A machine learning tool to predict mortality risk among patients with metastatic cancer in outpatient oncology care.

2021 ◽  
Vol 39 (15_suppl) ◽  
pp. 1560-1560
Author(s):  
Brandon Butler ◽  
Nadaa Tayiab ◽  
Serra Phu ◽  
Susan Nga Hoang ◽  
Brian Turnwald ◽  
...  
Keyword(s):  
Machine Learning ◽  
End Of Life ◽  
Metastatic Cancer ◽  
External Validation ◽  
Selection Criterion ◽  
Study Cohort ◽  
Multiple Time ◽  
Care Model ◽  
Final Model ◽  
Oncology Care

1560 Background: End-of-life management is a well-known challenging aspect of cancer care. In particular, timely hospice enrollment is a leading quality metric in the Oncology Care Model that has substantial room for improvement. An automated algorithmic tool that can incorporate the wealth of available EHR data and rapidly identify patients with a high risk of imminent mortality could be a valuable asset to supplement important clinical decisions and improve timely hospice care. Methods: A retrospective study cohort was formed using patients with metastatic cancer from US Oncology Network (USON) practices participating in the Oncology Care Model (OCM) between January 1, 2017 and June 30, 2019. Patients were required to have at least one record for lab values and vital signs in the EHR database. Patients were excluded from the study cohort if they were not enrolled in the OCM program or did not have a diagnosis for metastatic cancer. The patients satisfying the selection criterion were used to train and optimize the model. The training dataset was also used for internal validation and hyperparameter tuning until the final model was produced. As external validation, the final model was independently tested on 3 separate holdout datasets including OCM patients between July 1, 2019 and March 31, 2020. To avoid bias, all holdout datasets used for validation were excluded from the model. Results: A multivariable model to predict 90-day mortality was developed using a retrospective dataset derived from EHR data and Medicare claims data. A logistic regression algorithm using L1 (lasso) regularization yielded the best performance compared to other model candidates. The performance on the training cohort was given by a cross-validated AUC score of 0.85 (95% CI, 0.84 to 0.86). Further, external validation conducted using 3 independent holdout datasets demonstrated impressive generalizability marked by stable performance scores across multiple time periods (AUC between 0.84 and 0.85). Conclusions: This study builds upon previous work and further establishes the utility of machine learning to predict risk of imminent mortality for advanced cancer patients using available EHR data. A data-driven tool that estimates the probability of 90-day mortality could be leveraged as a powerful supplementary aid to clinicians managing end-of-life care at oncology practices.

American Society of Clinical Oncology Provisional Clinical Opinion: The Integration of Palliative Care Into Standard Oncology Care

2012 ◽  
Vol 30 (8) ◽  
pp. 880-887 ◽  
Author(s):  
Thomas J. Smith ◽  
Sarah Temin ◽  
Erin R. Alesi ◽  
Amy P. Abernethy ◽  
Tracy A. Balboni ◽  
...  
Keyword(s):  
Health Care ◽  
Palliative Care ◽  
End Of Life ◽  
Clinical Oncology ◽  
Metastatic Cancer ◽  
Expert Consensus ◽  
Care Services ◽  
Palliative Care Services ◽  
Oncology Care ◽  
American Society

