Developing consensus-based indicators for intensive care unit capacity and capability in South Korea: a modified Delphi study

Article information

Acute Crit Care. 2026;41(2):282-294

Publication date (electronic) : 2026 May 28

doi : https://doi.org/10.4266/acc.000676

Jeehye Lee¹, Ho Kyung Sung^2,3, Min-hwan Oh⁴, Jin Young Kim⁵, KyoungHo Kim⁶, Jae Young Moon^,7

¹Department of Preventive Medicine, Konkuk University College of Medicine, Chungju, Korea

²National Emergency Medical Center, National Medical Center, Seoul, Korea

³Public Health Research Institute, National Medical Center, Seoul, Korea

⁴Graduate School of Data Science, Seoul National University, Seoul, Korea

⁵Division of Respiratory Allergy and Critical Care Medicine, Department of Internal Medicine, Soonchunhyang University Cheonan Hospital, Cheonan, Korea

⁶Division of Emergency and Critical Care Medicine, Department of Emergency Medicine, Chungnam National University Sejong Hospital, Sejong, Korea

⁷Department of Pulmonary and Critical Care Medicine, Chungnam National University Sejong Hospital, Sejong, Korea

Correspondence: Jae Young Moon Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, Chungnam National University Sejong Hospital, Chungnam National University College of Medicine, 20, Bodeum 7-ro, Sejong 30099, Korea Tel: +82-42-280-7142 Fax: +82-42-257-5753 Email: diffable@hanmail.net

Received 2026 January 13; Revised 2026 March 16; Accepted 2026 March 25.

Abstract

INTRODUCTION

Intensive care units (ICUs) are essential components of modern healthcare systems that provide life-sustaining treatment for patients with acute, life-threatening conditions. Critical illness has been defined as a state characterized by vital organ dysfunction, a high risk of imminent death if care is not provided, and potential reversibility [1]. As a multidisciplinary and interprofessional specialty dedicated to the comprehensive management of such patients, critical care requires highly specialized personnel, advanced equipment, and substantial resources [2].

Many developed countries have established standards for ICUs, classify and manage them accordingly, and have quality management programs at the professional society or national level [3,4]. This is important because intensive care requires highly specialized professionals and multidisciplinary teams, as well as expensive equipment for life support [2,3], which makes geographic inequalities highly likely [5-7]. Moreover, for the reasons stated above, the inefficient use of intensive care resources can negatively affect not only hospitals but also the entire healthcare system. The importance of ICU capacity management was highlighted during the coronavirus disease 2019 (COVID-19) pandemic and other crises [8-12]. To respond effectively and efficiently to future public health crises, it is crucial to establish standards and manage the capacity of ICUs.

Korea currently has no capability-based standards for classifying ICUs [13,14]. In practice, critically ill patients enter a system in which the availability of a bed often outweighs the suitability of care, and there is no established mechanism to match patient acuity with ICU capability [15]. As a result, critically ill patients are frequently assigned according to bed availability rather than clinical need, and transfers depend on informal arrangements within the emergency network, rather than a defined critical-care system [16]. This can leave the most severely ill patients vulnerable to delays and mismatches between acuity and capability. To address this system-level gap, we used a modified Delphi process to identify consensus-based indicators that define ICU capability and can serve as the foundation for a future capability-oriented classification and transfer framework.

MATERIALS AND METHODS

This study was approved by the Institutional Review Board of Chungnam National University Sejong Hospital (No. CNUSH 2025-01-003-004). All participants provided informed consent and were assured of anonymity throughout the Delphi process.

Delphi Rounds

Three online survey rounds were conducted between April and July 2025. Table 1 summarizes the survey rounds and response rates. In Round 3, borderline items were stratified into two evaluation tracks. Items related to safety infrastructure and mandatory operational requirements (e.g., negative pressure rooms, bed count indicators) were assessed using a paired-item design, in which both a positively phrased and a reverse-phrased question were presented to verify internal consistency. The remaining items were re-evaluated using single positively phrased questions.

Table 1.

Summary of Delphi survey rounds and response rates

RESULTS

Twenty-four experts provided written consent and agreed to participate in the Delphi panel. The response rates were high across all three rounds: 22 experts (91.7%) in Round 1, 20 (83.3%) in Round 2, and 19 (79.2%) in Round 3 (Table 1, Figure 1).

