Hiroshi Ogura

The inter-event times of frequent words in real written texts have been shown to fall into three distributional classes: exponential, q-exponential, and power-law. Here, we define the inter-event times as the number of sentences between successive appearances of a given word. In a previous study, we proposed a pseudo-text generation algorithm in which each sentence is composed by one of two word-selection mechanisms, i.e., randomized selection and priority-based selection, and we further showed that the algorithm can reproduce synthetic texts in which words exhibiting the three distributional classes coexist as in real written texts. The present paper develops a revised and mathematically clearer formulation of that algorithm and introduces approximate stochastic models corresponding the algorithm to explain why the three classes emerge. For the randomized selection mechanism involved in the algorithm, we propose a corresponding stochastic model that approximately brings geometric inter-event times which can be regarded as a discrete exponential law. In the model, the exponential law can be derived when the local occurrence probability of a given word is supposed to vary slowly. For the priority-based selection mechanism, we introduce a coarse-grained local occurrence probability in a corresponding stochastic model and show that a mixture of geometric distributions proposed by the model naturally yields heavy-tailed behavior, including power-law asymptotics. The q-exponential regime is then interpreted as an intermediate regime produced by temporal variation in local occurrence probabilities and by the coexistence of the two selection mechanisms. These findings provide a logically smoother connection between the previous empirical study and the present theoretical development of the text generation algorithms and the corresponding stochastic models.

Background: Severe trauma remains a leading cause of admission to the intensive care unit. The Trauma and Injury Severity Score (TRISS) is an established standard for predicting outcomes and benchmarking the quality of trauma care globally. However, the TRISS model has some limitations when used for benchmarking trauma care. Objective: This study aimed to determine whether machine learning–derived trauma phenotypes can complement the TRISS via multivariable modeling to improve in-hospital death prediction. We also introduce “Trauma-Vis,” a freely accessible web-based platform, to facilitate the availability of this integrated assessment approach to clinicians. Methods: In this retrospective cohort study using the nationwide Japan Trauma Data Bank (JTDB), which encompasses data from 303 hospitals in Japan, we divided the data chronologically into a derivation cohort (JTDB 2015-2018) and a temporal validation cohort (JTDB 2019-2022). An integrated model was developed using multivariable logistic regression, incorporating the logit-transformed TRISS-predicted mortality and the assigned trauma phenotypes. After applying the exclusion criteria, 87,882 patients with blunt trauma were analyzed in the derivation cohort and 80,964 in the validation cohort. Predictive performance was evaluated using the area under the receiver operating characteristic curve, Brier score, logarithmic loss, net reclassification improvement, integrated discrimination improvement, and decision curve analysis. Results: In the derivation cohort, multivariable modeling demonstrated that trauma phenotype classification significantly recalibrated mortality risk; multiple phenotypes exhibited significant independent associations with in-hospital death after adjusting for baseline TRISS predictions (eg, for phenotype 8: odds ratio 2.38, 95% CI 2.11-2.68; P<.001). In the temporal validation cohort, the integrated multivariable model yielded higher performance metrics than the baseline TRISS model: the area under the receiver operating characteristic curve increased from 0.889 to 0.897 (DeLong test, P<.001), Brier score improved from 0.0454 to 0.0394, and logarithmic loss decreased from 0.1670 to 0.1458. The integrated model demonstrated a calibration intercept of –0.152 and a slope of 0.965 and provided a higher net benefit in the decision curve analysis across evaluated threshold probabilities. Conclusions: Integrating machine learning–derived trauma phenotypes with the TRISS via multivariable modeling improved the accuracy and utility of in-hospital death prediction. The developed “Trauma-Vis” platform demonstrates the technical feasibility of providing real-time risk stratification at the bedside.

In this paper, we investigate the inter-event times of frequent words observed in 17 academic books, where the inter-event time is defined as the number of sentences between two successive appearances of a given word. Our results show that the distributions of inter-event times for frequent words can be classified into three types: the exponential distribution, the q-exponential distribution, and the power-law distribution. To examine the generative mechanisms underlying these three types, we conducted text generation simulations and found that combining two mechanisms for word selection—priority-based selection and randomized selection—is sufficient to reproduce the observed three distribution types. In particular, the priority-based selection mechanism, in which words for a constructed sentence are chosen mechanically according to predefined priorities assigned to each word, is identified as the underlying mechanism of the power-law distribution of inter-event times. We also discuss in detail the relationship between the priority-based selection mechanism and the multivariate Hawkes process, which effectively captures mutual correlations among occurrences of important words.

