Objectives: The aim of this work is to describe the multidisciplinary model of intervention applied and the characteristics of some COVID-19 patients assisted by the hospital palliative care unit (UCP-H) of an Italian hospital in Lombardy, the Italian region most affected by the COVID-19 pandemic.
Methods: A retrospective study was conducted on patients admitted to the A. Manzoni Hospital (Lecco, Lombardy Region, Italy) and referred to the UCP-H between 11 March 2020 and 18 April 2020, the period of maximum spread of COVID-19 in this area. Data were collected on the type of hospitalisation, triage process, modality of palliative care and psychological support provided.
Results: 146 COVID-10 patients were referred to the UCP-H. Of these, 120 died during the observation time (82%) while 15 (10.2%) improved and were discharged from the UCP-H care. 93 had less favourable characteristics (rapid deterioration of respiratory function, old age, multiple comorbidities) and an intensive clinical approach was considered contraindicated, while 48 patients had more favourable presentations. Mean follow-up was 4.8 days. A mean of 4.3 assessments per patient were performed. As to respiratory support, 94 patients were treated with oxygen only (at different volumes) and 45 with Continuous Positive Airway Pressure (CPAP).
Conclusion: The ongoing pandemic highlighted the need for dedicated palliative care teams and units for dying patients. This work highlights how palliative medicine specialist can make a fundamental contribution thanks to their ability and work experience in an organised multiprofessional context.
In end-of-life care, deprescribing practices may vary considerably from one practitioner to the next, although most published frameworks for evaluating medication appropriateness in advanced illness consider three key principles (1) patient and caregiver goals, (2) remaining life expectancy (LE), and (3) medication time to benefit (TTB). The objective of this article is to provide clinicians with a structured, consistent approach for deprescribing that does not replace clinical judgment or the preferences of patients and their families but enhances it through clinical data. The emphasis will be on the time component of published models, including how to estimate remaining LE and medication TTB. Through case examples of two new hospice admissions, LE and TTB will be estimated and applied to deprescribing decisions. This time-centric approach may satisfy the palliative and hospice clinicians' desire for clear clinical justification for medication discontinuation while at the same time providing a strategy for communicating deprescribing rationale to patients and families.
Background: Evaluating the need for palliative care and predicting its mortality play important roles in the emergency department.
Aim: We developed a screening model for predicting 1-year mortality.
Design: A retrospective cohort study was conducted to identify risk factors associated with 1-year mortality. Our risk scores based on these significant risk factors were then developed. Its predictive validity performance was evaluated using area under receiving operating characteristic analysis and leave-one-out cross-validation.
Setting and participants: Patients aged 15 years or older were enrolled from June 2015 to May 2016 in the emergency department.
Results: We identified five independent risk factors, each of which was assigned a number of points proportional to its estimated regression coefficient: age (0.05 points per year), qSOFA >= 2 (1), Cancer (4), Eastern Cooperative Oncology Group Performance Status score >= 2 (2), and Do-Not-Resuscitate status (3). The sensitivity, specificity, positive predictive value, and negative predictive value of our screening tool given the cutoff larger than 3 points were 0.99 (0.98–0.99), 0.31 (0.29–0.32), 0.26 (0.24–0.27), and 0.99 (0.98–1.00), respectively. Those with screening scores larger than 9 points corresponding to 64.0% (60.0–67.9%) of 1-year mortality were prioritized for consultation and communication. The area under the receiving operating characteristic curves for the point system was 0.84 (0.83–0.85) for the cross-validation model.
Conclusions: A-qCPR risk scores provide a good screening tool for assessing patient prognosis. Routine screening for end-of-life using this tool plays an important role in early and efficient physician-patient communications regarding hospice and palliative needs in the emergency department.
Objective: This report describes a pilot hospice inpatient unit dedicated to individuals experiencing distressing behaviors from dementia.
Background: Patients with dementia who experience distressing symptoms cannot be well managed on typical inpatient units. Hospice of the Valley selected one unit to dedicate to dementia care.
Methods: Data were analyzed from 237 patients admitted to the unit between May 2019 and April 2020. Behaviors were identified and rated for severity on admission, discharge, and postdischarge. Rates of inpatient death and associated behaviors were calculated.
Results: Fifty percent of patients had their behaviors sufficiently managed to allow discharge. The most common behavior exhibited was agitation; the most common symptom leading to death was pain.
