Implementation of 2HELPS2B Seizure Risk Score

Abstract

Background and Objectives

Continuous EEG (cEEG) has become a standard for monitoring critically ill patients, but it is resource-intensive with limited availability. The 2HELP2B seizure risk score can help stratify seizure risk and aid in clinical decision making to optimize duration of monitoring. This study aimed to incorporate the 2HELPS2B score to inform cEEG duration and provide cost-effective care without compromising seizure detection.

Methods

We conducted a quality improvement study that targeted clinical workflow and seizure risk stratification in the intensive care units of a tertiary academic hospital. The study included adult patients who underwent cEEG between June 2020 and December 2022 (n = 552), after excluding patients undergoing cEEG for management of status epilepticus, spell characterization, intracranial pressure monitoring, and post–cardiac arrest (n = 129). We performed a retrospective chart review to establish baseline cEEG volume, seizure incidence, and monitoring duration. We then introduced the 2HELPS2B risk score through multidisciplinary education and used published recommendations to suggest optimal cEEG duration. After the intervention, we analyzed the impact of integrating the 2HELPS2B score on cEEG duration and seizure detection rates.

Results

Of 552 patients, most were low risk (n = 311, 56.3%), followed by moderate risk (n = 189, 34.2%) and high risk (n = 52, 9.4%). Before the intervention, cEEG duration was similar for all risk groups. After implementation of the 2HELPSB score, there was a significant reduction in cEEG duration for low-risk and moderate-risk patients (low 36.3 vs 23.8 hours; p < 0.0001, moderate 36.5 vs 29.3 hours; p = 0.01) and no significant change for the high-risk group (41.3 vs 40.4 hours; p = 0.92). Seizure detection was low except for the high-risk group (1.3% vs 7.9% vs 39.1%). Reduction in cEEG duration after implementation of the 2HELPS2B score did not lead to a significant change in seizure detection (0.6% vs 9% vs 37.9%).

Discussion

Most critically ill patients had low or moderate seizure risk and, accordingly, a low incidence of seizures detected during cEEG. Implementing the 2HELPS2B seizure risk score allowed customization of cEEG duration for individual patients, applying the practice of precision medicine. This approach successfully improved cEEG utilization without compromising seizure detection. In conclusion, implementing seizure risk stratification can provide cost-effective monitoring and improve cEEG access.

Introduction

Neuromonitoring is essential in critically ill patients. Continuous EEG (cEEG) has historically been one of the best noninvasive tools to assess cerebral activity and monitor changes in brain function.¹ Critically ill patients, particularly those with altered mental status and acute brain injury, are at increased risk of seizures, many of which are acute symptomatic seizures.^2-6 Studies show that seizures occur in up to 19% of all critically ill patients, with most being electrographic only.^2,3,5,7 Electrographic status epilepticus (ESE) is observed in approximately 14% of patients in coma and is not always preceded by clinical seizures.^5,8 In one study, 75% of ESE cases were not preceded by clinical seizures.⁹ Both clinical and electrographic seizures are associated with poor neurologic outcomes and increased mortality.^10-12 While nonmodifiable risk factors, such as the underlying brain injury etiology, patient age, and comorbidities, have been identified as contributors to worse outcomes in SE, studies also highlight that time to initial treatment and duration of SE are major determinants of mortality.¹³ Therefore, timely detection and management of seizures is paramount. The American Clinical Neurophysiology Society (ACNS) and other international societies recommend at least 24 hours of cEEG to detect seizures in patients with altered mental status and >48 hours in comatose patients, which has led to a significant increase in the use of cEEG over the past decade.^14-18 However, cEEG is a labor-intensive and costly procedure, and only a limited number of institutions possess the resources to monitor patients for extended periods.^19,20 The high cEEG demand necessitates an appropriate turnaround time for cEEG.^20,21 Tailoring cEEG duration without compromising patient care in settings where the demand exceeds supply is crucial for appropriate resource utilization.

