A new palliative prognostic score: a first step for the staging of terminally ill cancer patients. Italian Multicenter and Study Group on Palliative Care

Comparison of Prognostic Abilities of Palliative Prognostic Index, Laboratory Prognostic Score, and Palliative Prognostic Score

CONTEXT

Few studies have compared the prognostic value of scoring systems based on physical and blood parameters in terminally ill patients with cancer.

OBJECTIVES

This study evaluated the prognostic abilities of Palliative Prognostic Index (PPI), Laboratory Prognostic Score (LPS), and Palliative Prognostic Score (PaP).

METHODS

We included 989 terminally ill patients with cancer who consulted for admission to our palliative care unit. We compared the discriminative abilities of PPI, LPS, and PaP for 7-, 14-, 30-, 60-, and 90-day mortality. Additionally, we compared the estimated median survival of PPI, LPS, and PaP with the actual survival (AS). The prediction accuracy was considered adequate if the ratio of estimated median survival in days to AS in days fell within the range of 0.66 to 1.33, optimistic when it exceeds 1.33, and pessimistic when it falls below 0.66.

RESULTS

The accuracies for 7-, 14-, 30-, 60-, and 90-day mortality were superior for PPI, LPS, LPS, PaP, and PaP (72%, 73%, 71%, 80%, and 82%), respectively, although the discriminative abilities for 7-, 14-, 30-, 60-, and 90-day mortality were similar among the three scoring systems. The prediction accuracy of survival (PAS) was similar among the three scoring systems with adequate, optimistic, and pessimistic rates of 36%-41%, 20%-46%, and 16%-38%, respectively. PAS was superior in actual survival for 14-59 days.

CONCLUSIONS

The prognostic abilities of PPI, LPS, and PaP were comparable. The most adequate estimation occurred for patients with AS for 14-59 days. A more accurate prognostic model is needed for patients with longer survival.

Factors related to accurate clinicians' prediction of survival: an international multicenter study in East Asia

PURPOSE

Recent guidelines for prognostic evaluation recommend clinicians' prediction of survival (CPS) for survival prediction in patients with advanced cancer. However, CPS is often inaccurate and optimistic. Studies on factors associated with overestimation or underestimation of CPS are limited. We aimed to investigate the factors associated with the overestimation and underestimation of CPS in patients with far-advanced cancer.

METHODS

The current study was a secondary analysis of an international multicenter prospective cohort study, which enrolled newly admitted patients with advanced cancer in palliative care units (PCUs) in Japan, Korea, and Taiwan from 2017 to 2018. We obtained the temporal CPS at enrollment and performed multivariate logistic regression analysis to identify the factors associated with "underestimation (less than 33% of actual survival)" and "overestimation (more than 33% of actual survival)."

RESULTS

A total of 2571 patients were assessed and admitted in 37 PCUs between January 2017 and September 2018. Older age (adjusted odds ratio [aOR] 1.01; 95% confidence interval [CI] 1.01-1.02; P < 0.01) and reduced oral intake (aOR 0.68; 95% CI 0.51-0.89; P < 0.01) were identified as significant factors associated with underestimation. Dyspnea (aOR 1.28; 95% CI 1.06-1.54; P = 0.01) and hyperactive delirium (aOR 1.34; 95% CI 1.05-1.72; P = 0.02) were identified as significant factors associated with overestimation.

CONCLUSION

Older age was related to underestimation, while dyspnea and hyperactive delirium were related to overestimation of CPS for patients with weeks of survival. However, reduced oral intake was less likely to lead to underestimation.

Palliative prognostic scores for survival prediction of cancer patients: a systematic review and meta-analysis

BACKGROUND

The palliative prognostic score is the most widely validated prognostic tool for cancer survival prediction, with modified versions available. A systematic evaluation of palliative prognostic score tools is lacking. This systematic review and meta-analysis aimed to evaluate the performance and prognostic utility of palliative prognostic score, delirium-palliative prognostic score, and palliative prognostic score without clinician prediction in predicting 30-day survival of cancer patients and to compare their performance.

METHODS

Six databases were searched for peer-reviewed studies and grey literature published from inception to June 2, 2023. English studies must assess palliative prognostic score, delirium-palliative prognostic score, or palliative prognostic score without clinician-predicted survival for 30-day survival in adults aged 18 years and older with any stage or type of cancer. Outcomes were pooled using the random effects model or summarized narratively when meta-analysis was not possible.

RESULTS

A total of 39 studies (n = 10 617 patients) were included. Palliative prognostic score is an accurate prognostic tool (pooled area under the curve [AUC] = 0.82, 95% confidence interval [CI] = 0.79 to 0.84) and outperforms palliative prognostic score without clinician-predicted survival (pooled AUC = 0.74, 95% CI = 0.71 to 0.78), suggesting that the original palliative prognostic score should be preferred. The meta-analysis found palliative prognostic score and delirium-palliative prognostic score performance to be comparable. Most studies reported survival probabilities corresponding to the palliative prognostic score risk groups, and higher risk groups were statistically significantly associated with shorter survival.

CONCLUSIONS

Palliative prognostic score is a validated prognostic tool for cancer patients that can enhance clinicians' confidence and accuracy in predicting survival. Future studies should investigate if accuracy differs depending on clinician characteristics. Reporting of validation studies must be improved, as most studies were at high risk of bias, primarily because calibration was not assessed.

Are Surprise Questions and Probabilistic Questions by Nurses Useful in Home Palliative Care? A Prospective Study

Background: Surprise questions (SQs) are used as screening tools in palliative care. Probabilistic questions (PQs) are more accurate than temporal predictions. However, no study has examined the usefulness of SQs and PQs assessed by nurses. Objectives: To examine the accuracy of nurses' SQ and PQ assessments in patients with advanced cancer receiving home palliative care. Design: A prospective single-center cohort study. Setting/Subjects: Adult patients with advanced cancer who received palliative care at home in South Korea between 2019 and 2020. Measurements: Palliative care specialized nurses were asked the SQ, "Would you be surprised if the patient died in a specific timeframe?" and PQ, "What is the probability that this patient will be alive (0 to 100%) within a specific timeframe?" at the 1-, 2-, 4-, and 6-week timeframes at enrollment. We calculated the sensitivities and specificities of the SQs and PQs. Results: 81 patients were recruited with 47 days of median survival. The sensitivity, specificity, and overall accuracy (OA) of the 1-week SQ were 50.0, 93.2, and 88.9%, respectively. The accuracies for the 1-week PQ were 12.5, 100.0, and 91.3%, respectively. The 6-week SQ showed sensitivity, specificity, and OA of 84.6, 42.9, and 62.9%, respectively; the accuracies for the 6-week PQ were 59.0, 66.7, and 63.0%, respectively.Conclusion: The SQ and PQ showed acceptable accuracy in home palliative care patients. Interestingly, PQ showed higher specificity than SQ at all timeframes. The SQ and PQ assessed by nurses may be useful in providing additional prognostic information for home palliative care.

The utility of the surprise question by nurses to identify hospitalised older patients nearing the end-of-life and promotion of advance care planning: An interventional study

AIMS AND OBJECTIVES

To assess the prognostic accuracy of the surprise question (SQ) when used by nurses working in hospital wards to determine 1-year mortality in acutely hospitalised older patients.

BACKGROUND

The predictive accuracy of the SQ, when used by general nurses caring for older hospitalised patients, has not been comprehensively studied.

DESIGN

A prospective cohort study.

METHODS

This cohort study recruited consecutive 10,139 older patients (aged ≥65 years) who were admitted to Taipei City Hospital and were evaluated for the needs of palliative care in 2015. All patients were followed up for 12 months or until their death. The c-statistic value was calculated to indicate the predictive accuracy of the SQ and Palliative Care Screening Tool (PCST).

RESULTS

Of all participants, 18.8% and 18.6% had a SQ response of 'no' and a PCST score ≥4, respectively. After controlling for other covariates, an SQ response of 'no' (adjusted hazard ratio [aHR], 2.05; 95% confidence interval [CI], 1.83-2.31) and a PCST score ≥4 (AHR = 1.50; 95% CI: 1.29-1.75) were found to be the independent predictors for patients' 12-month mortality. The C-statistic values of the SQ and the PCST at recognising patients in their last year of life were .663 and .670, respectively. Moreover, there was moderate concordance (k = .44) between the SQ and the PCST in predicting 12-month mortality.

CONCLUSIONS

SQ response of 'no' and a PCST score ≥4 were independent predictors of 12-month mortality in older patients.

RELEVANCE TO CLINICAL PRACTICE

The SQ, when used by nurses working in hospital wards, is effective in identifying older patients nearing the end of life, as well as in providing advance care planning for patients.

PATIENT OR PUBLIC CONTRIBUTION

Patients' palliative care needs at admission were assessed by general nurses using the SQ and PCST.

Changepoint Detection in Heart Rate Variability Indices in Older Patients Without Cancer at End of Life Using Ballistocardiography Signals: Preliminary Retrospective Study

BACKGROUND

In an aging society such as Japan, where the number of older people continues to increase, providing in-hospital end-of-life care for all deaths, and end-of-life care outside of hospitals, such as at home or in nursing homes, will be difficult. In end-of-life care, monitoring patients is important to understand their condition and predict survival time; this information gives family members and caregivers time to prepare for the end of life. However, with no clear indicators, health care providers must subjectively decide if an older patient is in the end-of-life stage, considering factors such as condition changes and decreased food intake. This complicates decisions for family members, especially during home-based care.

