AKI Risk Score (AKI-RiSc): Developing an Interpretable Clinical Score for Early Identification of Acute Kidney Injury at the Emergency Department

Acute kidney injury (AKI) in hospitalised patients is a common syndrome associated with poorer patient outcomes. Clinical risk scores can be used for the early identification of patients at risk of AKI. We conducted a retrospective study using electronic health records of Singapore General Hospital emergency department patients who were admitted from 2008 to 2016. The primary outcome was inpatient AKI of any stage within 7 days of admission based on Kidney Disease Improving Global Outcome (KDIGO) 2012 guidelines. A machine learning AutoScore algorithm was used to generate clinical scores from the study sample which was divided into training, validation and testing cohorts. Model performance was evaluated using area under the curve (AUC). Among the 119,468 admissions, 10,693 (9.0%) developed AKI. 8,491 were stage 1 (79.4%), 906 stage 2 (8.5%) and 1,296 stage 3 (12.1%). The AKI Risk Score (AKI-RiSc) was a summation of the integer scores of 6 variables: serum creatinine, serum bicarbonate, pulse, systolic blood pressure, and diastolic blood pressure. AUC of AKI-RiSc was 0.730 (95% CI: 0.713 – 0.747), outperforming an existing AKI Prediction Score model which achieved AUC of 0.665 (95% CI: 0.646 – 0.679) on the testing cohort. At a cut-off of 4 points, AKI-RiSc had a sensitivity of 82.5% and specificity of 46.7%. AKI-RiSc is a simple clinical score that can be easily implemented on the ground for early identification of AKI and potentially be applied in international settings.

Ambiguous Diagnosis of Nodal T-Cell Lymphomas with T-Follicular Helper (TFH) Phenotype and Hodgkin Lymphoma - a Multicenter Pilot Study

Background and aims: Angioimmunoblastic T-cell lymphoma (AITL) is currently classified amongst an umbrella group of diagnoses referred to as nodal peripheral T-cell lymphoma (PTCL) with T-follicular helper (TFH) phenotype. In this entity, it is recognized that the presence of atypical B-cell blasts mimicking Hodgkin-Reed-Sternberg cells may lead to misdiagnosis as Hodgkin lymphoma (HL), resulting in suboptimal management and poorer outcomes. This diagnostic complexity has been acknowledged by the World Health Organization classification of lymphoid neoplasms. In this study, we aim to investigate the clinical features and outcomes of such cases of diagnostic ambiguity (HL/PTCL). Methods: A total of 379 patients from the Singapore Lymphoma Study database across three tertiary cancer centers (National Cancer Centre Singapore, Singapore General Hospital, National University Cancer Institute) were included in this study, including those diagnosed with AITL (n=169), HL (n=178) and HL/PTCL (n=32). Median follow-up was 47.5 months. Relevant demographical and clinical characteristics were collected. Survival analyses were performed using the Kaplan-Meier method. Results: The 32 patients with HL/PTCL were identified from three distinct clinical scenarios. In the first group (n=18), an initial histological diagnosis of HL was revised upon independent pathological re-assessment following a second opinion consultation (revised diagnoses to AITL, n=10; PTCL-TFH, n=7; PTCL-NOS, n=1). In the second group (n=11), the histological diagnoses were different at diagnosis and at relapse (mostly HL relapsed as AITL/PTCL-TFH, n=9). The third group consisted of patients with synchronous or composite features of both HL and AITL/PTCL at diagnosis (n=3). In the overall cohort, most were male (62.5%), and the majority had excellent performance status with ECOG 0-1 (96.9%). The median age was 45 years (range, 20 to 77), which is older than the HL cohort (28 years, p=0.0012) but younger than the AITL cohort (62 years, p=0.0005). In terms of disease staging, most were Ann Arbor stage 3-4 (62.5%), which is comparable to the AITL cohort (78.3%), but significantly more advanced than the HL cohort (32%, p=0.001). HL/PTCL represented a group with worse prognostic risk scores by IPI as compared to HL (p=0.0038), but better as compared to AITL (p=0.0418). Accordingly, the median overall survival was 8.4 years for HL/PTCL, compared to 5.5 years (AITL) and not reached (HL) (logrank p<0.0001). Conclusion: We describe a significant and clinically distinct group of HL and AITL/PTCL-TFH cases which are challenging to diagnose. Further studies on the molecular characteristics of this ambiguous disease entity may be required to resolve the diagnostic difficulties. Figure 1 Figure 1. Disclosures No relevant conflicts of interest to declare.

