Abstract
Background: Prognosis discussion is an essential component of informed decision-making. However, many terminally ill patients have a limited awareness of their prognosis and the causes are unclear.
Objective: To explore the impact of physicians' propensity to discuss prognosis on advanced cancer patients' prognosis awareness.
Design: Cancer Care Outcomes Research and Surveillance Consortium (CanCORS) study, a prospective cohort study with patient and physician surveys.
Setting/Subjects: We investigated 686 patients with metastatic lung or colorectal cancer at diagnosis who participated in the CanCORS study and reported about their life expectancy. Data were linked to the physician survey from 486 physicians who were identified by these patients as filling important roles in their cancer care.
Results: Few patients with metastatic cancers (16.5%) reported an accurate awareness of their prognosis, defined as reporting a life expectancy of less than 2 years for lung cancer or less than 5 years for colorectal cancer. Patients whose most-important-doctor (in helping patient make decisions) reported discussing prognosis with terminally ill patients earlier were more likely than those whose doctors deferred these discussions to have an accurate prognosis awareness (adjusted proportion, 18.5% versus 7.6%; odds ratio, 3.23; 95% confidence interval, 1.39–7.52; p=0.006).
Conclusions: Few patients with advanced cancer could articulate an accurate prognosis estimate, despite most having received chemotherapy and many physicians reported they would discuss prognosis early. Physicians' propensity to discuss prognosis earlier was associated with more accurate patient reports of prognosis. Enhancing the communication skills of providers with important roles in cancer care may help to improve cancer patients' understanding of their prognosis.
Introduction
National guidelines recommend advance care planning, including clear and consistent discussions about prognosis, for patients with terminal illness and life expectancy of 1 year or less.1–3 Knowledge of prognosis is essential for terminally ill patients because prognosis awareness can influence their preferences for aggressive therapy versus supportive care.4–6 Other evidence suggests that truthful communication about prognosis is not harmful.7
Several studies have documented that patients with metastatic cancers tend to overestimate their prognosis.4,5,8 Limited prognosis awareness may reflect a patient's preference for not knowing prognosis, but it may also reflect the lack of an adequate discussion about prognosis with their physicians.8–15 In fact, physicians differ widely in their propensity to discuss prognosis with their patients with advanced cancer, and in the timing of these discussions.16,17
Although patients' understanding of prognosis is likely shaped by physician communication on this topic, prior work has generally examined patient and physician reports of communication separately. The Cancer Care Outcomes Research and Surveillance (CanCORS) Consortium,18,19 a large, multiregional, population- and health system-based cohort study of patients with lung and colorectal cancer, collected survey data from patients and their physicians, providing a unique opportunity to evaluate the potential association between physicians' reported behaviors and patients' perceptions of their prognosis. We studied CanCORS participants with metastatic cancer at diagnosis to test the hypothesis that patients of physicians who report generally discussing prognosis earlier have more accurate awareness of their life expectancy. We also examined whether other patient or physician factors were associated with better prognosis awareness.
Methods
Study design
The CanCORS study prospectively enrolled over 10,000 patients aged 21 years or older who were diagnosed with lung or colorectal cancer of any stage between 2003–2005 and lived in one of five areas (Northern California; Los Angeles County, California; North Carolina; Iowa; or Alabama) or received care in one of five large health maintenance organizations (HMOs) or 15 Veterans Affairs medical centers.18,19 Data were collected from baseline patient interviews performed approximately 3–6 months after diagnosis (or interviews of surrogates if patients were too ill or deceased; participation rate, 60%), medical record abstractions, and surveys of physicians with important roles in patients' cancer care (participation rate, 61%). This study was approved by the human subjects committee at all participating institutions.
Participants
Among 2670 patients newly diagnosed with stage IV lung or colorectal cancer, we focused on 1099 who were alive at the time of the baseline interview and completed the interview themselves, and then restricted the cohort to the 1066 who responded to the question about prognosis. We then linked these data with physician survey data from physicians who were identified by CanCORS participants (in baseline interviews) as filling key roles in their cancer care, including providing/discussing chemotherapy (chemotherapy doctor), surgery (surgeon), or radiation therapy; helping to make treatment decisions (“most-important-doctor”) or referring patients to these key providers.17 Because each patient may have more than one physician, we prespecified a hierarchy to identify the physician with whom to link based on physician responses from the: (1) most-important-doctor, (2) chemotherapy doctor, (3) primary care physician, and (4) surgeon.
Accordingly, we excluded 166 patients who did not identify any physician filling a key role in their care and 214 patients who did not have a completed physician survey from at least one physician filling one of these roles, leaving a final cohort of 686 patients for whom we had at least one physician survey (Fig. 1). Among these patients, 410 had a survey from the physician they identified as their most-important-doctor and that survey was used for analyses. For the remaining patients, we linked with surveys of other doctors based on the role of the physician, prioritizing physicians likely to be engaged in discussions about prognosis/treatment, including chemotherapy doctor (n=138), primary care physician (n=60), or surgeons (n=78).
FIG. 1.
