Variable importance analysis with interpretable machine learning for fair risk prediction.

Machine learning (ML) methods are increasingly used to assess variable importance, but such black box models lack stability when limited in sample sizes, and do not formally indicate non-important factors. The Shapley variable importance cloud (ShapleyVIC) addresses these limitations by assessing variable importance from an ensemble of regression models, which enhances robustness while maintaining interpretability, and estimates uncertainty of overall importance to formally test its significance. In a clinical study, ShapleyVIC reasonably identified important variables when the random forest and XGBoost failed to, and generally reproduced the findings from smaller subsamples (n = 2500 and 500) when statistical power of the logistic regression became attenuated. Moreover, ShapleyVIC reasonably estimated non-significant importance of race to justify its exclusion from the final prediction model, as opposed to the race-dependent model from the conventional stepwise model building. Hence, ShapleyVIC is robust and interpretable for variable importance assessment, with potential contribution to fairer clinical risk prediction.

Increasing neurologically intact survival after out-of-hospital cardiac arrest among elderly: Singapore Experience.

Objectives: With more elderly presenting with Out-of-Hospital Cardiac Arrests (OHCAs) globally, neurologically intact survival (NIS) should be the aim of resuscitation. We aimed to study the trend of OHCA amongst elderly in a large Asian registry to identify if age is independently associated with NIS and factors associated with NIS. Methods: All adult OHCAs aged ≥18 years attended by emergency medical services (EMS) from April 2010 to December 2019 in Singapore was extracted from the Pan-Asian Resuscitation Outcomes Study (PAROS) registry. Cases pronounced dead at scene, non-EMS transported, traumatic OHCAs and OHCAs in ambulances were excluded. Patient characteristics and outcomes were compared across four age categories (18-64, 65-79, 80-89, ≥90). Multivariable logistic regression analysis determined the factors associated with NIS. Results: 19,519 eligible cases were analyzed. OHCA incidence increased with age almost doubling in octogenarians (from 312/100,000 in 2011 to 652/100,000 in 2019) and tripling in those ≥90 years (from 458/100,000 in 2011 to 1271/100,000 in 2019). The proportion of patients with NIS improved over time for the 18-64, 65-79- and 80-89-years age groups, with the greatest improvement in the youngest group. NIS decreased with each increasing year of age and minute of response time. NIS increased in the arrests of presumed cardiac etiology, witnessed and bystander CPR. Conclusions: Survival with good outcomes has increased even amongst the elderly. Regardless of age, NIS is possible with good-quality CPR, highlighting its importance. End-of-life planning is a complex yet necessary decision that requires qualitative exploration with elderly, their families and care providers.

Impact of time-to-compression on out-of-hospital cardiac arrest survival outcomes: A national registry study.

OBJECTIVE: We aimed to quantify the association of no-flow interval in out-of-hospital cardiac arrests (OHCA) with the odds of neurologically favorable survival and survival to hospital discharge/ 30th day. Our secondary aim was to explore futility thresholds to guide clinical decisions, such as prehospital termination of resuscitation. METHODS: All OHCAs from 2012 to 2017 in Singapore were extracted. We examined the association between no-flow interval (continuous variable) and survival outcomes using univariate and multivariable logistic regressions. The primary outcome was survival with favorable cerebral performance (Glasgow-Pittsburgh Cerebral Performance Categories 1/2), the secondary outcome was survival to hospital discharge/ 30th day if not discharged. To determine futility thresholds, we plotted the adjusted probability of good neurological outcomes to no-flow interval. RESULTS: 12,771 OHCAs were analyzed. The per-minute adjusted OR when no-flow interval was incorporated as a continuous variable in the multivariable model was: good neurological function- aOR 0.98 (95%CI: 0.97-0.98); survival to discharge- aOR 0.98 (95%CI: 0.98-0.99). Taking the 1% futility of survival line gave a no-flow interval cutoff of 12 mins (NPV 99%, sensitivity 85% and specificity 42%) overall and 7.5 mins for witnessed arrests. CONCLUSION: We demonstrated that prolonged no-flow interval had a significant effect on lower odds of favorable neurological outcomes, with medical futility occurring when no-flow interval was >12 mins (>7.5 mins for witnessed arrest). Our study adds to the literature of the importance of early CPR and EMS response and provided a threshold beyond traditional 'down-times', which could aid clinical decisions in TOR or OHCA management.