Sepsis mortality due to drug-resistant organisms remains high. However, the effectiveness of empiric combination therapy for sepsis caused by potential drug-resistant gram‑negative bacilli remains debated. This was a nested case-control study using Japanese nationwide data from a medical reimbursement system between 2010 and 2017. The eligibility criteria were adult patients who were admitted to the intensive care unit, with new-onset sepsis at or after admission. We enrolled patients with combination therapy and monotherapy by propensity score matching using their baseline characteristics. We developed a logistic regression model including combination therapy, severity of organ dysfunction, and interaction terms as the adjusting variables. Overall, 7084 patients were included in this study, 3542 and 3542 of whom were enrolled in the combination therapy and monotherapy groups, respectively. The logistic regression model revealed combination therapy and moderate and severe organ dysfunction were significantly associated with in-hospital mortality (odds ratio [OR] 95% confidence interval [CI]: 1.52 [1.20–1.91], OR [95% CI]: 1.78 [1.43–2.23], and OR [95% CI]: 4.22 [2.61–6.81], respectively). Interaction between combination therapy and severe organ dysfunction was statistically significant (OR [95% CI]: 0.41 [0.21–0.80], P for interaction = .009). However, adjusted in-hospital mortality based on the model was not significantly different between the combination therapy and monotherapy groups among patients with severe organ dysfunction (44% [95% CI: 38–51] vs 47% [95% CI: 41–54]). Our results did not reveal any specific effectiveness of combination therapy; however, the significant interaction between combination therapy and severe organ dysfunction may indicate the possible survival benefit of combination therapy for severe sepsis subset.

BACKGROUND Severe trauma remains a leading cause of admission to the intensive care unit. The Trauma and Injury Severity Score (TRISS) is an established standard for predicting outcomes and benchmarking the quality of trauma care globally. However, the TRISS model has some limitations when used for benchmarking trauma care. OBJECTIVE This study aimed to determine whether machine learning–derived trauma phenotypes can complement the TRISS via multivariable modeling to improve in-hospital death prediction. We also introduce “Trauma-Vis,” a freely accessible web-based platform, to facilitate the availability of this integrated assessment approach to clinicians. METHODS In this retrospective cohort study using the nationwide Japan Trauma Data Bank (JTDB), which encompasses data from 303 hospitals in Japan, we divided the data chronologically into a derivation cohort (JTDB 2015-2018) and a temporal validation cohort (JTDB 2019-2022). An integrated model was developed using multivariable logistic regression, incorporating the logit-transformed TRISS-predicted mortality and the assigned trauma phenotypes. After applying the exclusion criteria, 87,882 patients with blunt trauma were analyzed in the derivation cohort and 80,964 in the validation cohort. Predictive performance was evaluated using the area under the receiver operating characteristic curve, Brier score, logarithmic loss, net reclassification improvement, integrated discrimination improvement, and decision curve analysis. RESULTS In the derivation cohort, multivariable modeling demonstrated that trauma phenotype classification significantly recalibrated mortality risk; multiple phenotypes exhibited significant independent associations with in-hospital death after adjusting for baseline TRISS predictions (eg, for phenotype 8: odds ratio 2.38, 95% CI 2.11-2.68; P<.001). In the temporal validation cohort, the integrated multivariable model yielded higher performance metrics than the baseline TRISS model: the area under the receiver operating characteristic curve increased from 0.889 to 0.897 (DeLong test, P<.001), Brier score improved from 0.0454 to 0.0394, and logarithmic loss decreased from 0.1670 to 0.1458. The integrated model demonstrated a calibration intercept of –0.152 and a slope of 0.965 and provided a higher net benefit in the decision curve analysis across evaluated threshold probabilities. CONCLUSIONS Integrating machine learning–derived trauma phenotypes with the TRISS via multivariable modeling improved the accuracy and utility of in-hospital death prediction. The developed “Trauma-Vis” platform demonstrates the technical feasibility of providing real-time risk stratification at the bedside.

The occurrence patterns of important words found in six texts (one historical pamphlet and five renowned academic books) are analyzed using both univariate and multivariate Hawkes processes. By treating the occurrence patterns as binary time-series data along the texts, we investigate how effectively univariate and multivariate Hawkes processes capture the characteristics of these word occurrence signals. Through maximum likelihood estimation and subsequent simulations, we found that the multivariate Hawkes process clearly outperforms the univariate Hawkes process in modeling word occurrence signals. Moreover, we found that the multivariate Hawkes process can provide a Hawkes graph, which serves as an intuitive representation of the relationships between concepts appearing in the analyzed text. Furthermore, our study demonstrates that the importance of concepts within a given text can be quantitatively estimated based on the optimized parameter values of the multivariate Hawkes process.

The occurrence patterns of important words found in six texts (one historical pamphlet and five renowned academic books) are analyzed using both univariate and multivariate Hawkes processes. By treating these occurrence patterns as binary time series data along the texts, we investigate how effectively univariate and multivariate Hawkes processes capture the characteristics of these word occurrence signals. Through maxi-mum likelihood estimation and subsequent simulations, we found that the multivariate Hawkes process clearly outperforms the univariate Hawkes process in modeling word occurrence signals. Moreover, we found that the multivariate Hawkes process can provide a Hawkes graph, which serves as an intuitive representation of the relation-ships between concepts appearing in the analyzed text. Furthermore, our study demonstrates that the importance of concepts within a given text can be quantitatively estimated based on the optimized parameter values of the multivariate Hawkes process.