34. Stemming the Rise in Antibiotic Prescription for Community Acquired Respiratory Infections (ARI) During COVID-19 Pandemic in Singapore General Hospital (SGH)

Background In early months of COVID-19 pandemic, SGH recorded a year-on-year increase in antibiotic (ABx) use for community acquired acute respiratory infection (CA ARI) from Feb-Apr 2019 (48.7 defined daily doses (DDD)/100 bed-days) to 2020 (50.8 DDD/100 bed-days). To address concerns of misuse, the antibiotic stewardship unit (ASU) expanded prospective audit feedback (PAF) to CA ARI patients admitted to ARI wards, with low procalcitonin (PCT). PAF was conducted on day 2-3 of ABx, on weekdays. Doctors received feedback to stop/modify when ABx was deemed inappropriate. Here, we describe the impact of ASU’s adaptive approach to curb rising ABx use in patients admitted for ARI during COVID-19 pandemic. Methods A Pre- & Post-intervention study was conducted. All patients started on ABx (ceftriaxone/co-amoxiclav/piptazo/carbapenems/levofloxacin) for CA ARI & PCT < 0.5µg/L were analysed. Those who died ≤48h of admission; admitted to intensive care; required ABx escalation; >1 infective sites; complex lung infection were excluded. Primary objective was to compare the proportion of ABx stopped ≤4 days (time to final infection diagnosis) Pre (22/3-18/4/20) & Post (21/4-13/7/20). Results 184 (Pre) & 528 (Post) ABx courses were analysed. ASU audited 51 (Pre) & 380 (Post) courses with the rest discontinued/discharged before review. Patients were largely similar in both periods; a third had low likelihood of bacterial infection (C reactive protein < 30mg/L). In Post, 73 feedback was given to stop ABx (often because symptoms suggested viral/fluid overload) & 18 to switch to oral ABx. 82 (90%) feedback was accepted. No ABx was restarted ≤48h or deaths ≤30 days due to ARI. 1 patient had C. difficile diarrhoea a day after ABx cessation as per ASU feedback. Proportion of all ABx stopped ≤4 days was higher in Post than Pre [27/184 (15%) vs 152/528 (29%), p< 0.01]. Median duration of therapy of IV ABx was reduced (6.5 vs 3 days, p< 0.01), with corresponding shorter median length of stay (10.5 vs 6 days, p< 0.01). Conclusion PAF directly and indirectly reduced ABx duration in patients treated for CA ARI as prescribers become more conscious about stopping ABx when investigations show low likelihood of bacterial infection. ASU must remain agile during pandemics to detect emerging problems and adapt processes to counter early. Disclosures All Authors: No reported disclosures

