Improved Time to Notification of Impending Brain Death and Increased Organ Donation Using an Electronic Clinical Decision Support System

American Journal of Transplantation 2017; 17: 2186 2191 Wiley Periodicals Inc. Brief Communication 2017 The American Society of Transplantation and the American Society of Transplant Surgeons doi: 10.1111/ajt.14312 Improved Time to Notification of Impending Brain Death and Increased Organ Donation Using an Electronic Clinical Decision Support System J. L. Zier 1, *, A. B. Spaulding 2, M. Finch 3, T. Verschaetse 4 and R. Tarrago 5 1 Pediatric Critical Care, Children s Minnesota, Minneapolis, MN 2 Center for Acute Care Outcomes, Children s Minnesota, Minneapolis, MN 3 Research and Sponsored Programs, Children s Minnesota, Minneapolis, MN 4 LifeSource, Minneapolis, MN 5 Pediatric Critical Care, Seattle Children s Hospital, Seattle, WA *Corresponding author: Judith L. Zier, judy.zier@childrensmn.org Early referral of patients to an organ procurement organization (OPO) may positively affect donation outcomes. We implemented an electronic clinic decision support (CDS) system to automatically notify our OPO of children meeting clinical triggers indicating impending brain death. Medical records of all patients who died in a pediatric critical care unit or were referred for imminent death for 3 years prior to installation of the initial CDS (pre-cds) and for 1 year after implementation of the final CDS (post-cds) were reviewed. Mean time to OPO notification decreased from 30.2 h pre-cds to 1.7 h post-cds (p = 0.015). Notification within 1 h of meeting criteria increased from 36% pre-cds to 70% post-cds (p = 0.003). Although an increase in donor conversion from 50% pre-cds to 90% post-cds did not reach statistical significance (p = 0.0743), there were more organ donors post-cds (11 of 24 deaths) than pre-cds (seven of 57 deaths; p = 0.002). Positive outcomes were achieved with the use of a fully automated CDS system while simultaneously realizing few false-positive notifications, low costs, and minimal workflow interruption. Use of an electronic CDS system in a pediatric hospital setting improved timely OPO notification and was associated with increased organ donation. Abbreviations: CDS, clinical decision support; DBD, donation after brain death; DCD, donation after circulatory death; EHR, electronic health record; GCS, Glasgow Coma Scale; LOC, level of consciousness; OPO, organ procurement organization Received 25 January 2017, revised 17 March 2017 and accepted for publication 05 April 2017 Introduction The American Academy of Pediatrics has recognized timely referral of potential organ donors to an organ procurement organization (OPO) as a means of enhancing the probability that families will consent to donation (1). Timely referral of a child with impending brain death allows ample time for evaluation of donation eligibility and for collaboration between the OPO and clinical staff to plan the optimal approach to the family. The Organ Donation Breakthrough Collaborative, an initiative with a goal of increasing organ donation, also recognized timely referral as the first step in implementation of recognized best practices in the donation process (2). The collaborative set a goal for referral within 1 h of meeting an established clinical trigger. Referral within 1 h of reaching this trigger also complies with the federal mandate requiring OPO notification of imminent death in a timely manner (3). Nevertheless, meeting these referral time goals can be challenging. Standard practice requires healthcare personnel to first recognize that a patient meets notification triggers and then contact an OPO by telephone to make the referral. Delayed notification may occur when healthcare providers fail to promptly recognize notification triggers or when other clinical duties take priority over a call to the OPO. Electronic clinical decision support (CDS) systems based on information garnered from the electronic health record (EHR) have successfully used clinical and laboratory triggers to promptly notify providers of situations requiring clinical intervention, including acute kidney injury and sepsis, and to improve compliance with clinical guidelines for antibiotic use and prevention of venous thromboembolism (4 7). Our hospital system previously implemented an electronic CDS system to comply with clinical practice guidelines limiting unnecessary imaging for pediatric patients with suspected appendicitis (8). We hypothesized that implementation of an electronic CDS system using EHR data to automate OPO notification would improve timely referral of children with impending brain death and that this, in turn, would increase donations. To our knowledge, this study is the first to report the implementation of a fully automated CDS system for OPO notification. 2186

