A comparison of the Full Outline of Unresponsiveness (FOUR) and Glasgow Coma Scale (GCS) Scores in Predicting Mortality Among Patients with Reduced Level of Consciousness in Uganda

The Glasgow Coma Scale and Full Outline of Unresponsiveness score evaluation to predict patient outcomes with neurological illnesses in intensive care units in West Bank: a prospective cross-sectional study

BACKGROUND

Determining the clinical neurological state of the patient is essential for making decisions and forecasting results. The Glasgow Coma Scale and the Full Outline of Unresponsiveness (FOUR) Scale are commonly used tools for measuring behavioral consciousness. This study aims to compare scales among patients with neurological disorders in intensive care units (ICUs) in the West Bank.

METHODS

A prospective cross-sectional design was employed. All patients admitted to ICUs who met inclusion criteria were involved in this study. Data were collected from from An-Najah National University, Al-Watani, and Rafedia Hospital. Both tools were used to collect data.

RESULTS

A total of 84 patients were assessed, 69.0% of the patients were male, and the average length of stay was 6.4 days. The mean score on the Glasgow Coma scale was 11.2 on admission 11.6 after 48 hours, and 12.2 on discharge. The mean FOUR Scale score was 12.2 on admission, 12.4 after 48 hours, and 12.5 at discharge.

CONCLUSIONS

This study indicates that both the Glasgow Coma Scale and the FOUR scale are effective in predicting outcomes for neurologically deteriorated critically ill patients. However, the FOUR scale proved to be more reliable when assessing outcomes in ICU patients.

Full Outline of Unresponsiveness Score versus Glasgow Coma Scale in Predicting Clinical Outcomes in Altered Mental Status

Introduction

Full outline of unresponsiveness (FOUR) score has advantages over Glasgow Coma Scale (GCS); as it can be used in intubated patients and provides greater neurological details. It has been studied mainly in the trauma and neuroscience setting. Our primary objective was to compare the FOUR versus GCS score as predictors of mortality at 30 days and poor functional outcome at 3 months among nontrauma patients in the emergency department (ED).

Methods

This prospective observational study was conducted on adult patients presenting with altered mental status (duration <7 days) in the ED (March 2019-November 2020). Data collection included demographic and clinical features, the GCS and FOUR scores, the feasibility of acquiring and interpreting FOUR on a Likert scale, duration of hospital stay, 30-day mortality, and functional outcome at 3 months on the modified Rankin Scale. Trained emergency medicine residents managing the patient collected the data. The area under receiver's operating characteristics curve (AUROC) was used to compare the accuracy of the GCS and FOUR scores in predicting outcomes. The FOUR score equivalent of GCS cutoffs for categorizing neurological impairment (mild, moderate, and severe) was also investigated.

Results

Two hundred and ninety-one patients were included, with a mean age of 50.3 years and 67.4% males. Most patients (40.2%) had altered mental status for 1-3 days and hepatic encephalopathy was the most common ED diagnosis. The mortality at 30 days was 66.7% (194 of 291), and 88% (256 of 291) of patients had poor functional outcomes at 3 months. The AUROCs for predicting 30-day mortality were similar for both the scores (GCS: 0.70, FOUR: 0.71, and the P value for difference: 0.9). Similarly, the AUROCs for predicting 3-month poor functional outcome were 0.683 and 0.669 using GCS and FOUR, respectively, with a nonsignificant difference (P = 0.82). The FOUR score strata of 14-16, 11-13, and 0-10 were found to be equivalent to the GCS scores of 13-15 (mild), 9-12 (moderate), and 3-8 (severe). The feasibility of acquiring and interpreting GCS and FOUR scores on the Likert scale was found to be "easy."

Conclusion

The FOUR score is similar to GCS in predicting mortality at 30 days and poor neurological outcomes at 3 months among nontrauma patients of ED. Moreover, it was found that the FOUR score is "easy" to assess and interpret by the emergency residents.

Explainable artificial intelligence model for mortality risk prediction in the intensive care unit: a derivation and validation study

BACKGROUND

The lack of transparency is a prevalent issue among the current machine-learning (ML) algorithms utilized for predicting mortality risk. Herein, we aimed to improve transparency by utilizing the latest ML explicable technology, SHapley Additive exPlanation (SHAP), to develop a predictive model for critically ill patients.