OBJECTIVE

The purpose of this preliminary retrospective study was to determine whether and how changes in heart rate variability (HRV) indices estimated from ballistocardiography (BCG) occur before the date of death in terminally ill older patients, and ultimately to predict the date of death from the changepoint.

METHODS

This retrospective pilot study assessed the medical records of 15 older patients admitted to a special nursing home between August 2019 and December 2021. Patient characteristics and time-domain HRV indices such as the average normal-to-normal (ANN) interval, SD of the normal-to-normal (SDNN) interval, and root mean square of successive differences (RMSSD) from at least 2 months before the date of death were collected. Overall trends of indices were examined by drawing a restricted cubic spline curve. A repeated measures ANOVA was performed to evaluate changes in the indices over the observation period. To explore more detailed changes in HRV, a piecewise regression analysis was conducted to estimate the changepoint of HRV indices.

RESULTS

The 15 patients included 8 men and 7 women with a median age of 93 (IQR 91-96) years. The cubic spline curve showed a gradual decline of indices from approximately 30 days before the patients' deaths. The repeated measures ANOVA showed that when compared with 8 weeks before death, the ratio of the geometric mean of ANN (0.90, 95% CI 0.84-0.98; P=.005) and RMSSD (0.83, 95% CI 0.70-0.99; P=.03) began to decrease 3 weeks before death. The piecewise regression analysis estimated the changepoints for ANN, SDNN, and RMSSD at -34.5 (95% CI -42.5 to -26.5; P<.001), -33.0 (95% CI -40.9 to -25.1; P<.001), and -35.0 (95% CI -42.3 to -27.7; P<.001) days, respectively, before death.

CONCLUSIONS

This preliminary study identified the changepoint of HRV indices before death in older patients at end of life. Although few data were examined, our findings indicated that HRV indices from BCG can be useful for monitoring and predicting survival time in older patients at end of life. The study and results suggest the potential for more objective and accurate prognostic tools in predicting end-of-life outcomes.

Malignancy-related ascites in palliative care units: prognostic factor analysis

OBJECTIVES

The prognostic factors in patients with malignancy-related ascites (MA) have been poorly investigated. This study aimed to evaluate both the prognostic impact of MA on terminally ill patients with cancer and the prognostic factors in those with MA.

METHODS

This was a post hoc analysis of a multicentre, prospective cohort study. Patients with advanced cancer admitted to palliative care units at 23 institutions and aged≥18 years were enrolled between January and December 2017. Overall survival (OS) was compared according to MA. A multivariate analysis was conducted to explore prognostic factors in patients with MA.

RESULTS

Of 1896 eligible patients, gastrointestinal and hepatobiliary pancreatic cancers accounted for 42.5%. 568 (30.0%) of the total had MA. Patients with MA had significantly shorter OS than those without MA (median, 14 vs 22 days, respectively; HR, 1.55; 95% CI, 1.39 to 1.72; p<0.01). A multivariate analysis showed that MA was a poor prognostic factor (HR, 1.30; 95% CI, 1.13 to 1.50; p<0.01) and that among patients with MA, significant poor prognostic factors were liver metastasis, moderately to severely reduced oral intake, delirium, oedema, gastric cancer, high serum creatinine, high serum C reactive protein, high serum total bilirubin, dyspnoea and fatigue, while significant good prognostic factors were female sex, good performance status, high serum albumin and colorectal cancer.

CONCLUSIONS

MA had a negative impact on survival in terminally ill patients with cancer. A multivariate analysis revealed several prognostic factors in patients with terminal cancer and MA.

Enzyme replacement in advanced pancreatic cancer: patient perceptions

PURPOSE

Advanced pancreatic cancer has a universally poor survival rate. Patients frequently develop malabsorption that requires pancreatic enzyme replacement therapy (PERT). This study explores the experience of patient engagement with PERT and how the medication is taken and tolerated.

METHODS

Participants with advanced pancreatic cancer requiring PERT were interviewed after referral to a specialist palliative care team. An inductive analysis was used to code the data. Theoretical sufficiency was reached after 12 participants.

RESULTS

Four themes emerged from the interviews-patient context, health literacy, relationship to food and experience of taking the pancreatic enzymes. Respondents brought their own life experiences into the clinical encounter when told of the diagnosis. Patients had high levels of understanding and engagement with the diagnosis and treatment, understood the benefits of PERT in digestion and tolerated the medication well.

CONCLUSIONS

Patients with metastatic pancreatic cancer understand the life-limiting nature of their illness. They want to participate in their healthcare decisions and are capable of complex medication titration when given good explanations and they experience benefits. PERT should be offered to these patients by a team of knowledgeable health professionals with good communication skills that can continue to support and review their needs.

Clinical Features of Patients With Hematological Malignancies Treated at the Palliative Care Unit

Background

In Japan, the number of patients with aggressive hematological malignancies (PHMs) admitted at the palliative care unit (PCU) in their end-of-life (EOL) stage was fewer than that of patients with solid tumors due to several reasons. The assessment of patient characteristics and the methods of survival prediction among PHMs in the EOL stage are warranted.

Objectives

This study aimed to identify the current medical status and the method of survival prediction among PHMs treated at the PCU.

Setting/Subjects/Measurements

We retrospectively analyzed the clinical data of 25 PHMs treated at our PCU between January 2017 and December 2020. The association between survival time and the palliative prognostic score (PAP) and palliative prognostic index (PPI) was analyzed.

Results

The average age of the PHMs was higher than that of patients with lung cancer as a control. The median survival time of the PHMs was shorter than the control group. Most PHMs could not receive standard chemotherapy, and the most common cause of death was disease-related organ failure. Significant associations were observed between the survival time and each PAP/PPI value in patients with malignant lymphoma, but not in those with leukemia.

Conclusion

The PHMs in the PCU had a lower median survival time than the control group. These results were induced by the result of patient selection to avoid treatment-related severe toxicity. The survival prediction using the PAP and PPI was less accurate in patients with leukemia.

Evaluation of factors predicting the benefit from systemic oncological treatment for severely ill hospitalized patients: a retrospective study

BACKGROUND

Patients with cancer in the disease's end-stage with poor performance represent a challenging clinical scenario, as they have high chance of a fatal outcome due to clinical conditions, oncological emergencies, and/or metastatic disease. This study examines the factors predicting the potential benefit of "urgent" chemotherapy during hospitalization in this setting, thus addressing a research gap.

METHODS

This retrospective observational study was conducted in the largest cancer center in the outskirts of São Paulo. It identified factors predicting the benefit from antineoplastic treatment in severe in-hospital patients admitted during 2019-2020, considering post-chemotherapy survival time as the main dependent variable. Data were retrieved from medical records. All patients aged ≥ 18 years, with an ECOG-PS score ≥ 2, and undergoing non-elective systemic cancer treatment were included.

RESULTS

This study evaluated 204 records, of which 89 were included in the final analysis. A statistically significant association with the worse outcome (death within 30 days of chemotherapy) was found with higher ECOG performance status; chemotherapy dose reduction; lower values of serum albumin, hemoglobin, and creatinine clearance; and higher values of leukocytes, neutrophils, direct bilirubin, urea, and C-reactive protein. In the multivariate analysis, only albumin remained statistically associated with the outcome (hazard ratio = 0.35; confidence interval: 0.14, 0.90; p = 0.034).

CONCLUSIONS

Serum albumin and other clinical and laboratory variables might be associated with early post-treatment deaths in patients with cancer. The study data might help guide the decision to administer systemic treatment in this scenario and manage critically ill patients. This study adds to our knowledge of the factors predicting the objective benefits from "heroic" or "urgent" chemotherapy for hospitalized and severely ill patients with cancer.

Clinical applications of graph neural networks in computational histopathology: A review

Pathological examination is the optimal approach for diagnosing cancer, and with the advancement of digital imaging technologies, it has spurred the emergence of computational histopathology. The objective of computational histopathology is to assist in clinical tasks through image processing and analysis techniques. In the early stages, the technique involved analyzing histopathology images by extracting mathematical features, but the performance of these models was unsatisfactory. With the development of artificial intelligence (AI) technologies, traditional machine learning methods were applied in this field. Although the performance of the models improved, there were issues such as poor model generalization and tedious manual feature extraction. Subsequently, the introduction of deep learning techniques effectively addressed these problems. However, models based on traditional convolutional architectures could not adequately capture the contextual information and deep biological features in histopathology images. Due to the special structure of graphs, they are highly suitable for feature extraction in tissue histopathology images and have achieved promising performance in numerous studies. In this article, we review existing graph-based methods in computational histopathology and propose a novel and more comprehensive graph construction approach. Additionally, we categorize the methods and techniques in computational histopathology according to different learning paradigms. We summarize the common clinical applications of graph-based methods in computational histopathology. Furthermore, we discuss the core concepts in this field and highlight the current challenges and future research directions.

