Abstract
Context
Survival prognostication is important during end-of-life. The accuracy of clinician prediction of survival (CPS) over time has not been well characterized.
Objectives
To examine changes in prognostication accuracy during the last 14 days of life in a cohort of patients with advanced cancer admitted to two acute palliative care units and to compare the accuracy between the temporal and probabilistic approaches.
Methods
Physicians and nurses prognosticated survival daily for cancer patients in two hospitals until death/discharge using two prognostic approaches: temporal and probabilistic. We assessed accuracy for each method daily during the last 14 days of life comparing accuracy at day −14 (baseline) with accuracy at each time point using a test of proportions.
Results
6718 temporal and 6621 probabilistic estimations were provided by physicians and nurses for 311 patients, respectively. Median (interquartile range) survival was 8 (4, 20) days. Temporal CPS had low accuracy (10–40%) and did not change over time. In contrast, probabilistic CPS was significantly more accurate (p<.05 at each time point) but decreased close to death.
Conclusion
Probabilistic CPS was consistently more accurate than temporal CPS over the last 14 days of life; however, its accuracy decreased as patients approached death. Our findings suggest that better tools to predict impending death are necessary.
Keywords: longitudinal, prognosis, advanced cancer, inpatients, accuracy
Introduction
Survival prognostication is important in patients with advanced cancer, particularly during the last few weeks of life. For patients and families, having prognostic information influences treatment preferences, decreases uncertainty and helps them to plan ahead for both personal and healthcare matters (1–3). For healthcare providers, particularly for oncologists, short term prognostication during the last few weeks of a patient’s life is relevant for clinical decision making because discharge planning, code status discussions, goals of care, hospice transfers and enrollment onto integrated care pathways are all dependent on prognosis (2, 4). For institutions, accuracy in prognostication may help to redirect the use of resources from aggressive end-of-life measures to patient comfort.
Clinical prediction of survival (CPS) can be expressed in two ways: (a) temporal CPS: providing an estimated duration of survival or (b) probabilistic CPS: providing the probability that a patient would survive for a pre-defined length of time (e.g. 90% chance of being alive in 48 hours, 20% chance of being alive at 1 month). Healthcare providers are inaccurate in estimating survival for cancer patients (5–10). Temporal CPS has reported accuracies between 20% and 30% (5, 6, 8, 11). Few studies have examined the accuracy of the probabilistic CPS in patients with advanced cancer (6, 12). In a prior study from our group, we showed that clinicians were more accurate in estimating survival with probabilistic CPS than with temporal CPS (6).
Prognostication is a dynamic process. It changes as a patient progress through the stages of the disease, particularly in the last weeks of life when patients deteriorate rapidly. Most studies evaluating accuracy of prognostication have assessed CPS at a single point in time (13). Few studies have serially measured accuracy of prognostication (14–16). However, the methodologies used in these studies are heterogeneous and the conclusions are diverse. A better understanding of how the accuracy of these different prognostication strategies vary over time may allow us to improve our ability to prognosticate. The aim of this study is to examine the changes in prognostication accuracy over time during the last 14 days of life in a cohort of patients with advanced cancer admitted to acute palliative care units using two prognostication strategies. Our secondary objective is to compare the accuracy between temporal CPS and probabilistic CPS and to compare between physicians and nurses over time.
Methods
Patients
We enrolled consecutive patients with a diagnosis of advanced cancer who were 18 years of age or older and were admitted to the Acute Palliative Care Units (APCU) at MD Anderson Cancer Center (MDACC) in the United States between May 5, 2010 and July 6, 2010, and Barretos Cancer Center (BCC) in Brazil between January 27, 2011 and July 1, 2011. Both APCUs are dedicated units staffed by an interdisciplinary team including physicians trained in palliative care, nurses, social workers and other professionals, that provide intensive symptom support and transition of care for patients with advanced cancer and their families. Nurses and physicians in both units rotated in the APCUs a few days at a time, and ensured continuity of care by signing over cases routinely.
The Institutional Review Boards at both institutions approved this study and provided waiver of consent for patient participation. This approach was adopted for this non-interventional study to minimize distress during the consent process and to ensure that we could collect data on consecutive patients. All clinicians who participated in this study signed the informed consent prior to patient enrollment. Patient demographics, including age, gender, race, education, religion, cancer diagnosis and length of stay, were obtained from chart review.
