Natural language processing to assess the epidemiology of delirium-suggestive behavioural disturbances in critically ill patients

2021 ◽  
Vol 23 (2) ◽  
pp. 144-153
Author(s):  
Marcus Young ◽  
◽  
Natasha Holmes ◽  
Raymond Robbins ◽  
Nada Marhoon ◽  
...  
Keyword(s):  
Natural Language Processing ◽  
Natural Language ◽  
Hospital Mortality ◽  
Critically Ill ◽  
Language Processing ◽  
Critically Ill Patients ◽  
Behavioural Disturbance ◽  
Risk Of Death ◽  
Progress Notes ◽  
Increased Risk

Background: There is no gold standard approach for delirium diagnosis, making the assessment of its epidemiology difficult. Delirium can only be inferred though observation of behavioural disturbance and described with relevant nouns or adjectives. Objective: We aimed to use natural language processing (NLP) and its identification of words descriptive of behavioural disturbance to study the epidemiology of delirium in critically ill patients. Study design: Retrospective study using data collected from the electronic health records of a university-affiliated intensive care unit (ICU) in Melbourne, Australia. Participants: 12 375 patients Intervention: Analysis of electronic progress notes. Identification using NLP of at least one of a list of words describing behavioural disturbance within such notes. Results: We analysed 199 648 progress notes in 12 375 patients. Of these, 5108 patients (41.3%) had NLP-diagnosed behavioural disturbance (NLP-Dx-BD). Compared with those who did not have NLP-Dx-DB, these patients were older, more severely ill, and likely to have medical or unplanned admissions, neurological diagnosis, chronic kidney or liver disease and to receive mechanical ventilation and renal replacement therapy (P < 0.001). The unadjusted hospital mortality for NLP-Dx-BD patients was 14.1% versus 9.6% for patients without NLP-Dx-BD. After adjustment for baseline characteristics and illness severity, NLP-Dx-BD was not associated with increased risk of death (odds ratio [OR], 0.94; 95% CI, 0.80–1.10); a finding robust to multiple sensitivity, subgroups and time of observation subcohort analyses. In mechanically ventilated patients, NLP-Dx-BD was associated with decreased hospital mortality (OR, 0.80; 95% CI, 0.65–0.99) after adjustment for baseline severity of illness and year of admission. Conclusions: NLP enabled rapid assessment of large amounts of data identifying a population of ICU patients with typical high risk characteristics for delirium. Moreover, this technique enabled identification of previously poorly understood associations. Further investigations of this technique appear justified.

scholarly journals Comparison of 2 Natural Language Processing Methods for Identification of Bleeding Among Critically Ill Patients

2018 ◽  
Vol 1 (6) ◽  
pp. e183451 ◽  
Author(s):  
Maxwell Taggart ◽  
Wendy W. Chapman ◽  
Benjamin A. Steinberg ◽  
Shane Ruckel ◽  
Arianna Pregenzer-Wenzler ◽  
...  
Keyword(s):  
Natural Language Processing ◽  
Natural Language ◽  
Critically Ill ◽  
Language Processing ◽  
Critically Ill Patients ◽  
Processing Methods

scholarly journals Biomarker-assisted identification of sepsis-related acute liver impairment: a frequent and deadly condition in critically ill patients

2019 ◽  
Vol 57 (9) ◽  
pp. 1422-1431 ◽  
Author(s):  
Jens-Ulrik Stæhr Jensen ◽  
Lars Peters ◽  
Theis S. Itenov ◽  
Morten Bestle ◽  
Katrin M. Thormar ◽  
...  
Keyword(s):  
Critically Ill ◽  
Critically Ill Patients ◽  
Prognostic Impact ◽  
Surgical Intensive Care ◽  
Positive Interaction ◽  
Liver Impairment ◽  
Risk Of Death ◽  
Increased Risk ◽  
Co Morbidity ◽  
All Cause Mortality