METHODS

We extracted data from the Medical Information Mart for Intensive Care IV database, encompassing all intensive care unit admissions. We employed nine different methods to develop the models. The most accurate model, with the highest area under the receiver operating characteristic curve, was selected as the optimal model. Additionally, we used SHAP to explain the workings of the ML model.

RESULTS

The study included 21 395 critically ill patients, with a median age of 68 years (interquartile range, 56-79 years), and most patients were male (56.9%). The cohort was randomly split into a training set (N = 16 046) and a validation set (N = 5349). Among the nine models developed, the Random Forest model had the highest accuracy (87.62%) and the best area under the receiver operating characteristic curve value (0.89). The SHAP summary analysis showed that Glasgow Coma Scale, urine output, and blood urea nitrogen were the top three risk factors for outcome prediction. Furthermore, SHAP dependency analysis and SHAP force analysis were used to interpret the Random Forest model at the factor level and individual level, respectively.

CONCLUSION

A transparent ML model for predicting outcomes in critically ill patients using SHAP methodology is feasible and effective. SHAP values significantly improve the explainability of ML models.

Comparison of the Prognostic Accuracy of Full Outline of Unresponsiveness (FOUR) Score with Glasgow Coma Scale (GCS) Score among Patients with Traumatic Brain Injury in a Tertiary Care Center

Objectives  The Glasgow Coma Scale (GCS) is widely used and considered the gold standard in assessing the consciousness of patients with traumatic brain injury. However, some significant limitations, like the considerable variations in interobserver reliability and predictive validity, were the reason for developing the Full Outline of Unresponsiveness (FOUR) score. The current study aims to compare the prognostic accuracy of the FOUR score with the GCS score for in-hospital mortality and morbidity among patients with traumatic brain injury. Materials and Methods  A prospective cohort study was conducted, where 237 participants were selected by consecutive sampling from a tertiary care center. These patients were assessed with the help of GCS and FOUR scores within 6 hours of admission, and other clinical parameters were also noted. The level of consciousness was checked every day with the help of GCS and FOUR scores until their last hospitalization day. Glasgow Outcome Scale was used to assess their outcome on the last day of hospitalization. The GCS and FOUR scores were compared, and data were analyzed by descriptive and inferential statistics. The chi-square test, independent Student's t -test, and receiver operating characteristic analysis were used for inferential analysis. Results  The area under the curve (AUC) for the GCS score at the 6th hour for predicting mortality was 0.865 with a cutoff value of 5.5, and it yields a sensitivity of 87% and a specificity of 64%. The AUC for FOUR scores at the 6th hour for predicting the mortality was 0.893, with a cutoff value of 5.5, and it yields a sensitivity of 87% and a specificity of 73%. Conclusion  The current study shows that, as per the AUC of GCS and FOUR scores, their sensitivity was equal, but specificity was higher in the FOUR score. So, the FOUR score has higher accuracy than the GCS score in the prediction of mortality among traumatic brain injury patients.

Effect of protocol-based family visitation on physiological indicators in ICU patients: a randomized controlled trial

BACKGROUND

Intensive care unit (ICU) patients often experience significant physiological stress. This study evaluated the effect of a defined family visitation protocol on physiological responses in the ICU.

METHODS

A randomized, block-randomized clinical trial was conducted on 78 ICU patients at Imam Reza Hospital between February 8, 2017, and August 8, 2017. The intervention group received protocol-based visits, and the control group continued with standard visitation. Block randomization was utilized for group assignments. The primary outcome was the measurement of physiological signs using designated monitoring devices. Data were analyzed using SPSS version 22, employing independent t-tests, Mann-Whitney U test, repeated measures analysis, and Friedman's test.

RESULTS

The results showed no significant differences in systolic blood pressure, diastolic blood pressure, mean arterial pressure, respiratory rate, and arterial blood oxygen levels between the two groups. However, heart rate in the intervention group was significantly lower in three stages before, during, and after the meaningful visiting (P = 0.008).

CONCLUSION

Protocol-based scheduled family visits in the ICU may reduce physiological stress, as evidenced by a decrease in patients' heart rate. Implementing tailored visitation protocols sensitive to patient preferences and clinical contexts is advisable, suggesting the integration of family visits into standard care practices for enhanced patient outcomes.