Outcomes
Physicians and nurses were asked to prognosticate daily or twice daily, respectively, for cancer patients from APCUs admission to death or discharge using both the temporal approach and the probabilistic approach. Temporal CPS was obtained by asking clinicians “What is the approximate survival of this patient (in days)?” Probabilistic CPS was obtained by asking clinicians “What is the probability that this patient will be alive (0% to 100%) in 24 hours?” and “What is the probability that this patient will be alive (0% to 100%) in 48 hours?” (6) Patients who were discharged alive were followed until death or lost to follow-up.
For probabilistic CPS, we asked clinicians to make daily predictions over the next 24 and 48 hours instead of longer time frames (e.g. weeks). This is because longer time frames would overlap considerably with daily predictions, and would not reflect the frequent changes in prognostication for patients during the last 14 days of life.
Accuracy of survival estimation was determined based on previously defined criteria: Temporal CPS was considered accurate if it fell within ±33% of the AS (5, 6, 17). Probabilistic CPS was considered accurate if the clinician endorsed a survival probability ≤30% and the patient died in the specific time frame (e.g. 24 h or 48 h), or the clinician selected a survival probability ≥70% and the patient survived beyond the specific time frame (6).
To determine the daily accuracy of each method during the last two weeks of life, the date of death is needed. Thus, we excluded patients who were either still alive at day of last follow-up or date of death was not available.
Statistical Analysis
To characterize the relationship between CPS and AS we estimated the correlation between the first temporal CPS estimated by clinicians and AS using the Spearman correlation test. To assess whether or not the accuracy of each test remained stable over time we compared the daily accuracy with the accuracy at day −14 (baseline) using a test of proportions. We compared the daily accuracy of temporal CPS and the 24-hrs and 48-hrs probabilistic CPS during the last two weeks of life using McNemar’s test. This same test was also used to compare the daily accuracy between physicians and nurses with each prognostic strategy over time during the last two weeks of life. The Statistical Analysis System (SAS version 9.3, SAS Institute, Cary, North Carolina) was used for statistical analysis. A P-value of <0.05 was considered significant in this study.
Results
Patient Characteristics
Among the 357 patients enrolled onto the study, 203 patients died during the APCU stay. Among the patients discharged alive, the date of death was available for 108 patients. 46 patients were loss to follow-up and excluded from this analysis.
The baseline characteristics of the 311 patients included in the analysis are described in Table 1. The median (interquartile range [IQR]) survival was 8 (4, 20) days. Included patients had similar demographic characteristics than those excluded, except for length of APCU stay ((median, IQR): 5 (3, 9) v/s 8 (5, 13), p < 0.001, rank sum test) and ethnicity (χ 2, p < 0.01).
Table 1.
Demographic characteristics
| Demographics N (%) |
|
|---|---|
| N | 311 |
| Female | 165 (53) |
| Mean Age (SD) | 58 (14) |
| Ethnicity | |
| White | 88 (28) |
| Black | 16 (5) |
| Hispanic | 204 (66) |
| Others | 3 (1) |
| Married | 180 (59) |
| Christian | 285 (92) |
| Education | |
| High School or less | 212 (77) |
| College | 50 (18) |
| Advanced Degree | 15 (5) |
| Cancer Type | |
| Breast | 34 (11) |
| Gastrointestinal | 92 (30) |
| Genitourinary | 32 (11) |
| Gynecological | 34 (11) |
| Head and Neck | 24 (8) |
| Hematological | 16 (5) |
| Respiratory | 40 (13) |
| Other | 39 (13) |
| Months from diagnosis [median (IQR)] | 15 (6, 34) |
| Length of APCU stay in days [median (IQR)] | 5 (3, 9) |
| Dead at APCU discharge | 203 (65) |
| Survival in days [median (IQR)] | 8 (4, 20) |
| Palliative Performance Scale on admission [median (IQR)] | 40 (25, 60) |
SD: standard deviation
IQR: interquartile range
APCU: Acute Palliative Care Unit.
Clinician Characteristics
Sixty-nine clinicians, 29 nurses and 40 physicians, participated in this study. Thirty clinicians participated at MDACC and 39 clinicians at BCC. Clinician characteristics are described in Table 2. Nurses’ and doctors’ years of palliative care experience were similar ((median, IQR): 3 (1, 7) v/s 1 (1, 5), p = 0.26, rank sum test) (Table 2).
Table 2.