Automated Referral for Impending Brain Death Methods Study design and study site We conducted a retrospective study to evaluate timeliness of OPO notification and number of donations before and after implementation of the electronic CDS system at Children s Minnesota, an urban tertiary care hospital with campuses in Minneapolis and St. Paul. It is the 10th largest standalone pediatric health system in the country. The pediatric critical care units have a total of 67 beds with an average of 2178 admissions per year during the study period. Children s Minnesota has had an EHR system (Millennium; Cerner Corporation, Kansas City, MO) in place since 1998. Electronic CDS system development A CDS system was developed to identify patients who meet OPO notification criteria for impending brain death. The system is based on elements charted in the EHR during routine patient care and clinical triggers that include concurrent (i) severe neurological injury, (ii) mechanical ventilation, and (iii) loss of two or more brainstem reflexes and/or a Glasgow Coma Scale (GCS) score 5. The system is activated when bilateral nonresponsive pupils are documented in the nursing assessment. This initiates the algorithm to electronically search for additional data elements to confirm the presence of mechanical ventilation and additional cranial nerve dysfunction (i.e. absent gag reflex) or unresponsive level of consciousness (Figure 1). Presence of confounders that would result in false-positive notification (e.g. dilating eye drops, neuromuscular blockade) stop the process. When the algorithm detects a patient who fulfills notification criteria, a system-generated email is sent directly to the OPO. A donation resource coordinator then calls the critical care unit to validate the referral and to collect additional information. The initial CDS system was implemented in 2013 and included an interruptive pop-up notification that required the bedside nurse to approve email notification to the OPO prior to sending. The initial algorithm included data elements that failed to exclude a significant number of false-positive patients. Several quality-improvement cycles resulted in the final CDS algorithm, which decreased the number of false-positive patients and removed the requirement for active healthcare provider interaction. This final, fully automated CDS system was implemented in February 2015. Data collection For this retrospective study, medical records of all patients who died in a pediatric critical care unit at Children s Minnesota or who were referred to the OPO for imminent death during the 3-year period (January 2010 through February 2013) prior to initiation of the initial CDS system (pre- CDS) and for 1 year (February 2015 through January 2016) after implementation of the final version (post-cds) were reviewed. Records were examined to verify the presence of mechanical ventilation and severe neurologic injury. Nursing assessments, physician narrative notes, and nursing narrative notes were reviewed for documentation of two concurrent abnormal brainstem reflexes or GCS score 5. If found, the medication administration record was queried for documentation of confounding medications, including dilating eye drops given within the previous 24 h or administration of a neuromuscular blocking agent within the previous 4 h. If notation of abnormal brainstem reflexes or GCS score 5 was found in a narrative note, the time of dictation was recorded as the time of documentation. Outcome definitions The primary outcome of interest for this analysis was time to referral, defined as the time between meeting the trigger for OPO notification and actual time of OPO notification. Secondary outcomes included donor conversion rate, defined as the percentage of eligible-patient deaths that become donors. The time of meeting trigger for OPO notification for impending brain death was defined as the first documentation of two or more concurrent abnormal brainstem reflexes or GCS score 5 without medication confounders. If the time of telephone or email notification of the OPO occurred prior to the documentation of abnormal brainstem reflexes or GCS score 5, time of notification was used as the time that the trigger was met. Time of OPO notification was defined as the time of the notification telephone call to the OPO or the time of email notification, whichever came first, because some post-cds patients were referred by telephone call in addition to email notification. The time of telephone call notification was recorded from telephone logs provided by the OPO. The time of email notification was recorded from the CDS data log. Statistical analysis Comparisons between pre-cds and post-cds groups were done using Student t-tests, Kruskal Wallis tests, and Fisher exact tests, as appropriate. Analyses were performed using IBM SPSS Statistics version 23 (IBM Corp, Armonk, NY). This study was approved by the Children s Minnesota institutional review board (no. 1211-122). Results Patient demographics In total, 90 patients met trigger criteria for OPO notification for impending brain death, 60 in the pre-cds group and 30 in the post-cds group. Two of these patients in the pre-cds period were excluded from analysis. The first met trigger criteria on admission, had a prolonged intensive care unit stay, and died when life-sustaining care was withdrawn. This child was excluded from