Purpose An American Society of Clinical Oncology (ASCO) provisional clinical opinion (PCO) offers timely clinical direction to ASCO's membership following publication or presentation of potentially practice-changing data from major studies. This PCO addresses the integration of palliative care services into standard oncology practice at the time a person is diagnosed with metastatic or advanced cancer. Clinical Context Palliative care is frequently misconstrued as synonymous with end-of-life care. Palliative care is focused on the relief of suffering, in all of its dimensions, throughout the course of a patient's illness. Although the use of hospice and other palliative care services at the end of life has increased, many patients are enrolled in hospice less than 3 weeks before their death, which limits the benefit they may gain from these services. By potentially improving quality of life (QOL), cost of care, and even survival in patients with metastatic cancer, palliative care has increasing relevance for the care of patients with cancer. Until recently, data from randomized controlled trials (RCTs) demonstrating the benefits of palliative care in patients with metastatic cancer who are also receiving standard oncology care have not been available. Recent Data Seven published RCTs form the basis of this PCO. Provisional Clinical Opinion Based on strong evidence from a phase III RCT, patients with metastatic non–small-cell lung cancer should be offered concurrent palliative care and standard oncologic care at initial diagnosis. While a survival benefit from early involvement of palliative care has not yet been demonstrated in other oncology settings, substantial evidence demonstrates that palliative care—when combined with standard cancer care or as the main focus of care—leads to better patient and caregiver outcomes. These include improvement in symptoms, QOL, and patient satisfaction, with reduced caregiver burden. Earlier involvement of palliative care also leads to more appropriate referral to and use of hospice, and reduced use of futile intensive care. While evidence clarifying optimal delivery of palliative care to improve patient outcomes is evolving, no trials to date have demonstrated harm to patients and caregivers, or excessive costs, from early involvement of palliative care. Therefore, it is the Panel's expert consensus that combined standard oncology care and palliative care should be considered early in the course of illness for any patient with metastatic cancer and/or high symptom burden. Strategies to optimize concurrent palliative care and standard oncology care, with evaluation of its impact on important patient and caregiver outcomes (eg, QOL, survival, health care services utilization, and costs) and on society, should be an area of intense research. NOTE. ASCO's provisional clinical opinions (PCOs) reflect expert consensus based on clinical evidence and literature available at the time they are written and are intended to assist physicians in clinical decision making and identify questions and settings for further research. Because of the rapid flow of scientific information in oncology, new evidence may have emerged since the time a PCO was submitted for publication. PCOs are not continually updated and may not reflect the most recent evidence. PCOs cannot account for individual variation among patients and cannot be considered inclusive of all proper methods of care or exclusive of other treatments. It is the responsibility of the treating physician or other health care provider, relying on independent experience and knowledge of the patient, to determine the best course of treatment for the patient. Accordingly, adherence to any PCO is voluntary, with the ultimate determination regarding its application to be made by the physician in light of each patient's individual circumstances. ASCO PCOs describe the use of procedures and therapies in clinical trials and cannot be assumed to apply to the use of these interventions in the context of clinical practice. ASCO assumes no responsibility for any injury or damage to persons or property arising out of or related to any use of ASCO's PCOs, or for any errors or omissions.

scholarly journals Cardiovascular risk prediction in healthy older people

GeroScience ◽  
2021 ◽  
Author(s):  
Johannes T. Neumann ◽  
Le T. P. Thao ◽  
Emily Callander ◽  
Enayet Chowdhury ◽  
Jeff D. Williamson ◽  
...  
Keyword(s):  
Machine Learning ◽  
Cox Regression ◽  
External Validation ◽  
Randomised Trial ◽  
Antihypertensive Agents ◽  
Major Adverse Cardiovascular Events ◽  
Final Model ◽  
Elderly Individuals ◽  
Healthy Elderly ◽  
Increased Risk

AbstractIdentification of individuals with increased risk of major adverse cardiovascular events (MACE) is important. However, algorithms specific to the elderly are lacking. Data were analysed from a randomised trial involving 18,548 participants ≥ 70 years old (mean age 75.4 years), without prior cardiovascular disease events, dementia or physical disability. MACE included coronary heart disease death, fatal or nonfatal ischaemic stroke or myocardial infarction. Potential predictors tested were based on prior evidence and using a machine-learning approach. Cox regression analyses were used to calculate 5-year predicted risk, and discrimination evaluated from receiver operating characteristic curves. Calibration was also assessed, and the findings internally validated using bootstrapping. External validation was performed in 25,138 healthy, elderly individuals in the primary care environment. During median follow-up of 4.7 years, 594 MACE occurred. Predictors in the final model included age, sex, smoking, systolic blood pressure, high-density lipoprotein cholesterol (HDL-c), non-HDL-c, serum creatinine, diabetes and intake of antihypertensive agents. With variable selection based on machine-learning, age, sex and creatinine were the most important predictors. The final model resulted in an area under the curve (AUC) of 68.1 (95% confidence intervals 65.9; 70.4). The model had an AUC of 67.5 in internal and 64.2 in external validation. The model rank-ordered risk well but underestimated absolute risk in the external validation cohort. A model predicting incident MACE in healthy, elderly individuals includes well-recognised, potentially reversible risk factors and notably, renal function. Calibration would be necessary when used in other populations.

scholarly journals Impact of palliative care consultation on end-of-life care measures: A retrospective analysis of patients in the Oncology Care Model.

2019 ◽  
Vol 37 (27_suppl) ◽  
pp. 45-45
Author(s):  
Alison Greidinger ◽  
Maria Vershvovsky ◽  
Evan Lapinsky ◽  
Alison Rhoades ◽  
Amy Leader ◽  
...  
Keyword(s):  
Palliative Care ◽  
End Of Life ◽  
Acute Care ◽  
Advanced Cancer ◽  
Cancer Center ◽  
Do Not Resuscitate ◽  
Care Model ◽  
Code Status ◽  
Palliative Care Consultation ◽  
Oncology Care

