Developing machine learning models to personalize care levels among emergency room patients for hospital admission

ObjectiveTo develop prediction models for intensive care unit (ICU) vs non-ICU level-of-care need within 24 hours of inpatient admission for emergency department (ED) patients using electronic health record data.Materials and MethodsUsing records of 41 654 ED visits to a tertiary academic center from 2015 to 2019, we tested 4 algorithms—feed-forward neural networks, regularized regression, random forests, and gradient-boosted trees—to predict ICU vs non-ICU level-of-care within 24 hours and at the 24th hour following admission. Simple-feature models included patient demographics, Emergency Severity Index (ESI), and vital sign summary. Complex-feature models added all vital signs, lab results, and counts of diagnosis, imaging, procedures, medications, and lab orders.ResultsThe best-performing model, a gradient-boosted tree using a full feature set, achieved an AUROC of 0.88 (95%CI: 0.87–0.89) and AUPRC of 0.65 (95%CI: 0.63–0.68) for predicting ICU care need within 24 hours of admission. The logistic regression model using ESI achieved an AUROC of 0.67 (95%CI: 0.65–0.70) and AUPRC of 0.37 (95%CI: 0.35–0.40). Using a discrimination threshold, such as 0.6, the positive predictive value, negative predictive value, sensitivity, and specificity were 85%, 89%, 30%, and 99%, respectively. Vital signs were the most important predictors.Discussion and ConclusionsUndertriaging admitted ED patients who subsequently require ICU care is common and associated with poorer outcomes. Machine learning models using readily available electronic health record data predict subsequent need for ICU admission with good discrimination, substantially better than the benchmarking ESI system. The results could be used in a multitiered clinical decision-support system to improve ED triage.     

Building longitudinal medication dose data using medication information extracted from clinical notes in electronic health records

ObjectiveTo develop an algorithm for building longitudinal medication dose datasets using information extracted from clinical notes in electronic health records (EHRs).Materials and MethodsWe developed an algorithm that converts medication information extracted using natural language processing (NLP) into a usable format and builds longitudinal medication dose datasets. We evaluated the algorithm on 2 medications extracted from clinical notes of Vanderbilt’s EHR and externally validated the algorithm using clinical notes from the MIMIC-III clinical care database.ResultsFor the evaluation using Vanderbilt’s EHR data, the performance of our algorithm was excellent; F1-measures were ≥0.98 for both dose intake and daily dose. For the external validation using MIMIC-III, the algorithm achieved F1-measures ≥0.85 for dose intake and ≥0.82 for daily dose.DiscussionOur algorithm addresses the challenge of building longitudinal medication dose data using information extracted from clinical notes. Overall performance was excellent, but the algorithm can perform poorly when incorrect information is extracted by NLP systems. Although it performed reasonably well when applied to the external data source, its performance was worse due to differences in the way the drug information was written. The algorithm is implemented in the R package, “EHR,” and the extracted data from Vanderbilt’s EHRs along with the gold standards are provided so that users can reproduce the results and help improve the algorithm.ConclusionOur algorithm for building longitudinal dose data provides a straightforward way to use EHR data for medication-based studies. The external validation results suggest its potential for applicability to other systems.     

Development of a predictive model for retention in HIV care using natural language processing of clinical notes

ObjectiveAdherence to a treatment plan from HIV-positive patients is necessary to decrease their mortality and improve their quality of life, however some patients display poor appointment adherence and become lost to follow-up (LTFU). We applied natural language processing (NLP) to analyze indications towards or against LTFU in HIV-positive patients’ notes.Materials and MethodsUnstructured lemmatized notes were labeled with an LTFU or Retained status using a 183-day threshold. An NLP and supervised machine learning system with a linear model and elastic net regularization was trained to predict this status. Prevalence of characteristics domains in the learned model weights were evaluated.ResultsWe analyzed 838 LTFU vs 2964 Retained notes and obtained a weighted F1 mean of 0.912 via nested cross-validation; another experiment with notes from the same patients in both classes showed substantially lower metrics. “Comorbidities” were associated with LTFU through, for instance, “HCV” (hepatitis C virus) and likewise “Good adherence” with Retained, represented with “Well on ART” (antiretroviral therapy).DiscussionMentions of mental health disorders and substance use were associated with disparate retention outcomes, however history vs active use was not investigated. There remains further need to model transitions between LTFU and being retained in care over time.ConclusionWe provided an important step for the future development of a model that could eventually help to identify patients who are at risk for falling out of care and to analyze which characteristics could be factors for this. Further research is needed to enhance this method with structured electronic medical record fields.     