analysis because of the protracted time (4 mo) from presentation with abnormal brainstem reflexes to time of OPO notification, which occurred at the time of withdrawal of support. The second patient was excluded because their initial trigger-met time occurred prior to commencement of the study. Of the 88 patients included in this analysis (n = 58 pre- CDS and n = 30 post-cds), 46% were male. Among the patients who died (n = 81, 92%), the average age was 4.5 years (median 1.9 years; range 1 mo to 23 years). One patient in the pre-cds group survived and was discharged in a vegetative state. Six patients in the post- CDS group survived to discharge. Five of the surviving post-cds patients were discharged with new neurologic impairments: two in a vegetative state, one with severe neurologic impairment, one with moderate impairment, and one with mild neurologic impairment. One post-cds patient with severe neurologic deficits at baseline was discharged at baseline level of function. Time to referral The mean time to OPO notification significantly decreased pre-cds to post-cds, from 30.2 h (range American Journal of Transplantation 2017; 17: 2186 2191 2187

Zier et al Figure 1: Algorithm used to identify patients for OPO notification. LOC, level of consciousness; OPO, organ procurement organization. 0 288.5 h) to 1.7 h (range 0 23.3 h; p = 0.015) (Table 1). The proportion of notifications occurring 1 h increased from 36% pre-cds to 70% post-cds (p = 0.003) (Figure 2). Median time to notification was 3.5 h pre- CDS compared with <0.01 h post-cds (p = 0.001). Mean and median differences remained significant when 2188 American Journal of Transplantation 2017; 17: 2186 2191

Automated Referral for Impending Brain Death outliers (notification times >51 h) were removed (mean p = 0.003; median p = 0.006). Organ donation Among patients who died, 31 patients were declared dead by neurologic criteria and considered for organ donation; of these, nine patients were disqualified for medical reasons (Table 1). Of those eligible for donation after brain death (DBD), 15 patients became organ donors (68%). The donor conversion rate increased from 50% pre-cds to 90% post-cds, although this difference did not reach statistical significance (p = 0.074). An additional three patients (one pre-cds, two post-cds) became organ donors after circulatory death (donation after circulatory death [DCD]). The total number of organ donors (DBD plus DCD) post-cds (11 of 24 deaths) was significantly greater than the number pre-cds (seven of 57 deaths; p = 0.002). The number of organs transplanted per patient did not differ pre-cds (3.14) to post- CDS (2.82; p = 0.897). Relationship of referral time to organ donation Eleven of 37 patients (30%) referred within the 1-h standard became donors (DBD or DCD) compared with seven of 44 (16%) who were referred >1 h after meeting notification triggers (p = 0.136). There was no correlation between time to referral and conversion of patients eligible for DBD (mean donor 11.1 h vs. nondonor 17.1 h; p = 0.593). Mean time to referral of all donors (DBD plus DCD) was 9.3 h compared with 22.3 h for all other patients (p = 0.157). All of the DCD patients were referred within 1 h of meeting notification criteria. CDS referral accuracy All patients who met criteria for notification during the post-cds period were referred to the OPO. The OPO received email notification for two patients during the post-cds period who did not meet criteria for notification. These false positives included one postoperative patient who received a neuromuscular blocking agent in the operating room that was recorded on paper rather Table 1: Timeliness of notification and organ donation rates pre- CDS and post-cds implementation Pre-CDS Post-CDS p-value Timeliness of notification 1 h notification (%) 36% 70% 0.003 Mean (h) 30.2 1.7 0.015 Organ donation Deceased due to 19/57 (33%) 12/24 (50%) 0.211 brain death Organ donation 6/12 (50%) 9/10 (89%) 0.074 conversion 1 Total organ donors/ critical care death 7/57 (12%) 11/24 (46%) 0.002 1 Among those eligible, defined as brain death and no medical contraindication. than in the EHR. The second patient had wound botulism-associated multiple cranial nerve palsies who went on to complete recovery. Discussion When a child is admitted to a pediatric critical care unit with a devastating neurologic injury, all efforts are focused on resuscitation and therapeutic interventions aimed at preservation or restoration of function. Although loss of brainstem reflexes at this early stage is surely recognized as likely heralding a poor neurologic prognosis, it is understandable that the requirement for timely referral to an OPO may be overlooked in these circumstances, particularly given the infrequency of pediatric organ donation. Variability in healthcare provider attitudes, knowledge, and behaviors are also significant barriers to timely organ donation referral (9). Although hospitals maintain established clinical triggers for OPO notification, the decision to call the OPO may be fraught with emotion, especially if the care team has developed an attachment