Clinician characteristics
| RN (N = 29) N, (%) |
MD (N = 40) N, (%) |
|
|---|---|---|
| Age (mean, SD) | 37 (12) | 35 (9) |
| Female | 24 (83) | 13 (33) |
| Ethnicity (%) | ||
| White | 5 (17) | 3 (7) |
| Black | 4 (14) | 1 (2) |
| Hispanic | 9 (31) | 31 (78) |
| Others | 11 (38) | 5 (13) |
| Christian (%) | 22 (76) | 25 (63) |
| Years of clinical experience [median, (IQR)] | 6 (3, 15) | 5 (4, 10) |
| Years of palliative care experience [median, (IQR)] | 3 (1, 7) | 1 (1, 5) |
Clinical Prediction of Survival
The first temporal CPS was significantly correlated with AS for physicians (r=0.58, p<0.0001) and for nurses (r=0.52, p<0.0001).
Clinicians overestimated survival using temporal CPS in our cohort of advanced cancer patients during the last 14 days of life. Median estimation of survival was slightly higher than AS through the observation period, even when the patient was near death (Figure 1a).
Figure 1.
Clinician prediction of survival using the (A) temporal approach, the (B) 24-hour probabilistic approach and the (C) 48-hour probabilistic approach by time to death. Black lines represent the perfect prognosis. Grey dashed lines are the thresholds for accuracy. Dark colored lines are used for physicians and light colored lines are used for nurses.
Clinician estimation of probability of survival using the 24-hour and 48-hour time frames decreased as death approached. Nurses were perfectly accurate in prognosticating using the probabilistic approach until the last five days of life (Figure 1b). Physicians were less optimistic than nurses at defining a probability of survival (Figures 1b and 1c).
Accuracy of Temporal and Probabilistic CPS
Physicians made 2199 temporal estimations of survival and 2186 probabilistic estimations of survival, with a median (IQR) of 5 (3–8) estimations of survival per patient with each method. Nurses made 4519 temporal estimations of survival and 4435 probabilistic estimations of survival, with a median (IQR) of 9 (5–16) estimations of survival per patient with each method. For physicians, the proportions of accurate predictions with the temporal, 24-hour and 48-hour probabilistic approaches were 23%, 73% and 67%, respectively. For nurses, the proportions of accurate predictions with the temporal, 24-hour and 48-hour probabilistic approaches were 24%, 90% and 83%, respectively.
Changes in Accuracy Over Time
Physicians’ accuracy with temporal CPS was relatively stable over time with accuracies between 10% and 35% (Table 3). Using day −14 as the baseline accuracy, statistically significant differences on physician’s accuracy with temporal CPS were found on some days during the last week, although this was not a consistent finding. Physicians’ accuracy with 24-hour probabilistic CPS was significantly lower during the last week of life compared to baseline (Table 3). Physicians’ accuracy with 48-hour probabilistic CPS was significantly lower during the last week of life compared to baseline, except for the last day of life (Table 3).
Table 3.
Comparison of baseline accuracy (day −14) versus accuracy at each time point by clinician.
| Accuracy of Prognostication by Physicians | Accuracy of Prognostication by Nurses | |||||||
|---|---|---|---|---|---|---|---|---|
| Days to death |
N | Temporal %a (p)b |
24-hour Probabilistic %a (p)b |
48-hour Probabilistic %a (p)b |
N | Temporal %a (p)b |
24-hour Probabilistic %a (p)b |
48-hour Probabilistic %a (p)b |
| −14 (baseline) | 45 | 31 | 88 | 79 | 43 | 29 | 98 | 95 |
| −13 | 47 | 22 (.40) | 85 (.59) | 76 (.78) | 46 | 26 (.85) | 100 (.48)c | 91 (.68)c |
| −12 | 53 | 20 (.38) | 82 (.39) | 70 (.35) | 52 | 24 (.59) | 98 (>.99)c | 90 (.35)c |