Discussion: An inpatient hospice unit dedicated to patients with dementia can be successful. The hospice agency gains admissions that would otherwise be diverted to behavioral care settings. This successful pilot may be a model for other hospices.
OBJECTIVES: To evaluate a machine learning model designed to predict mortality for Medicare beneficiaries aged >65 years treated for hip fracture in Inpatient Rehabilitation Facilities (IRFs).
DESIGN: Retrospective design/cohort analysis of Centers for Medicare & Medicaid Services Inpatient Rehabilitation Facility-Patient Assessment Instrument data.
SETTING AND PARTICIPANTS: A total of 17,140 persons admitted to Medicare-certified IRFs in 2015 following hospitalization for hip fracture.
MEASURES: Patient characteristics include sociodemographic (age, gender, race, and social support) and clinical factors (functional status at admission, chronic conditions) and IRF length of stay. Outcomes were 30-day and 1-year all-cause mortality. We trained and evaluated 2 classification models, logistic regression and a multilayer perceptron (MLP), to predict the probability of 30-day and 1-year mortality and evaluated the calibration, discrimination, and precision of the models.
RESULTS: For 30-day mortality, MLP performed well [acc = 0.74, area under the receiver operating characteristic curve (AUROC) = 0.76, avg prec = 0.10, slope = 1.14] as did logistic regression (acc = 0.78, AUROC = 0.76, avg prec = 0.09, slope = 1.20). For 1-year mortality, the performances were similar for both MLP (acc = 0.68, AUROC = 0.75, avg prec = 0.32, slope = 0.96) and logistic regression (acc = 0.68, AUROC = 0.75, avg prec = 0.32, slope = 0.95).
CONCLUSION AND IMPLICATIONS: A scoring system based on logistic regression may be more feasible to run in current electronic medical records. But MLP models may reduce cognitive burden and increase ability to calibrate to local data, yielding clinical specificity in mortality prediction so that palliative care resources may be allocated more effectively.
Objective: To increase earlier access to palliative care, and in turn increase documented goals of care and appropriate hospice referrals for seriously ill patients admitted to hospital medicine.
Background: Due to the growing number of patients with serious illness and the specialty palliative care workforce shortage, innovative primary palliative care models are essential to meet this population's needs.
Methods: Patients with serious illness admitted to hospital medicine at a quaternary urban academic medical center in New York City and received an embedded palliative care social worker consultation in 2017. We used univariate analyses of sociodemographic, clinical, and utilization data to describe the sample.
Results: Overall, 232 patients received a primary palliative care consultation (mean age of 69 years, 44.8% female, 34% white, median Karnofsky Performance Status of 40%), and 159 (69%) had capacity to participate in a goals-of -are conversation. Referrals were from palliative care solid tumor oncology trigger program (113 [49%]), specialty palliative care consultation team (42[18%]), and hospital medicine (34[14.6%]). Before the consultation, 10(4.3%) had documented goals of care and 207 (89%) did after the consultation. The percentage of those referred to hospice was 24.1%. Of those transferred to specialty palliative care consultation service, nearly half required symptom management.
Discussion: Patients who received a primary palliative care consultation were seen earlier in their illness trajectory, based on their higher functional impairment, and the majority had capacity to participate in goals-of-care discussions, compared with those who were seen by specialty palliative care. The consultation increased goals-of-care documentation and the hospice referral rate was comparable with that of the specialty palliative consultation team.
Background: Transitioning care from hospital to home is associated with risks of adverse events and poor continuity of care. These transitions are even more challenging when new approaches to care, such as palliative care, are introduced before discharge. Family caregivers (FCGs) are expected to navigate these transitions while also managing care. In addition to extensive caregiving responsibilities, FCGs often have their own health needs that can inhibit their ability to provide care. Those living in rural areas have even fewer resources to meet their self-care and caregiving needs. The purpose of this study is to test the efficacy and cost-effectiveness of an intervention to improve FCGs’ health and well-being. The intervention uses video visits to teach, guide, and counsel FCGs in rural areas during hospital-to-home transitions. The intervention is based on evidence of transitional and palliative care principles, which are individualized to improve continuity of care, provide caregiver support, enhance knowledge and skills, and attend to caregivers’ health needs. It aims to test whether usual care practices are similar to this technology-enhanced intervention in (1) caregiving skills (e.g., caregiving preparedness, communication with clinicians, and satisfaction with care), (2) FCG health outcomes (e.g., quality of life, burden, coping skills, depression), and (3) cost. We describe the rationale for targeting rural caregivers, the methods for the study and intervention, and the analysis plan to test the intervention’s effect.