Performance of the Palliative Prognostic Index for cancer patients: A systematic review and meta-analysis

BACKGROUND

Clinician predicted survival for cancer patients is often inaccurate, and prognostic tools may be helpful, such as the Palliative Prognostic Index (PPI). The PPI development study reported that when PPI score is greater than 6, it predicted survival of less than 3 weeks with a sensitivity of 83% and specificity of 85%. When PPI score is greater than 4, it predicts survival of less than 6 weeks with a sensitivity of 79% and specificity of 77%. However, subsequent PPI validation studies have evaluated various thresholds and survival durations, and it is unclear which is most appropriate for use in clinical practice. With the development of numerous prognostic tools, it is also unclear which is most accurate and feasible for use in multiple care settings.

AIM

We evaluated PPI model performance in predicting survival of adult cancer patients based on different thresholds and survival durations and compared it to other prognostic tools.

DESIGN

This systematic review and meta-analysis was registered in PROSPERO (CRD42022302679). We calculated the pooled sensitivity and specificity of each threshold using bivariate random-effects meta-analysis and pooled diagnostic odds ratio of each survival duration using hierarchical summary receiver operating characteristic model. Meta-regression and subgroup analysis were used to compare PPI performance with clinician predicted survival and other prognostic tools. Findings which could not be included in meta-analyses were summarised narratively.

DATA SOURCES

PubMed, ScienceDirect, Web of Science, CINAHL, ProQuest and Google Scholar were searched for articles published from inception till 7 January 2022. Both retrospective and prospective observational studies evaluating PPI performance in predicting survival of adult cancer patients in any setting were included. The Prediction Model Risk of Bias Assessment Tool was used for quality appraisal.

RESULTS

Thirty-nine studies evaluating PPI performance in predicting survival of adult cancer patients were included (n = 19,714 patients). Across meta-analyses of 12 PPI score thresholds and survival durations, we found that PPI was most accurate for predicting survival of <3 weeks and <6 weeks. Survival prediction of <3 weeks was most accurate when PPI score>6 (pooled sensitivity = 0.68, 95% CI 0.60-0.75, specificity = 0.80, 95% CI 0.75-0.85). Survival prediction of <6 weeks was most accurate when PPI score>4 (pooled sensitivity = 0.72, 95% CI 0.65-0.78, specificity = 0.74, 95% CI 0.66-0.80). Comparative meta-analyses found that PPI performed similarly to Delirium-Palliative Prognostic Score and Palliative Prognostic Score in predicting <3-week survival, but less accurately in <30-day survival prediction. However, Delirium-Palliative Prognostic Score and Palliative Prognostic Score only provide <30-day survival probabilities, and it is uncertain how this would be helpful for patients and clinicians. PPI also performed similarly to clinician predicted survival in predicting <30-day survival. However, these findings should be interpreted with caution as limited studies were available for comparative meta-analyses. Risk of bias was high for all studies, mainly due to poor reporting of statistical analyses. while there were low applicability concerns for most (38/39) studies.

CONCLUSIONS

PPI score>6 should be used for <3-week survival prediction, and PPI score>4 for <6-week survival. PPI is easily scored and does not require invasive tests, and thus would be easily implemented in multiple care settings. Given the acceptable accuracy of PPI in predicting <3- and <6-week survival and its objective nature, it could be used to cross-check clinician predicted survival especially when clinicians have doubts about their own judgement, or when clinician estimates seem to be less reliable. Future studies should adhere to the reporting guidelines and provide comprehensive analyses of PPI model performance.

Should the Surprise Question be Used as a Prognostic Tool for People With Life-limiting Illnesses?

The surprise question screening tool ("Would I be surprised if this person died within the next 12 months?") was initially developed to identify possible palliative care needs. One controversial topic regarding the surprise question is whether it should be used as a prognostic tool (predicting survival) for patients with life-limiting illnesses. In this "Controversies in Palliative Care" article, three groups of expert clinicians independently answered this question. All experts provide an overview of current literature, practical advice, and opportunities for future research. All experts reported on the inconsistency of the prognostic capabilities of the surprise question. Two of the three expert groups felt that the surprise question should not be used as a prognostic tool due to these inconsistencies. The third expert group felt that the surprise question should be used as a prognostic tool, particularly for shorter time frames. The experts all highlighted that the original rationale for the surprise question was to trigger a further conversation about future treatment and a potential shift in the focus of the care, identifying patients who many benefit from specialist palliative care or advance care planning; however, many clinicians find this discussion a difficult one to initiate. The experts agreed that the benefit of the surprise question comes from its simplicity: a one-question tool that requires no specific information about the patient's condition. More research is needed to better support the application of this tool in routine practice, particularly in noncancer populations.

Development and external validation of a predictive multivariable model for last-weeks survival of advanced cancer patients in the palliative home care setting (PACS)

PURPOSE

Various prognostic indexes have been proposed to improve physicians' ability to predict survival time in advanced cancer patients, admitted to palliative care (PC) with a survival probably to a few weeks of life, but no optimal score has been identified. The study aims therefore to develop and externally validate a new multivariable predictive model in this setting.

METHODS

We developed a model to predict short-term overall survival in cancer patients on the basis of clinical factors collected at PC admission. The model was developed on 1020 cancer patients prospectively enrolled to home palliative care at VIDAS Milan, Italy, between May 2018 and February 2020 and followed-up to June 2020, and validated in two separate samples of 544 home care and 247 hospice patients.

RESULTS

Among 68 clinical factors considered, five predictors were included in the predictive model, i.e., rattle, heart rate, anorexia, liver failure, and the Karnofsky performance status. Patient's survival probability at 5, 15, 30 and 45 days was estimated. The predictive model showed a good calibration and moderate discrimination (area under the receiver operating characteristic curve between 0.72 and 0.79) in the home care validation set, but model calibration was suboptimal in hospice patients.

CONCLUSIONS

The new multivariable predictive model for palliative cancer patients' survival (PACS model) includes clinical parameters routinely at patient's admission to PC and can be easily used to facilitate immediate and appropriate short-term clinical decisions for PC cancer patients in the home setting.

Evaluating the Potential of Machine Learning and Wearable Devices in End-of-Life Care in Predicting 7-Day Death Events Among Patients With Terminal Cancer: Cohort Study

BACKGROUND

An accurate prediction of mortality in end-of-life care is crucial but presents challenges. Existing prognostic tools demonstrate moderate performance in predicting survival across various time frames, primarily in in-hospital settings and single-time evaluations. However, these tools may fail to capture the individualized and diverse trajectories of patients. Limited evidence exists regarding the use of artificial intelligence (AI) and wearable devices, specifically among patients with cancer at the end of life.

OBJECTIVE

This study aimed to investigate the potential of using wearable devices and AI to predict death events among patients with cancer at the end of life. Our hypothesis was that continuous monitoring through smartwatches can offer valuable insights into the progression of patients at the end of life and enable the prediction of changes in their condition, which could ultimately enhance personalized care, particularly in outpatient or home care settings.

METHODS

This prospective study was conducted at the National Taiwan University Hospital. Patients diagnosed with cancer and receiving end-of-life care were invited to enroll in wards, outpatient clinics, and home-based care settings. Each participant was given a smartwatch to collect physiological data, including steps taken, heart rate, sleep time, and blood oxygen saturation. Clinical assessments were conducted weekly. The participants were followed until the end of life or up to 52 weeks. With these input features, we evaluated the prediction performance of several machine learning-based classifiers and a deep neural network in 7-day death events. We used area under the receiver operating characteristic curve (AUROC), F1-score, accuracy, and specificity as evaluation metrics. A Shapley additive explanations value analysis was performed to further explore the models with good performance.

RESULTS

From September 2021 to August 2022, overall, 1657 data points were collected from 40 patients with a median survival time of 34 days, with the detection of 28 death events. Among the proposed models, extreme gradient boost (XGBoost) yielded the best result, with an AUROC of 96%, F1-score of 78.5%, accuracy of 93%, and specificity of 97% on the testing set. The Shapley additive explanations value analysis identified the average heart rate as the most important feature. Other important features included steps taken, appetite, urination status, and clinical care phase.

CONCLUSIONS

We demonstrated the successful prediction of patient deaths within the next 7 days using a combination of wearable devices and AI. Our findings highlight the potential of integrating AI and wearable technology into clinical end-of-life care, offering valuable insights and supporting clinical decision-making for personalized patient care. It is important to acknowledge that our study was conducted in a relatively small cohort; thus, further research is needed to validate our approach and assess its impact on clinical care.

TRIAL REGISTRATION

ClinicalTrials.gov NCT05054907; https://classic.clinicaltrials.gov/ct2/show/NCT05054907.