| −11 | 61 | 32 (.55) | 81 (.33) | 63 (.10) | 60 | 32 (.56) | 98 (>.99)c | 96 (1)c |
| −10 | 67 | 18 (.17) | 81 (.33) | 70 (.35) | 60 | 30 (.82) | 100 (.41)c | 89 (.46)c |
| −9 | 76 | 17 (.25) | 83 (.29) | 76 (.73) | 73 | 27 (.92) | 99 (1)c | 94 (1)c |
| −8 | 78 | 29 (.87) | 82 (.37) | 65 (.12) | 79 | 26 (.79) | 97 (1)c | 93 (1)c |
| −7 | 92 | 20 (.17) | 72 (.02) | 59 (.04) | 84 | 20 (.35) | 95 (.66)c | 84 (.06) |
| −6 | 97 | 22 (.44) | 56 (<.001) | 45 (<.001) | 95 | 22 (.48) | 97 (1)c | 86 (.12) |
| −5 | 110 | 29 (.80) | 61 (.001) | 44 (<.001) | 106 | 21 (.38) | 98 (.5)c | 85 (.09) |
| −4 | 124 | 12 (.02) | 55 (<.001) | 40 (<.001) | 125 | 33 (.57) | 95 (0.2) | 82 (.04) |
| −3 | 151 | 10 (<.01) | 49 (<.001) | 35 (<.001) | 143 | 20 (.25) | 83 (.01) | 71 (.001) |
| −2 | 189 | 18 (.13) | 33 (<.001) | 55 (.006) | 170 | 19 (.19) | 79 (.003) | 60 (<.001) |
| −1 | 124 | 35 (.35) | 53 (<.001) | 68 (.20) | 193 | 14 (.16) | 62 (<.001) | 35 (<.001) |
Because survival was variable and some patients were discharged alive home we had different number of data points for each patient.
: (%) percent of accurate prognosis.
: p-value; Chi-square test of proportions, except when indicated.
: p-value; Fisher’s exact test
Using day −14 as the baseline accuracy, nurses’ accuracy with temporal CPS did not change over time, with accuracies between 14% and 33% (Table 3). Nurses’ accuracy with 24- hour and 48-hour probabilistic CPS significantly decreased during the last three and four days of life, respectively, compared to baseline (Table 3).
Differences in Accuracy Between Temporal and Probabilistic CPS
Physicians’ and nurses’ accuracy with each method over time are shown in Figures 2A and 2B. Physicians and nurses were consistently and significantly more accurate in prognosticating survival with 24-hour and 48-hour probabilistic CPS than with temporal CPS during the last 14 days of life (Figures 2A and 2B).
Figure 2.
Accuracy of (A) physicians and (B) nurses prognostication of survival using the temporal, 24-hour and 48-hour probabilistic CPS. Temporal CPS was considered accurate if it fell within ±33% of the AS. Probabilistic CPS was considered accurate if the clinician endorsed a survival probability ≤30% and the patient died in the specific time frame (e.g. 24 h or 48 h), or the clinician selected a survival probability ≥70% and the patient survived beyond the specific time frame. (A) At all time points the 24-hour and 48-hour probabilistic approaches were significantly higher than the temporal approach (McNemar’s test; p < 0.05). (B) At all time points the 24-hour probabilistic and the 48-hour probabilistic approaches were significantly higher than the temporal approach (McNemar’s test; p < 0.001 for each time point). In the comparison between the temporal and the 24-hrs probabilistic approach, the McNemar’s test could not be used for days −13 and −10 before death because of restrictions of the test for complete agreement.
Differences in Accuracy Between Physicians and Nurses
Physicians and nurses were similarly accurate in estimating survival using temporal CPS (Table 4). Although there are some significantly differences during the last few days before death the direction of the difference varies. Nurses were consistently more accurate than physicians in estimating survival using 24-hour probabilistic CPS except during the last day of patient’s life when physicians’ accuracy increased (Figures 2 and Table 4). Nurses were consistently more accurate than physicians in estimating survival using 48-hour probabilistic CPS except during the last two days of patient’s life when physicians’ accuracy increased whereas nurses’ accuracy rapidly decreased (Figures 2 and Table 4).
Table 4.