Studies suggest that a combination of EEG and clinical parameters can be used to assess a patient's seizure risk.^7,22 Clinical criteria alone are not sufficient, and incorporation of EEG findings can improve seizure prediction.^23-28 Periodic or rhythmic EEG patterns, including lateralized periodic discharges (LPDs), lateralized rhythmic delta activity (LRDA), and bilateral independent periodic discharges (BIPD), have been shown to be highly associated with seizures.^7,22,29 The 2HELPS2B scoring system is a validated point system developed by Struck et al. to stratify seizure risk based on EEG features and clinical factors.^24,30 This scoring system can be used to allocate EEG resources according to need and improve cost-effective care for critically ill patients.

We conducted a quality improvement project to incorporate the 2HELPS2B scoring system as a tool to improve EEG workflow, assist with seizure risk stratification, and individualize duration of cEEG in critically ill patients.

Methods

Standard Protocol Approvals, Registrations, and Patient Consents

This study was approved by the University of North Carolina University Institutional Review Board (IRB number 21–2557). This study follows the Standards for Quality Improvement Reporting Excellence reporting guidelines.³¹

Methodology

Using Lean methodology and the Plan-Do-Study-Act (PDSA) cycle, one of the central quality improvement tools,^32-34 we aimed to understand current cEEG utilization and workflow in our institution to improve cEEG access. Our objective for this project was to incorporate the 2HELP2B score to stratify seizure risk of critically ill patients and aid in clinical decision making to reduce cEEG duration and improve cEEG access without compromising seizure detection.

Study Population

Our study included adult patients older than 18 years undergoing cEEG in our institution's intensive care units (ICUs) (medical, cardiac, surgical, and neuroscience ICU [NSICU]) between June 2020 and December 2022. We excluded patients undergoing cEEG with indications other than solely for seizure detection: management of SE, intracranial pressure (ICP) monitoring, optimization of antiseizure medications (ASMs), capturing events concerning for seizures (spell characterization), and post–cardiac arrest.

Data Collection and Calculation of the 2HELPS2B Score

We conducted a retrospective chart review, analyzing patient demographics, occurrence of clinical seizures before cEEG, history of epilepsy or acute brain injury, and cEEG findings. As part of standard practice, cEEG reports were generated by board-certified clinical neurophysiologists, who interpreted EEGs using ACNS standardized critical care EEG terminology, which has high interrater reliability.^35,36 The study initially followed the 2013 version of the terminology and the 2021 version subsequent to publication in January 2021. Based on cEEG reports and patient history, we calculated the 2HELPS2B seizure risk score as summarized in Table 1.^24,30 We noted the overall duration of cEEG, incidence of seizures, time to initial seizure onset (when seizures were present), and cEEG reconnection rates. We then compared cEEG duration for various seizure risk groups with published recommendations (Table 2).³⁰ Patients who had seizures captured during cEEG were excluded from the calculation of monitoring duration because additional cEEG monitoring was indicated to ensure appropriate treatment and termination of seizures. However, these patients were included in the calculation of seizure incidence.

Table 1.

Calculation of the 2HELPS2B Risk Score³⁰

Abbreviation	Variable	Points
2H	>2-Hz pattern (except GRDA)	1
E	Epileptiform discharges, sporadic	1
L	Lateralized discharges (LPD, BIPD, LRDA)	1
P	Plus features (superimposed rhythmic, sharp or fast activity)	1
S	History of seizures	1
2B	Brief rhythmic discharges	2

Abbreviations: BIPD = bilateral independent periodic discharge; GRDA = generalized rhythmic delta activity; LPD = lateralized periodic discharges; LRDA = lateralized rhythmic delta activity.

Table 2.

Risk Stratification and Recommended cEEG Duration³⁰

2HELPS2B risk group	Seizure risk at 72 h, (%)	Duration of monitoring Seizure risk <5%	Duration of monitoring Seizure risk <2%
Low risk, score 0	3.1	1 hour	3.3 hrs
Medium risk, score 1	12	12 hours	29 hrs
High risk, score ≥2	>25	24–48 hours	>30 hrs

Abbreviation: cEEG = continuous EEG.

