Identification of patients at high risk for brain death using an automated digital screening tool: a prospective diagnostic accuracy study

Organ donation is crucial for patients with organ failure to improve quality of life and increase chances of survival [1]. However, the shortage of available organs has resulted in long waiting lists, leaving many patients without access to organ transplantation. In Europe alone, an estimated 20 patients per day die while waiting for an organ transplant [2]. In Germany, waiting lists for organ donation are particularly long compared to other countries with a recent organ donation rate of ten per one million inhabitants, reflecting the country’s consistently low organ donation rates for years [3, 4]. Deficiencies in identifying patients who may progress towards brain death after brain injury may have largely contributed to low organ donation rates in Germany, as indicated by a recent nationwide secondary analysis of 112 million hospitalized patients [5]. Despite a 13.9% increase in the number of potential organ donors (i.e., those who died in the presence of acute brain injury, were mechanically ventilated and had no medical contraindications for organ donation) between 2010 and 2015, there was a concurrent 18.7% decline in referrals to the German organ procurement organization (OPO) and a 32.3% decrease in the actual number of deceased organ donations. Furthermore, findings from an analysis of 7889 deceased patients with acute brain injuries raised similar concerns regarding the proper identification of potential organ donors in procurement hospitals [6]. Notably, 73 patients who were retrospectively classified as potentially at risk of developing brain death were ultimately not subjected to a corresponding neurological evaluation.

To address these shortcomings in potential donor identification, an automated digital screening tool (DETECT; screening for potential brain DEath in paTiEnts with severe brain damage and clinically asCerTained loss of cerebral functions) was developed to prospectively identify intensive care patients who are at risk of developing brain death [7, 8]. In an analysis of 309 deceased patients with severe brain injury, the pilot implementation of DETECT in intensive care units of a tertiary care hospital was associated with a 93% decrease in the risk of missing patients who were retrospectively classified as impending brain death and thus being potential donors, compared to a period before implementation.

In view of aforementioned considerations, we conducted a diagnostic accuracy study to prospectively determine the screening performance of DETECT in identifying neurocritically ill patients with impending brain death. We hypothesized that DETECT could provide reliable information that would increase clinicians’ awareness of these patients and facilitate their early referral to neurointensivists for further evaluation.

The findings of this diagnostic accuracy study demonstrate the ability of the DETECT screening tool to accurately identify patients with acute brain injury who are at high risk of progressing to brain death during neurocritical care. Notably, the tool did not miss patients who were eventually confirmed as brain dead, and the false detection rate among negatively confirmed patients was only one in ten. These results highlight the potential effectiveness of automated screening for key clinical triggers in identifying potential organ donors.

Timely identification and referral of potential organ donors is widely recognized as a key strategy for addressing the global shortage of life-saving transplantable organs [15‐17]. Digital support systems utilizing patient health records have been shown to enhance early recognition and clinical decision-making for various conditions, including sepsis, venous thromboembolism, acute kidney injury, and pediatric appendicitis [18‐21]. In a recent retrospective study of adult patients, electronically automated referrals to an OPO based on clinical triggers such as mechanical ventilation, acute neurocritical injury and a GCS score of ≤ 5 was associated with a 45% increase in monthly donor referrals and a 92% increase in monthly deceased donors compared to a pre-implementation period when referrals were made manually by the ICU healthcare team [22]. Another retrospective study evaluated the effects of an automated clinical decision support system that sent an email notification to the OPO when clinical findings suggestive of impending brain death were documented for pediatric patients [23]. The system was triggered by the presence of bilateral unresponsive pupils, mechanical ventilation and either an absent gag reflex or a GCS score of ≤ 5 within the previous 24 h in the electronic nursing assessments. The implementation of the system resulted in a shorter time to OPO notification and a higher rate of conversion from potential to actual donors compared to the period before implementation. However, there was no difference in the frequency of actual brain death determination between the two periods.

While these studies suggest that automated potential donor identification systems can be effective, their conclusions are limited by the lack of a parallel reference standard, such as neurological assessment and the possibility of external factors influencing the potential donor pool. Additionally, the donor referral technology used in the former study relied solely on clinical data available at the time of ICU admission [22]. In contrast, the findings of the present study suggest that repeat alerts during ICU stay may further improve the diagnostic accuracy of automated identification of potential donors. In fact, nearly 90% of eventual brain-dead patients in this study were repeatedly identified, with a median of seven notifications per patient and up to 10 days prior to brain death determination. Therefore, repeat detection of neurologic key symptoms may indicate genuine evolution of critical intracranial pressure due to devastating brain injury, rather than being attributed to anesthetic effects such as suppression of brainstem reflexes [24].

Variations in healthcare practices across different countries hinder the comparability of potential donor identification approaches [24]. In the United States, the standard procedure is to directly notify the OPO upon admission of a patient meeting clinical criteria suggestive of impending brain death to the ICU. In contrast, the DETECT screening tool was developed to align with the prevailing practice in Germany and other European countries, where the hospital’s transplant coordination team is informed about a potential donor and establishes contact with the OPO after confirming brain death and approaching the family. Notifying the hospital’s transplant coordinators instead of the OPO in patients suggestive of impending brain death may have various benefits, including a closer evaluation of potential donors and their families by the transplant coordinators in collaboration with the treating intensivists and reduced burden on the OPO with fewer futile notifications [25].