Does witness type affect the chance of receiving bystander CPR in out-of-hospital cardiac arrest?

OBJECTIVES: The relationship between the bystander witness type and receipt of bystander CPR (BCPR) is not well understood. Herein we compared BCPR administration between family and non-family witnessed out-of-hospital cardiac arrest (OHCA). BACKGROUND: In many communities, interventions in the past decade have contributed to an increased receipt of BCPR, for example in Singapore from 15% to 60%. However, BCPR rates have plateaued despite sustained and ongoing community-based interventions, which may be related to gaps in education or training for various witness types. The purpose of this study was to investigate the association between witness type and BCPR administration. METHODS: Singapore data from 2010-2020 was extracted from the Pan-Asian Resuscitation Outcomes Study (PAROS) network registry (n = 25,024). All adult, layperson witnessed, non-traumatic OHCAs were included in this study. RESULTS: Of 10,016 eligible OHCA cases, 6,895 were family witnessed and 3,121 were non-family witnessed. After adjustment for potential confounders, BCPR administration was less likely for non-family witnessed OHCA (OR 0.83, 95% CI 0.75, 0.93). After location stratification, non-family witnessed OHCAs were less likely to receive BCPR in residential settings (OR 0.75, 95% CI 0.66, 0.85). In non-residential settings, there was no statistically significant association between witness type and BCPR administration (OR 1.11, 95% CI 0.88, 1.39). Details regarding witness type and bystander CPR were limited. CONCLUSION: This study found differences in BCPR administration between family and non-family witnessed OHCA cases. Elucidation of witness characteristics may be useful to determine populations that would benefit most from CPR education and training.

A multi-institutional exploration of emergency medicine physicians' attitudes and behaviours on antibiotic use during the COVID-19 pandemic: a mixed-methods study.

BACKGROUND: The COVID-19 pandemic has changed the epidemiology of upper respiratory tract infections (URTI) and the disease profile of patients attending the emergency department (ED). Hence, we sought to explore the changes in ED physicians' attitudes and behaviours in four EDs in Singapore. METHODS: We employed a sequential mixed-methods approach (quantitative survey followed by in-depth interviews). Principal component analysis was performed to derive latent factors, followed by multivariable logistic regression to explore the independent factors associated with high antibiotic prescribing. Interviews were analysed using the deductive-inductive-deductive framework. We derive five meta-inferences by integrating the quantitative and qualitative findings with an explanatory bidirectional framework. RESULTS: We obtained 560 (65.9%) valid responses from the survey and interviewed 50 physicians from various work experiences. ED physicians were twice as likely to report high antibiotic prescribing rates pre-COVID-19 pandemic than during the pandemic (AOR = 2.12, 95% CI 1.32 to 3.41, p = 0.002). Five meta-inferences were made by integrating the data: (1) Less pressure to prescribe antibiotics due to reduced patient demand and more patient education opportunities; (2) A higher proportion of ED physicians self-reported lower antibiotic prescribing rates during the COVID-19 pandemic but their perception of the overall outlook on antibiotic prescribing rates varied; (3) Physicians who were high antibiotic prescribers during the COVID-19 pandemic made less effort for prudent antibiotic prescribing as they were less concerned about antimicrobial resistance; (4) the COVID-19 pandemic did not change the factors that lowered the threshold for antibiotic prescribing; (5) the COVID-19 pandemic did not change the perception that the public's knowledge of antibiotics is poor. CONCLUSIONS: Self-reported antibiotic prescribing rates decreased in the ED during the COVID-19 pandemic due to less pressure to prescribe antibiotics. The lessons and experiences learnt from the COVID-19 pandemic can be incorporated into public and medical education in the war against antimicrobial resistance going forward. Antibiotic use should also be monitored post-pandemic to assess if the changes are sustained.