Implementation

We conducted interviews with NSICU providers and staff, including epileptologists, neurointensivists, EEG technologists, advanced practice providers, and nurses, to better understand cEEG workflow and identify potential barriers to a standardized approach for determining cEEG duration. We identified patients for whom shorter cEEG duration might be adequate based on published recommendations for cEEG duration stratified by the 2HELPS2B risk score.²⁴ The resultant multiphasic PDSA quality improvement project focused on (1) increasing awareness of the 2HELPS2B score as a validated seizure risk scoring system, (2) improving communication of EEG findings including seizure risk score, and (3) using data to drive decisions around when to terminate monitoring.

PDSA Phase 1

Phase 1 emphasized education about the 2HELPS2B score and communication of EEG findings with the entire NSICU team (advanced practice providers, residents, fellows, and physicians). Educational efforts took the form of team meetings, interactive journal club sessions, and multidisciplinary grand rounds lectures. Relevant background information was introduced as well as the study's aims and the expectations from participating team members.

Optimizing communication centered on timely relay of cEEG findings and report of the 2HELPS2B score to the clinical team, which was followed by an interdisciplinary discussion of the optimal duration of monitoring based on the risk score. During the day, EEG technologists would immediately inform the clinical neurophysiologist of new cEEG studies and, within 2 hours, the clinical neurophysiologist would report preliminary EEG findings and the 2HELPS2B score. For overnight connections, EEG studies that started after 5 pm, the clinical neurophysiologist would report EEG findings and seizure risk score the following morning. A “daily touchpoint” was established to facilitate discussion of relevant EEG findings and enable joint decision making regarding the optimal duration of cEEG monitoring for individual patients. This daily touchpoint involved a live discussion conducted in person or through phone, between the NSICU attending and the clinical neurophysiologist. These discussions were scheduled in the mornings, after the analysis of overnight EEG studies. To ensure consistent communication among physicians, standardized discussion points were introduced through written and verbal instructions during the PDSA cycles of our quality improvement project. These discussion points included the presence of seizures or periodic and rhythmic patterns, changes in the patient's clinical status and adjustments in ASMs or sedatives, and whether future cEEG monitoring was clinically indicated. If a joint decision was reached to discontinue EEG monitoring, the decision was promptly communicated to the EEG technologists. Our institution provides 24/7 in-house EEG technician staffing, enabling timely disconnection of patients from EEG monitoring once clinical decision was made. However, because reporting of the 2HELPS2B score occurred during daytime hours only, EEG monitoring continued until the following morning and discussion during the “daily touchpoint.”

PDSA Phase 2

In the second phase of the project, we incorporated the 2HELPS2B risk score into NSICU workflow and communication. This was aided by the introduction of “smartphrases” that were used for documenting the seizure risk score in both EEG reports and ICU progress notes. The “smartphrases” outlined the components of the risk score, patient populations for which utilization of the score was not appropriate, and the recommended duration of cEEG based on the score (eTable 1).

PDSA Phase 3

After successfully implementing the scoring system in the NSICU, we expanded use to other ICUs during the intervention's third phase. The inpatient neurology consult team aided in communication with clinical teams and provided the “daily touchpoint.” To enable implementation of the project beyond the NSICU, we held didactics for both residents and physicians outside the NSICU staff. We optimized workflow between non-neurology clinical teams, the general neurology consult service, and the EEG team. Data were collected at various phases of the intervention, and a time line was built to assess the efficacy of the intervention and address any obstacles to the implementation of the project beyond the NSICU.

Statistical Analysis

We summarized and compared the clinical characteristics of the study populations using frequencies and percentages for categorical variables and mean values with SDs for continuous variables. We performed Welch 2-sample t tests to compare the average cEEG duration between preintervention and postintervention groups.

To ensure that reductions in cEEG duration did not compromise seizure detection rates or result in increased cEEG reconnections, we performed a Cochran-Mantel-Haenszel test to evaluate seizure incidence and cEEG reconnections between preintervention and postintervention groups stratified by 2HELPS2B score. In addition, we performed 2-way analysis of variance to compare whether time to detection of the first seizure was statistically different between preintervention and postintervention groups.

