Responsible clinical AI requires regionally anchored networks tailored to local contexts. We propose a primary-care-first approach to prioritize local governance, ensuring decisions are context-driven; data sovereignty, to build trust and protect privacy; and community-driven infrastructure, ensuring sustainability. With capacity building, these networks support equitable AI by aligning with local needs, overcoming barriers to implementation, and enabling collaboration across regions to scale solutions effectively.

Perceived acoustic environments, or soundscapes, of intensive care units (ICUs) can be stressful for patients. We developed an approach to enhance ICU soundscapes with soundscape interventions. Compositions of Natural, Human, or Technological sounds were designed to establish three types of sonic ambiances: Comfortable, Pleasurable, or Stimulating. The objective was to investigate the approach's effectiveness in a real-world ICU environment. In a controlled experiment conducted in a single-patient, sound-proofed ICU room, 26 healthy participants were exposed to simulated ICU soundscapes, including patient monitor alarm sounds and mechanical ventilator sounds. Nine soundscape interventions were played via speakers. Perceived pleasantness and eventfulness of resulting soundscapes and experienced pleasure and arousal of listeners were evaluated with questionnaires. Physiological indicators of stress were measured using electrocardiograms (ECGs). Pleasurable and Stimulating interventions significantly increased perceived pleasantness and eventfulness when introduced to the simulated ICU soundscape. Comfortable interventions had no significant effect, suggesting that Pleasurable and Stimulating ambiances better aligned with participants’ needs relative to the simulated soundscape. It emphasized the need to tailor ICU interventions to the preexisting acoustic environment and sound-related needs of listeners, such as comfort, distraction or reassurance. Participants reported positive emotional states while listening to the soundscape interventions, indicative of positive listener experiences. Preliminary insights regarding changes in heartrate variability hinted that soundscape interventions could potentially contribute to reduced stress levels. The effectiveness of interventions depended on their featured sound categories, highlighting the importance of personalization. Overall, our approach was found effective, showing promise for creating listener-centric, restorative soundscapes during ICU stays.

Background: Despite several randomised controlled trials (RCTs) on the use of adjuvant treatment with corticosteroids in patients with community-acquired pneumonia (CAP), the effect of this intervention on mortality remains controversial. We aimed to evaluate heterogeneity of treatment effect (HTE) of adjuvant treatment with corticosteroids on 30-day mortality in patients with CAP. Methods: In this individual patient data meta-analysis, we included RCTs published before July 1, 2024, comparing adjuvant treatment with corticosteroids versus placebo in patients hospitalised with CAP. The primary endpoint was 30-day all-cause mortality, collected across all trials, and analyses followed the intention-to-treat principle. We analysed HTE using risk and effect modelling. For risk modelling, patients were classified as having less severe or severe CAP based on the pneumonia severity index (PSI), comparing PSI class I–III versus class IV–V. For effect modelling, we trained a corticosteroid-effect model on six trials and externally validated it using data from two trials, received after model preregistration. This model classified patients into two groups: no predicted benefit and predicted benefit from adjuvant treatment with corticosteroids. The literature search was registered on PROSPERO, CRD42022380746. Findings: We included eight RCTs with 3224 patients. Across all eight trials, 246 (7·6%) patients died within 30 days (106 [6·6%] of 1618 in the corticosteroid group vs 140 [8·7%] of 1606 in the placebo group; odds ratio [OR] 0·72 [95% CI 0·56–0·94], p=0·017). The corticosteroid-effect model, which selected C-reactive protein (CRP), showed significant HTE during external validation in the two most recent trials. In these trials, 154 (11·4%) of 1355 patients died within 30 days (88 [13·1%] of 671 in the placebo group vs 66 [9·6%] of 684 in the corticosteroid group; OR 0·71 [95% CI 0·50–0·99], p=0·044). Among patients predicted to have no benefit (CRP ≤204 mg/L, n=725), no significant effect was observed (OR 0·98 [95% CI 0·63–1·50]), whereas for those with predicted benefit (CRP >204 mg/L, n=630), 39 (13·0%) of 301 patients died in the placebo group compared with 20 (6·1%) of 329 in the corticosteroid group (0·43 [0·25–0·76], p_interaction=0·026). No significant HTE was found between less severe CAP (PSI class I–III, n=229) and severe CAP (PSI class IV–V, n=1126). Corticosteroid therapy significantly increased hyperglycaemia risk (44 [12·8%] of 344 in the placebo group vs 84 [24·8%] of 339 in the corticosteroid group; OR 2·50 [95% CI 1·63–3·83], p<0·0001) and hospital re-admission risk (30 [3·7%] of 814 in the placebo group vs 57 [7·0%] of 819 in the corticosteroid group; 1·95 [1·24–3·07], p=0·0038). Interpretation: Overall, adjuvant therapy with corticosteroids significantly reduces 30-day mortality in patients hospitalised with CAP. The treatment effect varied significantly among subgroups based on CRP concentrations, with a substantial mortality reduction observed only in patients with high baseline CRP. Funding: None.