Methods: The study uses a randomized controlled trial design, with FCGs assigned to the control condition or the caregiver intervention by computer-generated lists. The intervention period continues for 8 weeks after care recipients are discharged from the hospital. Data are collected at baseline, 2 weeks, 8 weeks, and 6 months. Time and monetary costs from a societal perspective are captured monthly.
Discussion: This study addresses 2 independent yet interrelated health care foci—transitional care and palliative care—by testing an intervention to extend palliative care practice and improve transition management for caregivers of seriously ill patients in rural areas. The comprehensive cost assessment will quantify the commitment and financial burden of FCGs.
Trial registration: ClinicalTrials.gov NCT03339271. Registered on 13 November 2017.
Background: This study was to investigate the prognostic factors of patients with advanced gastric cancer and described a sample model to better differentiate the patients who could better benefit from palliative chemotherapy.
Patients and methods: In this retrospective study, 112 gastric cancer patients at stage IV following first-line chemotherapy were enrolled from July 2013 to September 2019. The clinical factors including age, sex, ECOG, pathologic types, metastatic sites, blood indexes, response of first-line chemotherapy, and survival were collected. The treatment responses were evaluated using the response evaluation criteria in solid tumors (RECIST). The survival curves were drawn by the Kaplan-Meier method, and the independent prognostic factors of overall survival (OS) were analyzed by Cox proportional hazards regression model.
Results: In this study, the median overall survival (mOS) of gastric cancer patients was 10.5 months, the disease remission rate (PR) was 21.4%, and the disease control rate (DCR) was 86.6%. Multivariate analysis identified 5 independent prognostic factors: peritoneal metastasis [P = 0.002; hazard risk (HR), 2.394; 95% CI 1.394-4.113], hemoglobin <90g/L [P = 0.001; hazard risk (HR), 2.674; 95% CI 1.536-4.655], LDH =225 U/L [P = 0.033; hazard risk (HR), 1.818; 95% CI 1.409-3.150], and 3 times higher level of CEA [P = 0.006; hazard risk (HR), 2.123; 95% CI 1.238-3.640] along with CA199 [P = 0.005; hazard risk (HR), 2.544; 95% CI 1.332-4.856] than upper limit of normal. Based on the obtained data, a prognostic index was constructed, dividing the patients into three risk groups: low (n = 67), intermediate (n = 35), and high-risk group (n = 10). The mOS for low, intermediate, and high-risk groups was 13.9 months (95% CI 10.7-17.1), 8.1 months (95% CI 5.7-10.4), and 3.9 months (95% CI 2.6-5.3), respectively, whereas the 1-year survival rate was 56.4%, 20.0%, and 0.0%, respectively (P < 0.001).
Conclusion: This model should facilitate the prediction of treatment outcomes and then individualized treatment of advanced gastric cancer patients.
BACKGROUND: In the palliative care setting, infection control measures implemented due to COVID-19 have become barriers to end-of-life care discussions (eg, discharge planning and withdrawal of life-sustaining treatments) between patients, their families, and multidisciplinary medical teams. Strict restrictions in terms of visiting hours and the number of visitors have made it difficult to arrange in-person family conferences. Phone-based telehealth consultations may be a solution, but the lack of nonverbal cues may diminish the clinician-patient relationship. In this context, video-based, smartphone-enabled family conferences have become important.
OBJECTIVE: We aimed to establish a smartphone-enabled telehealth model for palliative care family conferences. Our model integrates principles from the concept of shared decision making (SDM) and the value, acknowledge, listen, understand, and elicit (VALUE) approach.
METHODS: Family conferences comprised three phases designed according to telehealth implementation guidelines-the previsit, during-visit, and postvisit phases. We incorporated the following SDM elements into the model: "team talk," "option talk," and "decision talk." The model has been implemented at a national cancer treatment center in Taiwan since February 2020.
RESULTS: From February to April 2020, 14 telehealth family conferences in the palliative care unit were analyzed. The patients' mean age was 73 (SD 10.1) years; 6 out of 14 patients (43%) were female and 12 (86%) were married. The primary caregiver joining the conference virtually comprised mostly of spouses and children (n=10, 71%). The majority of participants were terminally ill patients with cancer (n=13, 93%), with the exception of 1 patient with stroke. Consensus on care goals related to discharge planning and withdrawal of life-sustaining treatments was reached in 93% (n=13) of cases during the family conferences. In total, 5 families rated the family conferences as good or very good (36%), whereas 9 were neutral (64%).