Data Availability

Data sets are available from the corresponding author for any qualified investigator.

Results

Of the 681 adult patients who underwent cEEG in ICUs in our institution during the period, 552 patients met the inclusion criteria. A total of 129 patients were excluded based on primary indication for cEEG: SE (n = 68), post–cardiac arrest (n = 26), ICP monitoring (n = 12), ASM optimization or sedation wean (n = 16), and spell characterization (n = 7).

Of the patients who met the inclusion criteria, 269 patients were in the preintervention group and 283 patients were in the postintervention group. Most patients were in the NSICU (78.8%), followed by medical ICU (15.6%) and other ICUs (5.6%). The preintervention and postintervention groups were similar regarding demographics and location (eTable 2).

Among the entire cohort, most patients fell into the low-risk category (n = 311, 56.3%), followed by moderate risk (n = 189, 34.2%) and high risk (n = 52, 9.4%). The percentage of patients in each seizure risk group did not differ significantly between the preintervention and postintervention groups (Table 3).

Table 3.

Duration of cEEG Monitoring, Seizure Incidence, and Time To Initial Seizure

2HELPS2B risk group	Preintervention	Postintervention	p Value
	Average duration of cEEG, ha (n = 251, std 25.1)	Average duration of cEEG, ha (n = 262, std 17.5)
Low-risk score = 0N = 308	36.3 (n = 155, std 27.7)	23.8 (n = 153, std 14.9)	<0.01
Moderate-risk score = 1N = 173	36.5 (n = 82, std 20.6)	29.3 (n = 91, std 18.1)	0.01
Higher risk score ≥2N = 32	41.3 (n = 14, std 19.4)	40.4 (n = 18, std 25.7)	0.92
	Seizure detection rate (n = 269)	Seizure detection rate (n = 283)
Low-risk score = 0N = 311	2/157 = 1.3%	1/154 = 0.6%	0.97
Moderate-risk score = 1N = 189	7/89 = 7.9%	9/100 = 9.0%
Higher risk score ≥2N = 52	9/23 = 39.1%	11/29 = 37.9%
	Average time to first seizure, h (n = 18)	Average time to first seizure onset, h (n = 21)
Low-risk score = 0	14 (n = 2, std 18.4)	24 (n = 1, std NA)	0.37
Moderate-risk score = 1N = 16	15.1 (n = 7, std 6.72)	6.72 (n = 9, std 7.86)
Higher risk score ≥2N = 20	10.4 (n = 9, std 11.0)	6.82 (n = 11, std 7.69)

Abbreviation: cEEG = continuous EEG.

After intervention, monitoring duration was reduced for all groups with statistically significant reduction in the low-risk and moderate-risk groups. Seizure detection rates did not differ significantly between the preintervention and postintervention groups. The time to detect first seizure varied, and the difference was not statistically significant between seizure risk and preintervention and postintervention groups.

^aAssessment of duration excludes patients with seizures detected during cEEG.

Before the intervention, the durations of monitoring for low-risk, moderate-risk, and high-risk groups were 36.3, 36.5, and 41.3 hours, respectively. After the completion of the project, monitoring duration for low-risk and moderate-risk groups was significantly reduced to 23.8 hours (p < 0.01) and 29.3 hours (p = 0.01). The cEEG duration did not change significantly for the high-risk group (41.3 vs 40.4 hours; p = 0.92) (Table 3 and Figure A).

Figure. Duration of cEEG Monitoring (Hours) per Risk Group.

(A) Monitoring duration was reduced for all groups with overall statistically significant reduction for the low-risk and moderate-risk groups. (B) The change in cEEG monitoring duration (hours) over time during various phases of the plan-do-study-act (PDSA) cycle per risk group. During phase 1, there was a statistically significant reduction in duration of monitoring for the low-risk group only. During subsequent phases II and III of the cycle, there was a statistically significant reduction in duration of monitoring for both low-risk and moderate-risk groups. cEEG = continuous EEG.