Singapore SPICE: sedation practices in intensive care evaluation in Singapore – a prospective cohort study of the public healthcare system

INTRODUCTIONA study was conducted to describe the sedation practices of intensive care units (ICUs) in Singapore in terms of drug use, sedation depth and the incidence of delirium in both early (< 48 hours) and late (> 48 hours) periods of ICU admission.METHODSA prospective multicentre cohort study was conducted on patients who were expected to be sedated and ventilated for over 24 hours in seven ICUs (surgical ICU, n = 4; medical ICU, n = 3) of four major public hospitals in Singapore. Patients were followed up to 28 days or until ICU discharge, with four-hourly sedation monitoring and daily delirium assessment by trained nurses. The Richmond Agitation and Sedation Scale (RASS) and Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) were used.RESULTSWe enrolled 198 patients over a five-month period. The mean Acute Physiology and Chronic Health Evaluation (APACHE) II score was 25.3 ± 9.2, and 90.9% were emergency hospital admissions. Patients were followed up for 1,417 ICU patient days, of which 396 days were in the early period and 1,021 days were in the late period. 7,354 RASS assessments were performed. Propofol and fentanyl were the sedative agents of choice in the early and late periods, respectively. Patients were mostly in the light sedation range, especially in the late period. At least one episode of delirium was seen in 23.7% of patients.CONCLUSIONSedation practices in Singapore ICUs are characterised by light sedation depth and low incidence of delirium, possibly due to the drugs used.     

Automated model versus treating physician for predicting survival time of patients with metastatic cancer

ObjectiveBeing able to predict a patient’s life expectancy can help doctors and patients prioritize treatments and supportive care. For predicting life expectancy, physicians have been shown to outperform traditional models that use only a few predictor variables. It is possible that a machine learning model that uses many predictor variables and diverse data sources from the electronic medical record can improve on physicians’ performance. For patients with metastatic cancer, we compared accuracy of life expectancy predictions by the treating physician, a machine learning model, and a traditional model.Materials and MethodsA machine learning model was trained using 14 600 metastatic cancer patients’ data to predict each patient’s distribution of survival time. Data sources included note text, laboratory values, and vital signs. From 2015–2016, 899 patients receiving radiotherapy for metastatic cancer were enrolled in a study in which their radiation oncologist estimated life expectancy. Survival predictions were also made by the machine learning model and a traditional model using only performance status. Performance was assessed with area under the curve for 1-year survival and calibration plots.ResultsThe radiotherapy study included 1190 treatment courses in 899 patients. A total of 879 treatment courses in 685 patients were included in this analysis. Median overall survival was 11.7 months. Physicians, machine learning model, and traditional model had area under the curve for 1-year survival of 0.72 (95% CI 0.63–0.81), 0.77 (0.73–0.81), and 0.68 (0.65–0.71), respectively.ConclusionsThe machine learning model’s predictions were more accurate than those of the treating physician or a traditional model.     

Feasibility of using the predisposition,insult/infection,physiological response,and organ dysfunction concept of sepsis to predict the risk of deterioration and unplanned intensive care unit transfer after emergency department admission