Figure 1.

Study flowchart.

Round 1

In Round 1, 42 candidate indicators were evaluated for importance using a 5-point Likert scale (n=22). Thirty-six indicators met the predefined consensus threshold (CVR ≥0.455 and ≥75% agreement). Items that failed to reach consensus included the presence of a physician on-call room and the absolute number of ICU beds, both of which were judged to insufficiently reflect ICU capability as standalone measures. Following a panel review of ambiguous responses, four items were excluded, two were revised, and four new items were added based on expert suggestions, resulting in a revised pool of 42 items carried forward to Round 2 (Table 2).

Table 2.

Candidate indicators evaluated across Delphi Rounds 1 and 2

Round 3

In Round 3 (n=19), the eight borderline items were evaluated through two tracks. Three items were assessed using a paired-item design, and five were re-evaluated using single positively phrased questions. Those results are summarized in Table 3. All three paired items were rejected. The item "24/7 availability of emergency CABG" showed a logical inconsistency: the positive phrasing received high agreement (CVR 0.789), but the reverse phrasing ("not important") also showed substantial agreement (CVR 0.579), indicating that experts could not consistently determine its importance. "Negative pressure isolation room" was excluded due to unstable consensus, i.e., the CVR difference between the positive and reverse phrasings exceeded the acceptable threshold of 0.2. "Number of operational ICU beds" failed to meet the CVR threshold in both phrasings.

Table 3.

Results of Delphi Round 3: consistency check and re-evaluation of borderline items

Of the five single items, four reached consensus and were accepted: "Central monitoring system" (CVR 0.895), "Single-bed ICU room with central monitoring station" (CVR 0.895), "Isolation bed proportion" (CVR 0.579) and "Hospital-to-ICU bed ratio" (CVR 0.474). One item, "Accredited training institution for critical care medicine," did not meet the CVR threshold (CVR 0.368) and was excluded.

Final Framework

The Delphi process yielded a final consensus-based framework of 29 indicators across four domains (Table 4). The Ward Support Services domain and the Education, Training, Research, and Quality Management domain did not achieve consensus on any of their candidate indicators and were therefore excluded from the final framework. The four retained domains—Capacity (12 indicators), Capability (6 indicators), Human Resources (3 indicators), and Space and Equipment Infrastructure (8 indicators)—constitute the final set of consensus-based indicators for ICU capability assessment in the Korean healthcare context.

Table 4.

Final consensus-based framework for ICU capability assessment (29 indicators)

DISCUSSION

The aim of this study was to develop consensus-based indicators for assessing ICU capacity and capability in South Korea through a three-round modified Delphi process [17-19]. The survey had two distinct research aims: (1) gathering expert consensus on ICU capability indicators, evaluated through Capacity and capability domains; and (2) gathering expert consensus on ICU grading and classification criteria, evaluated through four additional domains—Human resources, Space and equipment infrastructure, Ward support services, and Education, training, research, and quality management. Of the 42 candidate indicators evaluated across the six domains, which were informed by international frameworks such as SCCM, ESICM, and ANZICS [3,20-22], 29 ultimately reached consensus. The Ward support services and Education, training, research, and quality management domains failed to achieve agreement on any of their candidate indicators, resulting in a final four-domain framework: Capacity (12 indicators), Capability (6 indicators), Human resources (3 indicators), and Space and equipment infrastructure (8 indicators). This outcome reflects not merely a statistical finding but a substantive statement by Korean ICU experts about what constitutes the essential identity of an ICU in the current healthcare context.

To characterize the expert agreement beyond binary consensus classification, a post-hoc analysis classified all Round 2 indicators into four evidence-quality types based on their mean scores and CVR values (Supplementary Table 1). Type A (mean ≥8.5, CVR ≥0.9) represented strong, unambiguous consensus; type B (mean ≥8.0, CVR ≥0.5) represented stable moderate-to-high importance with sufficient agreement; type C (7.0≤ mean <8.0, CVR ≥0.5) represented moderate consensus; and type D (mean <7.0 or CVR <0.5) indicated low priority or insufficient agreement.