We also calculated cEEG monitoring duration for each individual phase of the project. The Figure B shows the monitoring duration for each seizure risk group over time during the various phases of the project. After 5 months (after phase I), monitoring duration for the low-risk group was significantly reduced (25.1 hours; p < 0.01). In the subsequent 9 months (after phases II and III), the monitoring duration was further reduced in both low-risk and moderate-risk groups, 21.6 hours (p < 0.01) and 25.3 hours (p = 0.01), respectively.

The incidence of seizures detected during cEEG in the total cohort was 7.1%. When stratified by risk category, the seizure incidence was 1%, 8.5%, and 38.5% for low-risk, medium-risk, and high-risk patients, respectively. Before the intervention, the seizure detection rate was 1.3%, 7.9%, and 39.1% with no significant change after implementation of the score, 0.6% vs 9% vs 37.9% for low-risk, medium-risk, and high-risk patients, respectively (p = 0.97) (Table 3).

Furthermore, we analyzed the time to detect first seizure within risk groups. The average time to detect first seizure (hours) was 17.3 (n = 3, std 14.2), 10.4 (n = 16, std 8.34), and 8.45 (n = 20, std 9.26), for low-risk, moderate-risk, and high-risk patients, respectively, and there was no significant change when comparing preintervention and postintervention groups (p = 0.37) (Table 3).

After 2HELP2BS score implementation, cEEG reconnection rates decreased from 20.1% to 17.7% with the largest decrease in the high-risk group from 39.1% to 24.1%. However, there was no statistically significant difference in reconnection rates between preintervention and postintervention groups when stratified by 2HELPS2B risk score (p = 0.47) (Table 4). Seizure incidence among patients who were reconnected to EEG remained low except for the high-risk group (31.3%), and there was no statistically significant difference in reconnection rates between preintervention and postintervention groups when stratified by 2HELPS2B risk score (p = 0.88) (Table 4).

Table 4.

cEEG Reconnection Rates per Risk Group

2HELPS2B risk group	Preintervention reconnection rates (n = 269), (%)	Postintervention reconnection rates (n = 283), (%)	Overall reconnection rates, (%)	p Value
Low-risk score = 0N = 311	27/157 = 17.2	23/154 = 14.9	50/311 = 16.1	0.47
Moderate-risk score = 1N = 189	18/89 = 20.2	20/100 = 20.0	38/189 = 20.1
Higher risk score ≥2N = 52	9/23 = 39.1	7/29 = 24.1	16/52 = 30.8
All risk groups	54/269 = 20.1	50/283 = 17.7	104/552 = 18.8
	Seizure incidence in preintervention reconnection of EEGs (n = 54) (%)	Seizure incidence in postintervention reconnection of EEGs (n = 50) (%)	Overall seizure incidence in reconnection of EEGs (%)
Low-risk score = 0N = 50	1/27 = 3.7	0/23 = 0	1/50 = 2.0	0.88
Moderate-risk score = 1N = 38	1/18 = 5.6	0/20 = 0	1/38 = 2.6
Higher risk score ≥2N = 16	2/9 = 22.2	3/7 = 42.9	5/16 = 31.3

Abbreviation: cEEG = continuous EEG.

Reduction in duration of monitoring did not result in an increase in repeat cEEG sessions, and the seizure incidence was low except for the high-risk group. There was no statistical difference in overall reconnection rates and seizure incidence between preintervention and postintervention groups when stratified by the 2HELPS2B score.

The most common indications for repeat cEEG monitoring were altered mental status (43%, n = 45), clinical activity concerning for seizures (25%, n = 26), and focal examination changes (24%, n = 25). In addition, in 3 patients (2.9%), cEEG was disconnected to prevent skin breakdown (scalp break) and later, cEEG was reconnected. In 5 patients (4.8%), the EEG was reconnected for unknown reasons.