Flowchart of the study cohort from Cancer Care Outcomes Research and Surveillance (CanCORS) Database. For each patient, a prespecified hierarchy was used to identify the physician with whom to link based on physician responses from the: (1) most-important-doctor, (2) chemotherapy doctor, (3) primary care physician, and (4) surgeon.
Prognosis awareness
Our primary outcome of interest was whether patients with metastatic cancer reported an accurate awareness of their prognosis (accurate versus not). Patients were asked, “Based on your understanding about what your doctors have told you about your cancer, your health in general, and the treatments you are receiving, how long do you think you have to live?” Patients' open-ended answers were coded as less than 1 year; 1 year or more but less than 2 years; 2 years or more but less than 5 years; 5 years or more. In pilot testing the instrument, some patients were unable to provide a time horizon despite prompting, and based on patients' responses, interviewers also coded responses of “in God's hands,” and “do not know.” Recognizing the median survival after the detection of distant metastases for lung cancer is approximately 4–10 months20–22 and for colorectal cancer is less than 2 years23,24 and at the time of the survey participants had already survived 3–6 months, we defined a relatively accurate perception of prognosis as less than 2 years for stage IV lung cancer and less than 5 years for stage IV colorectal cancer (Table 1). Patients who responded “in God's hands” or “do not know” were considered not to have an accurate prognosis awareness in the primary analysis.
Table 1.
Responses about Life Expectancy for Six Hundred Eighty-Six Patients with Stage IV Lung or Colorectal Cancer
| Lung Cancer | Colorectal Cancer | |
|---|---|---|
| Response about Life Expectancy | n (%) | n (%) |
| Less than 1 year | 40 (10.3)a | 9 (3.0)a |
| At least 1 year, but less than 2 years | 32 (8.2)a | 16 (5.4)a |
| At least 2 years, but less than 5 years | 41 (10.6) | 16 (5.4)a |
| At least 5 years | 76 (19.6) | 131 (44.0) |
| “Do not know” | 154 (39.7) | 95 (31.9) |
| “In God's hands” | 45 (11.6) | 31 (10.4) |
| Total | 388 (100) | 298 (100) |
The responses about life expectancy of less than 2 years for stage IV lung cancer and under 5 years for stage IV colorectal cancer were defined as a relatively accurate perception of prognosis.
Physician's propensity to discuss prognosis
To understand each physician's propensity for timing of discussions regarding prognosis, physicians were asked: “Assume you are caring for a patient who is newly diagnosed with metastatic cancer, but is currently feeling well. You estimate that the patient has 4 to 6 months to live. When, in the course of the typical patient's illness, are you most likely, for the first time, to discuss prognosis with this patient or family?” Response options included “now,” “when the patient first has symptoms,” “when there are no more nonpalliative treatments,” “only if the patient is hospitalized,” and “only if the patient or family bring it up.” We classified physicians responding “now” as likely to have earlier discussions of prognosis.17
Covariates
Covariates included variables specified a priori as potential confounders between physician propensity for timing of discussion and patients' prognosis awareness. Patient-level covariates included cancer type, age at diagnosis, gender, self-reported race/ethnicity, being married/living with a partner, education, income, primary insurer, HMO enrollment, speaking English at home, geographic region, self-reported comorbidity, receipt of chemotherapy before the interview, number of days from diagnosis to interview (quintiles), and number of days from interview until death. We also included physician characteristics, including whether the physician surveyed was named as the patient's most-important-doctor, self-reported specialty, gender, age, race/ethnicity, and the number of terminally ill patients cared for in the past year. Variables were categorized as in Table 2.
Table 2.
Characteristics and Associations with Prognosis Awareness for Patients with Stage IV Lung or Colorectal Cancera
| Characteristics | n (%) | Unadjusted proportion with accurate prognosis awareness | Adjusted proportion with accurate prognosis awareness | Odds Ratio | (95% CI) | p value |
