scholarly journals NLP Methods for Extraction of Symptoms from Unstructured Data for Use in Prognostic COVID-19 Analytic Models

2021 ◽  
Vol 72 ◽  
pp. 429-474
Author(s):  
Greg M. Silverman ◽  
Himanshu S. Sahoo ◽  
Nicholas E. Ingraham ◽  
Monica Lupei ◽  
Michael A. Puskarich ◽  
...  
Keyword(s):  
Hospital Mortality ◽  
Clinical Decision Making ◽  
Quality Care ◽  
Vital Signs ◽  
Clinical Decision ◽  
Prognostic Models ◽  
P Value ◽  
Rule Based ◽  
Presenting Symptoms ◽  
Increased Risk

Statistical modeling of outcomes based on a patient's presenting symptoms (symptomatology) can help deliver high quality care and allocate essential resources, which is especially important during the COVID-19 pandemic. Patient symptoms are typically found in unstructured notes, and thus not readily available for clinical decision making. In an attempt to fill this gap, this study compared two methods for symptom extraction from Emergency Department (ED) admission notes. Both methods utilized a lexicon derived by expanding The Center for Disease Control and Prevention's (CDC) Symptoms of Coronavirus list. The first method utilized a word2vec model to expand the lexicon using a dictionary mapping to the Uni ed Medical Language System (UMLS). The second method utilized the expanded lexicon as a rule-based gazetteer and the UMLS. These methods were evaluated against a manually annotated reference (f1-score of 0.87 for UMLS-based ensemble; and 0.85 for rule-based gazetteer with UMLS). Through analyses of associations of extracted symptoms used as features against various outcomes, salient risks among the population of COVID-19 patients, including increased risk of in-hospital mortality (OR 1.85, p-value < 0.001), were identified for patients presenting with dyspnea. Disparities between English and non-English speaking patients were also identified, the most salient being a concerning finding of opposing risk signals between fatigue and in-hospital mortality (non-English: OR 1.95, p-value = 0.02; English: OR 0.63, p-value = 0.01). While use of symptomatology for modeling of outcomes is not unique, unlike previous studies this study showed that models built using symptoms with the outcome of in-hospital mortality were not significantly different from models using data collected during an in-patient encounter (AUC of 0.9 with 95% CI of [0.88, 0.91] using only vital signs; AUC of 0.87 with 95% CI of [0.85, 0.88] using only symptoms). These findings indicate that prognostic models based on symptomatology could aid in extending COVID-19 patient care through telemedicine, replacing the need for in-person options. The methods presented in this study have potential for use in development of symptomatology-based models for other diseases, including for the study of Post-Acute Sequelae of COVID-19 (PASC).

scholarly journals Hemoglobin A1c and preoperative glycemia as a decision tool to help minimise sternal wound complications: a retrospective study in OPCAB patients

2021 ◽  
Vol 16 (1) ◽  
Author(s):  
Jef Van den Eynde ◽  
Abel Van Vlasselaer ◽  
Annoushka Laenen ◽  
Delphine Szecel ◽  
Bart Meuris ◽  
...  
Keyword(s):  
Retrospective Study ◽  
Hemoglobin A1c ◽  
Clinical Decision Making ◽  
Artery Bypass ◽  
Clinical Decision ◽  
Wound Complications ◽  
Minimal Risk ◽  
Off Pump ◽  
Increased Risk ◽  
Diabetes Patients