Discussion

Although previously published data suggested an EEG duration of up to 48 hours to detect 95% of seizures in critically ill patients,⁷ our data demonstrate that individualized seizure risk stratification using the 2HELPS2B seizure risk score can reduce cEEG duration without compromising seizure detection rates. Reducing cEEG duration for select patients also improves EEG access and resource utilization that is most evident for patients with low and moderate seizure risk, which comprised 87% (481/552) of patients included in our study. Based on the calculated preintervention and postintervention cEEG durations, we reduced cEEG use by approximately 2,600 hours over the course of a year, ultimately increasing cEEG access to additional patients. Although we were unable to calculate the direct cost savings associated with implementing our project within our own institution, we can infer potential savings from previous studies. Abend et al. indicated that reducing the monitoring duration for low-risk patients from 48 hours to 24 hours decreases the incremental costs of detecting an additional seizure in this group from $75,000 to $5,000.¹⁹ This demonstrates that a risk-stratified approach to EEG monitoring is highly cost-effective.

There remains room for further improvement because, after this intervention, low-risk patients were still undergoing cEEG for an average duration of 24 hours, which is significantly longer than the recommended monitoring duration for this risk group.²⁸ The extended duration was partly due to communication of the 2HELP2B score being limited to daytime hours. For low-risk patients, the suggested recording time ranges from 1 to 3.3 hours. Had the 2HELPS2B score been assessed and communicated 24/7, significantly more resources could have been conserved.²⁸ There are additional potential reasons for the extended monitoring durations observed in our low-risk patient cohort compared with what is recommended. One is the adherence to older practices of 24–48 hour cEEG monitoring, which were based on earlier data that did not use risk stratification, following a “one size fits all” approach.^7,14 In addition, uncertainty exists about sedation or recent load of ASMs might affect the reliability of the 2HELPS2B score. This uncertainty was frequently discussed during the daily touchpoints, leading to a tendency to extend cEEG monitoring beyond the recommended duration to allow time for reduction in sedation and reevaluation of the risk score.

Our results further validate the use of the 2HELPSB score for seizure risk stratification, demonstrating that the seizure risk category strongly aligns with seizure incidence. Although the overall incidence of seizures in our cohort (7.1%) is lower compared with previous studies,^7,8,17,26 variability in seizure frequencies between studies may result from differences in study population and selection bias. In our study, the exclusion of high-risk patients being treated for SE or post–cardiac arrest is likely a significant factor. These patients were excluded from our evaluation, given that the indication for cEEG was not purely for seizure detection, and EEG duration is determined by additional factors including response to treatment and time-sensitive prognostication. However, seizure incidence stratified by risk group was similar to the predicted probability of seizures in the published validated cohort.^26,30 Therefore, the fact that most of our patient population was either in low-risk or moderate-risk categories would explain the overall low seizure incidence.

Furthermore, the implementation of the PDSA qualitative improvement model highlights the ability to improve workflow by enhancing communication between the clinical and EEG teams and expediting decision making. The morning “daily touchpoint” facilitated prompt decisions regarding EEG discontinuation, guided by the 2HELPS2B seizure risk score. As a result, cEEGs were discontinued earlier in the day, allowing for the allocation of EEG resources to other patients in need. In addition, our clinical teams across all ICUs of our hospital acquired valuable awareness of our institutional cEEG capacity, which led to further improvement in resource utilization.

Results from phase 1 of our PDSA cycle showed reduction in the duration of cEEG for the low-risk group simply by educating clinical teams about the seizure risk score and improving communication. Documenting the 2HELPS2B score in EEG reports and progress notes further enhanced implementation of the risk score in the NSICU (phase 2) and other ICUs later (phase 3), resulting in a reduction of cEEG duration for not only low-risk patients but also the moderate-risk group although the reduced EEG duration is also likely in part due to increased sample size.

