|Home | About | Journals | Submit | Contact Us | Français|
Palliative Performance Scale (PPS) scores have shown potential for prognosticating survival in Caucasian samples, but have not been studied for prognostic value in cancer patients from minority groups. Using data obtained from a retrospective chart audit of 492 cancer patients admitted over an 18-month period to a minority-serving home-based hospice and palliative care program, we examined the relationship between PPS scores and length of survival (survival days). Patients with PPS scores of 10% to 30% had fewer survival days than those with scores of 40% and those with scores of 50% to 100% (median = 6, 19, and 34 days, respectively; F = 25.02, P < 0.001). A PPS score of 40% serves as a reliable inclusion criterion for a study requiring two weeks for completion, while 50% to 100% is required for a three-week study. Findings from a predominantly minority sample are similar to those from predominantly Caucasian samples.
Prognosticating the length of time that a person with terminal cancer will survive is important for implementing feasible research studies that will have generalizable findings and contribute to effective end-of-life care (1, 2). Unfortunately, predicting survival is not easy (3), even after admission to hospice or palliative care programs. Palliative Performance Scale (PPS) scores have shown potential for prognosticating survival in predominately Caucasian samples, but have not been studied for prognostic value in cancer patients from racial and ethnic minority groups. From an instrument validity perspective, it is important to establish that the PPS prognosticates survival in racial and ethnic minority groups as well as it does for racial majority groups, but previous research studies included few cancer patients from racial and ethnic minority groups.
Although many cancer patients die soon after enrollment in hospice, the number of days they survive (survival days) varies greatly. Using Medicare data from 1990 with a sample of 6,451 patients, of whom 92% were Caucasian and 80% were diagnosed with cancer, investigators (4) found that the median survival after hospice enrollment was 36 days but that 16% of patients died within seven days and 15% lived longer than six months. More recently, the median survival has been reported as 28 days, and the average survival has been reported as 54 days (SD = 72) (5). The variability in survival days makes it difficult for researchers to predict the trajectory of survival for patients they could enroll in clinical studies designed to provide evidence for hospice practice. This difficulty is particularly important for randomized clinical trials, where attrition due to death poses threats to the internal validity of the study. Accurate prediction of survival days, therefore, is important for screening hospice patients eligible for participation in research that will generate valid results.
The PPS, a modified version of the Karnofsky Performance Scale,(6) has shown potential for survival prognostication of hospice patients (7). Five domains are assessed with the PPS: ambulation, activity and evidence of disease, self-care, intake and level of consciousness. Its scores range in 10% increments from 10% to 100%, with a score of 0% indicating death, 10% indicating a totally bedbound patient who is unable to do any activity and needs total assistance, and 100% indicating the patient is able to carry on normal activity and to work without any special care.
Anderson and colleagues (8) found that the majority of 213 Canadian patients admitted to an inpatient hospice unit had a PPS score of 30% to 50%. They also found that, of those with a PPS score of 30% when admitted, 82% died in a median of five days; 56% with a PPS score of 40% died in a median of eight days; and 37% with a PPS score of 50% died in a median of 11 days. Although not perfect prognostication, these findings indicated promise for the PPS scores to guide decisions related to survival days in hospice. The investigators, however, did not report sample characteristics such as diagnosis, gender, age, or race, which limits application of findings to the design of other studies.
Subsequent investigators have assessed the validity, reliability, and prognostic potential of the PPS in Japan (9), Australia (10), Canada (11), and three regions of the United States (12, 13). Findings from these studies support very strong validity with the KPS (9), adequate interrater reliability (12), and statistically significant prediction of survival with survival curve analysis methods (5, 10, 12, 13). Median length of survival from hospice admission to death, however, was variable in studies where median days were reported by PPS score and was less for samples from inpatient units than from home-based care. Sample characteristics such as diagnosis and care setting affected accuracy of the PPS to predict length of survival. None of the previous studies, however, included samples with sufficient representation of African American or Hispanic patients. The purpose of this study was to examine the prognostic value of the PPS in cancer patients admitted to a community-based hospice program that serves racial and ethnic minority groups. The specific aim was to examine relationships among PPS scores, survival days, and patient characteristics (age, gender, race/ethnicity, and cancer type). We also explored PPS scores and patient characteristics as predictors of survival days.
We conducted a retrospective chart audit of patients admitted to a not-for-profit, freestanding, minority-serving home-based hospice program between January 2005 and August 2006. This hospice program began in 1978 and now serves more than 500 patients per year, with a payer mix of 60% Medicare, 35% Medicaid, and 5% commercial insurance. In this organization, the admission nurses determine the PPS score as part of routine clinical care. We obtained permission from the privacy officer to receive and analyze information about deceased patients as part of Institutional Review Board approval at the University of Illinois at Chicago for a larger study (RO1 NR009092).
Our audit revealed 508 cancer patients who had died and thus were available for our study. The records of 16 patients did not have a PPS score. A sample of 492 patients was available for analysis.
