Personalised Dosing Using the CURATE.AI Algorithm: Protocol for a Feasibility Study in Patients with Hypertension and Type II Diabetes Mellitus.

Chronic diseases typically require long-term management through healthy lifestyle practices and pharmacological intervention. Although efficacious treatments exist, disease control is often sub-optimal leading to chronic disease-related sequela. Poor disease control can partially be explained by the 'one size fits all' pharmacological approach. Precision medicine aims to tailor treatments to the individual. CURATE.AI is a dosing optimisation platform that considers individual factors to improve the precision of drug therapies. CURATE.AI has been validated in other therapeutic areas, such as cancer, but has yet to be applied in chronic disease care. We will evaluate the CURATE.AI system through a single-arm feasibility study (n = 20 hypertensives and n = 20 type II diabetics). Dosing decisions will be based on CURATE.AI recommendations. We will prospectively collect clinical and qualitative data and report on the clinical effect, implementation challenges, and acceptability of using CURATE.AI. In addition, we will explore how to enhance the algorithm further using retrospective patient data. For example, the inclusion of other variables, the simultaneous optimisation of multiple drugs, and the incorporation of other artificial intelligence algorithms. Overall, this project aims to understand the feasibility of using CURATE.AI in clinical practice. Barriers and enablers to CURATE.AI will be identified to inform the system's future development.

Health Care Analytics With Time-Invariant and Time-Variant Feature Importance to Predict Hospital-Acquired Acute Kidney Injury: Observational Longitudinal Study.

BACKGROUND: Acute kidney injury (AKI) develops in 4% of hospitalized patients and is a marker of clinical deterioration and nephrotoxicity. AKI onset is highly variable in hospitals, which makes it difficult to time biomarker assessment in all patients for preemptive care. OBJECTIVE: The study sought to apply machine learning techniques to electronic health records and predict hospital-acquired AKI by a 48-hour lead time, with the aim to create an AKI surveillance algorithm that is deployable in real time. METHODS: The data were sourced from 20,732 case admissions in 16,288 patients over 1 year in our institution. We enhanced the bidirectional recurrent neural network model with a novel time-invariant and time-variant aggregated module to capture important clinical features temporal to AKI in every patient. Time-series features included laboratory parameters that preceded a 48-hour prediction window before AKI onset; the latter's corresponding reference was the final in-hospital serum creatinine performed in case admissions without AKI episodes. RESULTS: The cohort was of mean age 53 (SD 25) years, of whom 29%, 12%, 12%, and 53% had diabetes, ischemic heart disease, cancers, and baseline eGFR <90 mL/min/1.73 m2, respectively. There were 911 AKI episodes in 869 patients. We derived and validated an algorithm in the testing dataset with an AUROC of 0.81 (0.78-0.85) for predicting AKI. At a 15% prediction threshold, our model generated 699 AKI alerts with 2 false positives for every true AKI and predicted 26% of AKIs. A lowered 5% prediction threshold improved the recall to 60% but generated 3746 AKI alerts with 6 false positives for every true AKI. Representative interpretation results produced by our model alluded to the top-ranked features that predicted AKI that could be categorized in association with sepsis, acute coronary syndrome, nephrotoxicity, or multiorgan injury, specific to every case at risk. CONCLUSIONS: We generated an accurate algorithm from electronic health records through machine learning that predicted AKI by a lead time of at least 48 hours. The prediction threshold could be adjusted during deployment to optimize recall and minimize alert fatigue, while its precision could potentially be augmented by targeted AKI biomarker assessment in the high-risk cohort identified.

Factors and experiences associated with unscheduled 30-day hospital readmission: A mixed method study.