45 Background: Despite a 2016 ASCO recommendation that patients with advanced cancer receive dedicated palliative care (PC) services, many patients are not referred and continue to receive chemotherapy and utilize high-acuity services near the end of life (EOL). Studies suggest that early PC involvement is associated with lower spending, acute care utilization, and chemotherapy administration at the EOL. The Sidney Kimmel Cancer Center participates in the Oncology Care Model (OCM), a CMS episode-based alternative payment model promoting high-value care. Using OCM-generated data, we evaluated the effect of PC visits on EOL outcomes. Methods: We identified OCM patients with episodes starting April 1, 2016-July 1, 2018 with GI and head & neck malignancies who had died, and determined whether patients who saw a PC provider had greater documentation of a code status (CS) before death, as well as lower spending and utilization of chemotherapy or acute care in the last 30 days of life. CMS spending data and dates of death were derived from OCM quarterly feedback, while all other data was compiled via chart review. CS was recorded at the start of the episode and at the time of death. Results: The study included 126 patients (median age 66 years), of whom 38% had a PC visit. 24% had only an inpatient (IP) PC consult, 6% only an outpatient (OP) visit, and 9% both IP & OP visits. More patients who saw PC had an initial CS documented (85%, vs 46% for no PC), and had a greater proportional increase in CS documentation before death (96% vs 53%). Despite similar rates at baseline, the final CS was significantly more likely to be “Do Not Resuscitate/Intubate” (DNR/DNI) among PC patients (79%, vs 28% for no PC). An initial CS of DNR/DNI was associated with lower mean ICU and total non-hospice spending in the last 30 days of life. Conclusions: This retrospective study in OCM patients found that PC intervention is associated with improved documentation of a CS and higher rates of DNR/DNI documentation before death. There is an association between an initial DNR/DNI CS and lower acute care spending. This data suggests a beneficial effect of early PC on utilization at the EOL in advanced cancer patients.

Oncology care model: Early palliative care experiences and outcomes.

2017 ◽  
Vol 35 (31_suppl) ◽  
pp. 37-37 ◽  
Author(s):  
Pamela Spain ◽  
Nathan West ◽  
Stephanie Teixeira-Poit ◽  
Elizabeth Schaefer ◽  
Kerry Ketler
Keyword(s):  
Decision Making ◽  
Palliative Care ◽  
End Of Life ◽  
Shared Decision Making ◽  
Life Quality ◽  
Risk Scores ◽  
Shared Decision ◽  
Care Model ◽  
Early Palliative Care ◽  
Oncology Care

37 Background: Through the Oncology Care Model (OCM), the Center for Medicare & Medicaid Innovation at the Centers for Medicare & Medicaid Services aims to improve the effectiveness and efficiency of cancer care. OCM practices have committed to provide enhanced services to Medicare beneficiaries, including palliative care designed to optimize quality of life. This study examines if OCM practices engaging in early palliative care discussions have timely hospice referrals as well a lower aggressive end-of-life care and Medicare costs. Methods: During site visits to 30 OCM practices, we asked was How and when is palliative care introduced to patients? We used Medicare claims data to stratify the 30 practices into high or low performers based on 3 end-of-life quality measures scores. Next, we examined their scores on Cancer CAHPS shared decision making and Medicare expenditures, as well as what staff reported about the use of palliative care during site visits. Claims and CAHPS data encompass the first 6 months of OCM, July-December 2016. Site visits were conducted February - May 2017. Results: Patient risk scores were equal among practice groups. High Performers said: “Palliative care is introduced right off the bat. We provide information on hospice and palliative care so it’s not a word they find with end of life only. It’s a difficult conversation but you have to put it out there. As part of patient education, we say the differences between palliative care and hospice. Low Performers said: “Palliative care is introduced on a case by case basis. To transfer to an inpatient hospice, now you have to be on your last breath. Programs that get cut because they are not integral to the patients’ acute issues include palliative care. If hospice comes up, I pull palliative care at that point.” Conclusions: Early OCM results support growing evidence that palliative care shared decision making discussions are most beneficial near the time of cancer diagnosis. [Table: see text]

Impact of the Oncology Care Model on End‐of‐Life Care

2021 ◽  
Vol 56 (S2) ◽  
pp. 18-18
Author(s):  
Qing Zheng ◽  
Thomas Christian ◽  
Sean McClellan ◽  
Roberta Glass ◽  
Nancy Keating ◽  
...  
Keyword(s):  
End Of Life ◽  
End Of Life Care ◽  
Life Care ◽  
Care Model ◽  
Oncology Care

scholarly journals PCN209 INITIAL EVALUATION OF END-OF-LIFE MEASURES AMONG DECEDENT PATIENTS IN THE ONCOLOGY CARE MODEL

2019 ◽  
Vol 22 ◽  
pp. S95
Author(s):  
A. Kee ◽  
V. Csik ◽  
A. Leader ◽  
J. Minetola ◽  
K. Walsh ◽  
...  
Keyword(s):  
End Of Life ◽  
Initial Evaluation ◽  
Care Model ◽  
Oncology Care

scholarly journals A Clinical Decision Web to Predict ICU Admission or Death for Patients Hospitalised with COVID-19 Using Machine Learning Algorithms