BackgroundRecognizing patients at risk for deterioration and in need of critical care after emergency department (ED) admission may prevent unplanned intensive care unit (ICU) transfers and decrease the number of deaths in the hospital. The objective of this research was to study if the predisposition, insult, response, and organ dysfunction (PIRO) concept of sepsis can be used to predict the risk of unplanned ICU transfer after ED admission.MethodsThe ICU transfer group included 313 patients with unplanned transfer to the ICU within 48 hours of ED admission, and the control (non-transfer) group included 736 randomly sampled patients who were not transferred to the ICU. Two-thirds of the total 1049 patients in this study were randomly assigned to a derivation group, which was used to develop the PIRO model, and the remaining patients were assigned to a validation group.ResultsIndependent predictors of deterioration within 48 hours after ED admission were identified by the PIRO concept. PIRO scores were higher in the ICU transfer group than in the non-transfer group, both in the derivation group [median (mean ± SD), 5 (5.7 ± 3.7) vs. 2 (2.5 ± 2.5); p < 0.001], and in the validation group [median (mean ± SD), 6 (6.0 ± 3.4) vs. 2 (2.4 ± 2.6); p < 0.001]. The proportion of ICU transfer patients with a PIRO score of 0–3, 4–6, 7–9, and ≥10 was 14.1%, 46.5%, 57.3%, and 83.8% in the derivation group (p < 0.001) and 12.8%, 37.3%, 68.2%, and 70.0% in the validation group (p < 0.001), respectively. The proportion of inpatient mortality in patients with a PIRO score of 0–3, 4–6, 7–9, and ≥10 was 2.6%, 10.1%, 23.2%, and 45.9% in the derivation group (p < 0.001) and 3.3%, 12.0%, 18.2%, and 20.5% in the validation group (p < 0.001), respectively.ConclusionThe PIRO concept of sepsis may be used in undifferentiated medical ED patients as a prediction system for unplanned ICU transfer after admission.     

血清胱抑素C对ICU老年患者急性肾损伤及死亡的预测作用

目的评价血清胱抑素C是否可以预测ICU老年患者的急性肾损伤(AKI)及死亡风险。方法选取入住ICU时无AKI老年患者(年龄>60岁),68例患者符合条件入选,有13人发展为AKI,入院时有24人血清胱抑素C升高,分别按照有无AKI、血清胱抑素C浓度是否高于正常值分组,对各参数进行统计学分析。结果血清胱抑素C水平升高能预测AKI(1.82±0.58VS 1.33±0.34,P=0.01),而且是死亡的独立预测因素(OR=5.05,95%CI:1.14~9.47,P=0.02),而AKI与死亡率无关(P=0.241)。根据ROC曲线,血清胱抑素C与AKI相比较具有更大的曲线下面积(area=0.721,P=0.02 VS area=0.455,P=0.64)。结论高水平的血清胱抑素C能预测ICU老年患者的AKI,并且是死亡风险的独立预测因素,可以作为判断AKI及预后的血清标志物。     

Healthcare-associated Pneumonia: Clinical Features and Retrospective Analysis Over 10 Years

Background:

Healthcare-associated pneumonia (HCAP) is associated with drug-resistant pathogens and high mortality, and there is no clear evidence that this is due to inappropriate antibiotic therapy. This study was to elucidate the clinical features, pathogens, therapy, and outcomes of HCAP, and to clarify the risk factors for drug-resistant pathogens and prognosis.

Methods:

Retrospective observational study among hospitalized patients with HCAP over 10 years. The primary outcome was 30-day all-cause hospital mortality after admission. Demographics (age, gender, clinical features, and comorbidities), dates of admission, discharge and/or death, hospitalization costs, microbiological results, chest imaging studies, and CURB-65 were analyzed. Antibiotics, admission to Intensive Care Unit (ICU), mechanical ventilation, and pneumonia prognosis were recorded. Patients were dichotomized based on CURB-65 (low- vs. high-risk).

Results:

Among 612 patients (mean age of 70.7 years), 88.4% had at least one comorbidity. Commonly detected pathogens were Acinetobacter baumannii, Pseudomonas aeruginosa, and coagulase-negative staphylococci. Initial monotherapy with β-lactam antibiotics was the most common initial therapy (50%). Mean age, length of stay, hospitalization expenses, ICU admission, mechanical ventilation use, malignancies, and detection rate for P. aeruginosa, and Staphylococcus aureus were higher in the high-risk group compared with the low-risk group. CURB-65 ≥3, malignancies, and mechanical ventilation were associated with an increased mortality. Logistic regression analysis showed that cerebrovascular diseases and being bedridden were independent risk factors for HCAP.