Importance: Artificial intelligence (AI) presents transformative opportunities to address the increasing challenges faced by health care systems globally. Particularly, in data-rich environments, such as intensive care units (ICUs), AI could assist in enhancing clinical decision-making, streamline workflows, and improve patient outcomes. Despite these promising applications, the practical implementation of AI in clinical settings remains limited. Objective: To systematically evaluate AI system operationalization in the ICU, focusing on the AI field's progress over time, technical maturity, and risk of bias. Evidence Review: In this systematic review, 5 databases (Embase, MEDLINE ALL, Web of Science Core Collection, Cochrane Central Register of Controlled Trials, and Google Scholar) were searched for studies published from July 28, 2020, to June 10, 2024. Eligible studies evaluated AI applications designed for use within ICUs for adults (aged ≥16 years) and used data collected during ICU stays. Two reviewers independently screened titles and abstracts, with a third reviewer resolving disagreements. Data extraction included AI application aims, dataset origins, technology readiness level (TRL) categorization, and the use of reporting standards. Risk of bias was assessed using the PROBAST (Prediction Model Study Risk of Bias Assessment Tool). Findings: Of 17 401 screened records, 1263 studies met the inclusion criteria. A total of 936 studies (74% of all studies) were classified as TRL 4 or below, indicating early-stage development or initial validation. Among these, 447 (37%) used internal datasets, 562 (46%) used MIMIC (Medical Information Mart for Intensive Care) datasets (I-IV), and 78 (6%) used the open-source eICU Collaborative Research Database. External validation (TRL 5) was achieved by 24% of studies. Only 25 (2%) progressed to clinical integration (TRL≥6), with no studies reaching full implementation (TRL 9). Although approximately half of generative AI models reached a higher TRL (14 [47%] with TRL 5), none reached clinical integration. Additionally, only 207 studies (16%) referenced reporting standards, with adherence modestly increasing from 14% in 2021 to 23% in 2024. High risk of bias was identified in 581 of 1103 studies (53%), primarily due to methodologic shortcomings in the analysis domain. Conclusions and Relevance: Despite substantial growth in AI research within intensive care medicine in recent years, the transition from development to clinical implementation still remains limited and has made little progress over time. A paradigm shift is urgently required in the medical literature-one that moves beyond retrospective validation toward the operationalization and prospective testing of AI for tangible clinical impact.

Introduction: Although many preterm born infants require invasive mechanical ventilation, it is also associated with detrimental effects. Early extubation should be pursued, but extubation failure is yet common. The critical transition to noninvasive ventilation is characterized by respiratory physiological changes, warranting noninvasive monitoring. We aimed to determine whether electrical impedance tomography (EIT) could provide insights into the respiratory mechanics of neonates around extubation, and if findings were different between successful and failed extubation. Methods: Single-center observational study where EIT and transcutaneous CO2 measurements were performed in preterm born infants <32 weeks gestational age. Measurements were performed from 24 h before up to 48 h after extubation. EIT parameters extracted from the hour before and after extubation were analyzed to evaluate the shortterm physiological changes. Results: Twenty-one patients were included and 6 (29%) were reintubated. End-expiratory lung impedance and tidal impedance variation were stable around extubation (p = 0.86 and p = 0.47, respectively). Compared to successfully extubated patients, reintubated patients showed more lung inhomogeneity (GI index) after extubation (0.75 vs. 0.84, p = 0.03). The percentage of nondependent silent spaces decreased after extubation in successfully extubated patients (p < 0.001). Body position and ventilator mode influenced these findings. Conclusion: EIT measurements in preterm neonates provide valuable insight into the respiratory physiology during the transition from invasive to noninvasive ventilation, with significant differences in ventilation distribution and lung homogeneity between successfully extubated and reintubated patients. EIT has the potential to guide personalized respiratory support by assessing ventilation distribution and quantifying inhomogeneity, aiding in the optimization of ventilation settings.