TRIAL REGISTRATION

IRCT20161229031654N2; 25/01/2018; Iranian Registry of Clinical Trials ( https://en.irct.ir ).

Etiologies and factors associated with mortality in patients with non-traumatic coma in a tertiary hospital in Bukavu, eastern Democratic Republic of the Congo

Non-traumatic coma (NTC) is a common medical condition often associated with poor outcomes. Identifying underlying causes is crucial for effective management and prognostication, particularly in resource-poor settings. This study aimed to identify the most common causes and prognostic factors of NTC in a tertiary hospital in Bukavu, in the eastern Democratic Republic of the Congo (DRC), using the Glasgow Coma Scale (GCS) as well as other simple and affordable clinical and paraclinical tools. This retrospective observational study included 219 consecutive patients admitted to the intensive care unit of the Provincial General Hospital of Bukavu between January 2016 and December 2018. Sociodemographic, clinical, and laboratory data were also collected. Bivariate and multivariate analyses were performed to identify different causes and factors associated with poor outcomes in these patients. The median age of the patients was 49 (interquartile range [IQR]: 33-61) years, and they were predominantly men (62.8%). The most common causes of NTC were stroke (25.7%), acute metabolic complications of diabetes (21.9%), and primary brain infections (meningoencephalitis, 16.0%; and cerebral malaria, 14.2%). The NTC-related in-hospital mortality rate was 35.2%. A high mortality was significantly and independently associated with a GCS<7 (adjusted odds ratio [OR]: 4.30, 95% confidence interval [CI]: 1.73-10.71), the presence of meningismus at clinical evaluation (adjusted odds ratio [aOR] 3.86, 95%CI: 1.41-10.55), oxygen saturation <90% (aOR 3.99, 95%CI: 1.71-9.28), the consumption of traditional herbal medicines prior to hospital admission (aOR 2.82, 95%CI: 1.16-6.86), and elevated serum creatinine (aOR 1.64, 95%CI: 1.17-2.29). In conclusion, clinical neurological examinations, along with simple and affordable paraclinical investigations, can provide sufficient information to determine the etiology of NTC and evaluate the prognosis of comatose patients, even in resource-poor settings. Physicians may use the GCS as a simple, reliable, and affordable tool to identify patients who require special attention and care.

Prediction of early prognosis after traumatic brain injury by multifactor model

AIMS

To design a model to predict the early prognosis of patients with traumatic brain injury (TBI) based on parameters that can be quickly obtained in emergency conditions from medical history, physical examination, and supplementary examinations.

METHODS

The medical records of TBI patients who were hospitalized in two medical institutions between June 2015 and June 2021 were collected and analyzed. Patients were divided into the training set, validation set, and testing set. The possible predictive indicators were screened after analyzing the data of patients in the training set. Then prediction models were found based on the possible predictive indicators in the training set. Data of patients in the validation set and the testing set was provided to validate the predictive values of the models.

RESULTS

Age, Glasgow coma scale score, Apolipoprotein E genotype, damage area, serum C-reactive protein, and interleukin-8 (IL-8) levels, and Marshall computed tomography score were found associated with early prognosis of TBI patients. The accuracy of the early prognosis prediction model (EPPM) was 80%, and the sensitivity and specificity of the EPPM were 78.8% and 80.8% in the training set. The accuracy of the EPPM was 79%, and the sensitivity and specificity of the EPPM were 66.7% and 86.2% in the validation set. The accuracy of the early EPPM was 69.1%, and the sensitivity and specificity of the EPPM were 67.9% and 77.8% in the testing set.

CONCLUSION

Prediction models integrating general information, clinical manifestations, and auxiliary examination results may provide a reliable and rapid method to evaluate and predict the early prognosis of TBI patients.

Application of interpretable machine learning for early prediction of prognosis in acute kidney injury

Background

This study aimed to develop an algorithm using the explainable artificial intelligence (XAI) approaches for the early prediction of mortality in intensive care unit (ICU) patients with acute kidney injury (AKI).

Methods

This study gathered clinical data with AKI patients from the Medical Information Mart for Intensive Care IV (MIMIC-IV) in the US between 2008 and 2019. All the data were further randomly divided into a training cohort and a validation cohort. Seven machine learning methods were used to develop the models for assessing in-hospital mortality. The optimal model was selected based on its accuracy and area under the curve (AUC). The SHapley Additive exPlanation (SHAP) values and Local Interpretable Model-Agnostic Explanations (LIME) algorithm were utilized to interpret the optimal model.