|---|---|---|---|---|---|---|
| Role and response of the surveyed physicianb | 0.042 | |||||
| Most-important-doctor does not discuss “now” | 109 (15.9) | 9.2 | 7.6 | 1.00 | — | |
| Most-important-doctor discusses “now” | 301 (43.9) | 18.3 | 18.5 | 3.23 | (1.39–7.52) | |
| Not most-important-doctor does not discuss “now” | 79 (11.5) | 17.7 | 16.8 | 2.78 | (1.15–6.70) | |
| Not most-important-doctor discusses “now” | 197 (28.7) | 17.3 | 18.8 | 3.30 | (1.33–8.17) | |
| Characteristics of the surveyed physician | ||||||
| Specialty | 0.62 | |||||
| Primary care physician | 129 (18.8) | 14.7 | 16.4 | 1.00 | — | |
| Medical oncologist | 349 (50.9) | 17.8 | 14.9 | 0.87 | (0.40–1.89) | |
| Surgeon | 114 (16.6) | 14.9 | 20.8 | 1.42 | (0.61–3.26) | |
| Others | 94 (13.7) | 16.0 | 18.8 | 1.20 | (0.50–2.90) | |
| Physician gender | 0.099 | |||||
| Male | 567 (82.7) | 15.3 | 15.3 | 1.00 | — | |
| Female | 119 (17.3) | 21.9 | 21.8 | 1.68 | (0.91–3.13) | |
| Physician age, years | 0.192 | |||||
| 30–39 | 118 (17.2) | 20.8 | 19.8 | 1.00 | — | |
| 40–49 | 205 (29.9) | 12.1 | 11.2 | 0.46 | (0.22–0.95) | |
| 50–54 | 160 (23.3) | 17.7 | 17.5 | 0.84 | (0.39–1.81) | |
| 55–59 | 136 (19.8) | 20.0 | 20.0 | 1.02 | (0.47–2.17) | |
| ≥60 | 67 (9.8) | 12.2 | 17.4 | 0.83 | (0.32–2.14) | |
| Physician race | 0.156 | |||||
| White | 496 (72.3) | 16.4 | 16.0 | 1.00 | — | |
| Asian | 129 (18.8) | 19.6 | 20.9 | 1.48 | (0.82–2.67) | |
| Other | 61 (8.9) | 10.2 | 10.7 | 0.58 | (0.24–1.44) | |
| No. of terminally ill patients cared for in last year | 0.068 | |||||
| 0–10 | 168 (24.5) | 9.5 | 10.2 | 1.00 | — | |
| 11–20 | 118 (17.2) | 21.2 | 18.1 | 2.16 | (1.03–4.55) | |
| 21–60 | 194 (28.3) | 19.6 | 20.8 | 2.68 | (1.22–5.89) | |
| ≥61 | 206 (30.0) | 16.5 | 16.1 | 1.83 | (0.81–4.17) | |
| Patient characteristics | ||||||
| Cancer type | 0.165 | |||||
| Lung | 388 (56.6) | 18.6 | 14.7 | 1.00 | — | |
| Colorectal | 298 (43.4) | 13.8 | 20.0 | 1.57 | (0.83–2.97) | |
| Patient age, years | 0.183 | |||||
| 21–54 | 183 (26.7) | 13.7 | 17.1 | 1.00 | — | |
| 55–59 | 85 (12.4) | 23.5 | 26.9 | 2.02 | (0.94–4.34) | |
| 60–64 | 104 (15.2) | 21.1 | 23.3 | 1.59 | (0.78–3.23) | |
| 65–69 | 99 (14.4) | 11.1 | 10.0 | 0.48 | (0.18–1.27) | |
| 70–74 | 101 (14.7) | 18.8 | 13.7 | 0.73 | (0.27–1.99) | |
| 75–79 | 66 (9.6) | 15.2 | 14.8 | 0.82 | (0.27–2.51) | |
| ≥80 | 48 (7.0) | 12.5 | 9.3 | 0.45 | (0.14–1.40) | |
| Gender | 0.25 | |||||
| Male | 351 (51.2) | 16.5 | 14.9 | 1.00 | — | |
| Female | 335 (48.8) | 16.4 | 18.3 | 1.34 | (0.82–2.18) | |
| Race/ethnicity | 0.45 | |||||
| White | 496 (72.3) | 19.2 | 17.8 | 1.00 | — | |
| Black | 77 (11.2) | 5.2 | 8.3 | 0.37 | (0.11–1.21) | |
| Hispanic | 42 (6.1) | 11.9 | 13.2 | 0.66 | (0.17–2.50) | |
| Asian | 46 (6.7) | 13.0 | 17.0 | 0.94 | (0.30–2.88) | |
| Other | 25 (3.6) | 12.0 | 10.6 | 0.50 | (0.13–1.86) | |
| Married/live with partner | 0.094 | |||||
| Yes | 439 (64.0) | 17.7 | 18.5 | 1.00 | — | |
| No | 247 (36.0) | 14.3 | 13.0 | 0.61 | (0.34–1.09) | |
| Education | 0.33 | |||||
| <High school | 93 (13.6) | 9.7 | 13.2 | 1.00 | — | |
| High school/some college | 415 (60.5) | 16.1 | 15.4 | 1.23 | (0.57–2.68) | |
| ≥College degree | 178 (25.9) | 20.9 | 20.2 | 1.83 | (0.74–4.49) | |
| Income, $ | 0.71 | |||||
| <20 000 | 165 (23.6) | 16.3 | 19.2 | 1.00 | — | |
| 20,000–39,999 | 192 (29.0) | 13.8 | 14.6 | 0.67 | (0.34–1.35) | |
| 40,000–59,999 | 139 (19.8) | 18.6 | 16.8 | 0.82 | (0.40–1.70) | |
| ≥60,000 | 190 (27.6) | 17.6 | 15.7 | 0.75 | (0.33–1.70) | |
| Primary insurer | 0.090 | |||||
| Medicare | 307 (44.7) | 15.6 | 20.2 | 1.00 | — | |
| Private | 305 (44.5) | 16.7 | 12.8 | 0.52 | (0.24–1.14) | |
| Other | 74 (10.8) | 18.9 | 21.5 | 1.10 | (0.45–2.72) | |
| Health maintenance organization member | 0.23 | |||||
| No | 489 (71.3) | 14.3 | 15.1 | 1.00 | — | |
| Yes | 197 (28.7) | 21.8 | 19.2 | 1.41 | (0.81–2.46) | |
| Speaks English at home | 0.31 | |||||
| Yes | 666 (97.1) | 16.8 | 16.7 | 1.00 | — | |
| No | 20 (2.9) | 5.0 | 6.1 | 0.28 | (0.02–3.27) | |
| Region | 0.136 | |||||
| West | 421 (61.4) | 18.1 | 17.1 | 1.00 | — | |
| South | 156 (22.7) | 8.3 | 10.4 | 0.52 | (0.23–1.14) | |
| Other | 109 (15.9) | 22.0 | 20.9 | 1.34 | (0.69–2.61) | |
| Comorbidity | 0.41 | |||||