Abstract Background Poor glycemic control has been associated with an increased risk of wound complications after various types of operations. However, it remains unclear how hemoglobin A1c (HbA1c) and preoperative glycemia can be used in clinical decision-making to prevent sternal wound complications (SWC) following off-pump coronary artery bypass grafting (OPCAB). Methods We conducted a retrospective study of 1774 consecutive patients who underwent OPCAB surgery between January 2010 and November 2016. A new four-grade classification for SWC was used. The associations of HbA1c and preoperative glycemia with incidence and grade of SWC were analysed using logistic regression analysis and proportional odds models, respectively. Results During a median follow-up of 326 days (interquartile range (IQR) 21–1261 days), SWC occurred in 133/1316 (10%) of non-diabetes and 82/458 (18%) of diabetes patients (p < 0.001). Higher HbA1c was significantly associated with a higher incidence of SWC (odds ratio, OR 1.24 per 1% increase, 95% confidence interval, CI 1.04;1.48, p = 0.016) as well as a higher grade of SWC (OR 1.25, 95% CI 1.06;1.48, p = 0.010). There was no association between glycemia and incidence (p = 0.539) nor grade (p = 0.607) of SWC. Significant modifiers of these effects were found: HbA1c was associated with SWC in diabetes patients younger than 70 years (OR 1.41, 95% CI 1.17;1.71, p < 0.001), whereas it was not in those older than 70 years. Glycemia was associated with SWC in patients who underwent non-urgent surgery (OR 2.48, 95% CI 1.26;4.88, p = 0.009), in diabetes patients who received skeletonised grafts (OR 4.83, 95% CI 1.28;18.17, p = 0.020), and in diabetes patients with a BMI < 30 (OR 2.19, 95% CI 1.01;4.76, p = 0.047), whereas it was not in the counterparts of these groups. Conclusions Under certain conditions, HbA1c and glycemia are associated SWC following OPCAB. These findings are helpful in planning the procedure with minimal risk of SWC.

scholarly journals Choice of faecal immunochemical test matters: comparison of OC-Sensor and HM-JACKarc, in the assessment of patients at high risk of colorectal cancer

2020 ◽  
Vol 0 (0) ◽  
Author(s):  
Caroline J. Chapman ◽  
Ayan Banerjea ◽  
David J Humes ◽  
Jaren Allen ◽  
Simon Oliver ◽  
...  
Keyword(s):  
Colorectal Cancer ◽  
Clinical Decision Making ◽  
Full Range ◽  
Clinical Decision ◽  
Detection Rates ◽  
Faecal Immunochemical Test ◽  
Specimen Collection ◽  
Increased Risk ◽  
Immunochemical Test

AbstractObjectivesCurrently, NICE recommends the use of faecal immunochemical test (FIT) at faecal haemoglobin concentrations (f-Hb) of 10 μg Hb/g faeces to stratify for colorectal cancer (CRC) risk in symptomatic populations. This f-Hb cut-off is advised across all analysers, despite the fact that a direct comparison of analyser performance, in a clinical setting, has not been performed.MethodsTwo specimen collection devices (OC-Sensor, OC-S; HM-JACKarc, HM-J) were sent to 914 consecutive individuals referred for follow up due to their increased risk of CRC. Agreement of f-Hb around cut-offs of 4, 10 and 150 µg Hb/g faeces and CRC detection rates were assessed. Two OC-S devices were sent to a further 114 individuals, for within test comparisons.ResultsA total of 732 (80.1%) individuals correctly completed and returned two different FIT devices, with 38 (5.2%) CRCs detected. Median f-Hb for individuals diagnosed with and without CRC were 258.5 and 1.8 µg Hb/g faeces for OC-S and 318.1 and 1.0 µg Hb/g faeces for HM-J respectively. Correlation of f-Hb results between OC-S/HM-J over the full range was rho=0.74, p<0.001. Using a f-Hb of 4 µg Hb/g faeces for both tests found an agreement of 88.1%, at 10 µg Hb/g faeces 91.7% and at 150 µg Hb/g faeces 96.3%. A total of 114 individuals completed and returned two OC-S devices; correlation across the full range was rho=0.98, p<0.001.ConclusionsWe found large variations in f-Hb when different FIT devices were used, but a smaller variation when the same FIT device was used. Our data suggest that analyser-specific f-Hb cut-offs are applied with regard to clinical decision making, especially at lower f-Hb.

A Prediction Model for Positive Infant Meconium and Urine Drug Tests

2020 ◽  
Author(s):  
Elizabeth A. Simpson ◽  
David A. Skoglund ◽  
Sarah E. Stone ◽  
Ashley K. Sherman
Keyword(s):  
Prediction Model ◽  
Drug Testing ◽  
Clinical Decision Making ◽  
Drugs Of Abuse ◽  
Clinical Decision ◽  
P Value ◽  
Test Characteristics ◽  
Drug Test ◽  
Urine Drug ◽  
Urine Drug Test

Objective This study aimed to determine the factors associated with positive infant drug screen and create a shortened screen and a prediction model. Study Design This is a retrospective cohort study of all infants who were tested for drugs of abuse from May 2012 through May 2014. The primary outcome was positive infant urine or meconium drug test. Multivariable logistic regression was used to identify independent risk factors. A combined screen was created, and test characteristics were analyzed. Results Among the 3,861 live births, a total of 804 infants underwent drug tests. Variables associated with having a positive infant test were (1) positive maternal urine test, (2) substance use during pregnancy, (3) ≤ one prenatal visit, and (4) remote substance abuse; each p-value was less than 0.0001. A model with an indicator for having at least one of these four predictors had a sensitivity of 94% and a specificity of 69%. Application of this screen to our population would have decreased drug testing by 57%. No infants had a positive urine drug test when their mother's urine drug test was negative. Conclusion This simplified screen can guide clinical decision making for determining which infants should undergo drug testing. Infant urine drug tests may not be needed when a maternal drug test result is negative. Key Points

scholarly journals Task Shifting: The Use of Laypersons for Acquisition of Vital Signs Data for Clinical Decision Making in the Emergency Room Following Traumatic Injury

2017 ◽  
Vol 41 (12) ◽  
pp. 3066-3073 ◽  
Author(s):  
Bryce E. Haac ◽  
Jared R. Gallaher ◽  
Charles Mabedi ◽  
Andrew J. Geyer ◽  
Anthony G. Charles
Keyword(s):  
Decision Making ◽  
Emergency Room ◽  
Clinical Decision Making ◽  
Traumatic Injury ◽  
Vital Signs ◽  
Clinical Decision ◽  
Task Shifting

scholarly journals Procedural and remote outcome among patients undergoing urgent trans-catheter aortic valve implantation

2020 ◽  
Vol 41 (Supplement_2) ◽  
Author(s):  
A Berkovitch ◽  
A Segev ◽  
A Finkelstein ◽  
R Kornowski ◽  
H Danenberg ◽  
...  
Keyword(s):  
Multivariate Analysis ◽  
Hospital Mortality ◽  
Mortality Rates ◽  
P Value ◽  
Complication Rates ◽  
Increased Risk ◽  
Procedural Complication ◽  
Elective Procedure ◽  
Aortic Valve Implantation ◽  
Valve Implantation

Abstract Background Severe aortic stenosis patients suffer frequent heart failure decompensations events often requiring hospitalization. In extreme situations patients can be found with pulmonary edema and cardiogenic shock, unresponsive to medical treatment. Urgent trans-catheter aortic valve implantation (TAVI) has emerged as a treatment option for these high-risk patients. Methods We investigated 3,599 patients undergoing TAVI. Subjects were divided into two groups based on procedure urgency: patients who were electively hospitalized for the procedure (N=3,448) and those who had an urgent TAVI (N=151). Peri-procedural complications were documented according to the VARC-2 criteria. In hospital and 1-year mortality rates were prospectively documented. Results Mean age of the study population was 82±7, of whom 52% were female. Peri-procedural complication rates was significantly higher among patients with an urgent indication for TAVI compared to those having an elective procedure: valve malposition 3.6% vs. 0.6% (p-value=0.023), valve migration 3.2% vs. 0.9% (p-value=0.016), post procedure myocardial infarction 3.7% vs. 0.3% (p-value=0.004), and stage 3 acute kidney injury 2.6% vs. 0.5%, (p-value=0.02). Univariate analysis found that patients with urgent indication for TAVI had significantly higher in hospital mortality (5.8% vs. 1.4%, p-value&lt;0.001). similarly, multivariate analysis adjusted for age, gender and cardio-vascular risk factors found that patients with urgent indication had more than 5-folds increased risk of in-hospital mortality (OR 5.94, 95% CI 2.28–15.43, p-value&lt;0.001). Kaplan-Meier's survival analysis showed that patients undergoing urgent TAVI had higher 1-year mortality rates compared to patients undergoing an elective TAVI procedure (p-value log-rank&lt;0.001, Figure). Multivariate analysis found they had more than 2-folds increased risk of mortality at 1-year (HR 2.27, 95% CI 1.53–3.38, p&lt;0.001 compared to those having an elective procedure. Conclusions Patients with urgent indication for TAVI have higher in-hospital mortality and higher peri-procedural complication rates. However, if these patients survive the index hospitalization, they enjoy good prognosis. Kaplan-Meier's survival analysis Funding Acknowledgement Type of funding source: None

scholarly journals Understanding the usefulness of prognostic models in clinical decision-making

2017 ◽  
Vol 63 (2) ◽  
pp. 121-125 ◽  
Author(s):  
Adrian C Traeger ◽  
Markus Hübscher ◽  
James H McAuley
Keyword(s):  
Decision Making ◽  
Clinical Decision Making ◽  
Clinical Decision ◽  
Prognostic Models

scholarly journals Inpatient Outcomes and Rates of Venous Thromboembolism and Gastrointestinal Bleeding in Patients with Hepatocellular Carcinoma: Results from Nationwide Inpatient Sample Database 2007-2016

Blood ◽  
2019 ◽  
Vol 134 (Supplement_1) ◽  
pp. 2162-2162
Author(s):  
Kamelah Abushalha ◽  
Sawsan Abulaimoun ◽  
Ryan Walters ◽  
Sara Albagoush ◽  
Hussain I Rangoonwala ◽  
...  
Keyword(s):  
Hepatocellular Carcinoma ◽  
Venous Thromboembolism ◽  
Gastrointestinal Bleeding ◽  
Hospital Mortality ◽  
Hospital Cost ◽  
Clinical Decision Making ◽  
Conflicts Of Interest ◽  
Increased Risk ◽  
Increasing Trend ◽  
Increasing Trends

Background: Patients with hepatocellular carcinoma (HCC) are at an increased risk for developing venous thromboembolism (VTE)- mainly portal venous thrombosis (PVT). Malignancy and liver cirrhosis ( 80%-90% of HCC cases are related to cirrhosis) are conditions that can perturb the hemostatic balance towards a prothrombotic state. Also, these patients with HCC are at high risk for gastrointestinal bleeding (GIB), making thromboprophylaxis and anticoagulation a treatment challenge. Additional information regarding the outcomes and severity of both VTE and GIB in patients with HCC would be useful to guide clinical decision-making Aim: To determine the rates, inpatient mortality, length of stay (LOS) and hospital cost of VTE and GIB-related admissions in patients with hepatocellular carcinoma. Method: We used ICD-9-CM and ICD-10-CM codes to identify hospitalizations from 2007 to 2016 that included HCC with primary discharge diagnoses of GIB or VTE. Linear trends in the rate of GIB and VTE, as well as in-hospital mortality, LOS, and inflation-adjusted hospital cost (in 2016 US dollars), were evaluated using Daniel's test; piecewise slopes were used as needed. All analyses accounted for the NIS sampling design with updated hospital trend weights used as appropriate. SAS v. 9.4 was used for all analyses. Results: Between 2007 and 2016, we identified 6,527,871 hospitalizations with HCC and a primary discharge diagnosis of GIB (3,517,059; 53.9%) or VTE (3,010,812; 46.1%). From 2007 to 2010, a decreasing trend was observed in the rate of GIB diagnoses (55.5% to 51.6%, ptrend < .001), whereas an increasing trend was observed for VTE diagnoses (44.5% to 48.4%, ptrend < .001). By contrast, from 2010 to 2016, an increasing trend was observed in GIB (51.6% to 55.2%, ptrend < .001), whereas a decreasing trend was observed in VTE (48.4% to 44.8%, ptrend < .001). For in-hospital mortality, a decreasing trend was observed for GIB (2.3% to 1.9%, ptrend < .001), whereas a decreasing trend was observed in VTE until 2012 (1.8% to 1.5%, ptrend < .001), after which no trend was indicated (1.5% to 1.6%, ptrend = .337). Although decreasing trends in LOS were observed for GIB (3.4 days to 3.2 days, ptrend < .001) and VTE (4.3 days to 3.3 days, ptrend < .001), increasing trends were observed for inflation-adjusted hospital cost for both GIB ($6,996 to $7,707, ptrend < .001) and VTE ($7,283 to $7,584, ptrend = .048). Conclusion: In this NIS cohort of hospitalized patients with HCC, GIB was more frequently observed than VTE. Trends observed in the rates of GIB and VTE went in opposite directions. In general decreasing trends were observed in in-hospital mortality and LOS for both VTE and GIB. By contrast, increasing trends were observed in the hospital cost for both diagnoses. Clinicians should balance benefits against risks when deciding VTE prophylaxis and treatment in patients with HCC. Future studies are needed to determine the ideal agent and specific dosages to treat HCC-associated VTE. Disclosures No relevant conflicts of interest to declare.

scholarly journals Validation of parsimonious prognostic models for patients infected with COVID-19

2021 ◽  
Vol 28 (1) ◽  
pp. e100267
Author(s):  
Keerthi Harish ◽  
Ben Zhang ◽  
Peter Stella ◽  
Kevin Hauck ◽  
Marwa M Moussa ◽  
...  
Keyword(s):  
Clinical Decision Making ◽  
Validation Cohort ◽  
External Validation ◽  
Clinical Decision ◽  
Prognostic Models ◽  
Clinical Prediction ◽  
Inclusion Criteria ◽  
Local Data ◽  
External Data ◽  
Operator Curve

ObjectivesPredictive studies play important roles in the development of models informing care for patients with COVID-19. Our concern is that studies producing ill-performing models may lead to inappropriate clinical decision-making. Thus, our objective is to summarise and characterise performance of prognostic models for COVID-19 on external data.MethodsWe performed a validation of parsimonious prognostic models for patients with COVID-19 from a literature search for published and preprint articles. Ten models meeting inclusion criteria were either (a) externally validated with our data against the model variables and weights or (b) rebuilt using original features if no weights were provided. Nine studies had internally or externally validated models on cohorts of between 18 and 320 inpatients with COVID-19. One model used cross-validation. Our external validation cohort consisted of 4444 patients with COVID-19 hospitalised between 1 March and 27 May 2020.ResultsMost models failed validation when applied to our institution’s data. Included studies reported an average validation area under the receiver–operator curve (AUROC) of 0.828. Models applied with reported features averaged an AUROC of 0.66 when validated on our data. Models rebuilt with the same features averaged an AUROC of 0.755 when validated on our data. In both cases, models did not validate against their studies’ reported AUROC values.DiscussionPublished and preprint prognostic models for patients infected with COVID-19 performed substantially worse when applied to external data. Further inquiry is required to elucidate mechanisms underlying performance deviations.ConclusionsClinicians should employ caution when applying models for clinical prediction without careful validation on local data.

scholarly journals Vital signs assessed in initial clinical encounters predict COVID-19 mortality in an NYC hospital system

2020 ◽  
Vol 10 (1) ◽  
Author(s):  
Elza Rechtman ◽  
Paul Curtin ◽  
Esmeralda Navarro ◽  
Sharon Nirenberg ◽  
Megan K. Horton
Keyword(s):  
Machine Learning ◽  
New York ◽  
Clinical Decision Making ◽  
Care Delivery ◽  
Vital Signs ◽  
Clinical Decision ◽  
Rapid Identification ◽  
Gradient Boosting ◽  
Hospital System ◽  
Response Strategies

AbstractTimely and effective clinical decision-making for COVID-19 requires rapid identification of risk factors for disease outcomes. Our objective was to identify characteristics available immediately upon first clinical evaluation related COVID-19 mortality. We conducted a retrospective study of 8770 laboratory-confirmed cases of SARS-CoV-2 from a network of 53 facilities in New-York City. We analysed 3 classes of variables; demographic, clinical, and comorbid factors, in a two-tiered analysis that included traditional regression strategies and machine learning. COVID-19 mortality was 12.7%. Logistic regression identified older age (OR, 1.69 [95% CI 1.66–1.92]), male sex (OR, 1.57 [95% CI 1.30–1.90]), higher BMI (OR, 1.03 [95% CI 1.102–1.05]), higher heart rate (OR, 1.01 [95% CI 1.00–1.01]), higher respiratory rate (OR, 1.05 [95% CI 1.03–1.07]), lower oxygen saturation (OR, 0.94 [95% CI 0.93–0.96]), and chronic kidney disease (OR, 1.53 [95% CI 1.20–1.95]) were associated with COVID-19 mortality. Using gradient-boosting machine learning, these factors predicted COVID-19 related mortality (AUC = 0.86) following cross-validation in a training set. Immediate, objective and culturally generalizable measures accessible upon clinical presentation are effective predictors of COVID-19 outcome. These findings may inform rapid response strategies to optimize health care delivery in parts of the world who have not yet confronted this epidemic, as well as in those forecasting a possible second outbreak.

scholarly journals Prediction modelling of inpatient neonatal mortality in high-mortality settings

2020 ◽  
pp. archdischild-2020-319217
Author(s):  
Jalemba Aluvaala ◽  
Gary Collins ◽  
Beth Maina ◽  
Catherine Mutinda ◽  
Mary Waiyego ◽  
...  
Keyword(s):  
Hospital Mortality ◽  
High Mortality ◽  
Clinical Decision Making ◽  
Clinical Signs ◽  
Prognostic Models ◽  
Routine Practice ◽  
Low Resource ◽  
C Statistic ◽  
Temporal Validation ◽  
Time Of Admission

ObjectivePrognostic models aid clinical decision making and evaluation of hospital performance. Existing neonatal prognostic models typically use physiological measures that are often not available, such as pulse oximetry values, in routine practice in low-resource settings. We aimed to develop and validate two novel models to predict all cause in-hospital mortality following neonatal unit admission in a low-resource, high-mortality setting.Study design and settingWe used basic, routine clinical data recorded by duty clinicians at the time of admission to derive (n=5427) and validate (n=1627) two novel models to predict in-hospital mortality. The Neonatal Essential Treatment Score (NETS) included treatments prescribed at the time of admission while the Score for Essential Neonatal Symptoms and Signs (SENSS) used basic clinical signs. Logistic regression was used, and performance was evaluated using discrimination and calibration.ResultsAt derivation, c-statistic (discrimination) for NETS was 0.92 (95% CI 0.90 to 0.93) and that for SENSS was 0.91 (95% CI 0.89 to 0.93). At external (temporal) validation, NETS had a c-statistic of 0.89 (95% CI 0.86 to 0.92) and SENSS 0.89 (95% CI 0.84 to 0.93). The calibration intercept for NETS was −0.72 (95% CI −0.96 to −0.49) and that for SENSS was −0.33 (95% CI −0.56 to −0.11).ConclusionUsing routine neonatal data in a low-resource setting, we found that it is possible to predict in-hospital mortality using either treatments or signs and symptoms. Further validation of these models may support their use in treatment decisions and for case-mix adjustment to help understand performance variation across hospitals.

Sign in / Sign up

Export Citation Format

Share Document