Results

A total of 22,360 patients with AKI were finally enrolled in this study (median age, 69.5 years; female, 42.8%). They were randomly split into a training cohort (16770, 75%) and a validation cohort (5590, 25%). The eXtreme Gradient Boosting (XGBoost) model achieved the best performance with an AUC of 0.890. The SHAP values showed that Glasgow Coma Scale (GCS), blood urea nitrogen, cumulative urine output on Day 1 and age were the top 4 most important variables contributing to the XGBoost model. The LIME algorithm was used to explain the individualized predictions.

Conclusions

Machine-learning models based on clinical features were developed and validated with great performance for the early prediction of a high risk of death in patients with AKI.

TRADUÇÃO, ADAPTAÇÃO E VALIDAÇÃO DA ESCALA FULL OUTLINE OF UNRESPONSIVENESS PARA O PORTUGUÊS DO BRASIL

RESUMO Objetivo: traduzir, adaptar culturalmente e validar a escala Full Outline of UnResponsiveness para o português do Brasil. Método: estudo metodológico realizado no Hospital de Clínicas da Universidade Federal do Triângulo Mineiro, Uberaba, Brasil, por meio das etapas: tradução, síntese, avaliação pelo comitê de especialistas, retrotradução, consenso, avaliação semântica e pré-teste. Alcançou-se uma amostra de 188 pacientes adultos. A coleta de dados ocorreu entre agosto e dezembro de 2020. Analisou-se a validade de critério concorrente comparando a escala Full Outline of UnResponsiveness com a Escala de Coma de Glasgow por meio dos coeficientes de correlação de Spearman e Pearson, e a validade preditiva com a Regressão de Cox, Sensibilidade e Especificidade e Área Sob a Curva Receiver Operating Characteristic. Adotaram-se, também, o alfa de Cronbach e os coeficientes Kappa ponderado e de Correlação Intraclasse para a confiabilidade interobservador. Resultados: o teste de Spearman para os itens resposta motora e ocular, respectivamente, resultou-se em 0,81 e 0,96, e o de Pearson para o escore total em 0,97. Obteve-se um risco relativo de 0,80, especificidade de 95,5%, sensibilidade de 51,6% e acurácia de 0,80 (IC95%: 0,688-0,905, p<0,001). O alfa de Cronbach foi de 0,94, o Kappa ponderado variou entre 0,89 e 1,0 e o ICC resultou em 0,99. Conclusão: a escala Full Outline of UnResponsiveness - versão brasileira, manteve quatro domínios e os 20 itens da escala original, tornando-se apropriada para utilização no Brasil e contribuindo para a avaliação do nível de consciência e prognóstico de pacientes adultos em condição grave.

TRANSLATION, ADAPTATION AND VALIDATION OF THE FULL OUTLINE OF UNRESPONSIVENESS SCALE INTO BRAZILIAN PORTUGUESE

ABSTRACT Objective: to translate, culturally adapt and validate the Full Outline of UnResponsiveness scale into Brazilian Portuguese. Method: a methodological study carried out at the Clinical Hospital of Universidade Federal do Triângulo Mineiro, Uberaba, Brazil, through the following stages: translation, synthesis, evaluation by the experts' committee, back-translation, consensus, semantic evaluation and pre-test. A sample of 188 adult patients was reached. Data collection took place between August and December 2020. Concurrent criterion validity was analyzed by comparing the Full Outline of UnResponsiveness scale with the Glasgow Coma Scale by means of Spearman's and Pearson's correlation coefficients; and predictive validity analysis was performed with Cox Regression, Sensitivity and Specificity and Area Under the Receiver Operating Characteristic Curve. The Cronbach's alpha, weighted Kappa and Intraclass Correlation coefficients were also adopted for interobserver reliability. Results: Spearman’s test for the motor and eye response items, respectively, resulted in 0.81 and 0.96, and Pearson's test for the total score was 0.97. A relative risk of 0.80, 95.5% specificity, 51.6% sensitivity and accuracy of 0.80 (95% CI: 0.688-0,905, p<0.001) were obtained. Cronbach's alpha was 0.94, weighted Kappa varied from 0.89 to 1.0, and ICC resulted in 0.99. Conclusion: the Full Outline of UnResponsiveness scale (Brazilian version), maintained four domains and the 20 items from the original scale, making it appropriate for use in Brazil and contributing to the assessment of the level of consciousness and prognosis of adult patients in severe conditions.