| None | 320 (46.6) | 13.1 | 14.1 | 1.00 | — | |
| One | 216 (31.5) | 19.5 | 18.8 | 1.51 | (0.86–2.64) | |
| Two | 98 (14.3) | 18.2 | 15.8 | 1.17 | (0.57–2.39) | |
| Three and more | 52 (7.6) | 21.1 | 20.7 | 1.74 | (0.76–3.97) | |
| Received chemotherapy before interview | 0.85 | |||||
| Yes | 601 (87.7) | 16.5 | 16.3 | 1.00 | — | |
| No | 85 (12.3) | 16.5 | 17.1 | 1.07 | (0.54–2.13) | |
| Days from diagnosis to interview | 0.35 | |||||
| 76–105 | 171 (24.9) | 13.5 | 16.8 | 1.00 | — | |
| 106–128 | 175 (25.5) | 14.9 | 13.5 | 0.74 | (0.38–1.45) | |
| 129–162 | 169 (24.7) | 20.1 | 20.3 | 1.32 | (0.69–2.53) | |
| 163–666 | 171 (24.9) | 17.5 | 15.7 | 0.91 | (0.45–1.82) | |
| Days from interview until death | <0.001 | |||||
| 1–89 | 75 (10.9) | 32.0 | 28.6 | 1.00 | — | |
| 90–179 | 97 (14.1) | 20.6 | 21.7 | 0.65 | (0.32–1.35) | |
| 180–269 | 77 (11.2) | 27.3 | 28.1 | 0.98 | (0.42–2.27) | |
| 270–364 | 52 (7.6) | 15.4 | 17.7 | 0.49 | (0.18–1.31) | |
| 1–2 years | 137 (20.0) | 10.9 | 10.8 | 0.25 | (0.11–0.59) | |
| ≥2 years | 248 (36.2) | 10.1 | 9.8 | 0.23 | (0.11–0.48) | |
Odds ratios and adjusted rates were calculated using the logistic regression adjusted for all variables listed in the table. Confidence intervals take into account the clustering of patients with physicians. The p value reflects a global test of the significance of the variable.
Physicians reported their favored timing of prognosis discussion with a terminally ill patient who they believed to have 4–6 months to live.
CI, confidence interval.
Statistical analysis
We first described the proportion of patients reporting an accurate prognosis awareness by patient and physician characteristics. We used multivariable logistic regression to identify if physicians' report of discussing prognosis “now” was associated with patients' prognosis awareness, adjusting for all other physician and patient characteristics described above. Because we set a prespecified hierarchy of the role of physicians (most-important-doctor or not) assuming that most-important-doctors should be particularly influential, an interaction term between most-important-doctor and physicians' propensity to discuss prognosis was included in the model. Generalized estimating equations with robust variance estimators were used to account for the clustering of patients with physicians. We then calculated adjusted rates of prognosis awareness for each covariate by direct standardization using the regression model.25
In a series of sensitivity analyses, we repeated analyses for four subpopulations: (1) patients for whom we had the survey response from their most-important-doctor; (2) patients who died less than 2 years after diagnosis; (3) excluding patients who answered “in God's hands” to the prognosis question (rather than coding this as inaccurate prognosis awareness); and (4) further excluding patients who answered “do not know” from the above subpopulation. We also conducted the multinomial logistic model with a three-category dependent variable (aware, not aware, and “in God's hands”/“do not know”). Item nonresponse was infrequent (<3% for most variables except income at 8%); we used multiple imputation26 to impute missing data for all items except the dependent variable. All analyses were performed on the multiply imputed datasets with SAS statistical software, version 9.2 (SAS Institute Inc., Cary, NC); all statistical tests were two-sided.
Results
Cohort characteristics
The study cohort included 686 patients with stage IV lung or colorectal cancer with a median age of 63 years; 49% were women, 57% had lung cancer, and 88% had received chemotherapy before the interview (Table 2). The median time from diagnosis to interview was 4.3 months (interquartile range, 3.5–5.4 months). Overall, 589 (86%) identified a most-important-doctor, and for 410 (70%) of these patients, we had a completed survey from that most-important-doctor.
Compared with the 380 patients excluded because they could not identify an important physician or the physician did not complete the survey, the 686 included patients had more women, fewer black patients, and more patients with higher education or incomes (p<0.05 for all comparisons) but the groups had similar prognosis awareness and duration from interview until death.