2021 ◽  
Vol 18 (16) ◽  
pp. 8677
Author(s):  
Rocío Aznar-Gimeno ◽  
Luis M. Esteban ◽  
Gorka Labata-Lezaun ◽  
Rafael del-Hoyo-Alonso ◽  
David Abadia-Gallego ◽  
...  
Keyword(s):  
Machine Learning ◽  
Decision Making ◽  
External Validation ◽  
Laboratory Data ◽  
Machine Learning Algorithms ◽  
Gradient Boosting ◽  
Study Cohort ◽  
Icu Admission ◽  
Extreme Gradient Boosting ◽  
User Friendly

The purpose of the study was to build a predictive model for estimating the risk of ICU admission or mortality among patients hospitalized with COVID-19 and provide a user-friendly tool to assist clinicians in the decision-making process. The study cohort comprised 3623 patients with confirmed COVID-19 who were hospitalized in the SALUD hospital network of Aragon (Spain), which includes 23 hospitals, between February 2020 and January 2021, a period that includes several pandemic waves. Up to 165 variables were analysed, including demographics, comorbidity, chronic drugs, vital signs, and laboratory data. To build the predictive models, different techniques and machine learning (ML) algorithms were explored: multilayer perceptron, random forest, and extreme gradient boosting (XGBoost). A reduction dimensionality procedure was used to minimize the features to 20, ensuring feasible use of the tool in practice. Our model was validated both internally and externally. We also assessed its calibration and provide an analysis of the optimal cut-off points depending on the metric to be optimized. The best performing algorithm was XGBoost. The final model achieved good discrimination for the external validation set (AUC = 0.821, 95% CI 0.787–0.854) and accurate calibration (slope = 1, intercept = −0.12). A cut-off of 0.4 provides a sensitivity and specificity of 0.71 and 0.78, respectively. In conclusion, we built a risk prediction model from a large amount of data from several pandemic waves, which had good calibration and discrimination ability. We also created a user-friendly web application that can aid rapid decision-making in clinical practice.

Randomized trial of a collaborative palliative and oncology care intervention to improve communication about end-of-life care in patients with metastatic breast cancer.

2020 ◽  
Vol 38 (15_suppl) ◽  
pp. 1008-1008
Author(s):  
Jennifer S. Temel ◽  
Beverly Moy ◽  
Areej El-Jawahri ◽  
Vicki A. Jackson ◽  
Mihir Kamdar ◽  
...  
Keyword(s):  
Palliative Care ◽  
Usual Care ◽  
End Of Life ◽  
Poor Prognosis ◽  
Metastatic Breast ◽  
Care Model ◽  
Mood Symptoms ◽  
Care Intervention ◽  
Oncology Care ◽  
Eol Care

1008 Background: Studies have demonstrated the benefits of early, integrated palliative care in improving quality of life (QOL) and end-of-life (EOL) care for patients with poor prognosis cancers. However, the optimal timing and outcomes of delivering palliative care for those with advanced cancers that have longer disease trajectories, such as metastatic breast cancer (MBC), remains unknown. We tested the effect of a collaborative palliative and oncology care model on communication about EOL care in patients with MBC. Methods: Patients with MBC and clinical indicators of poor prognosis (N=120) were randomized to receive collaborative palliative and oncology care or usual care between 05/02/2016 and 12/26/2018. The intervention entailed five structured palliative care visits, including a joint visit with oncology when possible, which focused on symptom management, coping, prognostic awareness, decision-making, and planning for EOL. The primary outcome was documentation of EOL care discussions in the electronic health record. Patients also completed questionnaires at baseline and 6, 12, 18, and 24 weeks regarding communication with clinicians about EOL care, QOL, and mood symptoms. Results: The sample included only women (100.0%) who mostly identified as white (87.5%), with a mean age of 56.91 years (SD=11.24). The rate of EOL care discussions documented in the health record was higher among intervention patients versus those receiving usual care (67.2% vs 40.7%, p=0.006), including a higher completion rate of a Medical Orders for Life Sustaining Treatment form (39.3% vs 13.6%, p=0.002). Intervention patients were also more likely to report discussing their EOL care wishes with their doctor compared to usual care patients (OR=3.10, 95% CI: 1.21, 7.94, p=0.019). Study groups did not differ in reported QOL or mood symptoms. Conclusions: This novel collaborative palliative care intervention significantly improved communication and documentation regarding EOL care for women with MBC. Further work is needed to examine the effect of this care model on healthcare utilization at the end of life.

Sign in / Sign up

Export Citation Format

Share Document