Conclusion:

Initial treatment of HCAP with broad-spectrum antibiotics could be an appropriate approach. CURB-65 ≥3, malignancies, and mechanical ventilation may result in an increased mortality.     

Serial evaluation of the SOFA score to predict outcome in critically ill patients

CONTEXT: Evaluation of trends in organ dysfunction in critically ill patients may help predict outcome. OBJECTIVE: To determine the usefulness of repeated measurement the Sequential Organ Failure Assessment (SOFA) score for prediction of mortality in intensive care unit (ICU) patients. DESIGN: Prospective, observational cohort study conducted from April 1 to July 31, 1999. SETTING: A 31-bed medicosurgical ICU at a university hospital in Belgium. PATIENTS: Three hundred fifty-two consecutive patients (mean age, 59 years) admitted to the ICU for more than 24 hours for whom the SOFA score was calculated on admission and every 48 hours until discharge. MAIN OUTCOME MEASURES: Initial SOFA score (0-24), Delta-SOFA scores (differences between subsequent scores), and the highest and mean SOFA scores obtained during the ICU stay and their correlations with mortality. RESULTS: The initial, highest, and mean SOFA scores correlated well with mortality. Initial and highest scores of more than 11 or mean scores of more than 5 corresponded to mortality of more than 80%. The predictive value of the mean score was independent of the length of ICU stay. In univariate analysis, mean and highest SOFA scores had the strongest correlation with mortality, followed by Delta-SOFA and initial SOFA scores. The area under the receiver operating characteristic curve was largest for highest scores (0.90; SE, 0.02; P<.001 vs initial score). When analyzing trends in the SOFA score during the first 96 hours, regardless of the initial score, the mortality rate was at least 50% when the score increased, 27% to 35% when it remained unchanged, and less than 27% when it decreased. Differences in mortality were better predicted in the first 48 hours than in the subsequent 48 hours. There was no significant difference in the length of stay among these groups. Except for initial scores of more than 11 (mortality rate >90%), a decreasing score during the first 48 hours was associated with a mortality rate of less than 6%, while an unchanged or increasing score was associated with a mortality rate of 37% when the initial score was 2 to 7 and 60% when the initial score was 8 to 11. CONCLUSIONS: Sequential assessment of organ dysfunction during the first few days of ICU admission is a good indicator of prognosis. Both the mean and highest SOFA scores are particularly useful predictors of outcome. Independent of the initial score, an increase in SOFA score during the first 48 hours in the ICU predicts a mortality rate of at least 50%.     

Functional mitral regurgitation combined with increased early diastolic transmitral velocity to early mitral annulus diastolic velocity ratio is associated with a poor prognosis in patients with shock

Background:Functional mitral regurgitation (FMR) is common in critically ill patients and may cause left atrial (LA) pressure elevation. This study aims to explore the prognostic impact of synergistic LA pressure elevation and FMR in patients with shock.Methods:We retrospectively screened 130 consecutive patients of 175 patients with shock from April 2016 to June 2017. The incidence and impact of FMR and early diastolic transmitral velocity to early mitral annulus diastolic velocity ratio (E/e’) ≥ 4 within 6 h of shock on the prognosis of patients were evaluated. Finally, the synergistic effect of FMR and E/e’ were assessed by combination, grouping, and trend analyses.Results:Forty-four patients (33.8%) had FMR, and 15 patients (11.5%) had E/e’ elevation. A multivariate analysis revealed FMR and E/e’ as independent correlated factors for 28-day mortality (P = 0.043 and 0.028, respectively). The Kaplan-Meier survival analysis revealed a significant difference in survival between patients with and without FMR (χ² = 7.672, P = 0.006) and between the E/e’ ≥ 14 and E/e’ < 14 groups (χ² = 19.351, P < 0.010). Twenty-eight-day mortality was significantly different among the four groups (χ² = 30.141, P < 0.010). The risk of 28-day mortality was significantly higher in group 4 (E/e’ ≥ 14 with FMR) compared with groups 1 (E/e’ < 14 without FMR) and 2 (E/e’ < 14 with FMR) (P = 0.001 and 0.046, respectively).Conclusions:Patients with shock can be identified by the presence of FMR. FMR and E/e’ are independent risk factors for a poor prognosis in these patients, and prognosis is worst when FMR and E/e’ ≥ 14 are present. It may be possible to improve prognosis by reducing LA pressure and E/e’.Trial Registration:ClinicalTrials.gov, {"type":"clinical-trial","attrs":{"text":"NCT03082326","term_id":"NCT03082326"}}NCT03082326.     

