Study setting

The trial was set in Adelaide, South Australia and based at Southern Adelaide Palliative Services. GPs are the primary point of care for palliative patients. Specialized palliative care services funded by the state government provide consultative specialist medical and nursing support for GPs and community nurses. Referrals to the palliative care service come from health professionals, family and patients, and nearly all palliative care patients within the region are referred to the same geographic service. The indication for, and timing of referral varies; indications include physical symptom control, and the need for coordinated and multi-disciplinary care. Patients frequently receive active therapy such as radiotherapy and chemotherapy in parallel with palliative care services. The model of care is consistent with the definition of palliative care described in the 2004 United States (US) National Consensus Project Clinical Practice Guidelines for Quality Palliative Care [12].

Southern Adelaide Palliative Services is a comprehensive program with medical specialists, nursing specialists, social work, inpatient care, community and outpatient visits, home care, nursing home consultations, a bereavement program, volunteers, and complementary care. Venues of care include community, acute inpatient, sub-acute inpatient, inpatient hospice, respite, nursing home, and hostel settings. There were 1,094 new referrals in 2003, 85% of whom had cancer. The mean time from referral to death was 119 days with a median of 47 days. Ten percent of referrals were from GPs, 50% from local hospitals, 20% from medical specialists, and 10% from district nurses. Ninety percent of patients spent some time in the hospital in the last year of life and less than one third spent any time in the inpatient hospice.

Study participants

All adult patients referred to Southern Adelaide Palliative Services with any form of pain in the preceding three months were eligible for the Palliative Care Trial. This definition was very broadly applied and could refer to any type of pain. This pain could have been temporary and not related to the predominate illness, but would provide a reference point for pain assessments during the research study. Patients who did not live within the geographic region served by the palliative care service were excluded. Patients who were expected to die within 48 hours of referral were also excluded, since the recruitment process of consenting both the patient and GP was expected to take two days. Participants must have been mentally competent at enrollment as documented by a Folstein Mini-mental Status Examination (MMSE) Score = 24 [13], or have a GP-identified caregiver or legal healthcare proxy who could adequately provide informed consent [14]. Patients unwilling to provide their contact information to the trial staff in order to learn more about the study were not enrolled. All GPs and GP practices of consenting eligible patients were eligible. Both patient and GP consent were required for enrollment and randomization.

Measures

Patient participant functional status was assessed by the KPS, TKPS and AKPS (Table 1) simultaneously. For those patients who died during the trial, survival was defined as time from consent to participate in the trial until death.

Study procedures

KPS, TKPS and AKPS were collected at every clinical encounter and data collection time point (minimally baseline and every 2 weeks for the first 12 weeks then monthly until death or exit from the trial) for the first 300 participants randomized in the trial. Based on our sample size assumptions and recruitment estimates, this would provide data for over 120 participants who exited the trial through either death or withdrawal from the study. Palliative care clinical nurses collected all of the data within the study. Data collection was incorporated into regular palliative clinical visits irrespective of location of care (hospital- or community-based care). In total, 26 full or part-time nurses collected data.

All clinical nurses participated in training in performance status assessment. This included at least a 30-minute review of the scales, data collection forms, and reasons for using all three measures. In addition, written instructions were prepared that included a list of example questions that could be used to determine performance status classification. Nurses were provided laminated cards with each of the scales. Responses were captured on a standardized set of study forms. Nurses also underwent more specific training on data collection methods, the importance of data quality, communication and research ethics as part of the Palliative Care Trial [1].

Analysis

Descriptive statistics were used to summarize the participant population and performance status scores observed. Agreement between the KPS, TKPS and AKPS was assessed by a variety of methods. Percentage agreements and disagreements were calculated; overall agreement and proportion of disagreements at one, two or three levels on the scales were reported in the method of Nikoletti et al [3]. Kappa statistics were calculated in order to exclude agreement due to chance. Simple kappa statistics were reported and subjected to the following interpretation: 0.81 to 1.00 almost perfect agreement, 0.61 to 0.80 substantive agreement, 0.41–0.60 moderate agreement, 0.21–0.40 fair agreement, 0.00–0.20 slight agreement, and less than 0.00 poor or no agreement [15]. Bland-Altman plots of the difference between paired scores versus the mean of the scores were used to assess agreement [16]. The association between performance status and survival was determined using Spearman correlation coefficients [5]. Actuarial survival was estimated using Kaplan-Meier methods and reported as median (interquartile range (IQR)) survival [17]. Between group comparisons were made using the log rank test [18]. All analyses were conducted with the SAS System (Version 9.1, Cary, North Carolina, USA).