INTRODUCTION: Analysis of risk factors can pave the way for reducing unscheduled hospital readmissions and improve resource utilisation. METHODS: This was a concurrent nested, mixed method study. Factors associated with patients readmitted within 30 days between 2011 and 2015 at the National University Hospital, Singapore (N=104,496) were examined. Fifty patients were sampled in 2016 to inform an embedded qualitative study. Narrative interviews explored the periods of readmissions and related experiences, contrasted against those of non-readmitted patients. RESULTS: Neoplastic disease (odds ratio [OR] 1.91, 95% confidence interval [CI] 1.70-2.15), number of discharged medications (5 to 10 medications OR 1.21, 95% CI 1.14-1.29; ≥11 medications OR 1.80, 95% CI 1.66-1.95) and length of stay >7 days (OR 1.46, 95% CI 1.36-1.58) were most significantly associated with readmissions. Other factors including number of surgical operations, subvention class, number of emergency department visits in the previous year, hospital bill size, gender, age, Charlson comorbidity index and ethnicity were also independently associated with hospital readmissions. Although readmitted and non-readmitted patients shared some common experiences, they reported different psychological reactions to their illnesses and viewed hospital care differently. Negative emotions, feeling of being left out by the healthcare team and perception of ineffective or inappropriate treatment were expressed by readmitted patients. CONCLUSION: Patient, hospital and system-related factors were associated with readmissions, which may allow early identification of at-risk patients. Qualitative analysis suggested several areas of improvement in care including greater empowerment and involvement of patients in care and decision making.

Electronic health records accurately predict renal replacement therapy in acute kidney injury.

BACKGROUND: Electronic health records (EHR) detect the onset of acute kidney injury (AKI) in hospitalized patients, and may identify those at highest risk of mortality and renal replacement therapy (RRT), for earlier targeted intervention. METHODS: Prospective observational study to derive prediction models for hospital mortality and RRT, in inpatients aged ≥18 years with AKI detected by EHR over 1 year in a tertiary institution, fulfilling modified KDIGO criterion based on serial serum creatinine (sCr) measures. RESULTS: We studied 3333 patients with AKI, of 77,873 unique patient admissions, giving an AKI incidence of 4%. KDIGO AKI stages at detection were 1(74%), 2(15%), 3(10%); corresponding peak AKI staging in hospital were 61, 20, 19%. 392 patients (12%) died, and 174 (5%) received RRT. Multivariate logistic regression identified AKI onset in ICU, haematological malignancy, higher delta sCr (sCr rise from AKI detection till peak), higher serum potassium and baseline eGFR, as independent predictors of both mortality and RRT. Additionally, older age, higher serum urea, pneumonia and intraabdominal infections, acute cardiac diseases, solid organ malignancy, cerebrovascular disease, current need for RRT and admission under a medical specialty predicted mortality. The AUROC for RRT prediction was 0.94, averaging 0.93 after 10-fold cross-validation. Corresponding AUROC for mortality prediction was 0.9 and 0.9 after validation. Decision tree analysis for RRT prediction achieved a balanced accuracy of 70.4%, and identified delta-sCr ≥ 148 µmol/L as the key factor that predicted RRT. CONCLUSION: Case fatality was high with significant renal deterioration following hospital-wide AKI. EHR clinical model was highly accurate for both RRT prediction and for mortality; allowing excellent risk-stratification with potential for real-time deployment.

A prediction tool for nosocomial multi-drug Resistant Gram-Negative Bacilli infections in critically ill patients - prospective observational study.

