Value of hemophagocytosis in the diagnosis of hemophagocytic lymphohistiocytosis in critically ill patients.

BACKGROUND: Ferritin is an established biomarker in the diagnosis of secondary hemophagocytic lymphohistiocytosis (HLH), which is diagnosed by the HLH-2004 criteria. Among these criteria, detection of hemophagocytosis through invasive procedures may delay early life saving treatment. Our aim was to investigate the value of hemophagocytosis in diagnosing HLH in critically ill patients. METHODS: In this secondary analysis of a retrospective observational study, we included all patients aged ≥18 years and admitted to any adult ICU at Charité-Universitätsmedizin Berlin between January 2006 and August 2018, who had hyperferritinemia (≥500 µg/L) and underwent bone marrow biopsy during their ICU course. RESULTS: Two hundred fifty-two patients were included, of whom 31 (12.3%) showed hemophagocytosis. In multivariable logistic regression analysis, maximum ferritin was independently associated with hemophagocytosis. By removing hemophagocytosis from HLH-2004 criteria and HScore, prediction accuracy for HLH diagnosis was only marginally decreased compared to the original scores. CONCLUSIONS: Our results strengthen the diagnostic value of ferritin and underline the importance of considering HLH diagnosis in patients with high ferritin but only four fulfilled HLH-2004 criteria, when hemophagocytosis was not assessed or not detectable. Proof of hemophagocytosis is not required for a reliable HLH diagnosis.

A Standardized Clinical Data Harmonization Pipeline for Scalable AI Application Deployment (FHIR-DHP): Validation and Usability Study.

BACKGROUND: Increasing digitalization in the medical domain gives rise to large amounts of health care data, which has the potential to expand clinical knowledge and transform patient care if leveraged through artificial intelligence (AI). Yet, big data and AI oftentimes cannot unlock their full potential at scale, owing to nonstandardized data formats, lack of technical and semantic data interoperability, and limited cooperation between stakeholders in the health care system. Despite the existence of standardized data formats for the medical domain, such as Fast Healthcare Interoperability Resources (FHIR), their prevalence and usability for AI remain limited. OBJECTIVE: In this paper, we developed a data harmonization pipeline (DHP) for clinical data sets relying on the common FHIR data standard. METHODS: We validated the performance and usability of our FHIR-DHP with data from the Medical Information Mart for Intensive Care IV database. RESULTS: We present the FHIR-DHP workflow in respect of the transformation of "raw" hospital records into a harmonized, AI-friendly data representation. The pipeline consists of the following 5 key preprocessing steps: querying of data from hospital database, FHIR mapping, syntactic validation, transfer of harmonized data into the patient-model database, and export of data in an AI-friendly format for further medical applications. A detailed example of FHIR-DHP execution was presented for clinical diagnoses records. CONCLUSIONS: Our approach enables the scalable and needs-driven data modeling of large and heterogenous clinical data sets. The FHIR-DHP is a pivotal step toward increasing cooperation, interoperability, and quality of patient care in the clinical routine and for medical research.

Utilizing Intensive Care Alarms for Machine Learning.

Alarms help to detect medical conditions in intensive care units and improve patient safety. However, up to 99% of alarms are non-actionable, i.e. alarm that did not trigger a medical intervention in a defined time frame. Reducing their amount through machine learning (ML) is hypothesized to be a promising approach to improve patient monitoring and alarm management. This retrospective study presents the technical and medical pre-processing steps to annotate alarms into actionable and non-actionable, creating a basis for ML applications.

MIMIC-IV as a Clinical Data Schema.

Routinely collected electronic health records (EHR) in clinical information systems (CIS) are often heterogeneous, have inconsistent data formats and lack of documentation. We use the well-known open-source database schema of MIMIC-IV to address this issue aiming to support collaborative secondary analysis. Over 154 million data records from a German ICU have already been mapped and inserted into the schema successfully. However, discrepancies between the German and US health systems as well as specifics in our clinical source data hinder the direct translation to MIMIC. Evaluating and improving mapping completeness is part of the ongoing research.

Differential Diagnosis of Hyperferritinemia in Critically Ill Patients.