Mbarara University Research Training Initiative: Experiences and Accomplishments of the MEPI Junior D43 TW010128 Award in Uganda

OBJECTIVE

In 2015, Mbarara University of Science and Technology was awarded the Mbarara University Research Training Initiative (MURTI) under grant number D43 TW010128 to build capacity of junior faculty to become the next generation of researchers in Africa. In this paper, we document the experiences and achievements of the research capacity building efforts at MUST.

METHODS

We conducted a descriptive evaluation study which involved document review and in-depth interviews. We used "Reach" and 'Effectiveness' from the RE-AIM framework to guide the document review, and the organizational theory of implementation effectiveness to guide the in-depth interviews.

RESULTS

In the MURTI program, we conducted 17 short courses between August 2015 and July 2021, a total of 6597 attendances were recorded. The most attended courses were responsible conduct of research (n = 826), qualitative research methods (n = 744), and data management (n = 613). Thirty-three fellows were recruited and funded to conduct mentored research leading to 48 publications and 14 extramural grant applications were yielded. From the in-depth interviews, the participants appreciated the research training program, the enhanced research skills attained, and the institutional capacity built. They attributed the success of the program to the training approach of using short courses, readiness of the junior faculty to change, and the supportive environment by the mentors and trainers in the program.

CONCLUSION

The D43 TW010128 research training grant-built capacity for the junior faculty at MUST, enhanced their research skills, promoted research capacity institutionally and provided career development for the junior faculty. This was possible due to the positive attitude of the junior faculty (organizational readiness) to change and the supportive environment (mentors and trainers) during implementation of the grant. These two factors provided a favorable institutional climate that guaranteed success of the funding goals.

Prognostic Utility of Daily Changes in Glasgow Coma Scale and the Full Outline of Unresponsiveness Score Measurement in Patients with Metabolic Encephalopathy, Central Nervous System Infections and Stroke in Uganda

BACKGROUND

Metabolic encephalopathy (ME), central nervous system (CNS) infections, and stroke are common causes of reduced level of consciousness in Uganda. However, the prognostic utility of changes in the daily measurements of the Full Outline of Unresponsiveness (FOUR) score and Glasgow Coma Scale (GCS) score in these specific disorders is not known.

METHODS

We conducted secondary analyses of data from patients who presented with reduced level of consciousness due to CNS infections, stroke, or ME to a tertiary hospital in Uganda. Patients had FOUR/GCS scores at admission and at 24 and 48 h. We calculated a change in FOUR score (ΔFOUR) and change in GCS score (ΔGCS) at 24 and 48 h and used logistic regression models to determine whether these changes were predictive of 30-day mortality. In addition, we determined the prognostic utility of adding the admission score to the 24-h ΔFOUR and 24-h ΔGCS on mortality.

RESULTS

We analyzed data from 230 patients (86 with ME, 79 with CNS infections, and 65 with stroke). The mean (SD) age was 50.8 (21.3) years, 27% (61 of 230) had HIV infection, and 62% (134 of 230) were peasant farmers. ΔFOUR at 24 h was predictive of mortality among those with ME (odds ratio [OR] 0.64 [95% confidence interval {CI} 0.48-0.84]; p = 0.001) and those with CNS infections (OR 0.65 [95% CI 0.48-0.87]; p = 0.004) but not in those with stroke (OR 1.0 [95% CI 0.73-1.38]; p = 0.998). However, ΔGCS at 24 h was only predictive of mortality in the ME group (OR 0.69 [95% CI 0.56-0.86]; p = 0.001) and not in the CNS or stroke group. This 24-h ΔGCS and ΔFOUR pattern was similar at 48 h in all subgroups. The addition of an admission score to either 24-h ΔFOUR or 24-h ΔGCS significantly improved the predictive ability of the scores in those with stroke and CNS infection but not in those with ME.

CONCLUSIONS

Twenty-four-hour and 48-h ΔFOUR and ΔGCS are predictive of mortality in Ugandan patients with CNS infections and ME but not in those with stroke. For individuals with stroke, the admission score plays a more significant predictive role that the change in scores.