The average time between initial automated notification to the transplant team and determination of brain death was 3.6 days in our study, allowing for sufficient time to conduct comprehensive neurological assessments, communicate with the family, and obtain necessary authorizations while ensuring timely determination of brain death. Prolonging brain death determination may jeopardize the overall function of viable organs and their procurement, as brain death can initiate severe systemic inflammation with cytokine storms similar to those seen in sepsis, hemodynamic instability, and unfavorable hormonal changes [26‐28].

The automated screening algorithm implemented by DETECT employs simple yet specific neurologic criteria [11]. In contrast to other electronic support systems, mechanical ventilation was not mandated as a criterion for screening positivity, thus potentially expanding the pool of eligible donors beyond those who are mechanically ventilated. Although still uncommon in Germany, early identification and referral of possible organ donors from the emergency department to elective non-therapeutic intensive care is thought to be a contributing factor to Spain’s organ donation rate of 40 donors per one million, positioning it as the global leader in organ donation [17, 29]. The lower specificity of DETECT observed in mechanically ventilated patients may be attributed to the concurrent use of analgosedatives, which were administered to over 90% of these patients. Notably, patients receiving analgosedatives were six times more likely to trigger a positive result from the screening tool. Moreover, analysis of the entire study cohort unveiled distinct characteristics that were more frequently observed in falsely detected cases than in true negatives—any vertebrobasilar stroke, intracranial pressure-related complications, mechanical ventilation and analgosedation usage—conditions that could potentially explain compromised pupillary light reflexes and impaired vigilance, without necessarily indicating progressing to brain death. However, it is important to note that false positive screening for potential donors is considered less concerning than false negative screening. The former does not harm those who are mistakenly identified, whereas the latter can result in failure to identify potential donors and a missed opportunity to prevent death or disability in transplant candidates. While relying solely on routinely documented clinical assessments for automated screening saves ICU personnel resources, the potential enhancement of specificity through the integration of additional clinical data from bispectral index monitoring, intracranial pressure monitoring, biomarker studies, or machine learning techniques warrants further investigation [30‐33].

This study has several strengths. Firstly, it is the first to prospectively validate an automated digital screening approach for identifying key clinical triggers of impending brain death, using simultaneous and blinded neurological assessment as the reference standard. Unlike prior research on automated donor identification technology, this diagnostic accuracy study builds on a pilot exploration of DETECT at a tertiary care hospital. Our previous pilot study among 114 ICU beds revealed that DETECT was associated with a 93% lower risk of failing to identify potential brain-dead donors since its implementation in 2018, which prompted a subsequent investigation into its internal validity [7]. In the current study, we therefore focused on a neurocritical care cohort with an expected higher incidence of brain death, a rare outcome, which adds to the significance of our findings. Yet this approach may have limited the generalizability of our findings to non-neurological ICU cohorts. The low positive predictive value we observed could be due to the inherent rarity of brain death defined as the target condition in this study. Nonetheless, implementing DETECT in non-neurological ICUs that naturally lack neurocritical care expertise could be particularly beneficial in terms of potential donor identification.

Limitations of this study include the monocentric design, and the fact that we did not investigate the extent to which healthcare providers effectively utilized the information conveyed by the automated notifications or whether repeated notifications contributed to alarm fatigue, as these aspects were beyond the scope of this diagnostic accuracy study. However, it is crucial to consider the potential implications of alarm fatigue in real-world clinical practice when using DETECT. An increased number of notifications could desensitize transplant coordination teams and intensivists, leading to missed responses to critical information on potential organ donors. Moreover, the subsequent steps following potential donor identification are largely reliant on the initiative of the transplant coordination team and treating intensivists, potentially affecting the actual donation rate achieved by the DETECT screening tool. An upcoming multicenter randomized controlled trial on the efficacy of DETECT will therefore gain insights into how healthcare providers receive and manage notifications. Furthermore, future enhancements to DETECT aim to alleviate alarm fatigue and enhance its clinical utility, including customization options enabling selected healthcare professionals to tailor daily notifications—such as excluding patients under deep sedation for medical reasons or those with pre-existing tumors ineligible for organ donation—from the daily screening process. Additionally, our study might have been susceptible to bias arising from competing risks, which may have impacted the frequency of brain death in our study population due to the occurrence of death from other causes prior to progression to brain death (e.g. withdrawal of life-sustaining therapies for perceived poor neurologic prognosis). As a result, this bias could have resulted in an underestimation of the screening tool’s sensitivity, as potential brain death cases might have been excluded, and an overestimation of specificity, by considering potential brain death cases as true negatives. Lastly, DETECT was initially designed to be compatible with the electronic medical records systems employed in the study center restricting its external validity. However, its technical compatibility has been expended to ensure cost-neutral compliance with the majority of widely used electronic medical records systems in intensive care [10].

Our findings suggest that implementing an automated digital screening tool to monitor routine clinical data can facilitate the early detection of patients who may be at risk of developing brain death during neurocritical care. To assess the efficacy of DETECT in improving potential donor identification and increasing organ donation rates, a multicenter randomized controlled trial is currently underway.