Applying Machine Learning to Blood Count Data Predicts Sepsis with ICU Admission.

BACKGROUND: Timely diagnosis is crucial for sepsis treatment. Current machine learning (ML) models suffer from high complexity and limited applicability. We therefore created an ML model using only complete blood count (CBC) diagnostics. METHODS: We collected non-intensive care unit (non-ICU) data from a German tertiary care centre (January 2014 to December 2021). Using patient age, sex, and CBC parameters (haemoglobin, platelets, mean corpuscular volume, white and red blood cells), we trained a boosted random forest, which predicts sepsis with ICU admission. Two external validations were conducted using data from another German tertiary care centre and the Medical Information Mart for Intensive Care IV database (MIMIC-IV). Using the subset of laboratory orders also including procalcitonin (PCT), an analogous model was trained with PCT as an additional feature. RESULTS: After exclusion, 1 381 358 laboratory requests (2016 from sepsis cases) were available. The CBC model shows an area under the receiver operating characteristic (AUROC) of 0.872 (95% CI, 0.857-0.887). External validations show AUROCs of 0.805 (95% CI, 0.787-0.824) for University Medicine Greifswald and 0.845 (95% CI, 0.837-0.852) for MIMIC-IV. The model including PCT revealed a significantly higher AUROC (0.857; 95% CI, 0.836-0.877) than PCT alone (0.790; 95% CI, 0.759-0.821; P < 0.001). CONCLUSIONS: Our results demonstrate that routine CBC results could significantly improve diagnosis of sepsis when combined with ML. The CBC model can facilitate early sepsis prediction in non-ICU patients with high robustness in external validations. Its implementation in clinical decision support systems has strong potential to provide an essential time advantage and increase patient safety.

Development, Design and Utilization of a CDSS for Refeeding Syndrome in Real Life Inpatient Care-A Feasibility Study.

BACKGROUND: The refeeding syndrome (RFS) is an oftentimes-unrecognized complication of reintroducing nutrition in malnourished patients that can lead to fatal cardiovascular failure. We hypothesized that a clinical decision support system (CDSS) can improve RFS recognition and management. METHODS: We developed an algorithm from current diagnostic criteria for RFS detection, tested the algorithm on a retrospective dataset and combined the final algorithm with therapy and referral recommendations in a knowledge-based CDSS. The CDSS integration into clinical practice was prospectively investigated for six months. RESULTS: The utilization of the RFS-CDSS lead to RFS diagnosis in 13 out of 21 detected cases (62%). It improved patient-related care and documentation, e.g., RFS-specific coding (E87.7), increased from once coded in 30 month in the retrospective cohort to four times in six months in the prospective cohort and doubled the rate of nutrition referrals in true positive patients (retrospective referrals in true positive patients 33% vs. prospective referrals in true positive patients 71%). CONCLUSION: CDSS-facilitated RFS diagnosis is possible and improves RFS recognition. This effect and its impact on patient-related outcomes needs to be further investigated in a large randomized-controlled trial.

GestaltMatcher Database - a FAIR database for medical imaging data of rare disorders.

The value of computer-assisted image analysis has been shown in several studies. The performance of tools with artificial intelligence (AI), such as GestaltMatcher, is improved with the size and diversity of the training set, but properly labeled training data is currently the biggest bottleneck in developing next-generation phenotyping (NGP) applications. Therefore, we developed GestaltMatcher Database (GMDB) - a database for machine-readable medical image data that complies with the FAIR principles and improves the openness and accessibility of scientific findings in Medical Genetics. An entry in GMDB consists of a medical image such as a portrait, X-ray, or fundoscopy, and machine-readable meta information such as a clinical feature encoded in HPO terminology or a disease-causing mutation reported in HGVS format. In the beginning, data was mainly collected by curators gathering images from the literature. Currently, clinicians and individuals recruited from patient support groups provide their previously unpublished data. For this patient-centered approach, we developed a digital consent form. GMDB is a modern publication medium for case reports that complements preprints, e.g., on medRxiv. To enable inter-cohort comparisons, we implemented a research feature in GMDB that computes the pairwise syndromic similarity between hand-picked cases. Through a community-driven effort, we compiled an image collection of over 7,533 cases with 792 disorders in GMDB. Most of the data was collected from 2,058 publications. In addition, about 1,018 frontal images of 498 previously unpublished cases were obtained. The web interface enables gene- and phenotype-centered queries or infinite scrolls in the gallery. Digital consent has led to increasing adoption of the approach by patients. The research app within GMDB was used to generate syndromic similarity matrices to characterize two novel phenotypes (CSNK2B, PSMC3). GMDB is the first FAIR database for NGP, where data are findable, accessible, interoperable, and reusable. It is a repository for medical images that cannot be included in medRxiv. That means GMDB connects clinicians with a shared interest in particular phenotypes and improves the performance of AI.