Physiciańs versus Nursés accuracy for survival estimation using the temporal and probabilistic approaches.
| Days to death |
Accuracy of Temporal CPS | Accuracy of 24-hour Probabilistic CPS |
Accuracy of 48-hour Probabilistic CPS |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| # of pairs |
MDs (%)a |
RNs (%)a |
MDs vs. RNs pb |
# of pairs |
MDs (%)a |
RNs (%)a |
MDs vs. RNs pb |
# of pairs |
MDs (%)a |
RNs (%)a |
MDs vs. RNs pb |
|
| −14 | 42 | 31 | 29 | 0.82 | 42 | 88 | 98 | 0.1 | 42 | 79 | 95 | .03 |
| −13 | 46 | 22 | 26 | 0.59 | 46 | 85 | 100 | - | 46 | 76 | 91 | .03 |
| −12 | 50 | 20 | 24 | 0.56 | 50 | 82 | 98 | 0.005 | 50 | 70 | 90 | .01 |
| −11 | 57 | 32 | 32 | >0.99 | 57 | 81 | 98 | 0.004 | 57 | 63 | 96 | <0.001 |
| −10 | 57 | 18 | 30 | 0.13 | 57 | 81 | 100 | - | 57 | 70 | 89 | 0.002 |
| −9 | 70 | 17 | 27 | 0.16 | 70 | 83 | 99 | 0.002 | 70 | 76 | 94 | <0.001 |
| −8 | 76 | 29 | 26 | 0.73 | 74 | 82 | 97 | 0.002 | 74 | 65 | 93 | <0.001 |
| −7 | 80 | 20 | 20 | >0.99 | 79 | 72 | 95 | <0.001 | 79 | 59 | 84 | <0.001 |
| −6 | 89 | 22 | 22 | >0.99 | 87 | 56 | 97 | <0.001 | 87 | 45 | 86 | <0.001 |
| −5 | 101 | 29 | 21 | 0.21 | 100 | 61 | 98 | <0.001 | 100 | 44 | 85 | <0.001 |
| −4 | 113 | 12 | 33 | <0.001 | 112 | 55 | 95 | <0.001 | 111 | 40 | 82 | <0.001 |
| −3 | 137 | 10 | 20 | 0.03 | 136 | 49 | 83 | <0.001 | 136 | 35 | 71 | <0.001 |
| −2 | 164 | 18 | 19 | 0.77 | 159 | 33 | 79 | <0.001 | 160 | 55 | 60 | 0.43 |
| −1 | 121 | 35 | 14 | <0.001 | 116 | 53 | 62 | 0.22 | 116 | 68 | 35 | <0.001 |
Because survival was variable and some patients were discharged alive home we had different number of data points for each patient.
: (% percent of accurate prognosis.
: p-value; Mc Nemar’s Test.
Discussion
We systematically examined CPS over time using two different approaches and by two healthcare professions. The accuracy of the probabilistic CPS was consistently higher than the accuracy of temporal CPS. Interestingly, the accuracy of temporal CPS remained poor over time, and probabilistic CPS decreased in accuracy as death approached. Similar patterns were observed in both physicians and nurses, although nurses were generally more accurate with probabilistic CPS. Our study highlights the changes in accuracy over time of each approach, and provides important insights into how we can predict survival during the last two weeks of life.
To our knowledge this is the first study to serially examine accuracy of prognostication over time. Our study provides a better understanding of the operating characteristics of these prediction tools. We found that the first temporal CPS and AS are highly and significantly correlated suggesting that clinicians have the ability to recognize that a patient is declining. This finding is consistent with prior reports (5, 14–16). However, the accuracy of the temporal approach in our study was below 35% for all daily predictions even among experienced clinicians, suggesting that clinicians are unable to accurately assign a temporal estimation of survival. According to epidemiologists and other specialists in predicting events, the accuracy of predictions has two different components: discrimination, the observation that the relative ranking of the individual predictions is higher in those with higher survival; and calibration, the predicted survival is neither too high nor too low compared to the AS (8, 18, 19). Our findings regarding temporal CPS support the idea that clinicians have good discriminatory ability to prognosticate, but poorly calibrated. Future research should explore strategies to improve clinician calibration of temporal CPS.
In contrast, the accuracies of the 24-hour and 48-hour probabilistic approaches were high and stable between days− 14 and −7 before death, but decreased during the last few days of life. Our findings suggest that clinicians were excellent at establishing that the patient will not die in the short term, but had lower accuracy when the patient is imminently dying. The last few days of life is usually marked by significant changes in performance status, mentation, and the emergence of clinical signs of impending death (e.g. death rattle, respiration with mandibular breathing). The exact time of death may be difficult to predict. Further complicating this is that even if clinicians were aware that death is imminent, they have not been able to state this with a high level of confidence (i.e. 30% or less of survival within the defined time period). Further studies on the predictive abilities of bedside clinical signs may facilitate prognostication in the last days of life. We also found that 24-hour prognostication was more accurate than 48-hour prognostication, which is not surprising given that clinician’s uncertainty increases with longer time periods. Future studies should examine longer time frames of prognostication over long periods of time.
