviving a cardiac arrest in the hospital, again confirming results in other studies23 30; and once informed about likely outcomes, some patients were less interested in receiving CPR. The relationship between increased understanding of survival probabilities and decreased preference for CPR has been previously noted in non-hospitalised persons,22 31 and its importance is underscored by evidence that doctors may not provide information about CPR or survival during code status discussions.32

Another barrier to appropriate code status decisions involves communication, and the implications of failures in this domain are reflected by the observation that for 8.1% of patients in our study there was a discrepancy between their code status preferences and their code status as documented in their medical record. Fewer than a third of patients recalled discussing their code status with their doctor, which is within the range of other inpatient studies (16–47%)33 34 and higher than the 23% of patients in the SUPPORT study.2 13 As a remedy, some may see living wills as a means to improve communication of patients’ end-of-life preferences, but only 24.4% of our patients reported having a living will (consistent with other studies showing rates of 20–24%)33 35 and copies of living wills were found in only 4.4% of medical records in our study.

Goals of care represent an approach to code status discussions that may improve communication, and our results suggest that patients can understand and articulate their goals of care in this context. The finding that patients affirmed an average of 4.9 different goals supports the observation that patients may have multiple goals simultaneously,6 16 and the heterogeneity of patients’ single most important goals of care emphasises the importance of soliciting patients’ goals rather than presuming to understand them on the basis of general clinical impressions.

We did not find significant associations between particular goals of care and code status preferences. The absence of such findings may be due to the small proportion of patients whose single most important goal was to be comfortable (5.9%). Further investigation seems warranted, especially among patients with more advanced diseases such as the population in the SUPPORT study,2 of whom 30% preferred to forgo CPR compared with 14.1% in our study.

Our study had limitations. Our patient population represented a broad range of adults with acute diseases admitted to a general medical service. However, we did not study other clinical contexts (such as intensive, surgical or paediatric care), some important diseases were poorly represented (especially cancer and heart disease), only 14.1% of our participants were older than 64, and data were gathered from a single institution with little racial and ethnic diversity. Also, we did not assess near- or longer-term changes in patients’ code status preferences over time. Lastly, the outcome statistics we provided to patients contrast somewhat with more recent data on outcomes after inhospital resuscitation that report an 18.1% rate of survival to discharge and 13.9% rate of favourable neurological status.36 We also recognise that the outcome statistics we provided to patients were general rather than patient-specific; therefore, the statistics we provided may not reflect the individualised outcomes that might be predicted on the basis of each patient’s particular medical circumstances.

By providing a patient-centred telos, or end, a goal can define the worth of an intervention and guard against tendencies to isolate interventions from larger clinical realities. Addressing code status discussions in the context of such goals may help avoid focusing narrowly on resuscitation and, instead, may encourage consideration of important issues related to the trajectory of illness or the process of dying.37 Discussions of goals may also help resolve disagreements over code status orders, or at least clarify their sources. Admittedly, applying goals of care is not an automatic procedure. General goals require specification in a given clinical context,6 38 multiple goals may require prioritisation,6 16 and goals will shift over time as a patient’s disease progresses and personal preferences change.10 But our study suggests that addressing goals of care and the probable outcomes of CPR may help some patients during code status discussions. By situating these discussions within a goal-oriented framework, we may be able to increase the coherence between patients’ treatment choices, their medical conditions and the outcomes that are most important to them.