Abstract Background The prognostic impact of mild/moderate liver impairment among critically ill patients is not known. We aimed to determine whether acute liver impairment, as measured by several biomarkers, (i) is frequent, (ii) influences prognosis and (iii) to determine whether such an effect is specific for infected critically ill patients. Methods A biomarker and clinical cohort study based on a randomized controlled trial. All-cause mortality was the primary endpoint. Biomarkers hyaluronic acid (HA), bilirubin, albumin, alkaline phosphatase and the international normalized ratio (INR) were determined. Multivariable statistics were applied to estimate risk increase according to liver biomarker increase at baseline and the model was adjusted for age, APACHE II, severe sepsis/septic shock vs. milder infection, chronic alcohol abuse Charlson’s co-morbidity index, cancer disease, surgical or medical patient, body mass index, sex, estimated glomerular filtration rate, mechanical ventilation and the other biomarkers. Time-to-event graphs were used. The patients were critically ill patients (n = 1096) from nine mixed medical/surgical intensive care units without known hepatobiliary disease. Results HA levels differed between infected patients (median 210.8 ng/mL [IQR: 93.2–556.6]) vs. the non-infected (median 56.8 ng/mL [IQR: 31.9–116.8], p < 0.001). Serum HA quartiles 2, 3 and 4 were independent predictors of 90-day all-cause mortality for the entire population (infected and non-infected). However, the signal was driven by the infected patients (positive interaction test, no signal in non-infected patients). Among infected patients, HA quartiles corresponded directly to the 90-day risk of dying: 1st quartile: 57/192 = 29.7%, 2nd quartile: 84/194 = 43.3%, 3rd quartile: 90/193 = 46.6%, 4th quartile: 101/192 = 52.3 %, p for trend: <0.0001. This finding was confirmed in adjusted analyses: hazard ratio vs. 1st quartile: 2nd quartile: 1.3 [0.9–1.8], p = 0.14, 3rd quartile: 1.5 [1.1–2.2], p = 0.02, 4th quartile: 1.9 [1.3–2.6], p < 0.0001). High bilirubin was also an independent predictor of mortality. Conclusions Among infected critically ill patients, subtle liver impairment, (elevated HA and bilirubin), was associated with a progressive and highly increased risk of death for the patient; this was robust to adjustment for other predictors of mortality. HA can identify patients at high risk.

Increased risk of death and readmission after hospital discharge of critically ill patients in a developing country: a retrospective multicenter cohort study

2018 ◽  
Vol 44 (7) ◽  
pp. 1090-1096 ◽  
Author(s):  
Vanessa Chaves Barreto Ferreira de Lima ◽  
Ana Luiza Bierrenbach ◽  
Gizelton Pereira Alencar ◽  
Ana Lucia Andrade ◽  
Luciano Cesar Pontes Azevedo
Keyword(s):  
Cohort Study ◽  
Developing Country ◽  
Hospital Discharge ◽  
Critically Ill ◽  
Critically Ill Patients ◽  
Multicenter Cohort Study ◽  
Risk Of Death ◽  
Multicenter Cohort ◽  
Increased Risk

scholarly journals Association between Malnutrition and Hospital Mortality and Duration of Intensive Care Unit Admission in the Critically Ill: A Prospective Cohort Study

2017 ◽  
Author(s):  
Charles Chin Han Lew ◽  
Gabriel Jun Yung Wong ◽  
Ka Po Cheung ◽  
Ai Ping Chua ◽  
Mary Foong Fong Chong ◽  
...  
Keyword(s):  
Intensive Care Unit ◽  
Cohort Study ◽  
Intensive Care ◽  
Hospital Mortality ◽  
Critically Ill ◽  
Prospective Cohort Study ◽  
Critically Ill Patients ◽  
Prospective Cohort ◽  
Unit Length ◽  
Increased Risk

There is limited evidence for the association between malnutrition and hospital mortality as well as Intensive Care Unit length-of-stay (ICU-LOS) in critically ill patients. We aimed to examine the aforementioned associations by conducting a prospective cohort study in an ICU of a Singapore tertiary hospital. Between August 2015 and October 2016, all adult patients with &ge;24 h of ICU-LOS were included. The 7-point Subjective Global Assessment (7-point SGA) was used to determine patients&rsquo; nutritional status within 48 hours of ICU admission. Multivariate analyses were conducted in two ways: 1) presence versus absence of malnutrition, and 2) dose-dependent association for each 1-point decrease in the 7-point SGA. There were 439 patients of which 28.0% were malnourished, and 29.6% died before hospital discharge. Malnutrition was associated with an increased risk of hospital mortality [adjusted-RR 1.39 (95%CI: 1.10&ndash;1.76)], and this risk increased with a greater degree of malnutrition [adjusted-RR 1.09 (95%CI: 1.01&ndash;1.18) for each 1-point decrease in the 7-point SGA]. No significant association was found between malnutrition and ICU-LOS. Conclusion: There was a clear association between malnutrition and higher hospital mortality in critically ill patients. The association between malnutrition and ICU-LOS could not be replicated and hence requires further evaluation.