Prior studies have suggested that clinicians are more accurate at predicting survival closer to death (the “horizon effect”) and that repeated measures may improve prognostication accuracy of CPS. These observations have been studied to a limited extent for CPS, commonly for longer time frames, and results have not been consistent (6, 8, 13, 14, 19–21). Taken together, our study suggests that prognostication does not improve over time with the three questions during the last two weeks of life, even with serial observations and routine sign over among clinicians.
Physicians and nurses were similarly poor at prognosticating with temporal CPS. However, both were much more accurate with probabilistic CPS. This suggests that both clinicians are aware of the patients’ prognosis. Interestingly, nurses were generally more accurate than physicians in predicting that the patient will be alive in the next two days, except during the last few days of life, suggesting that nurses are better at predicting that the patient will not die, whereas physicians are more confident at indicating that the patient is actively dying.
Our study has several weaknesses. First, we only included patients with advanced cancer admitted to APCUs. This may decrease the applicability of these findings for patients with other diseases or patients admitted to other units. Second, we did not have information of the accuracy of these tools during the last two weeks of life for patients outside the hospital, which may have implications on the generalizability of our findings. Third, we focused our analysis on the last 14 days of life and used short term probabilistic estimation of survival. Future studies should examine serial prognostication over a longer time period and the operating characteristics of the probabilistic estimation of survival for longer time frames such as weeks and months. The readers should also use caution when interpreting the apparent higher accuracy with probabilistic approach compared to temporal approach because different criteria were used to define accuracy. However, we adopted highly stringent criteria for both approaches defined a priori and based on prior studies. Finally, a large number of clinicians – with variable years of palliative care experience - participated in this study, raising the question of generalizability of the ability to accurately predict survival. In this study we estimated overall accuracy per day, no accounting for clinician-specific accuracy. Although it is possible to expect differences in accuracy comparing experienced versus non-experienced clinicians, prior studies from our group found that clinician characteristics did not predict prognostication accuracy (6). The issue of clinician specific accuracy needs to be examined further in larger samples.
We serially analyzed the daily accuracy of two different approaches for survival estimation in a cohort of cancer patients admitted to two APCUs during the last two weeks of life and demonstrated that probabilistic CPS is better than temporal CPS. We also showed that the accuracy of probabilistic CPS decreased as patient approached death and when timeframe for prognostication increased. Probabilistic CPS for survival estimation should be preferred during the last two weeks of life of cancer patients. Further research into novel prognostic factors and signs of impending death may help clinicians prognosticate more accurately in the last days of life.
Acknowledgments
Dr. Bruera is supported in part by National Institutes of Health grants RO1NR010162- 01A1, RO1CA122292-01, and RO1CA124481-01. Dr. Hui is supported in part by an institutional startup grant (#18075582). This study also was supported by the M. D. Anderson Cancer Center Support Grant (CA 016672). The funding sources were not involved in the conduct of the study or development of the submission.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Disclosures
The authors have no financial disclosures or other potential conflicts of interest.
