Dataset Information

Improving palliative and end-of-life care with machine learning and routine data: a rapid review.

ABSTRACT: Introduction: Improving end-of-life (EOL) care is a priority worldwide as this population experiences poor outcomes and accounts disproportionately for costs. In clinical practice, physician judgement is the core method of identifying EOL care needs but has important limitations. Machine learning (ML) is a subset of artificial intelligence advancing capacity to identify patterns and make predictions using large datasets. ML approaches have the potential to improve clinical decision-making and policy design, but there has been no systematic assembly of current evidence. Methods: We conducted a rapid review, searching systematically seven databases from inception to December 31st, 2018: EMBASE, MEDLINE, Cochrane Library, PsycINFO, WOS, SCOPUS and ECONLIT. We included peer-reviewed studies that used ML approaches on routine data to improve palliative and EOL care for adults. Our specified outcomes were survival, quality of life (QoL), place of death, costs, and receipt of high-intensity treatment near end of life. We did not search grey literature and excluded material that was not a peer-reviewed article. Results: The database search identified 426 citations. We discarded 162 duplicates and screened 264 unique title/abstracts, of which 22 were forwarded for full text review. Three papers were included, 18 papers were excluded and one full text was sought but unobtainable. One paper predicted six-month mortality, one paper predicted 12-month mortality and one paper cross-referenced predicted 12-month mortality with healthcare spending. ML-informed models outperformed logistic regression in predicting mortality but poor prognosis is a weak driver of costs. Models using only routine administrative data had limited benefit from ML methods. Conclusion: While ML can in principle help to identify those at risk of adverse outcomes and inappropriate treatment near EOL, applications to policy and practice are formative. Future research must not only expand scope to other outcomes and longer timeframes, but also engage with individual preferences and ethical challenges.

SUBMITTER: Storick V

PROVIDER: S-EPMC6973530 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Improving palliative and end-of-life care with machine learning and routine data: a rapid review.

Storick Virginia V O'Herlihy Aoife A Abdelhafeez Sarah S Ahmed Rakesh R May Peter P

HRB open research 20190715

<b>Introduction:</b> Improving end-of-life (EOL) care is a priority worldwide as this population experiences poor outcomes and accounts disproportionately for costs. In clinical practice, physician judgement is the core method of identifying EOL care needs but has important limitations. Machine learning (ML) is a subset of artificial intelligence advancing capacity to identify patterns and make predictions using large datasets. ML approaches have the potential to improve clinical decision-makin ...[more]

PMID: 32002512

Similar Datasets

Project description:The increasing demand for palliative care in New Zealand presents a potential threat to the quality of service delivery. One strategy to overcome this is through the implementation of valid and reliable patient-reported outcome measures. This mixed-methods study aimed to (1) implement measurement-based palliative care (MBPC) in a community palliative care service in Auckland, New Zealand; (2) evaluate the clinical utility of MBPC perceived by clinicians; (3) describe patient characteristics as measured by the Integrated Palliative Care Outcome Scale (IPOS), the Australasian Modified Karnofsky Performance Scale (AKPS), and Phase of Illness (POI); and (4) evaluate the internal consistency of the IPOS. Participants were over 18 years of age from a community outpatient palliative care service. In a phased approach to implementation, healthcare staff were educated on each instrument used for patient assessment. Uptake and internal consistency were evaluated through descriptive statistics. An interpretive descriptive methodology was used to explore the clinical utility of MBPC through semi-structured interviews with seven clinical staff members. Individual patient assessments (n = 1507) were undertaken predominantly on admission, with decreasing frequency as patients advanced through to the terminal phase of their care. Mean total IPOS scores were 17.97 (SD = 10.39, α = 0.78). The POI showed that 65% of patients were in the stable phase, 20% were in the unstable phase, 9% were in the deteriorating phase, and 2% were in the terminal phase. Clinicians reported that MBPC facilitated holistic and comprehensive assessments, as well as the development of a common interdisciplinary language. Clinicians expressed discomfort using the psychosocial and spiritual items. Measurement-based palliative care was only partially implemented but it was valued by staff and perceived to increase the quality of service delivery. Future research should determine the optimal timing of assessments, cultural responsivity for Māori and Pacific patients, and the role of MBPC in decision support for clinicians.

Project description:Background/aimsThe use and costs of health care rise substantially in the months prior to death, and although the use of palliative care services may be expected to lead to less costly care, the evidence is mixed. We analysed the costs of care over the last year of life and the extent to which these are associated with the use and duration of specialist palliative care (SPC) for decedents who died from cancer or another life-limiting illness.MethodsThe decedents were participants in a cohort study of older residents of the state of New South Wales, Australia. Using linked survey and administrative health data from 2007 to 2016, two cohorts were identified: n = 10,535 where the cause of death was cancer; and n = 11,179 where the cause of death was another life-limiting illness. Costs of various types were analysed with separate risk-adjusted linear regression models for the last 1, 3, 6, 9 and 12 months before death and for both cohorts. SPC was categorised according to time to death from first contact with the service as 1-7 days, 7-30 days, 30-180 days and more than 180 days.ResultsSPC use was higher among the cancer cohort (30.0%) relative to the non-cancer cohort (4.8%). The mean costs over the final year of life were AU$55,037 (SD 45,059) for the cancer cohort and AU$35,318 (SD 41,948) for the non-cancer cohort. Earlier use of SPC was associated with higher costs over the last year of life but lower costs in the last 1 and 3 months for both cohorts. Initiating SPC use more than 180 days before death was associated with a mean difference relative to the no SPC group of AU$15,590 (95% CI 10,617 to 20,562) and AU$13,739 (95% CI 733 to 26,746) over the last year of life for those dying from cancer and another illness, respectively. The same differences over the last month of life were - AU$2810 (95% CI - 3945 to - 1676) and - AU$4345 (95% CI - 6625 to - 2066). Admitted hospital care was the major driver of costs, with longer SPC associated with lower rates of death in hospital for both cohorts.ConclusionEarly initiation of SPC was associated with higher costs over the last year of life and lower costs over the last months of life. This was the case for both the cancer and non-cancer cohorts, and appeared to be largely attributed to reduced hospitalisation. Although further investigation is required, our results suggest that expanding the availability of SPC services to provide more equitable access could enable patients to spend more time at their usual place of residence, reduce pressure on inpatient services and facilitate death at home when that is preferred.