Ethics approval and trial registration

The Palliative Care Trial was approved by all twelve relevant independent Institutional Review Boards (IRBs) and Human Research and Ethics Committees (HRECs) including the Australian Department of Veterans Affairs and Health Insurance Commission, Canberra, Australia. This sub-study was included in the approved protocols and participant consent forms. The Palliative Care Trial is registered with the ISRCTN – ISRCTN81117481 http://www.controlled-trials.com/isrctn/trial/81117481/0/81117481.html.

Longitudinal assessment

Any of the KPS tools provide both an objective measure of the current status of the person being assessed and useful trends when used longitudinally. Longitudinal trends in performance status is an important aspect of prognostication for the longer term outlook of the patient including his or her anticipated health resource and service needs over time, as demonstrated by the relationship between performance status and phase of palliative care [20]. Such changes in level of function reflect the disease trajectories described by Lunney, Lynn and colleagues [21], irrespective of the underlying life-limiting illness.

Other measures of performance

KPS is not the only measure of performance status used in palliative care and oncology. The shorter Eastern Cooperative Oncology Group (ECOG) performance status scale was derived from the KPS [22]. It is only occasionally used as a main outcome in clinical trials in the palliative care setting since the 5-item scale inadequately differentiates between patients with poor functional status. Similar concerns about the KPS having limited sensitivity to monitor change when patients score at the low end of the scale have been reported by other authors [2]. In this current study, AKPS was superior to KPS and TKPS in the lower range of the scale and provided more categorical levels of performance status than the ECOG scale.

In 1996, Anderson et al described the Palliative Performance Scale (PPS), a modification of the KPS based on 5 observable parameters (ambulation, activity combined with evidence of disease, self-care, intake and conscious level) scored into 11 categories. [23]. PPS predicted time to death (mean 162 days) for a population of Australian patients admitted to a palliative care unit[24]. PPS was not as predictive of survival for a similar group of Japanese palliative care patients with a mean survival of 49 days [11]. AKPS is a less complex measure which is easier to use with each clinical encounter.

Other scales such as the Edmonton Functional Assessment Tool extend on the functional parameters described in the KPS, PPS and TKPS [2], however as these scales and their scoring becomes more complicated, their day-to-day applicability decreases. AKPS focuses on current functional abilities and on changes in function if used longitudinally; it provides an important parameter in the overall assessment of any person with a life-limiting illness.

Limitations

This study is representative by age and gender for palliative care in Australia. Because it was a sub-study of a larger randomized controlled trial where pain in the previous 3 months was an inclusion criterion, the population almost all had cancer as their life-limiting illness (92%) versus 85% seen in the general population referred to the palliative care service. Given that KPS was originally developed for people with cancer and had been extrapolated to other clinical settings (AIDS, end-stage organ failure), this should not be a major limit to generalizing these findings. In the early parts of the trial, measurement of KPS, TKPS and AKPS was predominantly in the community setting limiting the ability to observe its utility in other care venues. As more trial participants were hospitalized over time, data were collected from the inpatient settings therefore reflecting the range of settings in which palliative care is delivered. Many assessments were in the upper range of the scales; only a minority of patients were bedridden. An evaluation on a palliative care unit with more severely disabled patients might show other results. The missing correlation of KPS and TKPS with survival in the lower range of the scale may have been biased by small patient numbers in these clusters. Also, ideally none of the performance status measures would have any reference to the amount of health services required at any of the levels. AKPS has considerably less reference, but still states "requires... frequent medical care" in its description of AKPS 50.

Inter-rater reliability evaluation of the AKPS was originally planned as part of this sub-study. Ill palliative care patients were overly burdened by multiple visits on the same day for research data collection. An alternative plan was enacted with collection of measures after the informed consent document was signed and then comparing these results with those reported on the baseline assessment within 48 hours of the consent visit. Unfortunately many patients were too unstable or the timing of the baseline assessment was too far from the consent visit; there were not enough data available for these analyses. The inter-rater reliability and other psychometric properties of the KPS has previously been documented[3, 6, 7]. As AKPS was more predictive of survival outcomes than KPS, the reliability was expected to be better if it differed from that reported for the KPS.

In addition to being more predictive of survival at the lower end of the scale, the AKPS may be more appropriate than the other performance status scales in settings outside of cancer. Ninety-two percent of participants in this study had cancer, so evaluation of the AKPS in non-cancer diagnoses was limited. Further work should concentrate on disabled palliative care patients whose performance status is at the lower end of the scale. Future studies will focus on validity outside of the cancer setting and with more diverse palliative care populations. Also, the performance of the AKPS will be compared to other performance status measurement tools appropriate for palliative care such as the ECOG scale, PPS and Edmonton Functional Assessment Tool.