Antibiotic expectation, behaviour, and receipt among patients presenting to emergency departments with uncomplicated upper respiratory tract infection during the COVID-19 pandemic.

OBJECTIVES: Pre-COVID-19 pandemic, patients who attended the emergency department (ED) for upper respiratory tract infection (URTI) were more likely to receive antibiotics if they expected them. These expectations could have changed with the change in health-seeking behaviour during the pandemic. We assessed the factors associated with antibiotics expectation and receipt for uncomplicated URTI patients in four Singapore EDs during the COVID-19 pandemic. METHODS: We conducted a cross-sectional study on adult patients with URTI from March 2021 to March 2022 in four Singapore EDs and assessed the determinants of antibiotics expectation and receipt using multivariable logistic regression models. We also assessed the reasons patients expect antibiotics during their ED visit. RESULTS: Among 681 patients, 31.0% expected antibiotics while 8.7% received antibiotics during their ED visit. Factors (adjusted odds ratio [95% confidence interval]) that significantly influenced expectation for antibiotics include: 1) prior consultation for current illness with (6.56 [3.30-13.11]) or without (1.50 [1.01-2.23]) antibiotics prescribed; 2) anticipation for COVID-19 test (1.56 [1.01-2.41]); and 3) poor (2.16 [1.26-3.68]) to moderate (2.26 [1.33-3.84]) knowledge on antibiotics use and resistance. Patients expecting antibiotics were 10.6 times (10.64 [5.34-21.17]) more likely to receive antibiotics. Those with tertiary education were twice (2.20 [1.09-4.43]) as likely to receive antibiotics. CONCLUSION: In conclusion, patients with URTI who expected antibiotics to be prescribed remained more likely to receive it during the COVID-19 pandemic. This highlights the need for more public education on the non-necessity for antibiotics for URTI and COVID-19 to address the problem of antibiotic resistance.

Perspectives towards End-of-Life Care in the Emergency Department of Tertiary Public Hospitals-A Qualitative Analysis.

Background and Objectives: End-of-life care in the emergency department (ED) is gaining importance along with the growth in the ageing population and those with chronic and terminal diseases. To explore key stakeholders' perspectives and experiences regarding end-of-life care in the ED. Materials and Methods: A descriptive qualitative study was conducted from November 2019 to January 2020. Study participants were recruited from the EDs of three tertiary hospitals and community care settings in Singapore through purposive sampling. Data collection included focus group discussions with 36 ED staff, 16 community healthcare professionals, and one-on-one semi-structured interviews with seven family members. Results: Three main themes and several subthemes emerged from the data analysis. (1) Reasons for ED visits were attributed to patients' preferences, families' decisions, limited services and capabilities in the community, and ease of access. (2) Barriers to providing end-of-life management in the ED included: conflicting priorities of staff, cramped environment, low confidence, ineffective communication, and lack of standardised workflows. (3) Discussion about continuity of end-of-life care beyond the ED uncovered issues related to delayed transfer to inpatient wards, challenging coordination of terminal discharge from the ED, and limited resources for end-of-life care in the community. Conclusions: Key stakeholders reported challenges and shared expectations in the provision of end-of-life care in the ED, which could be optimised by multidisciplinary collaborations addressing environmental factors and workflows in the ED. Equipping ED physicians and nurses with the necessary knowledge and skills is important to increase competency and confidence in managing patients attending the ED at the end of their lives.

Long term survival and disease burden from out-of-hospital cardiac arrest in Singapore: a population-based cohort study.