to the patient and family. Some providers screen families for receptivity to organ donation before making the referral call, although evidence shows that healthcare providers incorrectly predict a family s initial response to the request for donation more than half of the time (9,10). The CDS system described in this study overcomes these barriers by automating the process for referral based on completely objective triggers. In fact, the popup notification that required approval of email notification prior to sending was removed specifically to eliminate variability observed during initial implementation. Using the fully automated CDS system, we were able to significantly improve OPO referral time for children in our pediatric critical care units. In addition to shortening time to referral, we also demonstrated a significant increase in organ donors per pediatric critical care unit death after implementation of the electronic CDS system. Spain, which has achieved the highest number of donors per capita in the world, considers early referral an essential part of the success in achieving a target goal of 40 deceased organ donors per million population (11). Although our institution does not have an intentional process of intensive care to facilitate organ donation, as described in the Spanish model, it is conceivable that early recognition of organ donation potential during the post-cds period increased the tendency of the care team to continue critical care for patients with devastating brain injury until progression to brain death rather than recommend withdrawal of intensive care support, thereby increasing the potential donor pool during this period (12). Although our study is limited by small sample size, it is likely that this trend toward a higher percentage of death by neurologic criteria American Journal of Transplantation 2017; 17: 2186 2191 2189

Zier et al 100% 3% 17% 80% 43% 10% Proportion of notifications 60% 40% 10% 10% 70% > 10 hours >3 to 10 hours >1 to 3 hours 1 hour 20% 36% 0% Pre-CDS Post-CDS Figure 2: Timing of notifications pre-cds and post-cds implementation. CDS, clinical decision support. post-cds, combined with our post-cds donor conversion rate of 90%, is responsible for the increased number of organ donors seen. The increased number of donors per critical care death may also be directly related to the improvement in OPO referral time with our automated system, as we did see a trend, although not a significant difference, toward shorter referral time between donors and nondonors. It would be interesting to see if this difference would be borne out in a larger study. Secular trends also may have contributed to the increased donation rate seen. Initiatives at other institutions aimed at increasing timely notification and improving the consent process have demonstrated improvement in both referral and organ donation conversion rates; however, these qualityimprovement processes require ongoing provider participation and have not resulted in sustained improvements (13). It would be expected that referrals generated by our fully automated system would continue to meet timely referral goals in most circumstances. It remains to be seen whether the improvement in donation rates demonstrated in this study will be sustained over time. In addition to the impact on OPO referral time and organ donation, the successful implementation of this CDS system is important for a number of reasons. First, it provides proof of concept that a fully automated CDS system for OPO notification of impending brain death can be implemented in a pediatric setting without missing appropriate referrals or generating a burdensome number of false-positive referrals. False-positive referrals are easily identified during the initial validation telephone call from the OPO donation resource coordinator to the provider, at which time the referral is closed without affecting patient care. Second, it is a low-cost intervention, with implementation costs limited to the technology time to integrate the CDS system into the existing EHR and the brief training time for healthcare providers. Third, this automated intervention provides an efficient method for OPO notification that occurs without provider workflow interruption. Finally, this system is not dependent on patient demographic factors and could be implemented easily in any health system with an EHR. Since our data were collected, the state of Michigan has begun investigating an automated OPO referral system (14). It should be noted that we also had more patients referred to the OPO who were discharged alive during the post-cds period compared with pre-cds. These patients all met the clinical triggers for referral but did not progress to brain death. An OPO representative approached one family during the post-cds period when a parent requested information on donation during contemplation of withdrawal of life-sustaining treatment. OPO personnel approached another family when they initially decided to withdraw life-sustaining treatment. Although both families opted to continue aggressive support, none of the narrative notes suggest that they were distressed by their discussions with the OPO. There was 2190 American Journal of Transplantation 2017; 17: 2186 2191