scholarly journals The Association between Nutritional Status and In-Hospital Mortality of COVID-19 in Critically-Ill Patients in the ICU

Nutrients ◽  
2021 ◽  
Vol 13 (10) ◽  
pp. 3302
Author(s):  
Michał Czapla ◽  
Raúl Juárez-Vela ◽  
Vicente Gea-Caballero ◽  
Stanisław Zieliński ◽  
Marzena Zielińska
Keyword(s):  
Intensive Care Unit ◽  
Intensive Care ◽  
Nutritional Status ◽  
Hospital Mortality ◽  
Critically Ill ◽  
Critically Ill Patients ◽  
Invasive Mechanical Ventilation ◽  
Nutritional Risk ◽  
Risk Of Death ◽  
The Impact

Background: Coronavirus disease 2019 (COVID-19) has become one of the leading causes of death worldwide. The impact of poor nutritional status on increased mortality and prolonged ICU (intensive care unit) stay in critically ill patients is well-documented. This study aims to assess how nutritional status and BMI (body mass index) affected in-hospital mortality in critically ill COVID-19 patients Methods: We conducted a retrospective study and analysed medical records of 286 COVID-19 patients admitted to the intensive care unit of the University Clinical Hospital in Wroclaw (Poland). Results: A total of 286 patients were analysed. In the sample group, 8% of patients who died had a BMI within the normal range, 46% were overweight, and 46% were obese. There was a statistically significantly higher death rate in men (73%) and those with BMIs between 25.0–29.9 (p = 0.011). Nonsurvivors had a statistically significantly higher HF (Heart Failure) rate (p = 0.037) and HT (hypertension) rate (p < 0.001). Furthermore, nonsurvivors were statistically significantly older (p < 0.001). The risk of death was higher in overweight patients (HR = 2.13; p = 0.038). Mortality was influenced by higher scores in parameters such as age (HR = 1.03; p = 0.001), NRS2002 (nutritional risk score, HR = 1.18; p = 0.019), PCT (procalcitonin, HR = 1.10; p < 0.001) and potassium level (HR = 1.40; p = 0.023). Conclusions: Being overweight in critically ill COVID-19 patients requiring invasive mechanical ventilation increases their risk of death significantly. Additional factors indicating a higher risk of death include the patient’s age, high PCT, potassium levels, and NRS ≥ 3 measured at the time of admission to the ICU.

scholarly journals The triglyceride-glucose index played an important role in prognosis in critically ill patients with heart disease

2021 ◽  
Author(s):  
Guangyao Zhai ◽  
Biyang Zhang ◽  
Jianlong Wang ◽  
Yuyang Liu ◽  
Yujie Zhou
Keyword(s):  
Heart Disease ◽  
Hospital Mortality ◽  
Critically Ill ◽  
Critically Ill Patients ◽  
Binary Logistic Regression ◽  
Independent Effect ◽  
Binary Logistic Regression Analysis ◽  
Triglyceride Glucose Index ◽  
Increased Risk ◽  
Patients With Diabetes

Abstract Background As an alternative method to evaluate insulin resistance (IR), triglyceride-glucose index (TyG) was shown to be related to the severity and prognosis of cardiovascular diseases. The main objective of this study was to explore the association between TyG and in-hospital mortality in critically ill patients with heart disease Method: TyG was calculated as previously reported: ln [fasting TGs (mg/dL) * FBG (mg/dL)/2]. All patients were divided into four different categories based on TyG quartiles. Primary outcome was in-hospital mortality. Binary logistic regression analysis was performed to determine the independent effect of TyG. Result 4839 critically ill patients with heart disease were included. In-hospital mortality increased as TyG quartiles increased (Quartile 4 vs Quartile 1: 12.1 vs 5.3, P < 0.001). Even after adjusting for confounding variables, TyG was still independently associated with the increased risk of in-hospital mortality in critically ill patients with heart disease (Quartile 4 vs Quartile 1: OR, 95% CI: 2,43, 1.79–3.31, P < 0.001, P for trend < 0.001). However, we did not observe the association between increased TyG and the risk of mortality in patients with diabetes. Furthermore, as TyG quartiles increased, the length of intensive care unit (ICU) stay was prolonged (Quartile 4 vs Quartile 1: 2.3, 1.3–4.9 vs 2.1, 1.3–3.8, P = 0.007). And the significant interactions were not found in most subgroups. Conclusion TyG was independently correlated with in-hospital mortality in critically ill patients with heart disease.