Background: Understanding the long-term outcomes and disability-adjusted life years (DALY) after out-of-hospital cardiac arrest (OHCA) is important to understand the overall health and disease burden of OHCA respectively, but data in Asia remains limited. We aimed to quantify long-term survival and the annual disease burden of OHCA within a national multi-ethnic Asian cohort. Methods: We conducted an open cohort study linking the Singapore Pan-Asian Resuscitation Outcomes Study (PAROS) and the Singapore Registry of Births and Deaths from 2010 to 2019. We performed Cox regression, constructed Kaplan-Meier curves, and calculated DALYs and standardised mortality ratios (SMR) for each year of follow-up. Results: We analysed 802 cases. The mean age was 56.0 (SD 17.8). Most were male (631 cases, 78,7%) and of Chinese ethnicity (552 cases, 68.8%). At one year, the SMR was 14.9 (95% CI:12.5-17.8), decreasing to 1.2 (95% CI:0.7-1.8) at three years, and 0.4 (95% CI:0.2-0.8) at five years. Age at arrest (HR:1.03, 95% CI:1.02-1.04, p < 0.001), shockable presenting rhythm (HR:0.75, 95% CI:0.52-0.93, p = 0.015) and CPC category (HR:4.62, 95% CI:3.17-6.75, p < 0.001) were independently associated with mortality. Annual DALYs due to OHCA varied from 304.1 in 2010 to 849.7 in 2015, then 547.1 in 2018. Mean DALYs decreased from 12.162 in 2010 to 3.599 in 2018. Conclusions: OHCA survivors had an increased mortality rate for the first three years which subsequently normalised compared to that of the general population. Annual OHCA disease burden in DALY trended downwards from 2010 to 2018. Improved surveillance and OHCA treatment strategies may improve long-term survivorship and decrease its global burden. Funding: National Medical Research Council, Singapore, under the Clinician Scientist Award (NMRC/CSA-SI/0014/2017) and the Singapore Translational Research Investigator Award (MOH-000982-01).

Prevalence of intracranial hemorrhage amongst patients presenting with out-of-hospital cardiac arrest: A systematic review and meta-analysis.

INTRODUCTION: An unknown proportion of out-of-hospital cardiac arrest (OHCA) is caused by intracranial hemorrhage (ICH). There is uncertainty over the role of early head computed tomography (CT) in non-traumatic OHCA due to uncertain diagnostic yield and ways to identify high-risk patients. This study aimed to identify the prevalence of ICH in non-traumatic OHCA and possible predictors. METHODS: PubMed, EMBASE, and the Cochrane library were searched from inception to January 2022. Data extraction and quality assessment were independently reviewed by two authors. Meta-analyses estimated the prevalence of ICH amongst OHCA patients and pre-specified subgroups and geographical settings. Subgroup analysis were used to explore potential clinical predictors. RESULTS: 23 studies involving 54,349 patients were included. The pooled ICH prevalence was 4.28% (95%CI: 3.31-5.24). Asia had a significantly larger risk ratio (RR = 3.93, P value < 0.0001) than Europe. The ICH subgroup was significantly more likely to be female (OR: 2.16; 95%CI: 1.10-4.26), and less likely to experience shockable rhythms compared with non-shockable rhythms (OR: 0.22; 95% CI: 0.04-1.22), achieve ROSC prior to arrival (OR: 0.27; 95%CI: 0.10-0.77), and survive to discharge compared to those without ICH (OR: 0.26; 95%CI: 0.11-0.59). CONCLUSIONS: One in twenty OHCA have ICH at the time of presentation. An early head CT scan should be strongly considered after return of spontaneous circulation (ROSC), especially in patients who are female, with non-shockable rhythm and did not attain ROSC prior to arrival. These finding should influence clinical protocols to favor routine scans especially in Asia where prevalence is higher.

Validation of the CaRdiac Arrest Survival Score (CRASS) for predicting good neurological outcome after out-of-hospital cardiac arrest in an Asian emergency medical service system.