Validation of Modified Models of Objective Prognostic Score in Patients With Advanced Cancer

Background: The objective prognostic score (OPS) needs to be modified to reflect practical palliative care circumstances. Objectives: We aimed to validate modified models of OPS with few or no laboratory tests for patients with advanced cancer. Design: An observational study was performed. Setting/Subjects: A secondary analysis of an international, multicenter cohort study of patients in East Asia was performed. The subjects were inpatients with advanced cancer in the palliative care unit. Measurements: We developed two modified OPS (mOPS) models to predict two-week survival: mOPS-A consisted of two symptoms, two objective signs, and three laboratory results, while mOPS-B consisted of three symptoms, two signs, and no laboratory data. We compared the accuracy of the prognostic models using sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC). Calibration plots for two-week survival and net reclassification indices (NRIs) were compared for the two models. Survival differences between higher and lower score groups of each model were identified by the log-rank test. Results: We included a total of 1796 subjects having median survival of 19.0 days. We found that mOPS-A had higher specificity (0.805-0.836) and higher AUROCs (0.791-0.797). In contrast, mOPS-B showed higher sensitivity (0.721-0.725) and acceptable AUROCs (0.740-0.751) for prediction of two-week survival. Two mOPSs showed good concordance in calibration plots. Considering NRIs, replacing the original OPS with mOPSs improved overall reclassification (absolute NRI: 0.47-4.15%). Higher score groups of mOPS-A and mOPS-B showed poorer survival than those of lower score groups (p < 0.001). Conclusions: mOPSs used reduced laboratory data and had relatively good accuracy for predicting survival in advanced cancer patients receiving palliative care.

Liver-specific metastases as an independent prognostic factor in cancer patients receiving hospice care in hospital

BACKGROUND

Survival prediction is important in cancer patients receiving hospice care. Palliative prognostic index (PPI) and palliative prognostic (PaP) scores have been used to predict survival in cancer patients. However, cancer primary site with metastatic status, enteral feeding tubes, Foley catheter, tracheostomy, and treatment interventions are not considered in aforementioned tools. The study aimed to investigate the cancer features and potential clinical factors other than PPI and PaP to predict patient survival.

METHODS

We conducted a retrospective study for cancer patients admitted to a hospice ward between January 2021 and December 2021. We examined the correlation of PPI and PaP scores with survival time since hospice ward admission. Multiple linear regression was used to test the potential clinical factors other than PPI and PaP for predicting survival.

RESULTS

A total of 160 patients were enrolled. The correlation coefficients for PPI and PaP scores with survival time were -0.305 and -0.352 (both p < 0.001), but the predictabilities were only marginal at 0.087 and 0.118, respectively. In multiple regression, liver metastasis was an independent poor prognostic factor as adjusted by PPI (β = -8.495, p = 0.013) or PaP score (β = -7.139, p = 0.034), while feeding gastrostomy or jejunostomy were found to prolong survival as adjusted by PPI (β = 24.461, p < 0.001) or PaP score (β = 27.419, p < 0.001).

CONCLUSIONS

Association between PPI and PaP with patient survival in cancer patients at their terminal stages is low. The presence of liver metastases is a poor survival factor independent of PPI and PaP score.

Applications of Machine Learning in Palliative Care: A Systematic Review

Objective: To summarize the available literature on using machine learning (ML) for palliative care practice as well as research and to assess the adherence of the published studies to the most important ML best practices. Methods: The MEDLINE database was searched for the use of ML in palliative care practice or research, and the records were screened according to PRISMA guidelines. Results: In total, 22 publications using machine learning for mortality prediction (n = 15), data annotation (n = 5), predicting morbidity under palliative therapy (n = 1), and predicting response to palliative therapy (n = 1) were included. Publications used a variety of supervised or unsupervised models, but mostly tree-based classifiers and neural networks. Two publications had code uploaded to a public repository, and one publication uploaded the dataset. Conclusions: Machine learning in palliative care is mainly used to predict mortality. Similarly to other applications of ML, external test sets and prospective validations are the exception.

Prediction of Survival Time in Advanced Lung Cancer: A Retrospective Study in Home-Based Palliative Care Unit

Background: There is a large group of patients suffering from lung cancer and receiving home hospice care in China. However, little is known about the prediction of their survival time. The purpose of this study was to determine whether quality of life independently predicts survival among advanced lung cancer survivors who are receiving home-based palliative care. Methods: In this retrospective study, we analyzed data from 937 advanced lung cancer patients who had received home-based palliative care between March 2010 and March 2020. We used Kaplan-Meier survival curves to determine the factors associated with survival time and applied the Cox proportional hazards model to examine the effect of quality of life on survival. Results: The study included 928 patients with a mean age of 63 years; and 72.1% of them were men. Factors associated with shortened survival included age, sex, place of residence, weight loss, anorexia, nausea, edema, quality of life, and Karnofsky performance status. After adjusting for other variables in a multivariate Cox proportional hazards model, we found that quality of life was an independent positive predictor of survival. Conclusions: As an independent factor predicting the survival of advanced lung cancer patients, quality of life should be taken seriously. Medical staff and healthcare workers need to pay special attention to this predictive factor since it may serve as early risk identification indicator for professionals who provide home-based palliative care, helping them to create effective personalized care plans.

Prediction of Survival in Patients with Advanced Cancer: A Narrative Review and Future Research Priorities

This paper aimed to summarize the current situation of prognostication for patients with an expected survival of weeks or months, and to clarify future research priorities. Prognostic information is essential for patients, their families, and medical professionals to make end-of-life decisions. The clinician's prediction of survival is often used, but this may be inaccurate and optimistic. Many prognostic tools, such as the Palliative Performance Scale, Palliative Prognostic Index, Palliative Prognostic Score, and Prognosis in Palliative Care Study, have been developed and validated to reduce the inaccuracy of the clinician's prediction of survival. To date, there is no consensus on the most appropriate method of comparing tools that use different formats to predict survival. Therefore, the feasibility of using prognostic scales in clinical practice and the information wanted by the end users can determine the appropriate prognostic tool to use. We propose four major themes for further prognostication research: (1) functional prognosis, (2) outcomes of prognostic communication, (3) artificial intelligence, and (4) education for clinicians.

Prognostic factors in patients in the terminal phase of haematological malignancies who are receiving home medical care

Although there are many prognostic models for patients in the terminal phase of solid tumours, a reliable prognostic scoring system in patients in the terminal phase of haematological malignancies (HM) has not been established. We retrospectively evaluated 180 patients in the terminal phase of HM who were receiving home medical care (HMC). Multivariate analyses revealed that clinician's estimate, consciousness, loss of appetite, dyspnoea, neutrophil count, lymphocyte count, and lactate dehydrogenase were associated with overall survival (OS). Based on this result, we developed a novel prognostic scoring system, the Japan palliative haematological oncology prognostic estimates, in which four risk groups were shown to clearly differ in survival (p < 0.001): a low-risk group (n = 41, median OS of 434 days), an intermediate-low-risk group (n = 80, median OS of 112 days), an intermediate-high-risk group (n = 38, median OS of 31.5 days), and a high-risk group (n = 21, median OS of 10 days). This is the first investigation of prognostic factors that influence the OS of patients in the terminal phase of HM who are receiving HMC. Providing patients with reliable information about their prognosis is important for them to consider how to spend their remaining life.

The Palliative Prognostic (PaP) Score without Clinical Evaluation Predicts Early Mortality among Advanced NSCLC Patients Treated with Immunotherapy

Background: An acceptable risk-benefit ratio may encourage the prescription of immune checkpoint inhibitors (ICI) near the late stage of life. The lung immune prognostic index (LIPI) was validated in advanced non-small cell lung cancer (NSCLC) patients treated with ICIs. The palliative prognostic (PaP) score without clinical prediction of survival (PaPwCPS) predicts early mortality probability in terminal cancer patients. Methods: We performed a retrospective study including 182 deceased advanced NSCLC patients, treated with single-agent ICI at our Institution. Two prognostic categories of high and low mortality risk were identified through ROC curve analysis for PaPwCPS and LIPI scores. Results: Most were >65 years of age (68.3%) and received second-line ICI (61.2%). A total of 29 (15.9%) and 131 (72.0%) patients died within 30 and 90 days from treatment start, respectively. A total of 81 patients (44.5%) received ICI during the last month of life. Baseline PaPwCPS and LIPI scores were assessable for 78 patients. The AUC of ROC curves was significantly increased for PaPwCPS as compared with LIPI score for both 30-day and 90-day mortality. A high PaPwCPS score was associated in multivariate analysis with increased 30-day (HR 2.69, p = 0.037) and 90-day (HR 4.01, p < 0.001) mortality risk. A high LIPI score was associated with increased 90-day mortality risk (p < 0.001). Conclusion: We found a tendency towards ICI prescription near the late stage of life. The PaPwCPS score was a reliable predictor of 30- and 90-day mortality.

Re-Whole Brain Radiotherapy May Be One of the Treatment Choices for Symptomatic Brain Metastases Patients

Generally, patients with multiple brain metastases receive whole brain radiotherapy (WBRT). Although, more than 60% of patients show complete or partial responses, many experience recurrence. Therefore, some institutions consider re-WBRT administration; however, there is insufficient information regarding this. Therefore, we aimed to review re-WBRT administration among these patients. Although most patients did not live longer than 12 months, symptomatic improvement was sometimes observed, with tolerable acute toxicities. Therefore, re-WBRT may be a treatment option for patients with symptomatic recurrence of brain metastases. However, physicians should consider this treatment cautiously because there is insufficient data on late toxicity, including radiation necrosis, owing to poor prognosis. A better prognostic factor for survival following radiotherapy administration may be the time interval of > 9 months between the first WBRT and re-WBRT, but there is no evidence supporting that higher doses lead to prolonged survival, symptom improvement, and tumor control. Therefore, 20 Gy in 10 fractions or 18 Gy in five fractions may be a reasonable treatment method within the tolerable total biological effective dose 2 ≤ 150 Gy, considering the biologically effective dose for tumors and normal tissues.