The mean age of the subjects was 67 (SD = 15.5), and about half of them were female (n = 273, 56%). The sample was 46% African American, 31% Caucasian, 16% Hispanic, and about 7% other races (including 2.2% Asian, 0.4% Middle Eastern, and 4.5% unknown). There were nearly 50 different cancer diagnoses, which we categorized into nine system-based groups (Table 1).
As part of routine practice, admission nurses obtained a PPS score for each patient. Their training in doing so was part of orientation to the admission procedures. Therefore, scores obtained in this study are a reflection of translational science wherein the PPS tool was used as part of usual practice rather than as part of a research protocol.
Experienced research team members entered all data into SPSS 10.0 for statistical analysis. We used descriptive statistics to determine patients' characteristics, PPS scores, and survival days. Consistent with previous research findings and based on the PPS score distributions in the sample, for some analyses we divided subjects into three categories: 1) PPS score 30% and lower, 2) PPS score 40%, and 3) PPS score 50% and higher. Inferential procedures included Spearman's correlation, Student's t-test, one-way ANOVA, Kaplan-Meier survival analysis, and Cox proportional hazards regression analysis.
The mean number of survival days for the entire sample was 38 (SD = 52.1), the median was 18 days, and the number of survival days ranged from 1 to 402. At the time of enrollment in hospice, the mean PPS score was 41.1% (SD = 14.3), the median was 40%, and scores ranged from 10% to 100%. More than half (62%) of all patients had a PPS score of 40% or higher.
The mean, SD, and median for survival days by PPS scores, gender, race/ethnicity groups, and cancer types are shown in Table 1. Given the distribution of PPS scores, also shown in Table 1 are descriptive statistics for survival by the three PPS groups. There were no statistical differences in PPS scores by gender. African American patients had higher PPS scores than did other racial or ethnic groups, but the difference was not statistically significant (F (3, 488) = 2.21, P = 0.08). Patients with PPS scores of 50% and higher, however, were younger than patients with PPS scores of 30% and lower (Table 2). Age and PPS scores were associated with a statistically significant but weak negative correlation (Spearman rho = -0.13, P < 0.01).
In contrast, age and survival days were not statistically correlated (Spearman rho=0.04, P=0.38). Female patients survived statistically more days (mean 44, SD 57.7) than did males (mean 30.9, SD = 43.3) (t (490) = 2.87, P < 0.01). African American patients survived more days than other ethnicity groups, but the difference was not statistically significant (F (3, 488) = 1.94, P = 0.12).
The relationship between the PPS scores and survival days was statistically significant (Spearman rho = 0.47, P <0.001). There were significant statistical differences in survival days by PPS groups (F (2,489) = 25.02, P < 0.001). Pairwise comparisons indicated that patients who had a PPS score of 30% and lower had a significantly lower length of survival than those in the 40% group and the 50% and higher group, and patients who had a PPS score of 40% had a significantly lower length of survival than those in the 50% and higher group (mean = 12.8, 39.8, and 53.5, respectively).
The Kaplan–Meier survival curves for the three PPS score groups are shown in Figure 1. The log-rank test results showed that the categorized PPS scores of 30% and lower, 40%, and 50% and higher had different survival curves (log-rank test P < 0.05). The Cox proportional hazards model was used to examine the relationship between survival days and age, gender, race/ethnicity with other used as the reference group, and categorized PPS score (Table 3). Age, gender, and categorized PPS score were significantly related to the hazard survival days, but race/ethnicity was not. Younger patients had significant lower hazard than older patients (95% CI 0.98-0.99, P = 0.013). Female patients had lower hazard than male patients (95% CI 0.60-0.86, P = 0.001). Patients who had higher PPS scores had lower hazard than patients with lower PPS scores (95% CI 0.95-0.97, P < 0.001) (Table 3).
We are the first to report PPS and survival findings from a large sample of African American and Hispanic hospice patients with cancer. PPS scores did not differ by racial/ethnicity group, a finding that indicates there was no racial/ethnic bias when the PPS tool was translated in usual hospice practice. A minimum PPS score of 40% is a reasonable eligibility criterion for research requiring participation by the patient for two weeks; however, a PPS score of 30% or lower is not and could result in excessive attrition in studies that require more than a few days of participation. These cut scores are equally appropriate for Caucasian, African American, and Hispanic patients. Survival days also were not different by race/ethnicity, but not unexpectedly younger patients and women were more likely to survive longer than older patients and men. These findings are important for researchers to help them design studies to minimize attrition rates. Also clinicians could use findings to tailor the hospice care plan within the expected length of survival.
Our findings are consistent with those from predominately Caucasian samples (5, 12) and those in which race/ethnicity was not reported (8, 10, 14) that support the prognostic potential of PPS scores. Our findings also extend the knowledge to cancer patients from two minority groups and those receiving hospice care in their homes. Other investigators studied hospice patients with a variety of terminal illnesses, such as cancer, dementia, lung disease, and heart disease (4, 12, 13), and their study sites included the home (12), nursing home (12), or inpatient unit (1, 10, 11, 13).