BACKGROUND: Survival with favorable neurological outcomes is an important indicator of successful resuscitation in out-of-hospital cardiac arrest (OHCA). We sought to validate the CaRdiac Arrest Survival Score (CRASS), derived using data from the German Resuscitation Registry, in predicting the likelihood of good neurological outcomes after OHCA in Singapore. METHODS: We conducted a retrospective population-based validation study among EMS-attended OHCA patients (≥18 years) in Singapore, using data from the prospective Pan-Asian Resuscitation Outcomes Study registry. Good neurological outcome was defined as a cerebral performance category of 1 or 2. To evaluate the CRASS score in light of the difference in patient characteristics, we used the default constant coefficient (0.8) and the adjusted coefficient (0.2) to calculate the probability of good neurological outcomes. RESULTS: Out of 11,404 analyzed patients recruited between April 2010 and December 2018, 260 had good and 11,144 had poor neurological function. The CRASS score demonstrated good discrimination, with an area under the curve of 0.963 (95% confidence interval: 0.952-0.974). Using the default constant coefficient of 0.8, the CRASS score consistently overestimated the predicted probability of a good outcome. Following adjustment of the coefficient to 0.2, the CRASS score showed improved calibration. CONCLUSION: CRASS demonstrated good discrimination and moderate calibration in predicting favorable neurological outcomes in the validation Singapore cohort. Our study established a good foundation for future large-scale, cross-country validations of the CRASS score in diverse sociocultural, geographical, and clinical settings.

Determinants of emergency department utilisation by older adults in Singapore: A systematic review.

INTRODUCTION: Adults aged ≥60 years contribute to disproportionately higher visits to the emergency departments (ED). We performed a systematic review to examine the reasons why older persons visit the ED in Singapore. METHODS: We searched Medline, Embase and Scopus from January 2000 to December 2021 for studies reporting on ED utilisation by older adults in Singapore, and included studies that investigated determinants of ED utilisation. Statistically significant determinants and their effect sizes were extracted. Determinants of ED utilisation were organised using Andersen and Newman's model. Quality of studies was evaluated using Newcastle Ottawa Scale and Critical Appraisal Skills Programme. RESULTS: The search yielded 138 articles, of which 7 were used for analysis. Among the significant individual determinants were predisposing (staying in public rental housing, religiosity, loneliness, poorer coping), enabling (caregiver distress from behavioural and psychological symptoms of dementia) and health factors (multimorbidity in patients with dementia, frailty, primary care visit in last 6 months, better treatment adherence). The 7 included studies are of moderate quality and none of them employed conceptual frameworks to organise determinants of ED utilisation. CONCLUSION: The major determinants of ED utilisation by older adults in Singapore were largely individual factors. Evaluation of societal determinants of ED utilisation was lacking in the included studies. There is a need for a more holistic examination of determinants of ED utilisation locally based on conceptual models of health seeking behaviours.

Incidence, characteristics and complications of dispatcher-assisted cardiopulmonary resuscitation initiated in patients not in cardiac arrest.

AIM: Dispatcher-assisted cardiopulmonary resuscitation (DA-CPR) can increase bystander CPR rates and improve outcomes in out-of-hospital cardiac arrest (OHCA). Despite the use of protocols, dispatchers may falsely recognise some cases to be in cardiac arrest. Hence, this study aimed to find the incidence of DA-CPR initiated for non-OHCA cases, its characteristics and clinical outcomes in the Singapore population. METHODS: This was a multi-centre, observational study of all dispatcher-recognised cardiac arrests cases between January to December 2017 involving three tertiary hospitals in Singapore. Data was obtained from the Pan-Asian Resuscitation Outcomes Study cohort. Audio review of dispatch calls from the national emergency ambulance service were conducted and information about patients' clinical outcomes were prospectively collected from health records. Univariate analysis was performed to determine factors associated with in-hospital mortality among non-OHCA patients who received DA-CPR. RESULTS: Of the 821 patients recognised as having OHCA 328 (40.0%) were not in cardiac arrest and 173 (52.7%) of these received DA-CPR. No complications from chest compressions were found from hospital records. The top diagnoses of non-OHCA patients were cerebrovascular accidents (CVA), syncope and infection. Only final diagnoses of CVA (aOR 20.68), infection (aOR 17.34) and myocardial infarction (aOR 32.19) were significantly associated with in-hospital mortality. CONCLUSION: In this study, chest compressions initiated on patients not in cardiac arrest by dispatchers did not result in any reported complications and was not associated with in-hospital mortality. This provides reassurance for the continued implementation of DA-CPR.