scholarly journals Natural language processing and recurrent network models for identifying genomic mutation-associated cancer treatment change from patient progress notes

JAMIA Open ◽  
2019 ◽  
Vol 2 (1) ◽  
pp. 139-149 ◽  
Author(s):  
Meijian Guan ◽  
Samuel Cho ◽  
Robin Petro ◽  
Wei Zhang ◽  
Boris Pasche ◽  
...  
Keyword(s):  
Machine Learning ◽  
Natural Language Processing ◽  
Natural Language ◽  
Cancer Patients ◽  
Language Processing ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Free Text ◽  
Treatment Change ◽  
Progress Notes

Abstract Objectives Natural language processing (NLP) and machine learning approaches were used to build classifiers to identify genomic-related treatment changes in the free-text visit progress notes of cancer patients. Methods We obtained 5889 deidentified progress reports (2439 words on average) for 755 cancer patients who have undergone a clinical next generation sequencing (NGS) testing in Wake Forest Baptist Comprehensive Cancer Center for our data analyses. An NLP system was implemented to process the free-text data and extract NGS-related information. Three types of recurrent neural network (RNN) namely, gated recurrent unit, long short-term memory (LSTM), and bidirectional LSTM (LSTM_Bi) were applied to classify documents to the treatment-change and no-treatment-change groups. Further, we compared the performances of RNNs to 5 machine learning algorithms including Naive Bayes, K-nearest Neighbor, Support Vector Machine for classification, Random forest, and Logistic Regression. Results Our results suggested that, overall, RNNs outperformed traditional machine learning algorithms, and LSTM_Bi showed the best performance among the RNNs in terms of accuracy, precision, recall, and F1 score. In addition, pretrained word embedding can improve the accuracy of LSTM by 3.4% and reduce the training time by more than 60%. Discussion and Conclusion NLP and RNN-based text mining solutions have demonstrated advantages in information retrieval and document classification tasks for unstructured clinical progress notes.

The dose - response association between fluid overload and hospital mortality in critically ill patients: a multicenter, prospective, observational cohort study

2020 ◽  
Author(s):  
Meiping Wang ◽  
Bo Zhu ◽  
Li Jiang ◽  
Ying Wen ◽  
Bin Du ◽  
...  
Keyword(s):  
Hospital Mortality ◽  
Critically Ill ◽  
Dose Response ◽  
Fluid Balance ◽  
Critically Ill Patients ◽  
Fluid Overload ◽  
Fluid Management ◽  
Management Strategies ◽  
Icu Admission ◽  
Increased Risk

Abstract Background Fluid management is important for ensuring hemodynamic stability in critically ill patients but easily leads to fluid overload. However, the optimal fluid balance plot or range for critically ill patients is unknown. This study aimed to explore the dose-response relationship between fluid overload (FO) and hospital mortality in critically ill patients.Methods Data were derived from the China Critical Care Sepsis Trial (CCCST). Patients with sequential fluid data for the first 3 days of admission to the ICU were included. FO was expressed as the ratio of the cumulative fluid balance (L) and initial body weight (kg) at ICU admission as a percentage. Maximum fluid overload (MFO) was defined as the peak FO value during the first 3 days of ICU admission. We used logistic regression models with restricted cubic splines to assess the relationship between MFO and the risk of hospital mortality.ResultsIn total, 3850 patients were included, 929 (24.1%) of whom died in hospital. For each 1% L/kg increase in the FO, the risk of hospital mortality increased by 4% (HR 1.04, 95% CI 1.03 - 1.05, P < 0.001). FO greater than 10% was associated with a 44% increased HR of hospital mortality compared with FO less than 5% (HR 1.44, 95% CI 1.27 - 1.67). Notably, we also found a non-linear dose-response association between MFO and hospital mortality.Conclusions Both higher and lower fluid balance were associated with an increased risk of hospital mortality. Further studies should explore this relationship and seek for the optimal fluid management strategies for critically ill patients.

scholarly journals The Prognostic Value of Brain Dysfunction in Critically Ill Patients with and without Sepsis: A Post Hoc Analysis of the ICON Audit