Laboratory Diagnostic of Acute Kidney Injury and Its Progression: Risk of Underdiagnosis in Female and Elderly Patients.

Acute kidney injury (AKI) is a common disease, with high morbidity and mortality rates. In this study, we investigated the potential influence of sex and age on laboratory diagnostics and outcomes. It is known that serum creatinine (SCr) has limitations as a laboratory diagnostic parameter for AKI due to its dependence on muscle mass, which may lead to an incorrect or delayed diagnosis for certain patient groups, such as women and the elderly. Overall, 7592 cases with AKI, hospitalized at the University of Leipzig Medical Center (ULMC) between 1st January 2017 and 31st December 2019, were retrospectively analyzed. The diagnosis and staging of AKI were performed according to the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines, based on the level and dynamics of SCr. The impact of sex and age was analyzed by the recalculation of a female to male and an old to young SCr using the CKD-EPI equation. In our study cohort progressive AKI occurred in 19.2% of all cases (n = 1458). Female cases with AKI were underrepresented (40.4%), with a significantly lower first (-3.5 mL/min) and last eGFR (-2.7 mL/min) (p < 0.001). The highest incidence proportion of AKI was found in the [61-81) age group in female (49.5%) and male (52.7%) cases. Females with progressive AKI were underrepresented (p = 0.04). By defining and staging AKI on the basis of relative and absolute changes in the SCr level, it is more difficult for patients with low muscle mass and, thus, a lower baseline SCr to be diagnosed by an absolute SCr increase. AKIN1 and AKIN3 can be diagnosed by a relative or absolute change in SCr. In females, both stages were less frequently detected by an absolute criterion alone (AKIN1 ♀ 20.2%, ♂ 29.5%, p < 0.001; AKIN3 ♀ 13.4%, ♂ 15.2%, p < 0.001). A recalculated SCr for females (as males) and males (as young males) displayed the expected increase in AKI occurrence and severity with age and, in general, in females. Our study illustrates how SCr, as the sole parameter for the diagnosis and staging of AKI, bears the risk of underdiagnosis of patient groups with low muscle mass, such as women and the elderly. A sex- and age-adapted approach might offer advantages.

Refeeding Syndrome.

BACKGROUND: Refeeding syndrome (RFS) can occur in malnourished patients when normal, enteral, or parenteral feeding is resumed. The syndrome often goes unrecognized and may, in the most severe cases, result in death. The diagnosis of RFS can be crucially facilitated by the use of clinical decision support systems (CDSS). METHODS: The literature in PubMed was searched for current treatment recommendations, randomized intervention studies, and publications on RFS and CDSS. We also took account of insights gained from the development and implementation of our own CDSS for the diagnosis of RFS. RESULTS: The identification of high-risk patients and the recognition of manifest RFS is clinically challenging due to the syndrome's unspecific symptoms and physicians' lack of awareness of the risk of this condition. The literature shows that compared to patients without RFS, malnourished patients with RFS have significantly greater 6-month mortality (odds ratio 1.54, 95% confidence interval: [1.04; 2.28]) and an elevated risk of admission to intensive care (odds ratio 2.71 [1.01; 7.27]). In a prospective testing program, use of our own CDSS led to correct diagnosis in two thirds of cases. CONCLUSION: RFS is difficult to detect and represents a high risk to the patients affected. Appropriate CDSS can identify such patients and ensure proper professional care.