104. Improving Efficiency of Antimicrobial Stewardship Reviews Using Artificial Intelligence Modelling

Background Antimicrobial stewardship programs (ASP) in hospitals improve antibiotic prescribing, slow antimicrobial resistance, reduce hospitalisation duration, mortality and readmission rates, and save costs. However, the strategy of prospective audit and feedback is laborious. In Singapore General Hospital (SGH), 10 reviews are required to identify 2 inappropriate cases. Limited manpower constraints ASP audits to only about 30% of antibiotics prescribed. This proof-of-concept study explored the feasibility of developing a predictive model to prioritise inappropriate antibiotic prescriptions for ASP review. Methods ASP-audited adult pneumonia patients from January 2016 to December 2018 in SGH were included. Patient data e.g., demographics, allergies, past medical history, and relevant laboratory investigations at each antibiotic use episode were extracted from electronic medical records and re-assembled through linking for analysis. Ground truth for model training was based on ASP-defined appropriateness for each encounter. The dataset was split into 80% and 20% for training and testing respectively. Three modelling techniques, XGBoost, decision tree and logistic regression, were assessed for their relative performance in terms of precision, sensitivity and specificity. Results There were 12471 unique patient encounters. Training was done on 10459 encounters and 39 data elements were included. When tested on 2012 encounters, the logistic regression model performed the best (86.7% sensitivity, 71.4% specificity). The model correctly classified 1377 out of 1388 (99.2%) encounters as “appropriate” (do not require ASP intervention). 624 antibiotic use encounters were classified as “inappropriate”, of which only 72 were truly inappropriate (positive predictive value for ASP intervention, PPV 11.5%). The low PPV was likely due to inadequate representation of “inappropriate” cases in the training dataset (4.1%). Applying this model would prioritise the number of immediate ASP reviews needed to identify cases for intervention by two-thirds, from 2012 to 624 (Figure 1). Figure 1. Illustration of AI benefits in ASP Conclusion ASPs can leverage on machine learning capabilities to improve audit efficiency. This can increase ASP’s productivity and staff’s job satisfaction as they are freed up to perform other work. Disclosures All Authors: No reported disclosures

Strategies for a successful hepatic artery anastomosis in liver transplantation: A review of 51 cases

ABSTRACT Introduction: Hepatic artery reconstruction is a critical aspect of liver transplantation. The microsurgeon faces several challenges when reconstructing the hepatic artery—the donor hepatic artery stalk is short and often a poor match for the usually hypertrophic recipient vessels. Previous inflammation impedes vessel dissection, and recipient vessels have a tendency to delaminate with manipulation. We review 51 consecutive liver transplantations to highlight these problems and propose strategies for a successful reconstruction of the hepatic artery. Methods: A prospective study involving all adult patients undergoing liver transplantation at the Singapore General Hospital from January 2015 to December 2018 was undertaken. All hepatic artery anastomoses were performed by 2 microsurgeons at 10x magnification. Patients were started on a standard immunosuppressive regimen. Postoperative ultrasound scans on days 1, 3, 5, 7, 9 and 14 were used to confirm arterial patency. Results: There were 51 patients who underwent liver transplantation during the study period. Of this number, 31 patients received deceased donor grafts and 20 received living donor grafts. A total of 61 anastomoses were performed (5 dual anastomosis, 4 radial artery interposition grafts) with 1 case of hepatic artery thrombosis that was successfully salvaged. The mean (range) postoperative resistive index and hepatic artery peak systolic velocity were 0.69 (0.68–0.69) and 1.0m/s (0.88–1.10m/s), respectively. Conclusion: Hepatic artery thrombosis after liver transplantation is poorly tolerated. The challenges of hepatic artery reconstruction in liver transplantation are related to vessel quality and length. The use of microsurgical technique, appropriate recipient vessel selection, minimisation of vessel manipulation with modified instruments, variation in anastomosis techniques, and use of radial artery interpositional grafts are useful strategies to maximise the chances of success. Keywords: Hepatic artery, hepatology, liver transplant, microsurgery, plastic surgery

Physical activity in patients with axial spondyloarthritis in a multi-ethnic south-east Asian country