Project description:BackgroundThe goal of palliative care is to prevent and alleviate a suffering of incurable ill patients. A continuous intersectoral palliative care is important. The aim of this study is to analyse the continuity of palliative care, particularly the time gaps between hospital discharge and subsequent palliative care as well as the timing of the last palliative care before the patient's death.MethodsThe analysis was based on claims data from a large statutory health insurance. Patients who received their first palliative care in 2015 were included. The course of palliative care was followed for 12 months. Time intervals between discharge from hospital and first subsequent palliative care as well as between last palliative care and death were analysed. The continuity in palliative care was defined as an interval of less than 14 days between palliative care. Data were analysed using descriptive statistics and Chi-Square.ResultsIn 2015, 4177 patients with first palliative care were identified in the catchment area of the statutory health insurance. After general inpatient palliative care, 415 patients were transferred to subsequent palliative care, of these 67.7% (n = 281) received subsequent care within 14 days. After a stay in a palliative care ward, 124 patients received subsequent palliative care, of these 75.0% (n = 93) within 14 days. Altogether, 147 discharges did not receive subsequent palliative care. During the 12-months follow-up period, 2866 (68.7%) patients died, of these 78.7% (n = 2256) received palliative care within the last 2 weeks of life. Of these, 1223 patients received general ambulatory palliative care, 631 patients received specialised ambulatory palliative care, 313 patients received their last palliative care at a hospital and 89 patients received it in a hospice.ConclusionsThe majority of the palliative care patients received continuous palliative care. However, there are some patients who did not receive continuous palliative care. After inpatient palliative care, each patient should receive a discharge management for a continuation of palliative care. Readmissions of patients after discharge from inpatients palliative care can be an indication for a lack of support in the ambulatory health care setting and for an insufficient discharge management. Palliative care training and possibilities for palliative care consultations by specialists should strengthen the GPs in palliative care.

Project description:ImportancePalliative care has the potential to improve care for patients and families undergoing high-risk surgery.ObjectiveTo characterize the use of perioperative palliative care and its association with family-reported end-of-life experiences of patients who died within 90 days of a high-risk surgical operation.Design, setting, and participantsThis secondary analysis of administrative data from a retrospective cross-sectional patient cohort was conducted in the Department of Veterans Affairs (VA) Healthcare System. Patients who underwent any of 227 high-risk operations between January 1, 2012, and December 31, 2015, were included.ExposuresPalliative-care consultation within 30 days before or 90 days after surgery.Main outcomes and measuresThe outcomes were family-reported ratings of overall care, communication, and support in the patient's last month of life. The VA surveyed all families of inpatient decedents using the Bereaved Family Survey, a valid and reliable tool that measures patient and family-centered end-of-life outcomes.ResultsA total of 95 204 patients underwent high-risk operations in 129 inpatient VA Medical Centers. Most patients were 65 years or older (69 278 [72.8%]), and the most common procedures were cardiothoracic (31 157 [32.7%]) or vascular (23 517 [24.7%]). The 90-day mortality rate was 6.0% (5740 patients) and varied by surgical subspecialty (ranging from 278 of 7226 [3.8%] in urologic surgery to 875 of 6223 patients [14.1%] in neurosurgery). A multivariate mixed model revealed that families of decedents who received palliative care were 47% more likely to rate overall care in the last month of life as excellent than those who did not (odds ratio [OR], 1.47 [95% CI, 1.14-1.88]; P = .007), after adjusting for patient's characteristics, surgical subspecialty of the high-risk operation, and survey nonresponse. Similarly, families of decedents who received palliative care were more likely to rate end-of-life communication (OR, 1.43 [95% CI, 1.09-1.87]; P = .004) and support (OR, 1.31 [95% CI, 1.01-1.71]; P = .05) components of medical care as excellent. Of the entire cohort, 3374 patients (3.75%) had a palliative care consultation, and 770 patients (0.8%) received it before surgery. Of all decedents, 1632 (29.9%) had a palliative care consultation, with 319 (5.6%) receiving it before surgery.Conclusions and relevanceReceipt of a palliative consultation was associated with better ratings of overall end-of-life care, communication, and support, as reported by families of patients who died within 90 days of high-risk surgery. Yet only one-third of decedents was exposed to palliative care. Expanding integration of perioperative palliative care may benefit patients undergoing high-risk operations and their families.

Dataset Information

Improving palliative and end-of-life care with machine learning and routine data: a rapid review.

Publications

Improving palliative and end-of-life care with machine learning and routine data: a rapid review.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets