Predicting abnormal laboratory blood test results in the intensive care unit using novel features based on information theory and historical conditional probability (Preprint)

2021 ◽  
Author(s):  
Camilo E. Valderrama ◽  
Daniel J. Niven ◽  
Henry T. Stelfox ◽  
Joon Lee
Keyword(s):  
Machine Learning ◽  
Intensive Care ◽  
Conditional Probability ◽  
Blood Test ◽  
Laboratory Tests ◽  
Conditional Entropy ◽  
Admission Diagnosis ◽  
Test Results ◽  
Abnormal Laboratory ◽  
Blood Tests

BACKGROUND Redundancy in laboratory blood tests is common in intensive care units (ICU), affecting patients' health and increasing healthcare expenses. Medical communities have made recommendations to order laboratory tests more judiciously. Wise selection can rely on modern data-driven approaches that have been shown to help identify redundant laboratory blood tests in ICUs. However, most of these works have been developed for highly selected clinical conditions such as gastrointestinal bleeding. Moreover, features based on conditional entropy and conditional probability distribution have not been used to inform the need for performing a new test. OBJECTIVE We aimed to address the limitations of previous works by adapting conditional entropy and conditional probability to extract features to predict abnormal laboratory blood test results. METHODS We used an ICU dataset collected across Alberta, Canada which included 55,689 ICU admissions from 48,672 patients with different diagnoses. We investigated conditional entropy and conditional probability-based features by comparing the performances of two machine learning approaches to predict normal and abnormal results for 18 blood laboratory tests. Approach 1 used patients' vitals, age, sex, admission diagnosis, and other laboratory blood test results as features. Approach 2 used the same features plus the new conditional entropy and conditional probability-based features. RESULTS Across the 18 blood laboratory tests, both Approach 1 and Approach 2 achieved a median F1-score, AUC, precision-recall AUC, and Gmean above 80%. We found that the inclusion of the new features statistically significantly improved the capacity to predict abnormal laboratory blood test results in between ten and fifteen laboratory blood tests depending on the machine learning model. CONCLUSIONS Our novel approach with promising prediction results can help reduce over-testing in ICUs, as well as risks for patients and healthcare systems. CLINICALTRIAL N/A

scholarly journals The use of routine blood tests to assist the diagnosis of COVID-19 in symptomatic hospitalized patients

2021 ◽  
pp. 000456322199907
Author(s):  
IT Parsons ◽  
AT Parsons ◽  
E Balme ◽  
G Hazell ◽  
R Gifford ◽  
...  
Keyword(s):  
Cell Count ◽  
Retrospective Review ◽  
Blood Test ◽  
Patient Flow ◽  
White Cell Count ◽  
Area Under The Curve ◽  
White Cell ◽  
Test Results ◽  
Rt Pcr ◽  
Blood Tests

Introduction Specific patterns of blood test results are associated with COVID-19 infection. The aim of this study was to identify which blood tests could be used to assist in diagnosing COVID-19. Method A retrospective review was performed on consecutive patients referred to hospital with a clinical suspicion of COVID-19 over a period of four weeks. The patient’s clinical presentation and severe acute respiratory syndrome coronavirus 2 reverse-transcription polymerase chain reaction (SARS-CoV-2 RT-PCR) were recorded. The patients were divided by diagnosis into COVID (COVID-19 infection) or CONTROL (an alternate diagnosis). A retrospective review of consecutive patients over a further two-week period was used for the purposes of validation. Results Overall, 399 patients (53% COVID, 47% CONTROL) were analysed. White cell count, neutrophils and lymphocytes were significantly lower, while lactate dehydrogenase and ferritin were significantly higher, in the COVID group in comparison to CONTROL. Combining the white cell count, lymphocytes and ferritin results into a COVID Combined Blood Test (CCBT) had an area under the curve of 0.79. Using a threshold CCBT of –0.8 resulted in a sensitivity of 0.85 and a specificity of 0.63. Analysing this against a further retrospective review of 181 suspected COVID-19 patients, using the same CCBT threshold, resulted in a sensitivity of 0.73 and a specificity of 0.75. The sensitivity was comparable to the SARS-CoV-2 RT PCR. Discussion Mathematically combining the blood tests has the potential to assist clinical acumen allowing for rapid streaming and more accurate patient flow pending definitive diagnosis. This may be of particular use in low-resource settings.

Why physicians don't pursue abnormal laboratory tests: An investigation of hypercalcemia and the follow-up of abnormal test results

1984 ◽  
Vol 15 (1) ◽  
pp. 75-78 ◽  
Author(s):  
Kurt Link ◽  
Robert Centor ◽  
David Buchsbaum ◽  
John Witherspoon
Keyword(s):  
Laboratory Tests ◽  
Test Results ◽  
Abnormal Laboratory

scholarly journals Prediction of intensive care unit mortality based on missing events

2021 ◽  
Author(s):  
Tatsuma Shoji ◽  
Hiroshi Yonekura ◽  
Yoshiharu Sato ◽  
yohei Kawashiki
Keyword(s):  
Machine Learning ◽  
Intensive Care Unit ◽  
Intensive Care ◽  
Missing Values ◽  
Integrated Model ◽  
Mortality Prediction ◽  
Sepsis Group ◽  
Admission Diagnosis ◽  
Complementary Subset ◽  
Imputation Approach

Abstract BackgroundThe increasing availability of electronic health records has made it possible to construct and implement models for predicting intensive care unit (ICU) mortality using machine learning. However, the algorithms used are not clearly described, and the performance of the model remains low owing to several missing values, which is unavoidable in big databases.MethodsWe developed an algorithm for subgrouping patients based on missing event patterns using the Philips eICU Research Institute (eRI) database as an example. The eRI database contains data associated with 200,859 ICU admissions from many hospitals (>400) and is freely available. We then constructed a model for each subgroup using random forest classifiers and integrated the models. Finally, we compared the performance of the integrated model with the Acute Physiology and Chronic Health Evaluation (APACHE) scoring system, one of the best known predictors of patient mortality, and the imputation approach-based model.ResultsSubgrouping and patient mortality prediction were separately performed on two groups: the sepsis group (the ICU admission diagnosis of which is sepsis) and the non-sepsis group (a complementary subset of the sepsis group). The subgrouping algorithm identified a unique, clinically interpretable missing event patterns and divided the sepsis and non-sepsis groups into five and seven subgroups, respectively. The integrated model, which comprises five models for the sepsis group or seven models for the non-sepsis group, greatly outperformed the APACHE IV or IVa, with an area under the receiver operating characteristic (AUROC) of 0.91 (95% confidence interval 0.89–0.92) compared with 0.79 (0.76–0.81) for the APACHE system in the sepsis group and an AUROC of 0.90 (0.89–0.91) compared with 0.86 (0.85–0.87) in the non-sepsis group. Moreover, our model outperformed the imputation approach-based model, which had an AUROC of 0.85 (0.83–0.87) and 0.87 (0.86–0.88) in the sepsis and non-sepsis groups, respectively.ConclusionsWe developed a method to predict patient mortality based on missing event patterns. Our method more accurately predicts patient mortality than others. Our results indicate that subgrouping, based on missing event patterns, instead of imputation is essential and effective for machine learning against patient heterogeneity.Trial registrationNot applicable.

scholarly journals Clinical Features of Early Stage COVID-19 in a Primary Care Setting

2021 ◽  
Vol 8 ◽  
Author(s):  
Yohei Kawatani ◽  
Kei Nakayama ◽  
Atsushi Sawamura ◽  
Koichi Fujikawa ◽  
Motoki Nagai ◽  
...  
Keyword(s):  
Primary Care ◽  
Blood Cell ◽  
White Blood Cell ◽  
Blood Test ◽  
Early Stage ◽  
Test Results ◽  
Negative Group ◽  
Initial Screening ◽  
Positive Group ◽  
Blood Tests

Background: The coronavirus disease 2019 (COVID-19) pandemic remains a global healthcare crisis. Nevertheless, the majority of COVID-19 cases involve mild to moderate symptoms in the early stages. The lack of information relating to these cases necessitates further investigation.Methods: Patients visiting the outpatient clinic at the Kamagaya General Hospital were screened by interview and body temperature check. After initial screening, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection was suspected in 481 patients who then underwent blood tests and the loop-mediated isothermal amplification (LAMP) test for SARS-CoV-2. Clinical characteristics between positive and negative SARS-CoV-2 groups were compared. Further, the novel predictive value of routine blood test results for SARS-CoV-2 infection was evaluated using ROC analysis.Results: A total of 15,560 patients visited our hospital during the study period. After exclusion and initial screening by interview, 481 patients underwent the LAMP test and routine blood tests. Of these patients, 69 (14.3%) were positive for SARS-CoV-2 and diagnosed with COVID-19 (positive group), and 412 (85.7%) were negative (negative group). The median period between the first onset of symptoms and visit to our hospital was 3.4 and 2.9 days in the negative and positive groups, respectively. Cough (p = 0.014), rhinorrhea (p = 0.039), and taste disorders (p < 0.001) were significantly more common in the positive group, while gastrointestinal symptoms in the negative group (p = 0.043). The white blood cell count (p < 0.001), neutrophil count (p < 0.001), and percentage of neutrophils (p < 0.001) were higher in the negative group. The percentage of monocytes (p < 0.001) and the levels of ferritin (p < 0.001) were higher in the positive group. As per the predictive values for COVID-19 using blood tests, the values for the area under the curve for the neutrophil-to-monocyte ratio (NMR), white blood cell-to-hemoglobin ratio (WHR), and the product of the two (NMWH) were 0.857, 0.837, and 0.887, respectively.Conclusion: Symptoms in early stage COVID-19 patients were similar to those in previous reports. Some blood test results were not consistent with previous reports. NMR, WHR, and NMWH are novel diagnostic scores in early-stage mild-symptom COVID-19 patients in primary care settings.

scholarly journals Diagnosing brain tumours by routine blood tests using machine learning

2019 ◽  
Vol 9 (1) ◽  
Author(s):  
Simon Podnar ◽  
Matjaž Kukar ◽  
Gregor Gunčar ◽  
Mateja Notar ◽  
Nina Gošnjak ◽  
...  
Keyword(s):  
Machine Learning ◽  
Brain Tumour ◽  
Brain Tumours ◽  
Blood Test ◽  
Neurological Diseases ◽  
Blood Tests ◽  
Routine Blood ◽  
Routine Blood Test ◽  
Tumour Diagnosis ◽  
Neurological Patients

Abstract Routine blood test results are assumed to contain much more information than is usually recognised even by the most experienced clinicians. Using routine blood tests from 15,176 neurological patients we built a machine learning predictive model for the diagnosis of brain tumours. We validated the model by retrospective analysis of 68 consecutive brain tumour and 215 control patients presenting to the neurological emergency service. Only patients with head imaging and routine blood test data were included in the validation sample. The sensitivity and specificity of the adapted tumour model in the validation group were 96% and 74%, respectively. Our data demonstrate the feasibility of brain tumour diagnosis from routine blood tests using machine learning. The reported diagnostic accuracy is comparable and possibly complementary to that of imaging studies. The presented machine learning approach opens a completely new avenue in the diagnosis of these grave neurological diseases and demonstrates the utility of valuable information obtained from routine blood tests.

The correlation of neutrophil/lymphocyte ratio and tumor reduction ratio in preoperative chermoradiotherapy for advanced rectal cancer.

2014 ◽  
Vol 32 (3_suppl) ◽  
pp. 633-633
Author(s):  
Koji Yasuda
Keyword(s):  
Rectal Cancer ◽  
Local Recurrence ◽  
Blood Test ◽  
Local Recurrence Rate ◽  
Lower Rectal Cancer ◽  
Test Results ◽  
Blood Tests ◽  
Tumor Reduction ◽  
Rectal Cancers ◽  
Reduction Effect

633 Background: Preoperative chemoradiotherapy (CRT) is widely used for the treatment of advanced lower rectal cancer and is considered to reduce the local recurrence rate and improve the anus preservation rate. However, the tumor reduction effect varies among patients, and thus far, no factor has yet been found to be effective in the prediction of therapeutic efficacy. Recent studies have shown tumor immunity in the tumor microenvironment to be involved in the anti-tumor effect of radiation and anti-cancer drugs. Among these studies, some show that the neutrophil/lymphocyte (NL) ratio in blood test results correlates with the therapeutic efficacy. We examined the correlation between the NL ratio based on blood test results and tumor-reducing effect of preoperative CRT on advanced lower rectal cancers. Methods: We included 30 advanced lower rectal cancer patients who underwent preoperative CRT during the period ranging from January 2008 to June 2013.For advanced lower rectal cancers classified as cT3orT4NXM0, preoperative CRT was conducted according to the following regimen: 50.4 Gy (1.8 Gy × 28 fr) + UFT (300 mg/day, LV [75 mg/day]) × 28 days. The neutrophil count and lymphocyte count were determined based on the blood tests conducted before and after CRT; the NL ratio was calculated, and the correlation between the value of the ratio and tumor reduction rate was examined. Results: Radical surgery was performed on 28 of the 30 patients who underwent preoperative CRT. One of the 28 patients showed local recurrence (rate, 3.6%). In addition, down stage tumors were found in 16 of the 28 patients, and the down stage rate was 57.1%. Significant differences (p < 0.0001) were found between the tumor reduction effect and NL ratio from blood tests before CRT, as well as between the tumor reduction effect and the ratio from blood tests after CRT. Conclusions: The above findings suggest anti-tumor immunity in the tumor microenvironment, particularly that lymphocytes might be involved in the antitumor effectiveness of preoperative CRT in advanced lower rectal cancers. In addition, NL ratio from blood tests could be a predictive factor of the tumor reduction effect.

Reduction of Laboratory Utilization in the Intensive Care Unit

2016 ◽  
Vol 32 (8) ◽  
pp. 500-507 ◽  
Author(s):  
Samih Raad ◽  
Rachel Elliott ◽  
Evan Dickerson ◽  
Babar Khan ◽  
Khalil Diab
Keyword(s):  
Intensive Care Unit ◽  
Intensive Care ◽  
Laboratory Test ◽  
Laboratory Tests ◽  
Routine Laboratory ◽  
Test Results ◽  
Staff Education ◽  
Arterial Blood ◽  
Laboratory Test Results ◽  
Order Sets

Objective: In our academic intensive care unit (ICU), there is excess ordering of routine laboratory tests. This is partially due to a lack of transparency of laboratory-processing costs and to the admission order plans that favor daily laboratory test orders. We hypothesized that a program that involves physician and staff education and alters the current ICU order sets will lead to a sustained decrease in routine laboratory test ordering. Design: Prospective cohort study. Setting: Academic closed medical ICU (MICU). Patients: All patients admitted to the MICU. Methods: We consistently educated residents, faculty, and staff about laboratory test costs. We removed the daily laboratory test option from the admission order sets and asked residents to order needed laboratory test results every day. We only allowed the G3+I-STAT (arterial blood gas only) cartridges in the MICU in hopes of decreasing duplicative laboratory test results. We added laboratory review to the daily rounding checklist. Measurement and Main Results: Total number of laboratory tests per patient-day decreased from 39.43 to an average of 26.74 ( P <.001) over a 9-month period. The number of iSTAT laboratory tests per patient-day decreased from 7.37 to an average of 1.16 ( P < .001) over the same time period. The number of iSTAT/central laboratory processing duplicative laboratory tests per patient-day decreased from 0.17 to an average of 0.01 ( P < .001). The percentage of patients who have daily laboratory test orders decreased from 100% to an average of 11.94% ( P <. 001). US$123 436 in direct savings and US$258 035 dollars in indirect savings could be achieved with these trends. Intensive care unit morbidity and mortality were not impacted. Conclusion: A simple technique of resident, nursing, and ancillary staff education, combined with alterations in order sets using electronic medical records, can lead to a sustained reduction in laboratory test utilization over time and to significant cost savings without affecting patient safety.

Project of Computer-aided Management of Therapeutical Apheresis

1993 ◽  
Vol 16 (5_suppl) ◽  
pp. 185-186
Author(s):  
C. Falco ◽  
N. Scarpato ◽  
G. Nappi ◽  
S. Formisano
Keyword(s):  
Data Base ◽  
Blood Test ◽  
Great Increase ◽  
Data Entry ◽  
Base System ◽  
Test Results ◽  
Blood Tests ◽  
Therapeutic Protocol ◽  
Computer Aided ◽  
Before And After

The great increase in hemapheresis units activity that occurred during the last years caused the need for a computer-aided management (1, 2). We present a project for a data base system able to manage therapeutical apheresis (3). The program consists of five sections. a) Patient's file card: it allows to record anamnesis, examination and blood test results easily and under computer's guidance. b) Choice of therapeutic protocol: Therapeutic protocol is fixed in this section (device to be used, apheretic method, plasma volume to be processed, blood tests before and after apheresis). c) Procedures: It provides procedure's data entry and guides the operator during the treatment on the ground of therapeutical protocol. d) Data processing: It allows statistics on data placed in the data base. e) Registers: It includes both a general register and the possibility of search by disease, device and method.

scholarly journals Prediction of intensive care unit mortality based on missing events

2021 ◽  
Author(s):  
Tatsuma Shoji ◽  
Hiroshi Yonekura ◽  
Sato Yoshiharu ◽  
Yohei Kawasaki
Keyword(s):  
Machine Learning ◽  
Intensive Care Unit ◽  
Intensive Care ◽  
Missing Values ◽  
Integrated Model ◽  
Mortality Prediction ◽  
Sepsis Group ◽  
Admission Diagnosis ◽  
Complementary Subset ◽  
Imputation Approach

AbstractBackgroundThe increasing availability of electronic health records has made it possible to construct and implement models for predicting intensive care unit (ICU) mortality using machine learning. However, the algorithms used are not clearly described, and the performance of the model remains low owing to several missing values, which is unavoidable in big databases.MethodsWe developed an algorithm for subgrouping patients based on missing event patterns using the Philips eICU Research Institute (eRI) database as an example. The eRI database contains data associated with 200,859 ICU admissions from many hospitals (>400) and is freely available. We then constructed a model for each subgroup using random forest classifiers and integrated the models. Finally, we compared the performance of the integrated model with the Acute Physiology and Chronic Health Evaluation (APACHE) scoring system, one of the best known predictors of patient mortality, and the imputation approach-based model.ResultsSubgrouping and patient mortality prediction were separately performed on two groups: the sepsis group (the ICU admission diagnosis of which is sepsis) and the non-sepsis group (a complementary subset of the sepsis group). The subgrouping algorithm identified a unique, clinically interpretable missing event patterns and divided the sepsis and non-sepsis groups into five and seven subgroups, respectively. The integrated model, which comprises five models for the sepsis group or seven models for the non-sepsis group, greatly outperformed the APACHE IV or IVa, with an area under the receiver operating characteristic (AUROC) of 0.91 (95% confidence interval 0.89–0.92) compared with 0.79 (0.76–0.81) for the APACHE system in the sepsis group and an AUROC of 0.90 (0.89–0.91) compared with 0.86 (0.85–0.87) in the non-sepsis group. Moreover, our model outperformed the imputation approach-based model, which had an AUROC of 0.85 (0.83–0.87) and 0.87 (0.86–0.88) in the sepsis and non-sepsis groups, respectively.ConclusionsWe developed a method to predict patient mortality based on missing event patterns. Our method more accurately predicts patient mortality than others. Our results indicate that subgrouping, based on missing event patterns, instead of imputation is essential and effective for machine learning against patient heterogeneity.Trial registrationNot applicable.

Sign in / Sign up

Export Citation Format

Share Document