Combined enteral feeding and total parenteral nutritional support improves outcome in surgical intensive care unit patients

BackgroundFor intensive care unit (ICU) patients with gastrointestinal dysfunction and in need of total parenteral nutrition (TPN) support, the benefit of additional enteral feeding is not clear. This study aimed to investigate whether combined TPN with enteral feeding is associated with better outcomes in surgical intensive care unit (SICU) patients.MethodsClinical data of 88 patients in SICU were retrospectively collected. Variables used for analysis included route and percentage of nutritional support, total caloric intake, age, gender, body weight, body mass index, admission diagnosis, surgical procedure, Acute Physiology and Chronic Health Evaluation (APACHE) II score, comorbidities, length of hospital stay, postoperative complications, blood glucose values and hospital mortality.ResultsWound dehiscence and central catheter infection were observed more frequently in the group of patients receiving TPN calories less than 90% of total calorie intake (p = 0.004 and 0.043, respectively). APACHE II scores were higher in nonsurvivors than in survivors (p = 0.001). More nonsurvivors received TPN calories exceeding 90% of total calorie intake and were in need of dialysis during ICU admission (p = 0.005 and 0.013, respectively). Multivariate analysis revealed that the percentage of TPN calories over total calories and APACHE II scores were independent predictors of ICU mortality in patients receiving supplementary TPN after surgery.ConclusionIn SICU patients receiving TPN, patients who could be fed enterally more than 10% of total calories had better clinical outcomes than patients receiving less than 10% of total calorie intake from enteral feeding. Enteral feeding should be given whenever possible in severely ill patients.     

Correlation between cerebral co-oximetry (rSO2) and outcomes in traumatic brain injury cases: A prospective,observational study

BackgroundTraumatic brain injury (TBI) is known to be an important reason for the increase in disabilities and deaths worldwide. Studies have demonstrated that brain tissue oxygen (PO2) monitoring reduces mortality significantly but is a invasive method of monitoring. Therefore, there is a need to monitor cerebral ischemia in TBI by noninvasive methods. The study aims to correlate cerebral co-oximetry and possible outcomes in patients with TBI.MethodsThe study included 78 patients with TBI admitted in intensive care unit (ICU) with glascow coma scale (GCS) of 8 or less than 8. Near-infrared spectroscopy monitor is applied to the patients immediately after admission to ICU; readings are noted every 4 hours up to first 48 hours, and outcomes studied as survival or neurological deficit are noted at 28 days.ResultsA total of 12 (15.4%) deaths were seen in this study. Survived patients were further divided into good recovery 33 (42.3%), moderate disability 21(26.9%), major disability 8 (10.3%), and persistent vegetative state 4 (5.1%). The rSO2 values in surviving patients were ranging from mean of 60.74% (standard deviation [SD] 4.38) to a mean of 64.98% (SD 5.01), and the mean rSO2 values in patients who died were ranging from a mean of 52.17% (SD 4.11) to a mean of 37.17% (SD 12.48). Lower rSO2 values were correlating significantly with worse neurological outcome or death by using two independent sample t-test (p < 0.001).ConclusionCerebral co-oximetry is a simple noninvasive method for predicting the outcomes in TBI and can be used to guide the management of these patients.     

Sequential Organ Failure Assessment (SOFA) scores differ between genders in a sepsis cohort: Cause or effect?

Abstract

Background. Controversy exists regarding the influence of gender on sepsis events and outcome. Epidemiological data from other countries may not always apply to local circumstances. The aim of this study was to identify gender differences in patient characteristics, treatment, and outcome related to the occurrence of sepsis at admission to the ICU.

Methods. A prospective observational cohort study on patients admitted to the ICU over a 3-year period fulfilling sepsis criteria during the first 24 hours. Demographic data, APACHE II score, SOFA score, TISS 76, aetiology, length of stay (LOS), mortality rate, and aspects of treatment were collected and then analysed with respect to gender differences.