Linked physician survey
These 686 patients were linked with 486 doctors who played a major role in their cancer care and responded to the CanCORS physician survey. On average, each doctor was linked to 1.4 patients (range, 1–8) although most physicians (77%) were linked to one patient. The median age of the 486 physicians was 50 years, 83% were men, 40% were medical oncologists, 26% were primary care physicians, 19% were surgeons, and 15% reported another specialty. Only 8% of the 388 patients with lung cancer were linked with a surgeon, versus 28% among patients with colorectal cancer.
Most physicians (70.8%) reported they would discuss prognosis “now” with a terminally ill cancer patient, whereas other physicians would wait until the patient has symptoms or there are no further treatments, and 12.8% would only discuss prognosis if the patient or family brings it up (Fig. 2).
FIG. 2.
Physician response about the timing of prognosis discussion with a terminally ill patient who has 4–6 months to live (n=486).
Patients' prognosis awareness
Of the 686 patients with stage IV lung or colorectal cancer, 113 (16.5%) had an accurate awareness of their prognosis, i.e., reporting a life expectancy of less than 2 years for lung cancer or less than 5 years for colorectal cancer (Table 1). The proportion reporting an accurate prognosis awareness was similar for patients with lung and colorectal cancer (19% versus 14%; p=0.10). Many patients were unable to provide an estimate, reporting without prompting that their prognosis was “in God's hands” (11%) or “do not know” (36%).
In the multivariable model adjusted for patient and physician characteristics, patients whose most-important-doctor reported generally discussing prognosis “now,” compared with patients whose most-important-doctor did not report discussing prognosis “now,” were more likely to report an accurate prognosis awareness (18.5% versus 7.6%; odds ratio [OR]=3.23; 95% confidence interval [CI]=1.39–7.52; p=0.006; Table 2). There was no significant association between physician specialty and patient prognosis awareness; however, patients whose physician cared for more than 10 terminally-ill patients in the last year were more likely to have an accurate prognosis awareness. Individuals who died within 3 months after the interview were most likely to be aware of their prognosis, and those who lived more than 2 years were least likely to be (28.6% versus 9.8%; p<0.001).
In sensitivity analyses when we repeated analyses for four subpopulations described above, results were similar, with patients whose most-important-doctor reported discussing prognosis “now” more likely to have an accurate prognosis awareness than those whose most-important-doctor did not report discussing prognosis “now” (Table 3). Moreover, when we used multinomial logistic regression to examine prognosis awareness categorized as aware, not aware, or do not know/“in God's hands,” we found that patients whose most-important-doctor reported discussing prognosis “now” were more likely to report an accurate prognosis awareness (OR=3.17; 95% CI=1.36–7.37; p=0.007) and less likely to answer “do not know”/“in God's hands” (OR=0.51; 95% CI=0.31–0.83; p=0.007), compared with patients whose most-important-doctor reported not discussing prognosis “now” (Table 4).
Table 3.
Sensitivity Analyses for the Association between Most-Important-Doctors' Propensity to Discuss Prognosis and Patients' Prognosis Awarenessa
| Adjusted proportion with accurate prognosis awareness | ||||||
|---|---|---|---|---|---|---|
| Subpopulations | n | Most-important-doctor discusses “now” | Most-important-doctor does not discuss “now” | Odds ratio | (95% CI) | p value |
| Patients having a survey response from their most-important-doctor | 410 | 19.0 | 7.7 | 3.33 | (1.17– 9.53) | 0.02 |
| Patients who died within 2 years after diagnosis | 399 | 24.6 | 7.7 | 4.23 | (1.55–11.53) | 0.005 |
| Excluding patients who answered “in God's hands” | 610 | 20.6 | 8.5 | 3.33 | (1.39–7.97) | 0.007 |
| Excluding patients who answered “in God's hands” or “do not know” | 361 | 31.4 | 17.6 | 2.74 | (1.03–7.31) | 0.04 |
Models were adjusted for physician specialty, age, sex, race, the number of terminally ill patients cared for by the surveyed physician in last year, cancer type, patient age, gender, race/ethnicity, marital status, speaking English at home, education, income, insurance, health maintenance organization enrollment, geographic region, receipt of chemotherapy, comorbidity, the number of days from diagnosis to interview, and the number of days from interview until death. Confidence intervals take into account the clustering of patients with physicians.
CI, confidence interval.
Table 4.