BACKGROUND: The widespread use of empiric broad spectrum antibiotics has contributed to the global increase of Resistant Gram-Negative Bacilli (RGNB) infections in intensive care units (ICU). The aim of this study was to develop a tool to predict nosocomial RGNB infections among ICU patients for targeted therapy. METHODS: We conducted a prospective observational study from August'07 to December'11. All adult patients who were admitted and stayed for more than 24 hours at the medical and surgical ICU's were included. All patients who developed nosocomial RGNB infections 48 hours after ICU admission were identified. A prediction score was formulated by using independent risk factors obtained from logistic regression analysis. This was prospectively validated with a subsequent cohort of patients admitted to the ICUs during the following time period of January-September 2012. RESULTS: Seventy-six patients with nosocomial RGNB Infection (31bacteremia) were compared with 1398 patients with Systemic Inflammatory Response Syndrome (SIRS) without any gram negative bacterial infection/colonization admitted to the ICUs during the study period. The following independent risk factors were obtained by a multivariable logistic regression analysis - prior isolation of Gram negative organism (coeff: 1.1, 95% CI 0.5-1.7); Surgery during current admission (coeff: 0.69, 95% CI 0.2-1.2); prior Dialysis with end stage renal disease (coeff: 0.7, 95% CI 0.1-1.1); prior use of Carbapenems (coeff: 1.3, 95% CI 0.3-2.3) and Stay in the ICU for more than 5 days (coeff: 2.4, 95% CI 1.6-3.2). It was validated prospectively in a subsequent cohort (n = 408) and the area-under-the-curve (AUC) of the GSDCS score for predicting nosocomial ICU acquired RGNB infection and bacteremia was 0.77 (95% CI 0.68-0.89 and 0.78 (95% CI 0.69-0.89) respectively. The GSDCS (0-4.3) score clearly differentiated the low (0-1.3), medium (1.4-2.3) and high (2.4-4.3) risk patients, both for RGNB infection (p:0.003) and bacteremia (p:0.009). CONCLUSION: GSDCS is a simple bedside clinical score which predicts RGNB infection and bacteremia with high predictive value and differentiates low versus high risk patients. This score will help clinicians to choose appropriate, timely targeted antibiotic therapy and avoid exposure to unnecessary treatment for patients at low risk of nosocomial RGNB infection. This will reduce the selection pressure and help to contain antibiotic resistance in ICUs.

Risk factors for infection/colonization caused by resistant Gram negative bacilli in critically ill patients (an observational study of 1633 critically ill patients).

OBJECTIVE: The objective of this study is to identify risk factors associated with multi-resistant Gram negative (RGNB) infection and colonization among critically ill patients. METHODS: A prospective cohort study of all patients aged 21-90 admitted for more than 24 hours in Medical and Surgical intensive care units (ICU) at a large teaching hospital in Singapore for the period of Aug '07-Dec '09 was conducted. Patient demographics, comorbidities, antibiotics, invasive devices, and culture results were collected. Forward stepwise logistic regression analyses were done to identify risk factors associated with RGNB infection and colonization. RESULTS: Of the 1373 patients included in the analysis, 13.5% developed RGNB infection. A logistic regression analysis including variables with a p value of <0.2 in the univariate analysis showed that recent surgery (OR 2.1, 95% CI 1.2-3.6), renal impairment (OR 2.9, 95% CI 1.5-5.4), liver disease (OR: 3.8, 95% CI 1.7-8.8), central line (OR 1.8, 95% CI 1.01-3.4) were independently associated with RGNB infection in the ICU. Surgery (OR 3.9, 95% CI 2.7-5.7), third-line antibiotics (carbapenem, vancomycin, linezolid) (OR 1.8, 95% CI 1.2-2.9) were independently associated with RGNB infection during their hospitalization. CONCLUSION: The major risk factors identified for RGNB infection and colonization in the ICU were mainly patient dependent. However, broad spectrum initial antibiotic treatment remains an important independent modifiable risk factor. Interventions aimed at reducing initial broad spectrum antibiotics are clearly needed to help control the spread of these difficult to treat infections.

End-of-life care in the general wards of a Singaporean hospital: an Asian perspective.