Results. There were no gender-related differences in mortality or length of stay. Early organ dysfunction assessed as SOFA score at admission was a stronger risk factor for hospital mortality for women than for men. This discrepancy was mainly associated with the coagulation sub-score. CRP levels differed between genders in relation to hospital mortality. Infection from the abdominopelvic region was more common among women, whereas infection from skin or skin structures were more common in men.

Conclusion. In this cohort, gender was not associated with increased mortality during a 2-year follow-up period. SOFA score at ICU admission was a stronger risk factor for hospital mortality for women than for men. The discrepancy was mainly related to the coagulation SOFA sub-score. Together with differences in CRP levels this may suggest differences in inflammatory response patterns between genders.     

Characteristics of distractions in the intensive care unit: how serious are they and who are at risk?

INTRODUCTION

Distractions and interruptions of doctor’s work, although common and potentially deleterious in the intensive care unit (ICU), are not well studied.

METHODS

We used a simple observational method to describe the frequency, sources and severity of such distractions, and explore at-risk situations in the ICU. Independent paired observers separately shadowed eight residents and three fellows for 38 sessions (over 100 hrs) in a 20-bed medical ICU.

RESULTS

In total, 444 distractions were noted. Interobserver agreement was excellent at 99.1%. The mean number of distractions/doctor/hr was 4.36 ± 2.27. Median duration of each distraction was 2 mins (interquartile range 2–4 mins; range 1–20 mins). The top three initiators of distractions were other doctors (35.1%), nurses (30.4%) and oneself (18.7%). Of the 444 distractions, 107 (24.1%) were prolonged (lasting ≥ 5 mins), 210 (47.3%) led to a complete pause of current activity and 85 (19.1%) led to complete abandonment of the current activity. On multivariate analysis, physician seniority, time of session and day of week did not predict frequency of distraction. After adjusting for time of session, day of week and type of current activity, urgent distractions (to see another patient, perform immediate procedures or administer medications) and physician juniority were associated with major distractions (complete interruption or termination of current activity), while only urgent distractions were associated with prolonged distractions.

CONCLUSION

Distractions are common in the ICU and junior doctors are particularly susceptible to major distractions.     

Risk factors for hospital admission among COVID-19 patients with diabetes: A study from Saudi Arabia

Objectives:To elucidate the risk factors for hospital admission among COVID-19 patients with type 2 diabetes mellitus (T2DM).Methods:This retrospective study was conducted at the Prince Sultan Military Medical City, Riyadh, Saudi Arabia between May 2020 and July 2020. Out of 7,260 COVID-19 patients, 920 were identified as T2DM. After the exclusion process, 806 patients with T2DM were included in this analysis. Patients’ data were extracted from electronic medical records. A logistic regression model was performed to estimate the risk factors of hospital admission.Results:Of the total of 806 COVID-19 patients with T2DM, 48% were admitted in the hospital, 52% were placed under home isolation. Older age between 70-79 years (OR [odd ratio] 2.56; p=0.017), ≥80 years (OR 6.48; p=0.001) were significantly more likely to be hospitalized compared to <40 years. Similarly, patients with higher HbA1c level of ≥9% compared to <7%; (OR 1.58; p=0.047); patients with comorbidities such as, hypertension (OR 1.43; p=0.048), cardiovascular disease (OR 1.56; p=0.033), cerebrovascular disease (OR 2.38; p=0.016), chronic pulmonary disease (OR 1.51; p=0.018), malignancy (OR 2.45; p=0.025), chronic kidney disease (CKD) IIIa, IIIb, IV (OR 2.37; p=0.008), CKD V (OR 5.07; p=0.007) were significantly more likely to be hospitalized. Likewise, insulin-treated (OR 1.46; p=0.03) were more likely to require hospital admission compared to non-insulin treated patients.Conclusion:Among COVID-19 patients with diabetes, higher age, high HbA1c level, and presence of other comorbidities were found to be significant risk factors for the hospital admission.     