The time to detect a first seizure varied between risk groups, although the difference was not statistically significant. However, our sample size of patients with electrographic seizures was small, especially for the low-risk group, which limits conclusions that can be drawn. Only 3 of 311 patients (0.9%) in the low-risk group experienced seizures during cEEG with an average detection time of 17.3 hours, much longer than the recommended monitoring time.³⁰ A potential explanation is that medications (ASMs and sedation) may have suppressed hyperexcitable patterns in the first few hours of EEG assessment, hence lowering the initial 2HELPS2B score. However, with less than 1% of our low-risk patients experiencing seizures, the recommended monitoring time of 3.3 hours to achieve less than 2% seizure risk is still appropriate.^19,30 Furthermore, detecting seizures with 100% sensitivity has been shown to incur high economic cost without a clear prognostic benefit, particularly among low-risk patients.^19,37 Prolonged cEEG monitoring of low-risk patients can also limit EEG access for higher risk patients. Therefore, a tailored approach to optimize cEEG based on seizure risk is necessary because benefit diminishes when patients who are at low seizure risk are monitored for extended periods.^19,38

Critically ill patients with acute brain injury often experience significant fluctuations in clinical status, which can raise concern for electrographic seizures despite previous EEG monitoring sessions that were unremarkable.^2,3,6,27 Despite the shorter initial cEEG durations, there was actually a trend toward decreased cEEG reconnection rates. Although the difference in reconnection rates between preintervention and postintervention groups was not statistically significant, this trend is noteworthy. Patients in the high-risk seizure category continued to have high incidence of seizures on reconnection (31.3%), indicating that this patient population should have a relatively low threshold for additional EEG monitoring sessions. On the contrary, the seizure incidence for low-risk and moderate-risk patients who were reconnected to EEG remained low, consistent with the original incidence. These findings suggest that the 2HELPS2B score could be used for cEEG resource allocation when considering recurrent EEGs. However, further prospective studies are needed to support this argument.

A few limitations should be noted. First, we calculated the 2HELPS2B score for the postintervention group prospectively, whereas the score was calculated retrospectively for the preintervention group and relied on data recorded in clinical EEG reports. This posed a limitation because not all clinical neurophysiologists incorporated ACNS terminology in their EEG reports before the intervention. Although they were the minority, we emphasized the importance of using the standardized terminology during the education and communication phase of the project. Second, it is uncertain what effect sedation and ASM loads may have on the validity of the 2HELPS2B score as noted previously, particularly for postoperative patients and patients requiring ongoing sedation for ventilatory management.

Despite these limitations, our results demonstrate that the 2HELPS2B seizure risk score can be successfully implemented at an institutional level to reduce duration of cEEG for patients at low or moderate risk of seizures, without affecting seizure detection or resulting in additional cEEG sessions. We provide a guide for effectively incorporating the 2HELPS2B seizure risk score by using the PDSA quality improvement model with a focus on education, communication, and step-by-step expansion throughout ICUs of the hospital system.

This quality improvement project reports the incorporation of the 2HELPSB score into daily workflow, demonstrating a reduction in cEEG monitoring for patients who are at low and moderate risk of seizures without compromising seizure detection. The multiphasic PDSA approach makes the implementation of the 2HELPS2B score feasible and easy to replicate. Widespread implementation of the seizure risk score by other institutions can improve cost-effectiveness, potentially leading to health care savings and provision of high-value care.

Author Contributions

F. Aseem: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. E. Fink: drafting/revision of the manuscript for content, including medical writing for content; study concept or design. C. Liu: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. J. Whalen: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. J. Werdel: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. P. Nanavati: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. F. Zou: analysis or interpretation of data. A. Wabulya: drafting/revision of the manuscript for content, including medical writing for content; study concept or design. C. Olm-Shipman: drafting/revision of the manuscript for content, including medical writing for content; study concept or design. S.M. LaRoche: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. C. Rubinos: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data.

Study Funding

The authors report no targeted funding.

Disclosure

C. Rubinos receives funding from the Physician-Scientist in training Program from the University of North Carolina School of Medicine, consultation fees for advising from Quatro Consulting Inc., Marinus and Azurity pharmaceuticals, speaker honoraria from American Epilepsy Society and Marinus pharmaceuticals for education content development. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/cp.