Nonurgent Patients' Preferences for Emergency Department Versus General Practitioner and Effects of Incentives: A Discrete Choice Experiment.

Objective. This study investigates potential of a new financial incentive policy, the GP-referral discount scheme introduced in Singapore, in reducing nonurgent emergency department (ED) visits, and compares it with alternative interventions. Methods. A discrete choice experiment (DCE) was designed to elicit patients' preferences for ED and general practitioner (GP) under hypothetical nonurgent medical conditions. Through latent class multinomial logistic regression, choice models were estimated to quantify how patients' choices are influenced by GP-referral discount, other ED/GP attributes (waiting time, test facilities, and payment), patient demographics, and their perception of severity. The choice models were used to predict uptake of the GP-referral discount scheme and other countermeasures suggested by these models. Results. Survey responses from 849 respondents recruited from a public hospital in Singapore were included in the study. The choice model identified two prominent classes of patients, one of which was highly sensitive to GP-referral discount and the other to test-facility-availability. Patients' perceptions of severity ("critical" v. "not critical" enough to go to ED directly) were highly significant in influencing preference heterogeneity. Predictive analysis based on the choice model showed that GP-referral discount is more effective when patients visit ED expecting "shorter" waits, as opposed to test-facility provision at GPs and perception-correction measures that showed stronger effects under "longer" expected waits. Conclusions. The new GP-referral financial incentive introduced in Singapore can be effective in reducing nonurgent ED visits, if it reasonably covers the (extra) cost of visiting a GP. It may serve as a complement to test-facility provision at GPs or perception-correction measures, as the financial incentive and the latter two measures appear to influence distinct classes (discount-sensitive and facility-sensitive) of patients.

Validation of "Care of the Dying Evaluation" in Emergency Medicine (CODE-EM): pilot phase of end-of-life management protocol offered within emergency room (EMPOWER) study.

BACKGROUND: An increasing number of patients who present to emergency departments are at their end-of-life phase and have significant palliative care needs such as in symptom control for pain and dyspnoea. Evaluating quality of care provided is imperative, yet there is no suitable tool validated in the emergency and Asian settings. We aim to examine the face and construct validity, and reliability of a newly developed questionnaire, Care of the Dying Evaluation - Emergency Medicine, for measuring the quality of end-of-life care in an Asian emergency context. METHODS: A mixed methods pilot study was conducted. Participants composed of the next-of-kin to thirty dying patients who presented to the emergency departments of three public hospitals in Singapore. Qualitative evaluation, using cognitive "think-aloud" interviews, and quantitative analysis were employed. Percentage agreement and κ statistic were measured to evaluate temporal stability of the questionnaire. Cronbach's α and item-total correlations were used to assess internal consistency within the constructs. Confirmatory factor analysis was performed for construct validity. RESULTS: All participants reported clear understanding of the questionnaire with no ambiguity; a minority felt the questions caused emotional distress (7/30, 23.3%). The questions showed moderate to good test-retest reliability. Internal consistencies within the constructs were good for "ENVIRONMENT" and "CARE", and moderate for "COMMUNICATION". Factor loadings range from 0.40 to 0.99. CONCLUSIONS: The Care of the Dying Evaluation - Emergency Medicine questionnaire may be valid and reliable for use in an Asian emergency setting. Our prospective multicentre study using this evaluation tool may provide more insight on the quality of care rendered to dying patients and identify areas for improvement. TRIAL REGISTRATION: ClinicalTrials.gov (NCT03906747).

Impact of bystander-focused public health interventions on cardiopulmonary resuscitation and survival: a cohort study.