Short- and long-term predictors of spontaneous bacterial peritonitis in Singapore

INTRODUCTIONSpontaneous bacterial peritonitis (SBP) is the commonest complication of liver cirrhosis. Timely and appropriate treatment of SBP is crucial, particularly with the rising worldwide prevalence of multidrug-resistant organisms (MDROs). We aimed to investigate the clinical outcomes of SBP in Singapore.METHODSAll cirrhotic patients with SBP diagnosed between January 2014 and December 2017 were included. Nosocomial SBP (N-SBP) was defined as SBP diagnosed more than 48 hours after hospitalisation. Clinical outcomes were analysed as categorical outcomes using univariate and multivariate analysis.RESULTSThere were 33 patients with 39 episodes of SBP. Their mean age was 64.5 years and 69.7% were male. The commonest aetiology of cirrhosis was hepatitis B (27.3%). The Median Model for End-stage Liver Disease (MELD) score was 17; 33.3% had acute-on-chronic liver failure and 60.6% had septic shock at presentation. N-SBP occurred in 25.6% of SBP cases. N-SBP was more commonly associated with MDROs, previous antibiotic use in the past three months (p = 0.014) and longer length of stay (p = 0.011). The 30-day and 90-day mortality among SBP patients was 30.8% and 51.3%, respectively. MELD score > 20 was a predictor for 30-day mortality. N-SBP and MELD score > 20 were predictors for 90-day mortality.CONCLUSIONN-SBP was significantly associated with recent antibiotic use, longer hospitalisation, more resistant organisms and poorer survival among patients with SBP. N-SBP and MELD score predict higher mortality in SBP. Judicious use of antibiotics may reduce N-SBP and improve survival among cirrhotic patients.     

Underdiagnosis of delirium on admission and prediction of patients who will develop delirium during their inpatient stay: a pilot study

INTRODUCTION

The study aimed to determine the prevalence and documentation of delirium among the elderly and if the Clock Drawing Test (CDT) can be used to predict which patients had delirium on admission and those who may develop delirium during their stay in acute medical wards.

METHODS

A single researcher performed the Mini-Mental State Examination (MMSE) and CDT on admission and discharge of 57 elderly adults at the National University Hospital, Singapore. Delirium was defined as a ≥ 3-point improvement or ≥ 2-point decline in MMSE scores from admission to discharge, where a fall denotes development of delirium and a rise denotes resolution. The case notes of the same patients were reviewed for documentation of delirium. All inpatients from two acute medical wards were examined. One CDT score and a pair of MMSE scores were collected from each patient.

RESULTS

A total of 57 patients (28 male, 29 female) were involved in the study. Their mean age was 76.0 ± 8.7 years. The prevalence of delirium based on MMSE scores was 40.4%; 16 patients had delirium on admission while seven developed delirium during their inpatient stay. However, delirium was documented in the case notes of only 7 (30%) of the 23 patients. CDT score was better than baseline MMSE score at predicting a decline in MMSE score.

CONCLUSION

The prevalence of delirium in the acute medical setting is high but underdiagnosed. The CDT may be a good screening tool to identify patients at risk of delirium during their inpatient stay. Baseline cognition screening should be performed in every elderly patient admitted to hospital.     

The impact of caregiving on caregivers of older persons and its associated factors: a cross-sectional study

INTRODUCTION:Many older people rely on caregivers for support. Caring for older people can pose significant burdens for caregivers yet may also have positive effects. This study aimed to assess the impact on the caregivers and to determine factors associated with caregivers who were burdened.METHODS:This was a cross-sectional study of 385 caregivers of older people who attended a community clinic in Malaysia. Convenience sampling was employed during the study period on caregivers who were aged ≥ 21 years and provided ≥ 4 hours of unpaid support per week. Participants were asked to complete a self-administered questionnaire, which included the Carers of Older People in Europe (COPE) index and the EASYCare Standard 2010 independence score. The COPE index was used to assess the impact of caregiving. A highly burdened caregiver was defined as one whose scores for all three COPE subscales were positive for burden. Care recipients’ independence was assessed using the independence score of the EASYCare Standard 2010 questionnaire. Multiple logistic regression was used to determine the factors associated with caregiver burden.RESULTS:73 (19.0%) caregivers were burdened, of whom two were highly burdened. Caregivers’ median scores on the positive value, negative impact and quality of support scales were 13.0, 9.0 and 12.0, respectively. Care recipients’ median independence score was 18.0. Ethnicity and education levels were found to be associated with caregiver burden.CONCLUSION:Most caregivers gained satisfaction and felt supported in caregiving. Ethnicity and education level were associated with a caregiver being burdened.     