Background Patients with axial spondyloarthritis (axSpA) may experience spinal stiffness and pain, leading to reduced physical function and quality of life. Despite the benefits of physical activity (PA) and exercise, previous studies have demonstrated lower levels of PA among patients with axSpA. This study aims to examine the patterns of PA among patients with axSpA compared to the general population in a multi-ethnic Asian country. Methods This was a cross-sectional study conducted between May 2016 and Jan 2017. Consecutive patients with axSpA were recruited at an outpatient rheumatology clinic at Singapore General Hospital, the largest tertiary hospital in Singapore. Controls were based on a previous cross-sectional study. PA was assessed using the Global Physical Activity Questionnaire (GPAQ). Results Seventy-four patients with axSpA were recruited and compared with 2679 controls. Lower proportion of patients with axSpA met the WHO recommendations for PA (axSpA = 77.0%, controls = 89.7%, p < 0.001). More patients with axSpA had high level of sedentary activity compared to controls (axSpA = 56.8%, controls = 36.1%, p < 0.001). Levels of PA did not differ between patients with inactive versus active axSpA disease (p = 0.91). Conclusions Proportion of patients with axSpA meeting the WHO recommendations for PA differed significantly from the general population, and level of PA did not differ between patients with active and inactive axSpA disease. Higher levels of sedentary activity were seen in patient with axSpA compared to the general population, highlighting the need for interventions to promote PA among patients with axSpA.

Non-steroidal anti-inflammatory drugs and risk of acute adverse renal outcomes in diabetes and diabetic kidney disease

BACKGROUND: Individuals with diabetes mellitus (DM) may be susceptible to non-steroidal anti-inflammatory drug (NSAID) – induced acute kidney injury (AKI) but data on NSAID-related adverse renal events is sparse. We aimed to evaluate the risk of acute kidney injury and/or hyperkalemia after systemic NSAID among individuals with DM and diabetic chronic kidney disease (CKD). METHODS: Retrospective cohort study of 3896 adults with DM with incident prescriptions between July 2015 and December 2017 from Singapore General Hospital and SingHealth Polyclinics. Laboratory, hospitalization and medication data were retrieved from electronic medical records. The primary outcome was the incidence of AKI and/ or hyperkalemia within 30 days after prescription. RESULTS: AKI and/or hyperkalemia occurred in 13.5% of all DM and 15.8% of diabetic CKD. The association between systemic NSAID >14 days and 30-day risk of AKI and/or hyperkalemia failed to reach statistical significance in unselected DM (adjusted OR 1.62, 95% CI 0.99–2.65, p = 0.05) and diabetic CKD (adjusted OR 0.64, 95% CI 0.15–2.82, p = 0.64), but the odds of AKI and/or hyperkalemia were markedly and significantly increased when NSAID was prescribed with renin-angiotensin-aldosterone system (RAAS) blocker (adjusted OR 4.17, 95% CI 1.74–9.98, p = 0.001) or diuretic (adjusted OR 3.31, 95% CI 1.09–10.08, p = 0.04) and in the absence of diabetic CKD (adjusted OR 1.98, 95% CI 1.16–3.36, p = 0.01). CONCLUSION: NSAID prescription >14 days in individuals with DM with concurrent RAAS blockers or diuretics was associated with higher 30-day risk of AKI and/or hyperkalemia.

Perceptions of healthcare professionals towards palliative care in internal medicine wards: a cross-sectional survey

Background The extension of palliative care services to meet the needs of patients with chronic non-malignant life-limiting conditions faces misconceptions amongst healthcare professionals. A study of prevailing perceptions of healthcare professionals on this wider palliative care service was thus conducted to identify current obstacles, guide the education of local healthcare professionals and improve service accessibility. Methods A cross-sectional study was carried out at the Singapore General Hospital. An anonymised and close-ended online questionnaire was disseminated to 120 physicians and 500 nurses in the Department of Internal Medicine. The online survey tool focused on participant demographics; perceptions of palliative care and its perceived benefits; roles and indications; and attitudes and behaviours towards palliative care referrals. Results Forty four physicians and 156 nurses suggested that care of terminally ill patients with chronic non-malignant life-limiting conditions are compromised by concerns over the role of palliative care in non-cancer care and lapses in their prognostication and communication skills. Respondents also raised concerns about their ability to confront sociocultural issues and introduce palliative care services to patients and their families. Conclusions Gaps in understanding and the ability of nurses and physicians to communicate end of life issues, introduce palliative care services to patients and their families and confront sociocultural issues suggest the need for a longitudinal training program. With similar concerns likely prevalent in other clinical settings within this island nation, a concerted national education program targeting obstacles surrounding effective palliative care should be considered.