Intensive Care Unit (ICU) nurses are burdened by excessive number of false and irrelevant alarms generated by patient monitoring systems. Nurses rely on these patient monitoring systems for timely and relevant medical information concerning patients. However, the systems currently in place are not sensitive to the perceptual and cognitive abilities of nurses and thus fail to communicate information efficiently. An efficient communication and an effective collaboration between patient monitoring systems and ICU nurses is only possible by designing systems sensitive to the abilities and preferences of nurses. In order to design these sensitive systems, we need to gain in-depth understanding of the user group through revealing their latent individual characteristics. To this end, we conducted a survey on individual characteristics involving nurses from two IC units. Our results shed light on the personality and other characteristics of ICU nurses. Subsequently, we performed hierarchical cluster analysis to develop data-driven nurse profiles. We suggest design recommendations tailored to four distinct user profiles to address their unique needs. By optimizing the system interactions to match the natural tendencies of nurses, we aspire to alleviate the cognitive burden induced by system use to ensure that healthcare providers receive relevant information, ultimately improving patient safety.

AI holds the potential to transform healthcare, promising improvements in patient care. Yet, realizing this potential is hampered by over-reliance on limited datasets and a lack of transparency in validation processes. To overcome these obstacles, we advocate the creation of a detailed registry for AI algorithms. This registry would document the development, training, and validation of AI models, ensuring scientific integrity and transparency. Additionally, it would serve as a platform for peer review and ethical oversight. By bridging the gap between scientific validation and regulatory approval, such as by the FDA, we aim to enhance the integrity and trustworthiness of AI applications in healthcare.

Despite artificial intelligence (AI) technology progresses at unprecedented rate, our ability to translate these advancements into clinical value and adoption at the bedside remains comparatively limited. This paper reviews the current use of implementation outcomes in randomized controlled trials evaluating AI-based clinical decision support and found limited adoption. To advance trust and clinical adoption of AI, there is a need to bridge the gap between traditional quantitative metrics and implementation outcomes to better grasp the reasons behind the success or failure of AI systems and improve their translation into clinical value.

Objective. Geometrical region of interest (ROI) selection in electrical impedance tomography (EIT) monitoring may lack sensitivity to subtle changes in ventilation distribution. Therefore, we demonstrate a new physiological method for ROI definition. This is relevant when using ROIs to compute subsequent EIT-parameters, such as the ventral-to-dorsal ratio during a positive end-expiratory pressure (PEEP) trial.Approach.Our physiological approach divides an EIT image to ensure exactly 50% tidal impedance variation in the ventral and dorsal region. To demonstrate the effects of our new method, EIT measurements during a decremental PEEP trial in 49 mechanically ventilated ICU-patients were used. We compared the center of ventilation (CoV), a robust parameter for changes in ventro-dorsal ventilation distribution, to our physiological ROI selection method and different commonly used ROI selection methods. Moreover, we determined the impact of different ROI selection methods on the PEEP level corresponding to a ventral-to-dorsal ratio closest to 1.Main results.The division line separating the ventral and dorsal ROI was closer to the CoV for our new physiological method for ROI selection compared to geometrical ROI definition. Moreover, the PEEP level corresponding to a ventral-to-dorsal ratio of 1 is strongly influenced by the chosen ROI selection method, which could have a profound clinical impact; the within-subject range of PEEP level was 6.2 cmH2O depending on the chosen ROI selection method.Significance.Our novel physiological method for ROI definition is sensitive to subtle ventilation-induced changes in regional impedance (i.e. due to (de)recruitment) during mechanical ventilation, similar to the CoV.

Objective. The respiratory rate (RR) is considered one of the most informative vital signals. A well-validated standard for RR measurement in mechanically ventilated patient is capnography; a noninvasive technique for expiratory CO2measurements. Reliable RR measurements in spontaneously breathing patients remains a challenge as continuous mainstream capnography measurements are not available. This study aimed to assess the accuracy of RR measurement using electrical impedance tomography (EIT) in healthy volunteers and intensive care unit (ICU) patients on mechanical ventilation and spontaneously breathing post-extubation. Comparator methods included RR derived from both capnography and bioimpedance electrocardiogram (ECG) measurements.Approach. Twenty healthy volunteers wore an EIT belt and ECG electrodes while breathing through a capnometer within a 10-40 breaths per minute (BPM) range. Nineteen ICU patients underwent similar measurements during pressure support ventilation and spontaneously breathing after extubation from mechanical ventilation. Stable periods with regular breathing and no artefacts were selected, and agreement between measurement methods was assessed using Bland-Altman analysis for repeated measurements.Main result. Bland-Altman analysis revealed a bias less than 0.2 BPM, with tight limits of agreement (LOA) ±1.5 BPM in healthy volunteers and ventilated ICU patients when comparing EIT to capnography. Spontaneously breathing ICU patients had wider LOA (±2.5 BPM) when comparing EIT to ECG bioimpedance, but gold standard comparison was unavailable. RR measurements were stable for 91% of the time for capnography, 68% for EIT, and 64% of the ECG bioimpedance signals. After extubation, the percentage of stable periods decreased to 48% for EIT signals and to 55% for ECG bioimpedance.Significance. In periods of stable breathing, EIT demonstrated excellent RR measurement accuracy in healthy volunteers and ICU patients. However, stability of both EIT and ECG bioimpedance RR measurements declined in spontaneously breathing patients to approximately 50% of the time.