Fourteen indicators (33%) were classified as type A—the non-negotiable core of ICU capability in the view of Korean experts. These included all life support interventions (LSIs) (LSI provision: mean 9.55, CVR 1.000; 24/7 cardiovascular intervention: mean 9.55, CVR 1.000; 24/7 extracorporeal membrane oxygenation [ECMO]: mean 9.30, CVR 1.000; advanced critical care equipment: CVR 1.000), key staffing indicators (nurse-to-bed ratio: mean 9.45, CVR 0.900; daytime specialist coverage: mean 9.10, CVR 0.900; 24/7 specialist coverage: mean 8.75, CVR 0.900), and full-time intensivist staffing (mean 8.90, CVR 0.900). Six indicators (14%) were classified as type B, including nighttime physician coverage, multidisciplinary rounding, proportion of patients receiving LSI, other critical procedures, and hemodynamic monitoring indicators. Five indicators (12%) fell into type C, and the remaining 17 (40%) into type D.

A notable structural finding is the clear polarization between type A and type D: no indicator simultaneously received high mean scores and low CVR—that is, no indicator that was rated as highly important was deeply contested. When Korean experts agreed something was critical, they agreed strongly; disagreement was concentrated precisely among items considered less essential. The emergence of a distinct type B cluster—moderate-to-high importance with adequate consensus—provides a practical basis for a two-tier policy framework: type A indicators as mandatory minimum standards, and type B indicators as recommended standards for higher-tier ICU designations.

The distribution of the adopted and excluded domains reveals a consistent orientation among Korean experts toward functional capability over structural and institutional attributes. This stands in meaningful contrast to many international ICU frameworks. The SCCM, ESICM, and ANZICS standards incorporate structural indicators—dedicated physical layouts, education program accreditation, and research infrastructure—as integral components of ICU tiering [3,20-22]. Korean experts, by contrast, concentrated their strongest consensus on functional process indicators: 24/7 availability of cardiovascular interventions, LSI provision, and ECMO access—all of which reflect what an ICU can do for a patient at any given moment, rather than what it has in terms of physical infrastructure.

This functional orientation was also evident within the staffing domain. Experts rated the nurse-to-bed ratio (mean 9.45, CVR 1.000) more highly than absolute physician numbers, reflecting recognition that continuous 24-hour patient monitoring depends primarily on nursing presence. Physician staffing can demonstrate a threshold effect [23,24], whereas nursing workload increases proportionally with patient assignment. Because nurses provide continuous bedside care and monitoring, their staffing levels are closely linked to patient safety events. Accordingly, multiple observational studies have identified registered nurse staffing as one of the organizational factors most strongly associated with patient outcomes [25]. Furthermore, experts consistently prioritized the availability of interventions over the number of personnel capable of performing them—illustrated by the contrast between "24/7 cardiovascular intervention availability" (type A, CVR 1.000) and "number of surgeons capable of emergency CABG" (type D, CVR 0.200). The complete exclusion of the Education, training, research, and quality management domain—despite its prominence in international frameworks such as ESICM [21] and ANZICS [22]—suggests that Korean experts drew a deliberate boundary between clinical competence and institutional function. This implies a preference for a classification system in which non-university hospitals with strong clinical capability are not disadvantaged relative to academic centers.

Several indicators that failed to achieve final consensus merit specific discussion, as their exclusion carries substantive meaning. First, "negative pressure isolation rooms," despite initially appearing on the borderline, were excluded in Round 3 due to inconsistent paired-item responses—the CVR difference between the positive and reverse phrasings exceeded 0.2, indicating genuine expert uncertainty rather than clear rejection. Although the 2024 SCCM guidelines explicitly recommend single-patient rooms in ICU design [26], the broader question of mandatory infection control infrastructure remains unsettled in the Korean context and might warrant reconsideration as post-pandemic airborne precaution standards continue to evolve. Second, "24/7 emergency CABG availability" demonstrated a logical inconsistency in Round 3: experts gave high agreement to both the positive phrasing (CVR 0.789) and the reverse phrasing ("not important," CVR 0.579), revealing that although the capability is valued when present, it is not viewed as a universal prerequisite—reflecting a realistic appraisal of the current distribution of cardiac surgical capacity across Korean hospitals. Third, indicators related to ward support services—including rapid response systems led by critical care staff—did not reach consensus, suggesting that experts considered those dimensions to be important at the hospital level but insufficiently specific to the ICU environment to serve as defining capability criteria.