2021 ◽  
Vol 11 (5) ◽  
pp. 530
Author(s):  
Ilaria Crippa ◽  
Fabio Taccone ◽  
Xavier Wittebole ◽  
Ignacio Martin-Loeches ◽  
Mary Schroeder ◽  
...  
Keyword(s):  
Hospital Mortality ◽  
Critically Ill ◽  
Critically Ill Patients ◽  
Brain Function ◽  
Brain Dysfunction ◽  
Hospital Death ◽  
Post Hoc Analysis ◽  
Increased Risk ◽  
Brain Failure ◽  
Post Hoc

Brain dysfunction is associated with poor outcome in critically ill patients. In a post hoc analysis of the Intensive Care over Nations (ICON) database, we investigated the effect of brain dysfunction on hospital mortality in critically ill patients. Brain failure was defined as a neurological sequential organ failure assessment (nSOFA) score of 3–4, based on the assumed Glasgow Coma Scale (GCS) score. Multivariable analyses were performed to assess the independent roles of nSOFA and change in nSOFA from admission to day 3 (ΔnSOFA) for predicting hospital mortality. Data from 7192 (2096 septic and 5096 non-septic) patients were analyzed. Septic patients were more likely than non-septic patients to have brain failure on admission (434/2095 (21%) vs. 617/4665 (13%), p < 0.001) and during the ICU stay (625/2063 (30%) vs. 736/4665 (16%), p < 0.001). The presence of sepsis (RR 1.66 (1.31–2.09)), brain failure (RR 4.85 (3.33–7.07)), and both together (RR 5.61 (3.93–8.00)) were associated with an increased risk of in-hospital death, but nSOFA was not. In the 3280 (46%) patients in whom ΔnSOFA was available, sepsis (RR 2.42 (1.62–3.60)), brain function deterioration (RR 6.97 (3.71–13.08)), and the two together (RR 10.24 (5.93–17.67)) were associated with an increased risk of in-hospital death, whereas improvement in brain function was not.

scholarly journals Dose–response association between fluid overload and in-hospital mortality in critically ill patients: a multicentre, prospective, observational cohort study

BMJ Open ◽  
2020 ◽  
Vol 10 (12) ◽  
pp. e039875
Author(s):  
Meiping Wang ◽  
Bo Zhu ◽  
Li Jiang ◽  
Ying Wen ◽  
Bin Du ◽  
...  
Keyword(s):  
Hospital Mortality ◽  
Critically Ill ◽  
Dose Response ◽  
Fluid Balance ◽  
Critically Ill Patients ◽  
Fluid Overload ◽  
Multivariable Analysis ◽  
Main Diagnosis ◽  
Icu Admission ◽  
Increased Risk

ObjectivesFluid management is important in ensuring haemodynamic stability in critically ill patients, but can easily lead to fluid overload (FO). However, the optimal fluid balance plot or range for critically ill patients is unknown. This study aimed to explore the dose–response relationship between FO and in-hospital mortality in critically ill patients.DesignMulticentre, prospective, observational study.SettingEighteen intensive care units (ICUs) of 16 tertiary hospitals in China.ParticipantsCritically ill patients in the ICU for more than 3 days.Primary outcome measures and analysesFO was defined as the ratio of the cumulative fluid balance (L) and initial body weight (kg) on ICU admission, expressed as a percentage. Maximum FO was defined as the peak value of FO during the first 3 days of ICU admission. Logistic regression models with restricted cubic splines were used to explore the pattern and magnitude of the association between maximum FO and risk of in-hospital mortality. Age, sex, Acute Physiology and Chronic Health Evaluation II score, Sequential Organ Failure Assessment score on admission, main diagnosis on admission to ICU, comorbidities, time of maximum FO, mechanical ventilation, renal replacement therapy, use of vasopressors and centres were adjusted in multivariable analysis.ResultsA total of 3850 patients were included in the study, 929 (24.1%) of whom died in the hospital. For each 1% L/kg increase in maximum FO, the risk of in-hospital mortality increased by 4% (adjusted HR (aHR) 1.04, 95% CI 1.03 to 1.05, p<0.001). A maximum FO greater than 10% was associated with a 44% increased HR of in-hospital mortality compared with an FO less than 5% (aHR 1.44, 95% CI 1.27 to 1.67). Notably, we found a non-linear dose–response association between maximum FO and in-hospital mortality.ConclusionsBoth higher and negative fluid balance levels were associated with an increased risk of in-hospital mortality in critically ill patients.Trial registration numberChiCTR-ECH-13003934.

Sign in / Sign up

Export Citation Format

Share Document