Home Artificial Nutrition and Energy Balance in Cancer Patients: Nutritional and Clinical Outcomes

Malnutrition is one of the main factors determining cachexia syndrome, which negatively impacts the quality of life and survival. In cancer patients, artificial nutrition is considered as an appropriate therapy when the impossibility of an adequate oral intake worsened nutritional and clinical conditions. This study aims to verify, in a home palliative care setting for cancer patients, if home artificial nutrition (HAN) supplies a patient’s energy requirement, improving nutritional and performance status. A nutritional service team performed counseling at a patient’s home and assessed nutritional status (body mass index, weight loss in the past 6 months), resting energy expenditure (REE), and oral food intake; Karnofsky Performance Status (KPS); cachexia degree; and survival. From 1990 to 2021, 1063 patients started HAN. Among these patients, 101 suspended artificial nutrition for oral refeeding. Among the 962 patients continuing HAN until death, 226 patients (23.5%) survived 6 weeks or less. HAN allowed to achieve a positive energy balance in 736 patients who survived more than 6 weeks, improving body weight and KPS when evaluated after 1 month of HAN. Advanced cancer and cachexia degree at the entry of the study negatively affected the positive impact of HAN.

Are Prognostic Scores Better Than Clinician Judgment? A Prospective Study Using Three Models

CONTEXT

Several prognostic models such as the Palliative Performance Scale (PPS), Palliative Prognostic Index (PPI), Palliative Prognostic Score (PaP) have been developed to complement clinician's prediction of survival (CPS). However, few studies with large scales have been conducted to show which prognostic tool had better performance than CPS in patients with weeks of survival.

OBJECTIVES

We aimed to compare the prognostic performance of the PPS, PPI, PaP, and CPS in inpatients admitted to palliative care units (PCUs).

METHODS

This study was part of a multi-center prospective observational study involving patients admitted to PCUs in Japan. We computed their prognostic performance using the area under the receiver operating characteristics curve (AUROC) and calibration plots for seven, 14-, 30- and 60-day survival.

RESULTS

We included 1896 patients with a median overall survival of 19 days. The AUROC was 73% to 84% for 60-day and 30-day survival, 75% to 84% for 14-day survival, and 80% to 87% for seven-day survival. The calibration plot demonstrated satisfactory agreement between the observational and predictive probability for the four indices in all timeframes. Therefore, all four prognostic indices showed good performance. CPS and PaP consistently had significantly better performance than the PPS and PPI from one-week to two-month timeframes.

CONCLUSION

The PPS, PPI, PaP, and CPS had relatively good performance in patients admitted to PCUs with weeks of survival. CPS and PaP had significantly better performance than the PPS and PPI. CPS may be sufficient for experienced clinicians while PPS may help to improve prognostic confidence for inexperienced clinicians.

Comparison of Objective Prognostic Score and Palliative Prognostic Score performance in inpatients with advanced cancer in Japan and Korea

OBJECTIVE

Accurate prognostication is important for patients and their families to prepare for the end of life. Objective Prognostic Score (OPS) is an easy-to-use tool that does not require the clinicians' prediction of survival (CPS), whereas Palliative Prognostic Score (PaP) needs CPS. Thus, inexperienced clinicians may hesitate to use PaP. We aimed to evaluate the accuracy of OPS compared with PaP in inpatients in palliative care units (PCUs) in three East Asian countries.

METHOD

This study was a secondary analysis of a cross-cultural, multicenter cohort study. We enrolled inpatients with far-advanced cancer in PCUs in Japan, Korea, and Taiwan from 2017 to 2018. We calculated the area under the receiver operating characteristics (AUROC) curve to compare the accuracy of OPS and PaP.

RESULTS

A total of 1,628 inpatients in 33 PCUs in Japan and Korea were analyzed. OPS and PaP were calculated in 71.7% of the Japanese patients and 80.0% of the Korean patients. In Taiwan, PaP was calculated for 81.6% of the patients. The AUROC for 3-week survival was 0.74 for OPS in Japan, 0.68 for OPS in Korea, 0.80 for PaP in Japan, and 0.73 for PaP in Korea. The AUROC for 30-day survival was 0.70 for OPS in Japan, 0.71 for OPS in Korea, 0.79 for PaP in Japan, and 0.74 for PaP in Korea.

SIGNIFICANCE OF RESULTS

Both OPS and PaP showed good performance in Japan and Korea. Compared with PaP, OPS could be more useful for inexperienced physicians who hesitate to estimate CPS.

End-of-Life Care for End-stage Heart Failure Patients

Although recent heart failure (HF) guidelines highlight integrative palliative care, including end-of-life (EOL), appropriate discussing EOL issues can be challenging due to possibility of unexpected deterioration throughout HF trajectory. Open communication and discussions with multidisciplinary team are important for setting patient and family expectations and establishing mutually agreed goals of care based firmly on the patient’s ‘human dignity’ and ‘right to self-determination.’ Especially when quality-of-life outweighs expanding quantity-of-life, transition to EOL care should be considered. Advanced care planning including resuscitation, device deactivation, site for last days, and bereavement support should focus on ensuring a good death, and be reviewed regularly.

Efforts to improve end-of-life (EOL) care have generally been focused on cancer patients, but high-quality EOL care is also important for patients with other serious medical illnesses including heart failure (HF). Recent HF guidelines offer more clinical considerations for palliative care including EOL care than ever before. Because HF patients can experience rapid, unexpected clinical deterioration or sudden death throughout the disease trajectory, choosing an appropriate time to discuss issues such as advance directives or hospice can be challenging in real clinical situations. Therefore, EOL issues should be discussed early. Conversations are important for understanding patient and family expectations and developing mutually agreed goals of care. In particular, high-quality communication with patient and family through a multidisciplinary team is necessary to define patient-centered goals of care and establish treatment based on goals. Control of symptoms such as dyspnea, pain, anxiety/depression, fatigue, nausea, anorexia, and altered mental status throughout the dying process is an important issue that is often overlooked. When quality-of-life outweighs expanding quantity-of-life, the transition to EOL care should be considered. Advanced care planning including resuscitation (i.e., do-not resuscitate order), device deactivation, site for last days and bereavement support for the family should focus on ensuring a good death and be reviewed regularly. It is essential to ensure that treatment for all HF patients incorporates discussions about the overall goals of care and individual patient preferences at both the EOL and sudden changes in health status. In this review, we focus on EOL care for end-stage HF patients.

Prognostication in palliative radiotherapy—ProPaRT: Accuracy of prognostic scores

Background

Prognostication can be used within a tailored decision-making process to achieve a more personalized approach to the care of patients with cancer. This prospective observational study evaluated the accuracy of the Palliative Prognostic score (PaP score) to predict survival in patients identified by oncologists as candidates for palliative radiotherapy (PRT). We also studied interrater variability for the clinical prediction of survival and PaP scores and assessed the accuracy of the Survival Prediction Score (SPS) and TEACHH score.

Materials and methods

Consecutive patients were enrolled at first access to our Radiotherapy and Palliative Care Outpatient Clinic. The discriminating ability of the prognostic models was assessed using Harrell’s C index, and the corresponding 95% confidence intervals (95% CI) were obtained by bootstrapping.

Results

In total, 255 patients with metastatic cancer were evaluated, and 123 (48.2%) were selected for PRT, all of whom completed treatment without interruption. Then, 10.6% of the irradiated patients who died underwent treatment within the last 30 days of life. The PaP score showed an accuracy of 74.8 (95% CI, 69.5–80.1) for radiation oncologist (RO) and 80.7 (95% CI, 75.9–85.5) for palliative care physician (PCP) in predicting 30-day survival. The accuracy of TEACHH was 76.1 (95% CI, 70.9–81.3) and 64.7 (95% CI, 58.8–70.6) for RO and PCP, respectively, and the accuracy of SPS was 70 (95% CI, 64.4–75.6) and 72.8 (95% CI, 67.3–78.3).

Conclusion

Accurate prognostication can identify candidates for low-fraction PRT during the last days of life who are more likely to complete the planned treatment without interruption.

All the scores showed good discriminating capacity; the PaP had the higher accuracy, especially when used in a multidisciplinary way.

Is spiritual well-being related to survival time of inpatients with advanced cancer? An East Asian cohort study

OBJECTIVES

It has been suggested that psychosocial factors are related to survival time of inpatients with cancer. However, there are not many studies examining the relationship between spiritual well-being (SWB) and survival time among countries. This study investigated the relationship between SWB and survival time among three East Asian countries.

METHODS

This international multicenter cohort study is a secondary analysis involving newly admitted inpatients with advanced cancer in palliative care units in Japan, South Korea, and Taiwan. SWB was measured using the Integrated Palliative Outcome Scale (IPOS) at admission. We performed multivariate analysis using the Cox proportional hazards model to identify independent prognostic factors.