BACKGROUND: Despite international differences in cultural perspectives on end-of-life issues, little is known of the care for the dying in the general wards of acute hospitals in Asia. METHODS: We performed a retrospective medical chart review of all 683 adult patients who died without intensive care unit (ICU) admission in our Singaporean hospital in 2007. We first evaluated the prevalence of do-not-resuscitate (DNR) orders and orders for or against life-sustaining therapies; second, if such orders were discussed with the patients and/or family members; and third, the actual treatments provided before death. RESULTS: There were DNR orders for 66.2% of patients and neither commitment for DNR nor cardiopulmonary resuscitation (CPR) for 28.1%. Orders to limit life-sustaining therapies, including ICU admission, intubation, and vasopressors/inotropes were infrequent. Only 6.2% of the alert and conversant patients with DNR orders were involved in discussions on these orders. In contrast, such discussions with their family members occurred 82.9% of the time. Interventions in the last 24 hours of life included CPR (9.4%), intubation (6.4%), vasopressors/inotropes (14.8%), tube feeding (24.7%), and antibiotics (44.9%). Analgesia was provided in 29.1% of patients. CONCLUSIONS: There was a lack of commitment by doctors on orders for DNR/CPR and to limit life-sustaining therapies, infrequent discussions with patients on end-of-life decisions, and excessive burdensome interventions with inadequate palliative care for the dying. These findings may reflect certain Asian cultural biases. More work is required to improve our quality of end-of-life care.

Noninvasive and invasive ventilation in acute respiratory failure associated with bronchiectasis.

PURPOSE: To describe the outcomes of patients with bronchiectasis and acute respiratory failure (ARF) treated with noninvasive ventilation (NIV) and invasive mechanical ventilation (IMV) after a failure of conservative measures, and to identify the predictors of hospital mortality and NIV failure. METHODS: Retrospective review of bronchiectatic patients on NIV (n = 31) or IMV (n = 26) for ARF over 8 years in a medical intensive care unit (ICU) experienced in NIV. RESULTS: At baseline, the NIV group had more patients with acute exacerbations without identified precipitating factors (87.1 vs. 34.6%, p < 0.001), higher pH (mean 7.25 vs. 7.18, p = 0.008) and PaO(2)/FiO(2) ratio (mean 249.4 vs. 173.2, p = 0.02), and a trend towards a lower APACHE II score (mean 25.3 vs. 28.4, p = 0.07) than the IMV group. There was no difference in hospital mortality between the two groups (25.8 vs. 26.9%, p > 0.05). The NIV failure rate (need for intubation or death in the ICU) was 32.3%. Using logistic regression, the APACHE II score was the only predictor of hospital mortality (OR 1.19 per point), and the PaO(2)/FiO(2) ratio was the only predictor of NIV failure (OR 1.02 per mmHg decrease). CONCLUSIONS: The hospital mortality of patients with bronchiectasis and ARF approximates 25% and is predicted by the APACHE II score. When selectively applied, NIV fails in one-third of the patients, and this is predicted by hypoxemia. Our findings call for randomised controlled trials to compare NIV versus IMV in such patients.

Transbronchial needle aspiration: initial experience in routine diagnostic bronchoscopy.

BACKGROUND: Transbronchial needle aspiration (TBNA) has been shown to be useful not only for the diagnosis and staging of lung cancer, its most widely studied indication, but also for many of other clinical indications. Despite this, it remains largely underutilized, mainly because of concerns with poor yield, safety, lack of experience of the bronchoscopist, and lack of cytopathological support. OBJECTIVE: To study the clinical utility and yield of TBNA as an adjunct to other conventional procedures in diagnostic bronchoscopy at a centre that was relatively inexperienced with this technique, but where there was availability of rapid on-site evaluation (ROSE). Most of the major indications for TBNA in both malignant as well as benign disease were included. SETTING: University Teaching Hospital naïve to the procedure. PATIENT AND METHODS: Forty-five consecutive patients who underwent TBNA as part of diagnostic bronchoscopy during a 2-year study period. RESULTS: TBNA gave a yield of 65% for evaluation of mediastinal disease, both benign and malignant. The overall diagnostic utility for all indications was 71% and there were no complications. CONCLUSIONS: We conclude that TBNA is a useful and safe adjunct to diagnostic bronchoscopy in routine clinical practice. It has a satisfactory yield even with an inexperienced team, if used with ROSE.