CONCLUSIONS: Smartphone-enabled telehealth for palliative care family conferences with SDM and VALUE integration demonstrated high satisfaction for families. In most cases, it was effective in reaching consensus on care decisions. The model may be applied to other countries to promote quality in end-of-life care in the midst of the COVID-19 pandemic.
BACKGROUND: End-of-life caregiving frequently is managed by friends and family. Studies on hastened death, including aid in dying or assisted suicide, indicate friends and family also play essential roles before, during, and after death. No studies have compared the experiences of caregivers in hastened and non-hastened death. The study aim is to compare end-of-life and hastened death caregiving experience using Hudson's modified stress-coping model for palliative caregiving.
METHOD: Narrative synthesis of qualitative studies for caregivers at end of life and in hastened death, with 9946 end-of life and 1414 hastened death qualitative, peer-reviewed research articles extracted from MEDLINE, CINAHL, Web of Science, and PsycINFO, published between January 1998 and April 2020.
RESULTS: Forty-two end-of-life caregiving and 12 hastened death caregiving articles met inclusion criteria. In both end-of-life and hastened death contexts, caregivers are motivated to ease patient suffering and may put their own needs or feelings aside to focus on that priority. Hastened death caregivers' expectation of impending death and the short duration of caregiving may result in less caregiver burden. Acceptance of the patient's condition, social support, and support from healthcare professionals all appear to improve caregiver experience. However, data on hastened death are limited.
CONCLUSION: Caregivers in both groups sought closeness with the patient and reported satisfaction at having done their best to care for the patient in a critical time. Awareness of anticipated death and support from healthcare professionals appear to reduce caregiver stress. The modified stress-coping framework is an effective lens for interpreting caregivers' experiences at end of life and in the context of hastened death.
PURPOSE: The Oncology Care Model (OCM) was developed to improve care while also supporting patient-centered practices. This model could significantly affect experiences of patients with cancer; however, previous studies have not explored patient perspectives.
PATIENTS AND METHODS: This cross-sectional study used focus group and survey methodology to explore patient experiences in the OCM. The sample included 213 patients (OCM patients, n = 130 recruited within OCM practices; non-OCM patients, n = 83 recruited via e-mail from the Cancer Support Community Cancer Experience Registry).
RESULTS: Findings suggest that patients in OCM practices were more likely to report that their cancer care team asked about social/emotional distress or concerns and more likely to have social/emotional resources offered. OCM patients were also more likely to have discussed advance directives with providers. They were also more likely to be satisfied with provider explanations of treatment benefits as well as treatment risks and adverse effects. Lastly, OCM patients were significantly more satisfied with discussion of treatment costs and provided higher ratings of preparation by their cancer care team for management of adverse effects.
CONCLUSION: Patients in this study reported experiences consistent with many of the key goals of the OCM. This is promising and may indicate the need to expand the model. However, because of the potential selection bias of our sampling method, more research is needed.
BACKGROUND: Most terminally ill cancer patients prefer to die at home, but a majority die in institutional settings. Research questions about this discrepancy have not been fully answered. This study applies artificial intelligence and machine learning techniques to explore the complex network of factors and the cause-effect relationships affecting the place of death, with the ultimate aim of developing policies favouring home-based end-of-life care.
METHODS: A data mining algorithm and a causal probabilistic model for data analysis were developed with information derived from expert knowledge that was merged with data from 116 deceased cancer patients in southern Switzerland. This data set was obtained via a retrospective clinical chart review.
RESULTS: Dependencies of disease and treatment-related decisions demonstrate an influence on the place of death of 13%. Anticancer treatment in advanced disease prevents or delays communication about the end of life between oncologists, patients and families. Unknown preferences for the place of death represent a great barrier to a home death. A further barrier is the limited availability of family caregivers for terminal home care. The family's preference for the last place of care has a high impact on the place of death of 51%, while the influence of the patient's preference is low, at 14%. Approximately one-third of family systems can be empowered by health care professionals to provide home care through open end-of-life communication and good symptom management. Such intervention has an influence on the place of death of 17%. If families express a convincing preference for home care, the involvement of a specialist palliative home care service can increase the probability of home deaths by 24%.
CONCLUSION: Concerning death at home, open communication about death and dying is essential. Furthermore, for the patient preference for home care to be respected, the family's decision for the last place of care seems to be key. The early initiation of family-centred palliative care and the provision of specialist palliative home care for patients who wish to die at home are suggested.