BACKGROUND: Bystander cardiopulmonary resuscitation (CPR) increases an individual's chance of survival from out-of-hospital cardiac arrest (OHCA), but the frequency of bystander CPR is low in many communities. We aimed to assess the cumulative effect of CPR-targeted public health interventions in Singapore, which were incrementally introduced between 2012 and 2016. METHODS: We did a secondary analysis of a prospective cohort study of adult, non-traumatic OHCAs, through the Singapore registry. National interventions introduced during this time included emergency services interventions, as well as dispatch-assisted CPR (introduced on July 1, 2012), a training programme for CPR and automated external defibrillators (April 1, 2014), and a first responder mobile application (myResponder; April 17, 2015). Using multilevel mixed-effects logistic regression, we modelled the likelihood of receiving bystander CPR with the increasing number of interventions, accounting for year as a random effect. FINDINGS: The Singapore registry contained 11 465 OHCA events between Jan 1, 2011, and Dec 31, 2016. Paediatric arrests, arrests witnessed by emergency medical services, and healthcare-facility arrests were excluded, and 6788 events were analysed. Bystander CPR was administered in 3248 (48%) of 6788 events. Compared with no intervention, likelihood of bystander CPR was not significantly altered by the addition of emergency medical services interventions (odds ratio [OR] 1·33 [95% CI 0·98-1·79]; p=0·065), but increased with implementation of dispatch-assisted CPR (3·72 [2·84-4·88]; p<0·0001), with addition of the CPR and automated external defibrillator training programme (6·16 [4·66-8·14]; p<0·0001), and with addition of the myResponder application (7·66 [5·85-10·03]; p<0·0001). Survival to hospital discharge increased after the addition of all interventions, compared with no intervention (OR 3·10 [95% CI 1·53-6·26]; p<0·0001). INTERPRETATION: National bystander-focused public health interventions were associated with an increased likelihood of bystander CPR, and an increased survival to hospital discharge. Understanding the combined impact of public health interventions might improve strategies to increase the likelihood of bystander CPR, and inform targeted initiatives to improve survival from OHCA. FUNDING: National Medical Research Council, Clinician Scientist Award, Singapore and Ministry of Health, Health Services Research Grant, Singapore.

End-of-life management protocol offered within emergency room (EMPOWER): study protocol for a multicentre study.

BACKGROUND: Patients at their end-of-life (EOL) phase frequently visit the emergency department (ED) due to their symptoms, yet the environment and physicians in ED are not traditionally equipped or trained to provide palliative care. This multicentre study aims to measure the current quality of EOL care in ED to identify gaps, formulate improvements and implement the improved EOL care protocol. We shall also evaluate healthcare resource utilisation and its associated costs. METHODS AND ANALYSIS: This study employs a quasiexperimental interrupted time series design using both qualitative and quantitative methods, involving the EDs of three tertiary hospitals in Singapore, over a period of 3 years. There are five phases in this study: (1) retrospective chart reviews of patients who died within 5 days of ED attendance; (2) pilot phase to validate the CODE questionnaire in the local context; (3) preimplementation phase; (4) focus group discussions (FGDs); and (5) postimplementation phase. In the prospective cohort, patients who are actively dying or have high likelihood of mortality this admission, and whose goal of care is palliation, will be eligible for inclusion. At least 140 patients will be recruited for each preimplementation and postimplementation phase. There will be face-to-face interviews with patients' family members, review of medical records and self-administered staff survey to evaluate existing knowledge and confidence. The FGDs will involve hospital and community healthcare providers. Data obtained from the retrospective cohort, preimplementation phase and FGDs will be used to guide prospective improvement and protocol changes. Patient, family and staff relevant outcomes from these changes will be measured using time series regression. ETHICS AND DISSEMINATION: The study protocol has been reviewed and ethics approval obtained from the National Healthcare Group Domain Specific Review Board, Singapore. The results from this study will be actively disseminated through manuscript publications and conference presentations. TRIAL REGISTRATION NUMBER: NCT03906747.