We found that the number of survival days for the 30% or lower category was significantly different from the 40% category and the 50% or higher category, but did not find a difference between the 40% category and the 50% or higher category. These results were similar to findings of other investigators (5, 14). Overall, findings from the U.S., Canada, Japan, and Australia are strikingly similar and provide evidence that the PPS is a valid and useful tool for predicting survival days in three cultural groups (Caucasians, African Americans, and Hispanics) with cancer. When completed by trained clinicians, the behavioral focus of the instrument with clearly defined indicators of each score is not biased for these three racial/ethnic groups. Whether the findings would be similar in other cultural groups and other countries is unknown.
In our study, the median length of survival (18 days) was longer than in Australia (13 days) (10), Canada (10 days) (11), and another region of the U.S. (9 days) (13). Our median length of survival was shorter than in Japan (27 days) (9) and another region of the U.S. (28 days) (5). Variability in survival was large in all the studies. Clearly, the length of survival in all the studies is less than desirable for patients to receive optimal care (15) and would decrease the likelihood that long-term longitudinal studies would recruit subjects rapidly unless multiple sites would be employed.
The frequency distributions of our PPS scores and those from previous research show minor differences. In the majority of the studies, the most frequent PPS scores have been either 30% or 40%. According to the PPS scoring definitions, a score of 30% indicated the patient was totally bedbound, unable to do any activity, had extensive disease and required total care; whereas a score of 40% indicated the patient was mainly in bed, unable to do most activity with extensive disease and required assistance for self-care. These characteristics identify a patient as weak, requiring assistance, and perhaps with some loss of autonomy but conscious. Regarding participation in research, at a PPS score of 30 or 40 the patient's functional ability would be sufficient to allow him/her to participate in relatively simple data collection or intervention protocols.
The median number of survival days differs among the three PPS groups. In our study, the median survival was six days for the 30% or lower group, 19 days for the 40% group, and 34 days for the 50% or higher group. These findings are lower than those of another recent study conducted in the U.S. (5), where the median survival was 9 to 20 days for the 30% and lower category, 29 days for the 40% category, and 43 to 44 days for the 50% and higher category. Differences in length of survival may be related to differences in sample characteristics and subjectivity of the clinicians who determined the PPS scores. We included only patients with a cancer diagnosis, whereas Head et al. (5) included patients with cancer and chronic non-cancer illnesses or diseases. In general, patients with a cancer diagnosis might be expected to die within a shorter period than patients with chronic non-cancer illnesses diseases. Moreover, Harold and colleagues (12) demonstrated that the percentage of cancer patients who die within seven days had a different PPS score. We found that 50% of patients with a PPS score of 10% to 20% died within seven days, whereas only 20% of patients at the 30% to 40% score died within that first week. Researchers employed survival curve analysis method and also found that patients with higher PPS scores also had increased survival (9-11, 13).
The relationship between the PPS score and length of survival was significantly correlated indicating a moderate positive linear association. The PPS score also predicted survival in our Cox proportional analysis. These results were similar to past studies and support the PPS score as a valid predictor of patients' survival (8, 11, 13). Harold et al. (12) indicated that the PPS score was more accurate in overall prediction of survival for patients with a non-cancer diagnosis. In our study, we found that the PPS score had limited but beneficial ability to predict longevity in cancer patients, but our sample was too heterogeneous to show significant differences in survival or membership in PPS groups by type of cancer.
There were several limitations in our study. First, the ability to generalize results could be compromised because we studied patients from only one site and included only cancer patients. Further research could compare patients from different hospice programs, and multicenter studies could be conducted to validate findings in other regions and nations. Second, this study was cross-sectional and did not address changing patterns of the PPS scores. Knowing how they change over time will be crucial for further research and clinical practice. In future studies, a longitudinal design will be needed to document the trajectory of PPS scores and may provide a better indication of the patient's functional status and longevity when enrolled in the hospice program.
In summary, we found that PPS scores and survival days did not differ by racial/ethnicity group. Younger patients and women were more likely to survive longer than older patients and men. Regardless of race or ethnicity, cancer patients with lower PPS scores had reduced number of survival days. Our findings from a minority-serving hospice support those from previous studies in the U.S., Canada, Australia, and Japan. Our findings support using the PPS score of 40% as an inclusion criterion for research protocols requiring two weeks so that attrition due to death, a threat to internal and external study validity, can be minimized. These findings allowed our group to make decisions about eligibility criteria for our ongoing randomized clinical trial of massage therapy (RO1 NR009092) based on population-specific evidence rather than opinion or intuition. Study findings also provide insights about PPS scores and survival that can inform decisions by other researchers and can help clinicians to target prioritized care based on needs as well as the number of days the patient is likely to survive. Given the variability in survival, however, researchers and clinicians are encouraged to use these survival day numbers as estimates rather than precise indicators.
The authors would like to thank Marilyn Ward and Kevin Grandfield for their editorial assistance in preparing this text, and Ed Wang, PhD, and Young Ok Kim, DrPH, for their statistical consultations.
This study was supported in part by Grant Number RO1 NR009092 from the National Institutes of Health, National Institute of Nursing Research (NINR). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NINR.
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.