BACKGROUND: Elevated serum ferritin is a common condition in critically ill patients. It is well known that hyperferritinemia constitutes a good biomarker for hemophagocytic lymphohistiocytosis (HLH) in critically ill patients. However, further differential diagnoses of hyperferritinemia in adult critically ill patients remain poorly investigated. We sought to systematically investigate hyperferritinemia in adult critically ill patients without HLH. METHODS: In this secondary analysis of a retrospective observational study, patients ≥18 years admitted to at least one adult intensive care unit at Charité-Universitätsmedizin Berlin between January 2006 and August 2018, and with hyperferritinemia of ≥500 µg/L were included. Patients with HLH were excluded. All patients were categorized into non-sepsis, sepsis, and septic shock. They were also classified into 17 disease groups, based on their ICD-10 codes, and pre-existing immunosuppression was determined. Uni- and multivariable linear regression analyses were performed in all patients. RESULTS: A total of 2583 patients were analyzed. Multivariable linear regression analysis revealed positive associations of maximum SOFA score, sepsis or septic shock, liver disease (except hepatitis), and hematological malignancy with maximum ferritin. T/NK cell lymphoma, acute myeloblastic leukemia, Kaposi's sarcoma, acute or subacute liver failure, and hepatic veno-occlusive disease were positively associated with maximum ferritin in post-hoc multivariable linear regression analysis. CONCLUSIONS: Sepsis or septic shock, liver disease (except hepatitis) and hematological malignancy are important differential diagnoses in hyperferritinemic adult critically ill patients without HLH. Together with HLH, they complete the quartet of important differential diagnoses of hyperferritinemia in adult critically ill patients. As these conditions are also related to HLH, it is important to apply HLH-2004 criteria for exclusion of HLH in hyperferritinemic patients. Hyperferritinemic critically ill patients without HLH require quick investigation of differential diagnoses.

Influence of transfusions, hemodialysis and extracorporeal life support on hyperferritinemia in critically ill patients.

BACKGROUND: Ferritin is the major iron storage protein and an acute phase reactant. Hyperferritinemia is frequently seen in the critically ill where it has been hypothesized that not only underlying conditions but also factors such as transfusions, hemodialysis and extracorporeal life support (ECLS) lead to hyperferritinemia. This study aims to investigate the influence of transfusions, hemodialysis, and ECLS on hyperferritinemia in a multidisciplinary ICU cohort. METHODS: This is a post-hoc analysis of a retrospective observational study including patients aged ≥ 18 years who were admitted to at least one adult ICU between January 2006 and August 2018 with hyperferritinemia ≥ 500 µg/L and of ≥ 14 days between two ICU ferritin measurements. Patients with hemophagocytic lymphohistiocytosis (HLH) were excluded. To identify the influence of transfusions, hemodialysis, and ECLS on ferritin change, multivariable linear regression analysis with ferritin change between two measurements as dependent variable was performed. RESULTS: A total of 268 patients was analyzed. Median duration between measurements was 36 days (22-57). Over all patients, ferritin significantly increased between the first and last measurement (p = 0.006). Multivariable linear regression analysis showed no effect of transfusions, hemodialysis, or ECLS on ferritin change. Changes in aspartate aminotransferase (ASAT) and sequential organ failure assessment (SOFA) score were identified as influencing factors on ferritin change [unstandardized regression coefficient (B) = 2.6; (95% confidence interval (CI) 1.9, 3.3); p < 0.001 and B = 376.5; (95% CI 113.8, 639.1); p = 0.005, respectively]. Using the same model for subgroups of SOFA score, we found SOFA platelet count to be associated with ferritin change [B = 1729.3; (95% CI 466.8, 2991.9); p = 0.007]. No association of ferritin change and in-hospital mortality was seen in multivariable analysis. CONCLUSIONS: The present study demonstrates that transfusions, hemodialysis, and ECLS had no influence on ferritin increases in critically ill patients. Hyperferritinemia appears to be less the result of iatrogenic influences in the ICU thereby underscoring its unskewed diagnostic value. TRIAL REGISTRATION: The study was registered with www.ClinicalTrials.gov (NCT02854943) on August 1, 2016.

Patient Monitoring Alarms in an Intensive Care Unit: Observational Study With Do-It-Yourself Instructions.