RESULTS

A total of 2,638 patients treated at 37 palliative care units from January 2017 to September 2018 were analyzed. The median survival time was 18.0 days (95% confidence interval [CI] 16.5-19.5) in Japan, 23.0 days (95% CI 19.9-26.1) in Korea, and 15.0 days (95% CI 13.0-17.0) in Taiwan. SWB was a significant factor correlated with survival in Taiwan (hazard ratio [HR] 1.27; 95% CI 1.01-1.59; p = 0.04), while it was insignificant in Japan (HR 1.10; 95% CI 1.00-1.22; p = 0.06), and Korea (HR 1.02; 95% CI 0.77-1.35; p = 0.89).

SIGNIFICANCE OF RESULTS

SWB on admission was associated with survival in patients with advanced cancer in Taiwan but not Japan or Korea. The findings suggest the possibility of a positive relationship between spiritual care and survival time in patients with far advanced cancer.

HAprog: A New Prognostic Application to Assist Oncologists in Routine Care

CONTEXT

More patients are seeing palliative care (PC) earlier in the disease trajectory. The Barretos Prognostic Nomogram (BPN) was designed to fill the gap of survival prognostication for patients with advanced cancer and months of life expectancy. However, its routine use is limited by the common need for a ruler and calculator. Additionally, the BPN requires blood tests.

OBJECTIVES

The aim is to refine the BPN and to create a prognostic application (App) for use on smartphones.

METHODS

This is a reanalysis of the two cohorts of advanced cancer patients (development, n=215 and validation, n=276). The variable 'metastasis' was revised (volume-site combinations) and 'KPS' replaced by 'ECOG-PS'. Prognostic variables were selected for multivariable Cox and Log-logistic parametric regression analyses; the most accurate final models were identified by backward variable elimination. Calibration and discrimination properties were evaluated in the validation sample.

RESULTS

The 'full version' model is composed of 6 parameters: sex, locoregional disease, sites of metastasis, ECOG-PS, WBC and albumin. In the 'clinical version' model (5 variables), the variable 'antineoplastic treatment' was included and the laboratory variables were excluded. At validation, both models were well calibrated and presented adequate c-Index values (0.778 and 0.739). HAprog is a freely downloadable offline App that is used by clinicians to calculate prognosis in less than 1 minute.

CONCLUSION

The new models that integrate HAprog are refined prognostic tools with adequate calibration and discrimination properties. It has potential practical impact for the oncologist dealing with outpatients with advanced cancer during the decision-making process.

The ostrich approach - Prognostic avoidance, strategies and barriers to assessing older hospital patients' risk of dying

BACKGROUND

Predicting older patients' life expectancy is an important yet challenging task. Hospital aged care assessment teams advise treating teams on older patients' type and place of care, directly affecting quality of care. Yet, little is known about their experiences with prognostication.

METHODS

Twenty semi-structured interviews were conducted with seven geriatricians/ registrars, ten nurses and three allied health staff from aged care assessment teams across two hospitals in Melbourne, Australia. Data were analysed thematically.

RESULTS

To generate prognoses, clinicians used analytical thinking, intuition, assessments from others, and pattern matching. Prognostic tools were an underutilised resource. Barriers to recognition of dying included: diffusion of responsibility regarding whose role it is to identify patients at end-of-life; lack of feedback about whether a prognosis was correct; system pressures to pursue active treatment and vacate beds; avoidance of end-of-life discussions; lack of confidence, knowledge and training in prognostication and pandemic-related challenges.

Development and Validation of the PaP Score Nomogram for Terminally Ill Cancer Patients

The validated Palliative Prognostic (PaP) score predicts survival in terminally ill cancer patients, assigning patients to three different risk groups according to a 30-day survival probability: group A, >70%; group B, 30−70%; and group C, <30%. We aimed to develop and validate a PaP nomogram to provide individualized prediction of survival at 15, 30 and 60 days. Three cohorts of consecutive terminally ill cancer patients were used: one (n = 519) for nomogram development and internal validation, and a second (n = 451) and third (n = 549) for external validation. Multivariate analyses included dyspnea, anorexia, Karnofsky performance status, clinical prediction of survival, total white blood count and lymphocyte percentage. The predictive accuracy of the nomogram was determined by Harrell’s concordance index (95% CI), and calibration plots were generated. The nomogram had a concordance index of 0.74 (0.72−0.75) and showed good calibration. The internal validation showed no departures from ideal prediction. The accuracy of the nomogram at 15, 30 and 60 days was 74% (70−77), 89% (85−92) and 72% (68−76) in the external validation cohorts, respectively. The PaP nomogram predicts the individualized estimate of survival and could greatly facilitate clinical care decision-making at the end of life.

Cost-saving prediction model of transfer to palliative care for terminal cancer patients in a Japanese general hospital

BACKGROUND

Although medical costs need to be controlled, there are no easily applicable cost prediction models of transfer to palliative care (PC) for terminal cancer patients.

OBJECTIVE

Construct a cost-saving prediction model based on terminal cancer patients' data at hospital admission.

STUDY DESIGN

Retrospective cohort study.

SETTING

A Japanese general hospital.

PATIENTS

A total of 139 stage IV cancer patients transferred to PC, who died during hospitalization from April 2014 to March 2019.

MAIN OUTCOME MEASURE

Patients were divided into higher (59) and lower (80) total medical costs per day after transfer to PC. We compared demographics, cancer type, medical history, and laboratory results between the groups. Stepwise logistic regression analysis was used for model development and area under the curve (AUC) calculation.

RESULTS

A cost-saving prediction model (AUC = 0.78, 95% CI: 0.70, 0.85) with a total score of 13 points was constructed as follows: 2 points each for age ≤ 74 years, creatinine ≥ 0.68 mg/dL, and lactate dehydrogenase ≤ 188 IU/L; 3 points for hemoglobin ≤ 8.8 g/dL; and 4 points for potassium ≤ 3.3 mEq/L.

CONCLUSION

Our model contains five predictors easily available in clinical settings and exhibited good predictive ability.

Comparison of intuitive assessment and palliative care screening tool in the early identification of patients needing palliative care

The intuitive assessment of palliative care (PC) needs and Palliative Care Screening Tool (PCST) are the assessment tools used in the early detection of patients requiring PC. However, the comparison of their prognostic accuracies has not been extensively studied. This cohort study aimed to compare the validity of intuitive assessment and PCST in terms of recognizing patients nearing end-of-life (EOL) and those appropriate for PC. All adult patients admitted to Taipei City Hospital from 2016 through 2019 were included in this prospective study. We used both the intuitive assessment of PC and PCST to predict patients’ 6-month mortality and identified those appropriate for PC. The c-statistic value was calculated to indicate the predictive accuracies of the intuition and PCST. Of 111,483 patients, 4.5% needed PC by the healthcare workers’ intuitive assessment, and 6.7% had a PCST score ≥ 4. After controlling for other covariates, a positive response ‘yes’ to intuitive assessment of PC needs [adjusted odds ratio (AOR) = 9.89; 95% confidence interval (CI) 914–10.71] and a PCST score ≥ 4 (AOR = 6.59; 95%CI 6.17–7.00) were the independent predictors of 6-month mortality. Kappa statistics showed moderate concordance between intuitive assessment and PCST in predicting patients' 6-month mortality (k = 0.49). The c-statistic values of the PCST at recognizing patients’ 6-month mortality was significantly higher than intuition (0.723 vs. 0.679; p < 0.001). As early identification of patients in need of PC could improve the quality of EOL care, our results suggest that it is imperative to screen patients’ palliative needs by using a highly accurate screening tool of PCST.

[Re-visiting the Pronopall score ten years later: A multicenter retrospective study]

INTRODUCTION

The Pronopall score, which distinguishes 3 prognostic groups in patients with advanced cancer, was initially proposed in 2008 and validated in a study published in 2018 but including patients between 2009 and 2010. Since the last decade, cancer management and therapeutic options have progressed. The objective of this study was to confirm the value of this score in patients with digestive and thoracic cancer.

METHODS

From July 2019 to November 2020, this retrospective multi-center study included patients with digestive or thoracic cancers who fulfilled the same inclusion criteria as those used in the initial study, and in whom the Pronopall score could be calculated using its four variables (albumin serum level, LDH level, ECOG score, number of metastatic sites). Survival curves were analyzed using the Kaplan-Meier method.

RESULTS

One hundred patients were included. According to the Pronopall score, patients were separated into group A (score 8-10, 7 patients), group B (score 4-7, 41 patients) and group C (score 0-3, 52 patients). Median overall survival was 73 days, CI [17-129], 228 days, CI [128-328] and 575 days, CI [432-718] for groups A, B and C, respectively. Survival at 2 months was 28 % for population A, 61 % for population B, and 94 % for population C.

CONCLUSION

This study confirms that the Pronopall score still allows clinically relevant discrimination of patients, score C being associated with a good prognosis compared to scores A and B.

Knowledge and use of prognostic scales by oncologists and palliative care physicians in adult patients with advanced cancer: A national survey (ONCOPRONO study)

BACKGROUND

Prognostic scales exist to estimate patient survival in advanced cancer. However, there are no studies evaluating their use and practice. The objective of this study was to evaluate in a nationwide study the proportion of oncologists and palliative care physicians who had knowledge of these scales.

METHODS

A descriptive, national, cross-sectional study was conducted via an online questionnaire to oncologists and palliative care physicians across France.