Prediction of patient disposition: comparison of computer and human approaches and a proposed synthesis

ObjectiveTo compare the accuracy of computer versus physician predictions of hospitalization and to explore the potential synergies of hybrid physician–computer models.Materials and MethodsA single-center prospective observational study in a tertiary pediatric hospital in Boston, Massachusetts, United States. Nine emergency department (ED) attending physicians participated in the study. Physicians predicted the likelihood of admission for patients in the ED whose hospitalization disposition had not yet been decided. In parallel, a random-forest computer model was developed to predict hospitalizations from the ED, based on data available within the first hour of the ED encounter. The model was tested on the same cohort of patients evaluated by the participating physicians.Results198 pediatric patients were considered for inclusion. Six patients were excluded due to incomplete or erroneous physician forms. Of the 192 included patients, 54 (28%) were admitted and 138 (72%) were discharged. The positive predictive value for the prediction of admission was 66% for the clinicians, 73% for the computer model, and 86% for a hybrid model combining the two. To predict admission, physicians relied more heavily on the clinical appearance of the patient, while the computer model relied more heavily on technical data-driven features, such as the rate of prior admissions or distance traveled to hospital.DiscussionComputer-generated predictions of patient disposition were more accurate than clinician-generated predictions. A hybrid prediction model improved accuracy over both individual predictions, highlighting the complementary and synergistic effects of both approaches.ConclusionThe integration of computer and clinician predictions can yield improved predictive performance.     

Clinical features and treatment outcomes of human immunodeficiency virus-associated cryptococcal meningitis: a 2-year retrospective analysis

Background:Cryptococcal meningitis (CM) is one of the most common opportunistic infections caused by Cryptococcus neoformans in human immunodeficiency virus (HIV)-infected patients, and is complicated with significant morbidity and mortality. This study retrospectively analyzed the clinical features, characteristics, treatment, and outcomes of first-diagnosed HIV-associated CM after 2-years of follow-up.Methods:Data from all patients (n = 101) of HIV-associated CM hospitalized in Shanghai Public Health Clinical Center from September 2013 to December 2016 were collected and analyzed using logistic regression to identify clinical and microbiological factors associated with mortality.Results:Of the 101 patients, 86/99 (86.9%) of patients had CD4 count <50 cells/mm³, 57/101 (56.4%) were diagnosed at ≥14 days from the onset to diagnosis, 42/99 (42.4%) had normal cerebrospinal fluid (CSF) cell counts and biochemical examination, 30/101 (29.7%) had concomitant Pneumocystis (carinii) jiroveci pneumonia (PCP) on admission and 37/92 (40.2%) were complicated with cryptococcal pneumonia, 50/74 (67.6%) had abnormalities shown on intracranial imaging, amongst whom 24/50 (48.0%) had more than one lesion. The median time to negative CSF Indian ink staining was 8.50 months (interquartile range, 3.25–12.00 months). Patients who initiated antiretroviral therapy (ART) before admission had a shorter time to negative CSF Indian ink compared with ART-naïve patients (7 vs. 12 months, χ² = 15.53, P < 0.001). All-cause mortality at 2 weeks, 8 weeks, and 2 years was 10.1% (10/99), 18.9% (18/95), and 20.7% (19/92), respectively. Coinfection with PCP on admission (adjusted odds ratio [AOR], 3.933; 95% confidence interval [CI], 1.166–13.269, P = 0.027) and altered mental status (AOR, 9.574; 95% CI, 2.548–35.974, P = 0.001) were associated with higher mortality at 8 weeks.Conclusion:This study described the clinical features and outcomes of first diagnosed HIV-associated CM with 2-year follow-up data. Altered mental status and coinfection with PCP predicted mortality in HIV-associated CM.