End-of-life (EOL) decision making in the intensive care unit (ICU) is challenging for both families and clinicians. This decision-making process is ideally framed around a shared understanding of a patient’s values and goals, all taken in the context of their critical illness and prognosis. However, clinicians commonly face uncertainty regarding prognosis and may have difficulty offering families an accurate assessment of the likely outcomes of treatment decisions. Adding to the complexity of these scenarios, clinicians, patients and families are each susceptible to unconscious but influential cognitive biases when making decisions under stress. Given these challenges, and a rapidly growing interest in data science to inform care in the ICU, investigators have explored the use of prediction models (eg, machine learning or ML algorithms) to assist with prognostication. Prediction models describe an outcome distribution among individuals with a particular set of characteristics, such as risk of acute kidney injury among individuals with particular laboratory values and clinical characteristics in a population. However, they do not compare how that outcome distribution would change were different treatment decisions made in that population—this requires causal effect estimation, rather than prediction modelling. Herein, we explain why prediction modelling alone is not sufficient to inform many ICU treatment decisions, including EOL decision making, and describe why causal effect estimation is necessary.
BACKGROUND: Health systems need evidence about how best to deliver home-based palliative care (HBPC) to meet the growing needs of seriously ill patients. We hypothesised that a tech-supported model that aimed to promote timely inter-professional team coordination using video consultation with a remote physician while a nurse is in the patient's home would be non-inferior compared with a standard model that includes routine home visits by nurses and physicians.
METHODS: We conducted a pragmatic, cluster randomised non-inferiority trial across 14 sites (HomePal Study). Registered nurses (n=111) were randomised to the two models so that approximately half of the patients with any serious illness admitted to HBPC and their caregivers were enrolled in each study arm. Process measures (video and home visits and satisfaction) were tracked. The primary outcomes for patients and caregivers were symptom burden and caregiving preparedness at 1-2 months.
RESULTS: The study was stopped early after 12 months of enrolment (patients=3533; caregivers=463) due to a combination of low video visit uptake (31%), limited substitution of video for home visits, and the health system's decision to expand telehealth use in response to changes in telehealth payment policies, the latter of which was incompatible with the randomised design. Implementation barriers included persistent workforce shortages and inadequate systems that contributed to scheduling and coordination challenges and unreliable technology and connectivity.
CONCLUSIONS: We encountered multiple challenges to feasibility, relevance and value of conducting large, multiyear pragmatic randomised trials with seriously ill patients in the real-world settings where care delivery, regulatory and payment policies are constantly shifting.
Background: For residential aged care facility (RACF) residents with dementia, lack of prognostic guidance presents a significant challenge for end of life care planning. In an attempt to address this issue, models have been developed to assess mortality risk for people with advanced dementia, predominantly using long-term care minimum data set (MDS) information from the USA. A limitation of these models is that the information contained within the MDS used for model development was not collected for the purpose of identifying prognostic factors. The models developed using MDS data have had relatively modest ability to discriminate mortality risk and are difficult to apply outside the MDS setting. This study will aim to develop a model to estimate 6- and 12-month mortality risk for people with dementia from prognostic indicators recorded during usual clinical care provided in RACFs in Australia.
Methods: A secondary analysis will be conducted for a cohort of people with dementia from RACFs participating in a cluster-randomized trial of a palliative care education intervention (IMPETUS-D). Ten prognostic indicator variables were identified based on a literature review of clinical features associated with increased mortality for people with dementia living in RACFs. Variables will be extracted from RACF files at baseline and mortality measured at 6 and 12 months after baseline data collection. A multivariable logistic regression model will be developed for 6- and 12-month mortality outcome measures using backwards elimination with a fractional polynomial approach for continuous variables. Internal validation will be undertaken using bootstrapping methods. Discrimination of the model for 6- and 12-month mortality will be presented as receiver operating curves with c statistics. Calibration curves will be presented comparing observed and predicted event rates for each decile of risk as well as flexible calibration curves derived using loess-based functions.
Discussion: The model developed in this study aims to improve clinical assessment of mortality risk for people with dementia living in RACFs in Australia. Further external validation in different populations will be required before the model could be developed into a tool to assist with clinical decision-making in the future.
CONTEXT: Previous studies suggest that clinicians' prediction of survival (CPS) may have reduce 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 far advanced cancer patients.
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 1,534 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.