Purpose: The purpose of this study was to describe the characteristics of pediatric patients who underwent nonoperative management (NOM) for blunt splenic and hepatic injuries and to explore factors associated with NOM failure. Methods: This was a secondary analysis of a multicenter cohort study of pediatric patients with blunt liver and spleen injuries in Japan. Participants included pediatric trauma patients aged 16 years or younger between 2008 and 2019 with NOM, which was defined as no surgery provided within 6 h of hospital arrival. NOM failure, defined as abdominal surgery performed after 6 h of hospital arrival, was the primary outcome. Descriptive statistics were provided and exploratory analysis to assess the associations with outcome using logistic regression. Results: During the study period, 1339 met our eligibility criteria. The median age was 9 years, with a majority being male. The median Injury Severity Score (ISS) was 10. About 14.0% required transfusion within 24 h, and 22.3% underwent interventional radiology procedures. NOM failure occurred in 1.0% of patients and the in-hospital mortality was 0.7%. Factors associated with NOM failure included age, positive focused assessment with sonography for trauma (FAST), contrast extravasation on computed tomography (CT), severe liver injury, concomitant pancreas injury, concomitant gastrointestinal injury, concomitant mesenteric injury, and ISS. Conclusions: In our study, NOM failure were rare. Older age, positive FAST, contrast extravasation on CT, severe liver injury, concomitant pancreas injury, concomitant gastrointestinal injury, concomitant mesenteric injury, and higher ISS were suggested as possible risk factors for NOM failure.

Background Japanese Association for Acute Medicine (JAAM) disseminated intravascular coagulation (DIC) criteria were launched nearly 20 years ago. Following the revised conceptual definition of sepsis and subsequent omission of systemic inflammatory response syndrome (SIRS) score from the latest sepsis diagnostic criteria, we omitted the SIRS score and proposed a modified version of JAAM DIC criteria, the JAAM-2 DIC criteria. Objectives To validate and compare performance between new JAAM-2 DIC criteria and conventional JAAM DIC criteria for sepsis. Methods We used three datasets containing adult sepsis patients from a multicenter nationwide Japanese cohort study (J-septic DIC, FORECAST, and SPICE-ICU registries). JAAM-2 DIC criteria omitted the SIRS score and set the cutoff value at ≥3 points. Receiver operating characteristic (ROC) analyses were performed between the two DIC criteria to evaluate prognostic value. Associations between in-hospital mortality and anticoagulant therapy according to DIC status were analyzed using propensity score weighting to compare significance of the criteria in determining introduction of anticoagulants against sepsis. Results Final study cohorts of the datasets included 2,154, 1,065, and 608 sepsis patients, respectively. ROC analysis revealed that curves for both JAAM and JAAM-2 DIC criteria as predictors of in-hospital mortality were almost consistent. Survival curves for the anticoagulant and control groups in the propensity score-weighted prediction model diagnosed using the two criteria were also almost entirely consistent. Conclusion JAAM-2 DIC criteria were equivalent to JAAM DIC criteria regarding prognostic and diagnostic values for initiating anticoagulation. The newly proposed JAAM-2 DIC criteria could be potentially alternative criteria for sepsis management.

BACKGROUND: Trauma-related deaths and posttraumatic sequelae are a global health concern, necessitating a deeper understanding of the pathophysiology to advance trauma therapy. Proteomics offers insights into identifying and analyzing plasma proteins associated with trauma and inflammatory conditions; however, current proteomic methods have limitations in accurately measuring low-abundance plasma proteins. This study compared plasma proteomics profiles of patients from different acute trauma subgroups to identify new therapeutic targets and devise better strategies for personalized medicine. METHODS: This prospective observational single-center cohort study was conducted between August 2020 and September 2021 in the intensive care unit of Osaka University Hospital in Japan. Enrolling 59 consecutive patients with blunt trauma, we meticulously analyzed plasma proteomics profiles in participants with torso or head trauma, comparing them with those of controls (mild trauma). Using the Olink Explore 3072 instrument (Olink Proteomics AB, Uppsala, Sweden), we identified five endotypes (α–ε) via unsupervised hierarchical clustering. RESULTS: The median time from injury to blood collection was 47 minutes [interquartile range, 36–64 minutes]. The torso trauma subgroup exhibited 26 unique proteins with significantly altered expression, while the head trauma subgroup showed 68 unique proteins with no overlap between the two. The identified endotypes included α (torso trauma, n = 8), β (young patients with brain injury, n = 5), γ (severe brain injury postsurgery, n = 8), δ (torso or brain trauma with mild hyperfibrinolysis, n = 18), and ε (minor trauma, n = 20). Patients with torso trauma showed changes in blood pressure, smooth muscle adaptation, hypermetabolism, and hypoxemia. Patients with traumatic brain injury had dysregulated blood coagulation and altered nerves regeneration and differentiation. CONCLUSION: This study identified unique plasma protein expression patterns in patients with torso trauma and traumatic brain injury, helping categorize five distinct endotypes. Our findings may offer new insights for clinicians, highlighting potential strategies for personalized medicine and improved trauma-related care. LEVEL OF EVIDENCE: Prognostic and Epidemiological; Level III.