RESULTS

Palliative care physicians had better knowledge of the scales than oncologists (42.3% (n = 74) vs. 27.8% (n = 33), p = 0.015). The Palliative Performance Status (PPS) and Pronopall Scale were the best-known (51.4% (n = 55) and 65.4% (n = 70), respectively) and the most widely used (35% (n = 28) and 60% (n = 48), respectively). Improved training in the use of these scales was requested by 85.4% (n = 251) of participants, while 72.8% (n = 214) reported that they did not use them at all. Limited training and lack of consensus on which scale to use were cited as the main obstacles to use.

CONCLUSION

This is the first national study on the use of prognostic scales in advanced cancer. Our findings highlight a need to improve training in these scales and to reach a consensus on scale selection.

Perimortem changes in clinical parameters in patients undergoing maintenance hemodialysis

Background

End-of-life medical care for patients receiving maintenance hemodialysis (HD) therapy has become an increasingly important issue worldwide. Thus far, no clear indicators and/or biomarkers exist regarding the timing of HD therapy withdrawal.

Methods

To clarify the perimortem circumstances, we examined temporal changes in multiple clinical parameters during the last 10 serial HD sessions of 65 terminal patients with end-stage renal disease who had undergone maintenance HD and died in our hospital.

Results

The results showed that, while most of the laboratory data were unaltered, the physical parameters, such as systolic blood pressure and consciousness levels, gradually and significantly deteriorated toward the last HD session prior to death. The frequency of the use of vasopressors and O2 inhalation tended to increase. The accumulation of such severe conditions was observed at the last HD session. Of interest, the accumulation of severe conditions at the last HD session in patients with malignancies was significantly less than those with cardiovascular diseases or infectious diseases. The accumulation of severe conditions at the last HD session did not differ between patients who withdrew HD versus those who continued HD.

Conclusion

The results of the present study suggest that predicting the timing of maintenance HD therapy withdrawal is likely to be difficult and that the timing of maintenance HD therapy termination may differ among patient groups with distinct comorbid conditions.

Deep-Learning Approach to Predict Survival Outcomes Using Wearable Actigraphy Device Among End-Stage Cancer Patients

Survival prediction is highly valued in end-of-life care clinical practice, and patient performance status evaluation stands as a predominant component in survival prognostication. While current performance status evaluation tools are limited to their subjective nature, the advent of wearable technology enables continual recordings of patients' activity and has the potential to measure performance status objectively. We hypothesize that wristband actigraphy monitoring devices can predict in-hospital death of end-stage cancer patients during the time of their hospital admissions. The objective of this study was to train and validate a long short-term memory (LSTM) deep-learning prediction model based on activity data of wearable actigraphy devices. The study recruited 60 end-stage cancer patients in a hospice care unit, with 28 deaths and 32 discharged in stable condition at the end of their hospital stay. The standard Karnofsky Performance Status score had an overall prognostic accuracy of 0.83. The LSTM prediction model based on patients' continual actigraphy monitoring had an overall prognostic accuracy of 0.83. Furthermore, the model performance improved with longer input data length up to 48 h. In conclusion, our research suggests the potential feasibility of wristband actigraphy to predict end-of-life admission outcomes in palliative care for end-stage cancer patients. Clinical Trial Registration: The study protocol was registered on ClinicalTrials.gov (ID: NCT04883879).

AutoScore-Survival: Developing interpretable machine learning-based time-to-event scores with right-censored survival data

BACKGROUND

Scoring systems are highly interpretable and widely used to evaluate time-to-event outcomes in healthcare research. However, existing time-to-event scores are predominantly created ad-hoc using a few manually selected variables based on clinician's knowledge, suggesting an unmet need for a robust and efficient generic score-generating method.

METHODS

AutoScore was previously developed as an interpretable machine learning score generator, integrating both machine learning and point-based scores in the strong discriminability and accessibility. We have further extended it to the time-to-event outcomes and developed AutoScore-Survival, for generating time-to-event scores with right-censored survival data. Random survival forest provided an efficient solution for selecting variables, and Cox regression was used for score weighting. We implemented our proposed method as an R package. We illustrated our method in a study of 90-day survival prediction for patients in intensive care units and compared its performance with other survival models, the random survival forest, and two traditional clinical scores.

RESULTS

The AutoScore-Survival-derived scoring system was more parsimonious than survival models built using traditional variable selection methods (e.g., penalized likelihood approach and stepwise variable selection), and its performance was comparable to survival models using the same set of variables. Although AutoScore-Survival achieved a comparable integrated area under the curve of 0.782 (95% CI: 0.767-0.794), the integer-valued time-to-event scores generated are favorable in clinical applications because they are easier to compute and interpret.

CONCLUSIONS

Our proposed AutoScore-Survival provides a robust and easy-to-use machine learning-based clinical score generator to studies of time-to-event outcomes. It gives a systematic guideline to facilitate the future development of time-to-event scores for clinical applications.

Comparison of the accuracy of clinicians' prediction of survival and Palliative Prognostic Score: an East Asian cross-cultural study

PURPOSE

No study has been conducted to compare the clinicians' prediction of survival (CPS) with Palliative Prognostic Scores (PaP) across countries. We aimed to compare the performance of the CPS in PaP (PaP-CPS), the PaP without the CPS, and the PaP total scores in patients with advanced cancer in three East Asian countries.

METHODS

We compared the discriminative accuracy of the three predictive models (the PaP-CPS [the score of the categorical CPS of PaP], the PaP without the CPS [sum of the scores of only the objective variables of PaP], and the PaP total score) in patients admitted to palliative care units (PCUs) in Japan, Korea, and Taiwan. We calculated the area under the receiver operating characteristic curve (AUROC) for 30-day survival to compare the discriminative accuracy of these three models.

RESULTS

We analyzed 2,072 patients from three countries. The AUROC for the PaP total scores was 0.84 in patients in Japan, 0.76 in Korea, and 0.79 in Taiwan. The AUROC of the PaP-CPS was 0.82 in patients in Japan, 0.75 in Korea, and 0.78 in Taiwan. The AUROC of the PaP without the CPS was 0.75 in patients in Japan, 0.66 in Korea, and 0.67 in Taiwan.

CONCLUSION

The PaP total scores and the PaP-CPS consistently showed similar discriminative accuracy in predicting 30-day survival in patients admitted to PCUs in Japan, Korea, and Taiwan. It may be sufficient for experienced clinicians to use the CPS alone for estimating the short-term survival (less than one month) of patients with far-advanced cancer. The PaP may help to improve prognostic confidence and further reduce subjective variations.

Improving Machine Learning 30-Day Mortality Prediction by Discounting Surprising Deaths

BACKGROUND

Machine learning (ML) is an emerging tool for predicting need of end-of-life discussion and palliative care, by using mortality as a proxy. But deaths, unforeseen by emergency physicians at time of the emergency department (ED) visit, might have a weaker association with the ED visit.

OBJECTIVES

To develop an ML algorithm that predicts unsurprising deaths within 30 days after ED discharge.

METHODS

In this retrospective registry study, we included all ED attendances within the Swedish region of Halland in 2015 and 2016. All registered deaths within 30 days after ED discharge were classified as either "surprising" or "unsurprising" by an adjudicating committee with three senior specialists in emergency medicine. ML algorithms were developed for the death subclasses by using Logistic Regression (LR), Random Forest (RF), and Support Vector Machine (SVM).

RESULTS

Of all 30-day deaths (n = 148), 76% (n = 113) were not surprising to the adjudicating committee. The most common diseases were advanced stage cancer, multidisease/frailty, and dementia. By using LR, RF, and SVM, mean area under the receiver operating characteristic curve (ROC-AUC) of unsurprising deaths in the test set were 0.950 (SD 0.008), 0.944 (SD 0.007), and 0.949 (SD 0.007), respectively. For all mortality, the ROC-AUCs for LR, RF, and SVM were 0.924 (SD 0.012), 0.922 (SD 0.009), and 0.931 (SD 0.008). The difference in prediction performance between all and unsurprising death was statistically significant (P < .001) for all three models.

CONCLUSION

In patients discharged to home from the ED, three-quarters of all 30-day deaths did not surprise an adjudicating committee with emergency medicine specialists. When only unsurprising deaths were included, ML mortality prediction improved significantly.

Comparing the performance of the palliative prognostic (PaP) score with clinical predictions of survival: A systematic review

BACKGROUND

In patients with advanced cancer, prognosis is usually determined using clinicians' predictions of survival (CPS). The palliative prognostic (PaP) score is a prognostic algorithm that was developed to predict survival in patients with advanced cancer. The score categorises patients into three risk groups in accordance with their probability of surviving for 30 days. The relative accuracy of PaP and CPS is unclear.

DESIGN

This was a systematic review of MEDLINE, Embase, AMED, CINAHL Plus and the Cochrane Database of Systematic Reviews and Trials from inception up to June 2021. The inclusion criteria were studies in adults with advanced cancer reporting data on performance of both PaP and CPS. Data were extracted on accuracy of prognoses and where available on discrimination (area under the receiver operating characteristic curve or C-index) and/or diagnostic performance (sensitivity, specificity).