Physicians' Propensity to Discuss Prognosis versus Patients' Prognosis Awareness in the Multinomial Logistic Modela
| Inaccurate prognosis awareness | Accurate prognosis awareness | Do not know or “In God's hands” | ||||
|---|---|---|---|---|---|---|
| Role and response of the surveyed physicianb | n | Adjusted proportion | Adjusted proportion | Odds ratio (95% CI) | Adjusted proportion | OddsrRatio (95% CI) |
| Most-important-doctor does not discuss “now” | 109 | 35.4 | 7.6 | — | 57.0 | — |
| Most-important-doctor discusses “now” | 301 | 38.6 | 18.4 | 3.17 (1.36–7.37) | 42.9 | 0.51 (0.31–0.83) |
| Not most-important-doctor does not discuss “now” | 79 | 39.4 | 17.1 | 2.85 (1.19–6.82) | 43.5 | 0.52 (0.29–0.94) |
| Not most-important-doctor discusses “now” | 197 | 30.6 | 18.9 | 3.29 (1.33–8.16) | 50.4 | 0.73 (0.43–1.25) |
Models were adjusted for physician specialty, age, gender, race, the number of terminally ill patients cared for by the surveyed physician in last year, cancer type, patient age, gender, race/ethnicity, marital status, speaking English at home, education, income, primary insurer, health maintenance organization enrollment, geographic region, receipt of chemotherapy, comorbidity, the number of days from diagnosis to interview, and the number of days from interview until death.
Physicians reported their favored timing of prognosis discussion with a terminally ill patient who they believed to have 4–6 months to live.
CI, confidence interval.
Discussion
Current guidelines recommend that physicians be open and honest in communicating about prognosis and treatment options with advanced cancer patients.1–3 In a large multiregional population- and health system-based cohort of patients with metastatic lung or colorectal cancer, we found that few (16%) patients who were alive 3–6 months after diagnosis had an accurate understanding of their prognosis although 88% had received chemotherapy. Nonetheless, we also found that patients of physicians who reported a greater propensity for earlier discussions of prognosis were more likely to have a relatively accurate awareness of life expectancy than patients whose physicians reported discussing prognosis later in their course of illness. This suggests that physicians' communication behaviors may play an important role in explaining the very low rate of prognostic understanding we observed among patients with incurable cancers.
Does physician communication behavior account for the whole of patient prognosis awareness? Over 70% of physicians in our study reported that they would discuss prognosis “now” with a terminally ill patient who they believed to have 4–6 months to live, yet only 16.5% of patients reported an accurate awareness of their prognosis. The low degree of prognosis awareness is consistent with other studies suggesting that prognostic estimates of advanced cancer patients tend to be overly optimistic.4–5,8 Other evidence suggests that patients may misinterpret or fail to absorb information provided by physicians, and some patients prefer to delay or even to evade prognostic discussions.10–12 Moreover, patients may have had prognosis discussions, but are unwilling to acknowledge their mortality and instead take avoidant coping strategies.
Many patients reported that their prognosis was “in God's hands” or that they did not know their prognosis. This may partly explain the low rates of prognosis awareness in our study. We cannot be certain if they were unable to articulate their prognosis because they had never discussed it, if they were unable to discuss it with their interviewer, or if they were referring to the inherent uncertainty of prognostication. Nonetheless, the association between most-important-doctors' propensity to discuss prognosis and patients' prognosis awareness was robust to several sensitivity analyses where we reclassified these responses.
Physicians' reports of their preferred timing of prognosis discussion based on the assumed vignette may be subject to social desirability bias and may not perfectly reflect the actual practices for their patients. The finding that few patients had prognosis awareness despite most physicians reported generally discussing prognosis early warrants further investigations on the quality of communications about prognosis between doctors and patients. Alternatively, some patients may have discussed prognosis early in the course of illness after diagnosis but not recall the discussion accurately, especially if they responded well to anticancer therapies.
The finding that patients who died within 90 days following their survey reported the highest prognosis awareness suggests that patients with advanced cancer may gradually perceive a poor prognosis as their physical condition deteriorates.27 Otherwise, this finding may reflect the fact that many physicians tend to delay discussions about prognosis or end-of-life issues until patients' clinical circumstances get direr.9,16,17 Physicians' reluctance or delay in discussing prognosis may have negative consequences for patients with advanced cancer, such as unwanted aggressive treatment and delays in advance care planning.4
Previous evidence suggested that surgeons and medical oncologists were more likely than noncancer specialists to report discussing prognosis earlier.17 In the CanCORS study, patients were asked to identify the physician who was most important in helping them to make decisions about their cancer. Although patients with cancer most often identified a medical oncologist as their most-important-doctor, physician specialty was not associated with patients' prognosis awareness after adjusting for other physician/patient factors, highlighting the importance of understanding better the roles rather than specialties of physicians in end-of-life care for patients with cancer. Nevertheless, we observed that patients whose physicians cared for more than 10 terminally ill patients in the last year were more likely to report an accurate prognosis awareness, suggesting that greater experience caring for terminally ill patients may improve physicians' prognosis communication.