BACKGROUND: As one of the most essential technical components of the intensive care unit (ICU), continuous monitoring of patients' vital parameters has significantly improved patient safety by alerting staff through an alarm when a parameter deviates from the normal range. However, the vast number of alarms regularly overwhelms staff and may induce alarm fatigue, a condition recently exacerbated by COVID-19 and potentially endangering patients. OBJECTIVE: This study focused on providing a complete and repeatable analysis of the alarm data of an ICU's patient monitoring system. We aimed to develop do-it-yourself (DIY) instructions for technically versed ICU staff to analyze their monitoring data themselves, which is an essential element for developing efficient and effective alarm optimization strategies. METHODS: This observational study was conducted using alarm log data extracted from the patient monitoring system of a 21-bed surgical ICU in 2019. DIY instructions were iteratively developed in informal interdisciplinary team meetings. The data analysis was grounded in a framework consisting of 5 dimensions, each with specific metrics: alarm load (eg, alarms per bed per day, alarm flood conditions, alarm per device and per criticality), avoidable alarms, (eg, the number of technical alarms), responsiveness and alarm handling (eg alarm duration), sensing (eg, usage of the alarm pause function), and exposure (eg, alarms per room type). Results were visualized using the R package ggplot2 to provide detailed insights into the ICU's alarm situation. RESULTS: We developed 6 DIY instructions that should be followed iteratively step by step. Alarm load metrics should be (re)defined before alarm log data are collected and analyzed. Intuitive visualizations of the alarm metrics should be created next and presented to staff in order to help identify patterns in the alarm data for designing and implementing effective alarm management interventions. We provide the script we used for the data preparation and an R-Markdown file to create comprehensive alarm reports. The alarm load in the respective ICU was quantified by 152.5 (SD 42.2) alarms per bed per day on average and alarm flood conditions with, on average, 69.55 (SD 31.12) per day that both occurred mostly in the morning shifts. Most alarms were issued by the ventilator, invasive blood pressure device, and electrocardiogram (ie, high and low blood pressure, high respiratory rate, low heart rate). The exposure to alarms per bed per day was higher in single rooms (26%, mean 172.9/137.2 alarms per day per bed). CONCLUSIONS: Analyzing ICU alarm log data provides valuable insights into the current alarm situation. Our results call for alarm management interventions that effectively reduce the number of alarms in order to ensure patient safety and ICU staff's work satisfaction. We hope our DIY instructions encourage others to follow suit in analyzing and publishing their ICU alarm data.

Hemophagocytic lymphohistiocytosis in critically ill patients: diagnostic reliability of HLH-2004 criteria and HScore.

BACKGROUND: Hemophagocytic lymphohistiocytosis (HLH) is a rare though often fatal hyperinflammatory syndrome mimicking sepsis in the critically ill. Diagnosis relies on the HLH-2004 criteria and HScore, both of which have been developed in pediatric or adult non-critically ill patients, respectively. Therefore, we aimed to determine the sensitivity and specificity of HLH-2004 criteria and HScore in a cohort of adult critically ill patients. METHODS: In this further analysis of a retrospective observational study, patients ≥ 18 years admitted to at least one adult ICU at Charité - Universitätsmedizin Berlin between January 2006 and August 2018 with hyperferritinemia of ≥ 500 µg/L were included. Patients' charts were reviewed for clinically diagnosed or suspected HLH. Receiver operating characteristics (ROC) analysis was performed to determine prediction accuracy. RESULTS: In total, 2623 patients with hyperferritinemia were included, of whom 40 patients had HLH. We found the best prediction accuracy of HLH diagnosis for a cutoff of 4 fulfilled HLH-2004 criteria (95.0% sensitivity and 93.6% specificity) and HScore cutoff of 168 (100% sensitivity and 94.1% specificity). By adjusting HLH-2004 criteria cutoffs of both hyperferritinemia to 3000 µg/L and fever to 38.2 °C, sensitivity and specificity increased to 97.5% and 96.1%, respectively. Both a higher number of fulfilled HLH-2004 criteria [OR 1.513 (95% CI 1.372-1.667); p <  0.001] and a higher HScore [OR 1.011 (95% CI 1.009-1.013); p <  0.001] were significantly associated with in-hospital mortality. CONCLUSIONS: An HScore cutoff of 168 revealed a sensitivity of 100% and a specificity of 94.1%, thereby providing slightly superior diagnostic accuracy compared to HLH-2004 criteria. Both HLH-2004 criteria and HScore proved to be of good diagnostic accuracy and consequently might be used for HLH diagnosis in critically ill patients. CLINICAL TRIAL REGISTRATION: The study was registered with www.ClinicalTrials.gov (NCT02854943) on August 1, 2016.