RESULTS

Eleven studies were included. One study reported a direct comparison between PaP risk groups and equivalent risk groups defined by CPS and found that PaP was as accurate as CPS. Five studies reported discrimination of PaP as a continuous total score (rather than using the previously validated risk categories) and reported C-statistics that ranged from 0.64 (95% confidence interval [CI] 0.54, 0.74) up to 0.90 (95% CI 0.87, 0.92). Other studies compared PaP against CPS using non-equivalent metrics (e.g. comparing probability estimates against length of survival estimates).

CONCLUSIONS

PaP risk categories and CPS are equally able to discriminate between patients with different survival probabilities. Total PaP scores show good discrimination between patients in accordance with their length of survival. The role of PaP in clinical practice still needs to be defined.

TRIAL REGISTRATION

PROSPERO (CRD42021241074, 5th March 2021).

Improving time to palliative care review with predictive modeling in an inpatient adult population: study protocol for a stepped-wedge, pragmatic randomized controlled trial

Background

Palliative care is a medical specialty centered on improving the quality of life (QOL) of patients with complex or life-threatening illnesses. The need for palliative care is increasing and with that the rigorous testing of triage tools that can be used quickly and reliably to identify patients that may benefit from palliative care.

Methods

To that aim, we will conduct a two-armed stepped-wedge cluster randomized trial rolled out to two inpatient hospitals to evaluate whether a machine learning algorithm accurately identifies patients who may benefit from a comprehensive review by a palliative care specialist and decreases time to receiving a palliative care consult in hospital. This is a single-center study which will be conducted from August 2019 to November 2020 at Saint Mary’s Hospital & Methodist Hospital both within Mayo Clinic Rochester in Minnesota. Clusters will be nursing units which will be chosen to be a mix of complex patients from Cardiology, Critical Care, and Oncology and had previously established relationships with palliative medicine. The stepped wedge design will have 12 units allocated to a design matrix of 5 treatment wedges. Each wedge will last 75 days resulting in a study period of 12 months of recruitment unless otherwise specified. Data will be analyzed with Bayesian hierarchical models with credible intervals denoting statistical significance.

Discussion

This intervention offers a pragmatic approach to delivering specialty palliative care to hospital patients in need using machine learning, thereby leading to high value care and improved outcomes. It is not enough for AI to be utilized by simply publishing research showing predictive performance; clinical trials demonstrating better outcomes are critically needed. Furthermore, the deployment of an AI algorithm is a complex process that requires multiple teams with varying skill sets. To evaluate a deployed AI, a pragmatic clinical trial can accommodate the difficulties of clinical practice while retaining scientific rigor.

Trial registration

ClinicalTrials.gov NCT03976297. Registered on 6 June 2019, prior to trial start.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13063-021-05546-5.

Prognostic Value of Palliative Prognostic Index for Hospitalized Patients With End-of-Life Hematologic Malignancies in a Japanese University Hospital

PURPOSE

Uncertainty of prognosis is one reason patients with hematologic malignancies receive aggressive therapy near end of life more often than those with advanced solid tumors. It is unknown whether end-of-life prognosis prediction models are useful for patients with hematologic malignancies, especially hospitalized patients receiving chemotherapy, because most prognostic models were developed for patients with solid tumors. The purpose of this study was to evaluate the prognostic accuracy of the Palliative Prognostic Index (PPI) for end-of-life patients with advanced hematologic malignancies.

METHODS

We retrospectively reviewed the records of 143 patients who became resistant to standard chemotherapy and died of disease progression in our university hospital hematology ward between May 2015 and November 2019. Patients were classified according to PPI scores (groups: A, PPI ≤ 2.0; B, 2.0 < PPI ≤ 4.0; and C, PPI > 4.0) based on their clinical charts at admission. The median overall survival for each patient (95% confidence interval) was calculated using the Kaplan-Meier method. Log-rank tests were used to determine significant differences between survival curves.

RESULTS

Median patient age was 76 years (range: 39-92 years), and 59% were men. Median overall survival times in the PPI groups A, B, and C were 58 days, 36 days, and 10 days, respectively. Statistically significant differences in survival time were observed between the groups (P < .01); prediction accuracy was similar to that for patients with different diagnoses.

CONCLUSION

The usefulness of PPI was validated for near-end-of-life hospitalized patients with hematologic malignancies.

Utility of Palliative Prognostic Index in Predicting Survival Outcomes in Patients With Hematological Malignancies in the Acute Ward Setting

BACKGROUND

The palliative prognostic index (PPI) predicts the life expectancy of patients with terminally ill cancer in hospice settings. This study aimed to evaluate PPI as a prognostic tool for predicting the life expectancy of patients with hematological malignancies admitted to the acute ward.

METHODS

A total of 308 patients with hematological malignancies admitted to the hematological ward at a medical center between January 2016 and December 2017 were consecutively enrolled. PPI was scored within 24 h of admission. All patients were categorized into 3 groups by PPI for comparing survival and in-hospital mortality rates.

RESULTS

The median survival times were 38.4, 3.6, and 1.1 months for patients with good, intermediate, and poor prognostic group, respectively. The hazard ratio was 2.31 (95% CI 1.59-3.35, p < 0.001) when comparing the intermediate and good prognosis groups, and 3.90 (95% CI 2.52-6.03, p < 0.001) when comparing the poor and good prognosis groups. Forty-five (14.6%) patients died at discharge; in-hospital mortality rates among the good, intermediate, and poor prognostic groups were 9.0%, 23.4%, and 46.4%, respectively. The adjusted odds ratio for in-hospital mortality was 1.96 (95% CI, 0.80-4.82, p = 0.14) and 5.25 (95% CI, 2.01-13.7, p < 0.001) for patients in the intermediate and poor prognostic groups compared to those in the good prognostic group.

CONCLUSION

PPI is an accurate prognostic tool for predicting survival times and in-hospital mortality rates in patients with hematological malignancies in an acute ward setting. PPI could assist clinicians in discussing end-of-life issues and in referring patients with hematological malignancies to palliative care.

"The surprise questions" using variable time frames in hospitalized patients with advanced cancer

OBJECTIVE

Several studies supported the usefulness of "the surprise question" in terms of 1-year mortality of patients. "The surprise question" requires a "Yes" or "No" answer to the question "Would I be surprised if this patient died in [specific time frame]." However, the 1-year time frame is often too long for advanced cancer patients seen by palliative care personnel. "The surprise question" with shorter time frames is needed for decision making. We examined the accuracy of "the surprise question" for 7-day, 21-day, and 42-day survival in hospitalized patients admitted to palliative care units (PCUs).

METHOD

This was a prospective multicenter cohort study of 130 adult patients with advanced cancer admitted to 7 hospital-based PCUs in South Korea. The accuracy of "the surprise question" was compared with that of the temporal question for clinician's prediction of survival.

RESULTS

We analyzed 130 inpatients who died in PCUs during the study period. The median survival was 21.0 days. The sensitivity, specificity, and overall accuracy for the 7-day "the surprise question" were 46.7, 88.7, and 83.9%, respectively. The sensitivity, specificity, and overall accuracy for the 7-day temporal question were 6.7, 98.3, and 87.7%, respectively. The c-indices of the 7-day "the surprise question" and 7-day temporal question were 0.662 (95% CI: 0.539-0.785) and 0.521 (95% CI: 0.464-0.579), respectively. The c-indices of the 42-day "the surprise question" and 42-day temporal question were 0.554 (95% CI: 0.509-0.599) and 0.616 (95% CI: 0.569-0.663), respectively.

SIGNIFICANCE OF RESULTS

Surprisingly, "the surprise questions" and temporal questions had similar accuracies. The high specificities for the 7-day "the surprise question" and 7- and 21-day temporal question suggest they may be useful to rule in death if positive.

Accuracy of the Palliative Prognostic Score With or Without Clinicians' Prediction of Survival in Patients With Far Advanced Cancer

CONTEXT

Previous studies suggest that clinicians' prediction of survival (CPS) may have reduced the accuracy of objective indicators for prognostication in palliative care.

OBJECTIVES

We aimed to examine the accuracy of CPS alone, compared to the original Palliative Prognostic Score (PaP), and five clinical/laboratory variables of the PaP in patients with far advanced cancer.

METHODS

We compared the discriminative accuracy of three prediction models (the PaP-CPS [the score of the categorical CPS of PaP], PaP without CPS [sum of the scores of only the objective variables of PaP], and PaP total score) across 3 settings: inpatient palliative care consultation team, palliative care unit, and home palliative care. We computed the area under receiver operating characteristic curve (AUROC) for 30-day survival and concordance index (C-index) to compare the discriminative accuracy of these three models.

RESULTS

We included a total of 1534 subjects with median survival of 34.0 days. The AUROC and C-index in the three settings were 0.816-0.896 and 0.732-0.799 for the PaP total score, 0.808-0.884 and 0.713-0.782 for the PaP-CPS, and 0.726-0.815 and 0.672-0.728 for the PaP without CPS, respectively. The PaP total score and PaP-CPS showed similar AUROCs and C-indices across the three settings. The PaP total score had significantly higher AUROCs and C-indices than the PaP without CPS across the three settings.

CONCLUSION

Overall, the PaP total score, PaP-CPS, and PaP without CPS showed good discriminative performances. However, the PaP total score and PaP-CPS were significantly more accurate than the PaP without CPS.