This scoping review focuses on the essential role of models for causal inference in shaping actionable artificial intelligence (AI) designed to aid clinicians in decision-making. The objective was to identify and evaluate the reporting quality of studies introducing models for causal inference in intensive care units (ICUs), and to provide recommendations to improve the future landscape of research practices in this domain. To achieve this, we searched various databases including Embase, MEDLINE ALL, Web of Science Core Collection, Google Scholar, medRxiv, bioRxiv, arXiv, and the ACM Digital Library. Studies involving models for causal inference addressing time-varying treatments in the adult ICU were reviewed. Data extraction encompassed the study settings and methodologies applied. Furthermore, we assessed reporting quality of target trial components (i.e., eligibility criteria, treatment strategies, follow-up period, outcome, and analysis plan) and main causal assumptions (i.e., conditional exchangeability, positivity, and consistency). Among the 2184 titles screened, 79 studies met the inclusion criteria. The methodologies used were G methods (61%) and reinforcement learning methods (39%). Studies considered both static (51%) and dynamic treatment regimes (49%). Only 30 (38%) of the studies reported all five target trial components, and only seven (9%) studies mentioned all three causal assumptions. To achieve actionable AI in the ICU, we advocate careful consideration of the causal question of interest, describing this research question as a target trial emulation, usage of appropriate causal inference methods, and acknowledgement (and examination of potential violations of) the causal assumptions.

Background: Timely identification of deteriorating COVID-19 patients is needed to guide changes in clinical management and admission to intensive care units (ICUs). There is significant concern that widely used Early warning scores (EWSs) underestimate illness severity in COVID-19 patients and therefore, we developed an early warning model specifically for COVID-19 patients. Methods: We retrospectively collected electronic medical record data to extract predictors and used these to fit a random forest model. To simulate the situation in which the model would have been developed after the first and implemented during the second COVID-19 ‘wave’ in the Netherlands, we performed a temporal validation by splitting all included patients into groups admitted before and after August 1, 2020. Furthermore, we propose a method for dynamic model updating to retain model performance over time. We evaluated model discrimination and calibration, performed a decision curve analysis, and quantified the importance of predictors using SHapley Additive exPlanations values. Results: We included 3514 COVID-19 patient admissions from six Dutch hospitals between February 2020 and May 2021, and included a total of 18 predictors for model fitting. The model showed a higher discriminative performance in terms of partial area under the receiver operating characteristic curve (0.82 [0.80–0.84]) compared to the National early warning score (0.72 [0.69–0.74]) and the Modified early warning score (0.67 [0.65–0.69]), a greater net benefit over a range of clinically relevant model thresholds, and relatively good calibration (intercept = 0.03 [− 0.09 to 0.14], slope = 0.79 [0.73–0.86]). Conclusions: This study shows the potential benefit of moving from early warning models for the general inpatient population to models for specific patient groups. Further (independent) validation of the model is needed.

Purpose: The healthcare sector is responsible for 6–7% of CO₂ emissions. The intensive care unit (ICU) contributes to these CO₂ emissions and a shift from a linear system to a circular system is needed. The aim of our research was to perform a material flow analysis (MFA) in an academic ICU. Secondary aims were to obtain information and numbers on mass, carbon footprint, agricultural land occupation and water usage and to determine so-called “environmental hotspots” in the ICU. Methods: A material flow analysis was performed over the year 2019, followed by an environmental footprint analysis of materials and environmental hotspot identification. Results: 2839 patients were admitted to our ICU in 2019. The average length of stay was 4.6 days. Our MFA showed a material mass inflow of 247,000 kg in 2019 for intensive care, of which 50,000 kg is incinerated as (hazardous) hospital waste. The environmental impact per patient resulted in 17 kg of mass, 12 kg CO₂ eq, 300 L of water usage and 4 m² of agricultural land occupation per day. Five hotspots were identified: non-sterile gloves, isolation gowns, bed liners, surgical masks and syringes (including packaging). Conclusion: This is the first material flow analysis that identified environmental risks and its magnitude in the intensive care unit.