AKI Risk Score (AKI-RiSc): Developing an Interpretable Clinical Score for Early Identification of Acute Kidney Injury for Patients Presenting to the Emergency Department

Background: Acute kidney injury (AKI) in hospitalised patients is a common syndrome associated with poorer patient outcomes. Clinical risk scores can be used for the early identification of patients at risk of AKI. Methods: We conducted a retrospective study using electronic health records of Singapore General Hospital emergency department patients who were admitted from 2008 to 2016. The primary outcome was inpatient AKI of any stage within 7 days of admission based on Kidney Disease Improving Global Outcome (KDIGO) 2012 guidelines. AutoScore, a machine learning based algorithm, was used to generate point based clinical scores from the study sample which was divided into training, validation and testing cohorts. Model performance was evaluated using area under the curve (AUC). Results: Among the 119,468 admissions, 10,693 (9.0%) developed AKI. 8,491 were stage 1 (79.4%), 906 stage 2 (8.5%) and 1,296 stage 3 (12.1%). The AKI Risk Score (AKI-RiSc) was a summation of the integer scores of 6 variables: serum creatinine, serum bicarbonate, pulse, systolic blood pressure, and diastolic blood pressure. AUC of AKI-RiSc was 0.730 (95% CI: 0.713 - 0.747), outperforming an existing AKI Prediction Score model which achieved AUC of 0.665 (95% CI: 0.646 - 0.679) when evaluated on the same test cohort. At a cut-off of 4 points, AKI-RiSc had a sensitivity of 82.5% and specificity of 46.7%. Conclusion: AKI-RiSc is a simple point based clinical score that can be easily implemented on the ground for early identification of AKI in high-risk patients and potentially be applied in healthcare settings internationally.

Healthcare workers as a sentinel surveillance population in the early phase of the COVID-19 pandemic

Introduction: Healthcare workers (HCWs) are a critical resource in the effort to control the COVID-19 pandemic. They are also a sentinel surveillance population whose clinical status reflects the effectiveness of the hospital’s infection prevention measures in the pandemic. Methods: This was a retrospective cohort study conducted in Singapore General Hospital (SGH), a 1,822-bed tertiary hospital. Participants were all HCWs working in SGH during the study period. HCW protection measures included clinical workflows and personal protective equipment developed and adapted to minimise the risk of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) transmission. HCW monitoring comprised staff contact logs in high-risk locations, twice-daily temperature monitoring, assessment of HCWs with acute respiratory illnesses (ARIs) in the staff clinic and, in the event of an exposure, extensive contact tracing, detailed risk assessment and risk-based interventions. HCW surveillance utilised monitoring data and ARI presentations and outcomes. Results: In the ten-week period between 6 January 2020 and 16 March 2020, 333 (17.1%) of 1,946 HCWs at risk of occupational COVID-19 presented with ARI. 32 (9.6%) screened negative for SARS-CoV-2 from throat swabs. Five other HCWs developed COVID-19 attributed to non-clinical exposures. From the nine COVID-19 exposure episodes investigated, 189 HCW contacts were identified, of whom 68 (36.2%) were placed on quarantine and remained well. Conclusion: Early in an emerging infectious disease outbreak, close monitoring of frontline HCWs is essential in ascertaining the effectiveness of infection prevention measures. HCWs are at risk of community disease acquisition and should be monitored and managed to prevent onward transmission.