Palliative care is a values-driven approach for providing holistic care for individuals and their families enduring serious life-limiting illness. Despite its proven benefits, access and acceptance is not uniform across society. The genesis of palliative care was developed through a traditional Western lens, which dictated models of interaction and communication. As the importance of palliative care is increasingly recognized, barriers to accessing services and perceptions of relevance and appropriateness are being given greater consideration. The COVID-19 pandemic and recent social justice movements in the United States, and around the world, have led to an important moment in time for the palliative care community to step back and consider opportunities for expansion and growth. This article reviews traditional models of palliative care delivery and outlines a modified conceptual framework to support researchers, clinicians, and staff in evaluating priorities for ensuring individualized patient needs are addressed from a position of equity, to create an actionable path forward.
OBJECTIVE: The aim of this project was to assess the value for money of a modified unit within a residential aged care facility (RACF) for people requiring palliative care at the end of life.
METHODS: A three-way comparison using a mixed-method costing was used to estimate the per day cost of the unit compared to care in a palliative care unit within a hospital and a standard RACF bed.
RESULTS: The cost of the unit was estimated at $242 per day (2015 Australian dollars). The palliative care hospital bed cost $1,664 per day. The cost of a standard RACF bed was $123 per day, indicating that an additional $120 per day is required to provide the higher level of care required by people with complex palliative care needs.
CONCLUSION: A modified RACF unit could provide substantial cost savings to the health budget for selected complex palliative care patients.
OBJECTIVES: To develop and validate a model for prediction of near-term in-hospital mortality among patients with COVID-19 by application of a machine learning (ML) algorithm on time-series inpatient data from electronic health records.
METHODS: A cohort comprised of 567 patients with COVID-19 at a large acute care healthcare system between 10 February 2020 and 7 April 2020 observed until either death or discharge. Random forest (RF) model was developed on randomly drawn 70% of the cohort (training set) and its performance was evaluated on the rest of 30% (the test set). The outcome variable was in-hospital mortality within 20-84 hours from the time of prediction. Input features included patients' vital signs, laboratory data and ECG results.
RESULTS: Patients had a median age of 60.2 years (IQR 26.2 years); 54.1% were men. In-hospital mortality rate was 17.0% and overall median time to death was 6.5 days (range 1.3-23.0 days). In the test set, the RF classifier yielded a sensitivity of 87.8% (95% CI: 78.2% to 94.3%), specificity of 60.6% (95% CI: 55.2% to 65.8%), accuracy of 65.5% (95% CI: 60.7% to 70.0%), area under the receiver operating characteristic curve of 85.5% (95% CI: 80.8% to 90.2%) and area under the precision recall curve of 64.4% (95% CI: 53.5% to 75.3%).
CONCLUSIONS: Our ML-based approach can be used to analyse electronic health record data and reliably predict near-term mortality prediction. Using such a model in hospitals could help improve care, thereby better aligning clinical decisions with prognosis in critically ill patients with COVID-19.
Context: Palliative care programs are typically evaluated using observational data, raising concerns about selection bias.
Objective: To quantify selection bias due to observed and unobserved characteristics in a palliative care demonstration program.
Methods: Program administrative data and 100% Medicare claims data in two-states and a 20% sample in eight-states (2013-2017). The sample included 2983 Medicare fee-for-service beneficiaries age 65+ participating in the palliative care program and three matched cohorts: 1) regional 2) two-states and 3) eight-states. Confounding due to observed factors was measured by comparing patient baseline characteristics. Confounding due to unobserved factors was measured by comparing days of follow-up, and six-month and one-year mortality rates.
Results: After matching, evidence for observed confounding included differences in observable baseline characteristics including race, morbidity and utilization. Evidence for unobserved confounding included significantly longer mean follow-up in the regional, two-state, and eight-state comparison cohorts, with 207 (p<.001), 192 (p<.001), and 187 (p<.001) days, respectively, compared to the 162 days for the palliative care cohort. The palliative care cohort had higher 6 month and 1-year mortality rates of 53.5% and 64.5% compared to 43.5% and 48.0% in the regional comparison, 53.4% and 57.4% in the two-state comparison, and 55.0% and 59.0% in the eight-state comparison.
Conclusions: This case study demonstrates that selection of comparison groups impacts the magnitude of measured and unmeasured confounding, which may change effect estimates. The substantial impact of confounding on effect estimates in this study raises concerns about the evaluation of novel serious illness care models in the absence of randomization. We present key lessons learned for improving future evaluations of palliative care using observational study designs.