References
- 1.de Miguel Sanchez C, Elustondo SG, Estirado A, et al. Palliative performance status, heart rate and respiratory rate as predictive factors of survival time in terminally ill cancer patients. J Pain Symptom Manage. 2006;31:485–492. doi: 10.1016/j.jpainsymman.2005.10.007. [DOI] [PubMed] [Google Scholar]
- 2.Mori M, Parsons HA, De la Cruz M, et al. Changes in symptoms and inpatient mortality: a study in advanced cancer patients admitted to an acute palliative care unit in a comprehensive cancer center. J Palliat Med. 2011;14:1034–1041. doi: 10.1089/jpm.2010.0544. [DOI] [PubMed] [Google Scholar]
- 3.Weeks JC, Cook EF, O'Day SJ, et al. Relationship between cancer patients' predictions of prognosis and their treatment preferences. JAMA. 1998;279:1709–1714. doi: 10.1001/jama.279.21.1709. [DOI] [PubMed] [Google Scholar]
- 4.Morita T, Ichiki T, Tsunoda J, Inoue S, Chihara S. A prospective study on the dying process in terminally ill cancer patients. Am J Hosp Palliat Care. 1998;15:217–222. doi: 10.1177/104990919801500407. [DOI] [PubMed] [Google Scholar]
- 5.Christakis NA, Lamont EB. Extent and determinants of error in doctors' prognoses in terminally ill patients: prospective cohort study. BMJ. 2000;320:469–472. doi: 10.1136/bmj.320.7233.469. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Hui D, Kilgore K, Nguyen L, et al. The accuracy of probabilistic versus temporal clinician prediction of survival for patients with advanced cancer: a preliminary report. Oncologist. 2011;16:1642–1648. doi: 10.1634/theoncologist.2011-0173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Selby D, Chakraborty A, Lilien T, et al. Clinician accuracy when estimating survival duration: the role of the patient's performance status and time-based prognostic categories. J Pain Symptom Manage. 2011;42:578–588. doi: 10.1016/j.jpainsymman.2011.01.012. [DOI] [PubMed] [Google Scholar]
- 8.Glare P, Virik K, Jones M, et al. A systematic review of physicians' survival predictions in terminally ill cancer patients. BMJ. 2003;327:195–198. doi: 10.1136/bmj.327.7408.195. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Maltoni M, Nanni O, Pirovano M, et al. Successful validation of the palliative prognostic score in terminally ill cancer patients. Italian Multicenter Study Group on Palliative Care. J Pain Symptom Manage. 1999;17:240–247. doi: 10.1016/s0885-3924(98)00146-8. [DOI] [PubMed] [Google Scholar]
- 10.Glare P, Sinclair C, Downing M, et al. Predicting survival in patients with advanced disease. Eur J Cancer. 2008;44:1146–1156. doi: 10.1016/j.ejca.2008.02.030. [DOI] [PubMed] [Google Scholar]
- 11.Glare P. Clinical predictors of survival in advanced cancer. J Support Oncol. 2005;3:331–339. [PubMed] [Google Scholar]
- 12.Forster LE, Lynn J. Predicting life span for applicants to inpatient hospice. Arch Intern Med. 1988;148:2540–2543. [PubMed] [Google Scholar]
- 13.Chow E, Harth T, Hruby G, et al. How accurate are physicians' clinical predictions of survival and the available prognostic tools in estimating survival times in terminally ill cancer patients? A systematic review. Clin Oncol. 2001;13:209–218. doi: 10.1053/clon.2001.9256. [DOI] [PubMed] [Google Scholar]
- 14.Evans C, McCarthy M. Prognostic uncertainty in terminal care: can the Karnofsky index help? Lancet. 1985;1:1204–1206. doi: 10.1016/s0140-6736(85)92876-4. [DOI] [PubMed] [Google Scholar]
- 15.Parkes CM. Accuracy of predictions of survival in later stages of cancer. BMJ. 1972;2:29–31. doi: 10.1136/bmj.2.5804.29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Oxenham D, Cornbleet MA. Accuracy of prediction of survival by different professional groups in a hospice. Palliat Med. 1998;12:117–118. doi: 10.1191/026921698672034203. [DOI] [PubMed] [Google Scholar]
- 17.Llobera J, Esteva M, Rifa J, et al. Terminal cancer. duration and prediction of survival time. Eur J Cancer. 2000;36:2036–2043. doi: 10.1016/s0959-8049(00)00291-4. [DOI] [PubMed] [Google Scholar]
- 18.Justice AC, Covinsky KE, Berlin JA. Assessing the generalizability of prognostic information. Ann Intern Med. 1999;130:515–524. doi: 10.7326/0003-4819-130-6-199903160-00016. [DOI] [PubMed] [Google Scholar]
- 19.Mackillop WJ, Quirt CF. Measuring the accuracy of prognostic judgments in oncology. J Clin Epidemiol. 1997;50:21–29. doi: 10.1016/s0895-4356(96)00316-2. [DOI] [PubMed] [Google Scholar]
- 20.Bruera E, Miller MJ, Kuehn N, MacEachern T, Hanson J. Estimate of survival of patients admitted to a palliative care unit: a prospective study. J Pain Symptom Manage. 1992;7:82–86. doi: 10.1016/0885-3924(92)90118-2. [DOI] [PubMed] [Google Scholar]
- 21.Vigano A, Dorgan M, Buckingham J, Bruera E, Suarez-Almazor ME. Survival prediction in terminal cancer patients: a systematic review of the medical literature. Palliat Med. 2000;14:363–374. doi: 10.1191/026921600701536192. [DOI] [PubMed] [Google Scholar]