We demonstrate the use of piecewise regression as a statistical technique to model ecological thresholds. Recommended procedures for analysis are illustrated with a case study examining the width of edge effects in two understory plant communities. Piecewise regression models are ''broken-stick'' models, where two or more lines are joined at unknown points, called ''breakpoints.'' Breakpoints can be used as estimates of thresholds and are used here to determine the width of edge effects. We compare a sharp-transition model with three models incorporating smooth transitions: the hyperbolic-tangent, benthyperbola, and bent-cable models. We also calculate three types of confidence intervals for the breakpoint estimate: an interval based on the computed standard error of the estimate from the fitting procedure, an empirical bootstrap confidence interval, and a confidence interval derived from an inverted F test. We recommend use of the inverted F test confidence interval when sample sizes are large, and cautious use of bootstrapped confidence intervals when sample sizes are smaller. Our analysis demonstrates the need for a careful study of the likelihood surface when fitting and interpreting the results from piecewise-regression models.

BackgroundThe Palliative Performance Scale (PPS) was first introduced in1996 as a new tool for measurement of performance status in palliative care. PPS has been used in many countries and has been translated into other languages.MethodsThis study evaluated the reliability and validity of PPS. A web-based, case scenarios study with a test-retest format was used to determine reliability. Fifty-three participants were recruited and randomly divided into two groups, each evaluating 11 cases at two time points. The validity study was based on the content validation of 15 palliative care experts conducted over telephone interviews, with discussion on five themes: PPS as clinical assessment tool, the usefulness of PPS, PPS scores affecting decision making, the problems in using PPS, and the adequacy of PPS instruction.ResultsThe intraclass correlation coefficients for absolute agreement were 0.959 and 0.964 for Group 1, at Time-1 and Time-2; 0.951 and 0.931 for Group 2, at Time-1 and Time-2 respectively. Results showed that the participants were consistent in their scoring over the two times, with a mean Cohen's kappa of 0.67 for Group 1 and 0.71 for Group 2. In the validity study, all experts agreed that PPS is a valuable clinical assessment tool in palliative care. Many of them have already incorporated PPS as part of their practice standard.ConclusionThe results of the reliability study demonstrated that PPS is a reliable tool. The validity study found that most experts did not feel a need to further modify PPS and, only two experts requested that some performance status measures be defined more clearly. Areas of PPS use include prognostication, disease monitoring, care planning, hospital resource allocation, clinical teaching and research. PPS is also a good communication tool between palliative care workers.

Our findings differ somewhat from earlier studies that suggested the presence of three distinct PPS survival profiles or bands, with diagnosis and noncancer as significant covariates. Such differences are likely attributed to the size and characteristics of the patient populations involved and further analysis with larger patient samples may help clarify PPS use in prognosis.

This paper aims to reconcile the use of Palliative Performance Scale (PPSv2) for survival prediction in palliative care through an international collaborative study by five research groups. The study involves an individual patient data meta-analysis on 1,808 patients from four original datasets to reanalyze their survival patterns by age, gender, cancer status, and initial PPS score. Our findings reveal a strong association between PPS and survival across the four datasets. The Kaplan-Meier survival curves show each PPS level as distinct, with a strong ordering effect in which higher PPS levels are associated with increased length of survival. Using a stratified Cox proportional hazard model to adjust for study differences, we found females lived significantly longer than males, with a further decrease in hazard for females not diagnosed with cancer. Further work is needed to refine the reporting of survival times/probabilities and to improve prediction accuracy with the inclusion of other variables in the models. Resume / Cet article vise a reconcilier !'usage de l'echelle de performance en soins palliatifs [EPSP] comme facteur predlctlf de survie grace aun projet international commun auquel ont particlpe cinq groupes de recherche. Cette etude comporte une meta-analyse des donnees recueillies aupres des 1,808 patients provenant de quatre cohortes differentes. On a done refait I'analyse du pronostic de survie selon I'age, Ie genre, Ie stade de la maladie et Ie score initial de I'EPSP. Nos resultats revelent qu'il y a une relation importante entre I'EPSP et Ie facteur predictif de survie atravers I'ensemble des donnees des quatre groupes. Les courbes de survie Kaplan-Meier demontrent que chaque niveau de I'EPSP est distinct et que les plus hauts scores de I'EPSP sont assocles aune plus longue periods de survie. A I'aide du modele de hasard proportionnel de Cox, nous avons realuste les differences qui existaient entre les etudes et nous avons decouvert que les femmes, de tacon significative, vivent plus longtemps que les hommes. De plus, ces probabilites sont encore plus elevees chez les femmes n'ayant pas ete diagnostiquees du cancer. II nous faudra d'autres etudes pour perfectionner nos previsions sur Ie temps de survie/probabilite et pour ameliorer "exactitude des predictions tout en y ajoutant d'autres variables.

This study examines the use of the Palliative Performance Scale (PPS) in end-of-life prognostication within a regional palliative care program in a Canadian province. The analysis was done on a prospective cohort of 513 patients assessed by a palliative care consult team as part of an initial community/hospital-based consult. The variables used were initial PPS score, age, gender, diagnosis, cancer type, and survival time. The findings revealed initial PPS to be a significant predictor of survival, along with age, diagnosis, cancer type and site, but not gender. The survival curves were distinct for PPS 10%, 20%, and 30% individually, and for 40%-60% and > or =70% as bands. This is consistent with earlier findings of the ambiguity and difficulty when assessing patients at higher PPS levels because of the subjective nature of the tool. We advocate the use of median survival and survival rates based on a local cohort where feasible, when reporting individual survival estimates.