Given the association between physicians' propensity to discuss prognosis and patients' prognosis awareness, the strikingly lower rate of reporting an accurate estimate of life expectancy among patients whose most-important-doctor did not report discussing prognosis “now” is particularly noteworthy. This suggests that some patients may be hindered from having an accurate understanding of their prognosis if the doctors whom they are relying on for their key decisions about their cancer tend to delay or are reluctant to discuss prognosis with them. It is also possible that patients who want to avoid prognosis discussions seek out physicians who avoid such discussions, implying “collusion” between patients with cancer and their physicians.28
Important strengths of our study include the rich clinical and sociodemographic data from a large multiregional representative population and the linking of patient and physician data. Nevertheless, our study has several limitations. First, we could not directly investigate communication processes between doctors and patients because it was not feasible to audiotape clinical encounters or collect patient-specific information in the physician surveys. Second, only patients who were alive and able to complete the interview themselves provided an estimate of their life expectancy. Therefore, the results cannot be generalized to patients who died soon after diagnosis or those who were too sick to complete the patient interview. Third, the multifaceted nature of the question on prognosis awareness may augment patients' overly optimistic responses. Fourth, we did not adjust for depression/anxiety, symptom burden, or response to prior chemotherapy, which could influence prognosis awareness. Finally, the linkage between patients and physicians was subject to nonresponse bias from patient and physician surveys, although prognosis awareness was similar between the study cohort and patients excluded because we had no survey data from their key physicians.
Conclusion
Prognosis awareness is an essential component of informed decision-making.4–6 Although increasingly, studies have documented that most patients want to receive prognostic information and that truthful communication is not harmful,7,11,12 in this large, population-based study of patients with metastatic lung or colorectal cancer, we found that many had a limited prognosis awareness soon after diagnosis, even though most of them had received chemotherapy. Our findings suggest that physician communication behaviors are associated with patients' perception of prognostic information. Identifying health care providers with important roles in cancer care and enhancing their communication skills may help to improve patients' understanding of their prognosis.13–15 Future research is needed to better understand the communication styles and content of prognosis discussions to maximize patients' understanding of their prognosis and improve the end-of-life care experiences for cancer patients.
Acknowledgments
This work of the Cancer Care Outcomes Research and Surveillance (CanCORS) Consortium was supported by the National Cancer Institute grants to the Statistical Coordinating Center (U01-CA093344) and Primary Data Collection and Research Centers (Dana-Farber Cancer Institute/Cancer Research Network U01-CA093332, Harvard Medical School/Northern California Cancer Center U01-CA093324, RAND/UCLA U01-CA093348, University of Alabama at Birmingham U01-CA093329, University of Iowa U01-CA093339, University of North Carolina U01-CA093326), and a Department of Veteran's Affairs grant to the Durham VA Medical Center CRS 02-164. The effort of Drs. Keating, Landrum, and Huskamp was also funded by 1R01CH164021-01A1.
These data were presented in abstract form as an oral presentation at the Society of General Internal Medicine 36th Annual Meeting, April 26, 2013, Denver, Colorado.
Author Disclosure Statement
No competing financial interests exist.
References
- 1.National Comprehensive Cancer Network: NCCN Clinical Practice Guidelines in Oncology: Palliative Care, Version 2.2013. www.nccn.org/professionals/physician_gls/pdf/palliative.pdf (Last accessed July31, 2013) [DOI] [PubMed]
- 2.National Consensus Project for Quality Palliative Care: Clinical Practice Guidelines for Quality Palliative, ThirdEdition. www.nationalconsensusproject.org (Last accessed July31, 2013)
- 3.Peppercorn JM, Smith TJ, Helft PR, Debono DJ, Berry SR, Wollins DS, Hayes DM, Von Roenn JH, Schnipper LE; American Society of Clinical Oncology: American Society of Clinical Oncology Statement: Toward individualized care for patients with advanced cancer. J Clin Oncol 2011;29:755–760 [DOI] [PubMed] [Google Scholar]
- 4.Weeks JC, Cook EF, O'Day SJ, Peterson LM, Wenger N, Reding D, Harrell FE, Kussin P, Dawson NV, Connors AF, Jr, Lynn J, Phillips RS: Relationship between cancer patients' predictions of prognosis and their treatment preferences. JAMA 1998;279:1709–1714 [DOI] [PubMed] [Google Scholar]
- 5.Temel JS, Greer JA, Admane S, Gallagher ER, Jackson VA, Lynch TJ, Lennes IT, Dahlin CM, Pirl WF: Longitudinal perceptions of prognosis and goals of therapy in patients with metastatic non-small-cell lung cancer: Results of a randomized study of early palliative care. J Clin Oncol 2011;29:2319–2326 [DOI] [PubMed] [Google Scholar]
- 6.Glare PA, Sinclair CT: Palliative medicine review: Prognostication. J Palliat Med 2008;11:84–103 [DOI] [PubMed] [Google Scholar]