Automated Referral for Impending Brain Death no indication in the EHR that families of the remaining patients were aware of the OPO referral, although this was not specifically studied. This study has a number of limitations. Our small sample size and single-hospital-system population may limit the generalizability of our study. It is also possible that increased awareness of the importance of organ donation accompanied the implementation of the CDS system and that this, rather than the improved time to referral, may have had an impact on the increased number of organ donors. If this is indeed the case, it may represent evidence of sustained positive outcome of the intervention because there was a 2-year interval between initial CDS implementation and the start of the post-cds study period, with periodic adjustments to the algorithm occurring without bedside provider knowledge. Finally, we had patients that did not meet the 1-h notification threshold, despite automation. This situation can occur, for example, if a nurse who is involved with the active resuscitation of a severely injured child does not input assessments into the EHR in real time, thereby delaying the initiation of the CDS algorithm. This situation can be addressed through caregiver training. Conclusions Use of an electronic CDS system in a pediatric hospital setting improved timely OPO notification and was associated with increased organ donation. Implementation and evaluation of CDS systems in other pediatric or adult settings is recommended, as are additional studies to address the direct impact of automated OPO notification on organ donation rates. Acknowledgments Thanks to Tricia Munter and Mary Mattson for their technical assistance with development and implementation of the clinical decision support system. Disclosure The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation. 2. Shafer TJ, Wagner D, Chessare J, et al. US organ donation breakthrough collaborative increases organ donation. Crit Care Nurs Q 2008; 31: 190 210. 3. Centers for Medicare Medicaid Services. Medicare and medicaid programs; hospital conditions of participation; identification of potential organ, tissue, and eye donors and transplant hospitals provision of transplant-related data HCFA. Final rule. Fed Regist 1998; 63: 33856 33875. 4. Colpaert K, Hoste EA, Steurbaut K, et al. Impact of real-time electronic alerting of acute kidney injury on therapeutic intervention and progression of RIFLE class. Crit Care Med 2012; 40: 1164 1170. 5. Amland RC, Lyons JJ, Greene TL, Haley JM. A two-stage clinical decision support system for early recognition and stratification of patients with sepsis: An observational cohort study. JRSM Open 2015; 6: 2054270415609004. 6. Okumura LM, Veroneze I, Burgardt CI, Fragoso MF. Effects of a computerized provider order entry and a clinical decision support system to improve cefazolin use in surgical prophylaxis: A cost saving analysis. Pharm Pract (Granada) 2016; 14: 717. 7. Amland RC, Dean BB, Yu H, et al. Computerized clinical decision support to prevent venous thromboembolism among hospitalized patients: Proximal outcomes from a multiyear quality improvement project. J Healthc Qual 2015; 37: 221 231. 8. Kharbanda AB, Madhok M, Krause E, et al. Implementation of electronic clinical decision support for pediatric appendicitis. Pediatrics 2016; 137: e20151745. 9. Traino HM, Alolod GP, Shafer T, Siminoff LA. Interim results of a national test of the rapid assessment of hospital procurement barriers in donation (RAPiD). Am J Transplant 2012; 12: 3094 3103. 10. Siminoff LA, Gordon N, Hewlett J, Arnold RM. Factors influencing families consent for donation of solid organs for transplantation. JAMA 2001; 286: 71 77. 11. Matesanz R, Domınguez-Gil B, Coll E, Mahıllo B, Marazuela R. How Spain reached 40 deceased organ donors per million population. Am J Transplant 2017. https://doi.org/10.1111/ajt.14104. [Epub ahead of print]. 12. Domınguez-Gil B, Coll E, Elizalde J, et al. Expanding the donor pool through intensive care to facilitate organ donation: Results of a spanish multicenter study. Transplantation 2017. https://doi. org/10.1097/tp.0000000000001701. [Epub ahead of print]. 13. Koh HK, Jacobson MD, Lyddy AM, et al. A statewide public health approach to improving organ donation: The Massachusetts Organ Donation Initiative. Am J Public Health 2007; 97: 30 36. 14. Gift of Life Michigan and Michigan Health Information Network Shared Services. Use Case Summary: Organ Donor Notifications. East Lansing, MI: Michigan Health Information Network Shared Services. 2016 Aug 18 [cited 2017 Mar 15]. Available from: https://mihin.org/wp-content/uploads/2013/07/mihin-ucs- Organ-Donor-Notifications-v10-08-18-16.pdf. References 1. Committee on Hospital Care. Section on Surgery, and Section on Critical Care. Policy statement pediatric organ donation and transplantation. Pediatrics 2010; 125: 822 828. American Journal of Transplantation 2017; 17: 2186 2191 2191