- 7.Apatira L, Boyd EA, Malvar G, Evans LR, Luce JM, Lo B, White DB: Hope, truth, and preparing for death: Perspectives of surrogate decision makers. Ann Intern Med 2008;149:861–868 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Weeks JC, Catalano PJ, Cronin A, Finkelman MD, Mack JW, Keating NL, Schrag D: Patients' expectations about effects of chemotherapy for advanced cancer. N Engl J Med 2012;367:1616–1625 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Bradley EH, Hallemeier AG, Fried TR, Johnson-Hurzeler R, Cherlin EJ, Kasl SV, Horwitz SM: Documentation of discussions about prognosis with terminally ill patients. Am J Med 2001;111:218–223 [DOI] [PubMed] [Google Scholar]
- 10.Fried TR, Bradley EH, O'Leary J: Prognosis communication in serious illness: Perceptions of older patients, caregivers, and clinicians. J Am Geriatr Soc 2003;51:1398–1403 [DOI] [PubMed] [Google Scholar]
- 11.Innes S, Payne S: Advanced cancer patients' prognostic information preferences: A review. Palliat Med 2009;23:29–39 [DOI] [PubMed] [Google Scholar]
- 12.Hagerty RG, Butow PN, Ellis PA, Lobb EA, Pendlebury S, Leighl N, Goldstein D, Lo SK, Tattersall MH: Cancer patient preferences for communication of prognosis in the metastatic setting. J Clin Oncol 2004;22:1721–1730 [DOI] [PubMed] [Google Scholar]
- 13.Back AL, Arnold RM: Discussing prognosis: “How much do you want to know?” Talking to patients who do not want information or who are ambivalent. J Clin Oncol 2006;24:4214–4217 [DOI] [PubMed] [Google Scholar]
- 14.Back AL, Arnold RM: Discussing prognosis: “How much do you want to know?” Talking to patients who are prepared for explicit information. J Clin Oncol 2006;24:4209–4213 [DOI] [PubMed] [Google Scholar]
- 15.Smith AK, White DB, Arnold RM: Uncertainty—The other side of prognosis. N Engl J Med 2013;368:2448–2450 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Daugherty CK, Hlubocky FJ: What are terminally ill cancer patients told about their expected deaths? A study of cancer physicians' self-reports of prognosis disclosure. J Clin Oncol 2008;26:5988–5993 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Keating NL, Landrum MB, Rogers SO, Jr., Baum SK, Virnig BA, Huskamp HA, Earle CC, Kahn KL: Physician factors associated with discussions about end-of-life care. Cancer 2010;116:998–1006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ayanian JZ, Chrischilles EA, Fletcher RH, Fouad MN, Harrington DP, Kahn KL, Kiefe CI, Lipscomb J, Malin JL, Potosky AL, Provenzale DT, Sandler RS, van Ryn M, Wallace RB, Weeks JC, West DW: Understanding cancer treatment and outcomes: The Cancer Care Outcomes Research and Surveillance Consortium. J Clin Oncol 2004;22:2992–2996 [DOI] [PubMed] [Google Scholar]
- 19.Catalano PJ, Ayanian JZ, Weeks JC, Kahn KL, Landrum MB, Zaslavsky AM, Lee J, Pendergast J, Harrington DP; Cancer Care Outcomes Research Surveillance Consortium: Representativeness of participants in the cancer care outcomes research and surveillance consortium relative to the surveillance, epidemiology, and end results program. Med Care 2013;51:e9–15 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kiely BE, Alam M, Blinman P, Tattersall MH, Stockler MR: Estimating typical, best-case and worst-case life expectancy scenarios for patients starting chemotherapy for advanced non-small-cell lung cancer: A systematic review of contemporary randomized trials. Lung Cancer 2012;77:537–544 [DOI] [PubMed] [Google Scholar]
- 21.NSCLC Meta-Analyses Collaborative Group: Chemotherapy in addition to supportive care improves survival in advanced non-small-cell lung cancer: A systematic review and meta-analysis of individual patient data from 16 randomized controlled trials. J Clin Oncol 2008;26:4617–4625 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Schiller JH, Harrington D, Belani CP, Langer C, Sandler A, Krook J, Zhu J, Johnson DH; Eastern Cooperative Oncology Group: Comparison of four chemotherapy regimens for advanced non-small-cell lung cancer. N Engl J Med 2002;346:92–98 [DOI] [PubMed] [Google Scholar]
- 23.Golfinopoulos V, Salanti G, Pavlidis N, Ioannidis JP: Survival and disease-progression benefits with treatment regimens for advanced colorectal cancer: A meta-analysis. Lancet Oncol 2007;8:898–911 [DOI] [PubMed] [Google Scholar]
- 24.Kopetz S, Chang GJ, Overman MJ, Eng C, Sargent DJ, Larson DW, Grothey A, Vauthey JN, Nagorney DM, McWilliams RR: Improved survival in metastatic colorectal cancer is associated with adoption of hepatic resection and improved chemotherapy. J Clin Oncol 2009;27:3677–3683 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Little RJ: Direct standardization: A tool for teaching linear models for unbalanced data. Am Stat 1982;36:38–43 [Google Scholar]
- 26.He Y, Zaslavsky AM, Landrum MB, Harrington DP, Catalano P: Multiple imputation in a large-scale complex survey: A practical guide. Stat Methods Med Res 2010;19:653–670 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Burns CM, Broom DH, Smith WT, Dear K, Craft PS: Fluctuating awareness of treatment goals among patients and their caregivers: A longitudinal study of a dynamic process. Support Care Cancer 2007;15:187–196 [DOI] [PubMed] [Google Scholar]
- 28.The AM, Hak T, Koeter G, van Der Wal G: Collusion in doctor-patient communication about imminent death: An ethnographic study. BMJ 2000;321:1376–1381 [DOI] [PMC free article] [PubMed] [Google Scholar]