Outcome indicators, such as risk-adjusted mortality, were deliberately excluded from the framework. In the Korean context, the absence of a standardized national database for severity adjustment and the lack of a systematized inter-hospital transfer network render comparative outcome metrics unreliable at present. Beyond data infrastructure, risk-adjusted mortality indicators are inherently susceptible to selection bias, potentially penalizing high-acuity referral centers that disproportionately accept the most critically ill patients [27,28]. By prioritizing objective structural and process indicators—such as 24/7 staffing availability and life support capacity—this framework offers a more equitable and immediately actionable foundation for policy decisions. This can be interpreted as consistent with the World Federation of Intensive and Critical Care framework [2], which first defines ICU structural capability levels and then uses them for benchmarking and capacity planning

This study has several strengths. It represents the first consensus-based effort to develop ICU capability indicators in South Korea, and it is grounded in a multidisciplinary panel of 24 experts—including members of the Korean Society of Critical Care Medicine Standardization Committee—who provided high and sustained response rates across three rounds (79.2%–91.7%). The modified Delphi design incorporated methodological features that strengthen the rigor of the consensus process, including a pre-structured item pool derived from a systematic review of international frameworks [3,20-22], varying rating scales across rounds to optimize discrimination [17], and reverse-phrased items in Round 3 to verify internal consistency [18,19].

Several limitations also warrant acknowledgment. Nursing representation on the expert panel was small (n=1 critical care nurse), which might have introduced bias in the relative weighting of nursing-related indicators; however, the nurse-to-bed ratio emerged as one of the highest-rated measures overall. As a consensus-based study, the indicators require empirical validation against clinical outcomes such as mortality, length of ICU stay, and transfer efficiency before their adoption as formal regulatory standards [17]. This study also reflects the state of Korean ICU infrastructure as of 2025; evolving post-pandemic norms—particularly regarding infection control requirements—might shift expert consensus on currently excluded indicators in future iterations. Also, the composition of the expert panel might have introduced selection bias and limited representativeness because the results can be influenced by participants’ institutional characteristics, specialty background, and healthcare system context. Finally, the framework does not yet incorporate patient- and family-centered outcome perspectives, and they are an important dimension of ICU quality that future studies should address.

The 29 indicators identified in this study provide a practical foundation for constructing a capability-oriented ICU classification system in South Korea. International experience offers important lessons in this regard. Ontario, Canada, through Critical Care Services Ontario, demonstrates the value of a centralized, real-time capability network in which patient transfers are coordinated based on standardized tiering of staffing, life support equipment, and ventilation capacity [29,30]. The type A indicators identified in our study—particularly LSI provision, 24/7 cardiovascular and ECMO availability, and nurse-to-bed ratios—map directly onto the data fields required to build a monitoring infrastructure that connects institutional capability data to regional dashboards to enable transparent patient-acuity matching [31,32]. Conversely, Japan's experience illustrates the risks of linking ICU classification to reimbursement without functional oversight: a tiered fee structure based on staffing ratios incentivized increases in invasive procedures without corresponding improvements in mortality or occupancy rates [33,34]. South Korea, which shares a fee-for-service model with Japan, faces analogous risks if classification is tied to reimbursement prematurely or without functional validation. The indicators defined in this study should therefore be operationalized as a standardized data dictionary for a real-time ICU monitoring system—one that supports patient-acuity matching and transfer coordination at both the institutional and regional levels.

Future research should proceed in three directions. First, the indicators should be validated against clinical outcomes in a prospective cohort or registry-based study to confirm that higher capability scores correspond to improved patient survival, shorter ICU lengths of stay, and more efficient inter-hospital transfers [27,28]. Second, a pilot implementation in a defined geographic region—analogous to Ontario's provincial rollout [31,33]—would allow testing of the data dictionary, monitoring dashboard, and classification algorithm in real-world conditions before national adoption. Third, the framework should be revisited periodically as the Korean ICU infrastructure evolves: indicators currently in type D or excluded (such as negative pressure isolation and rapid response systems) might achieve stronger consensus as implementation conditions improve. Together, these steps would transform the present expert consensus into an operational classification system that can improve patient safety and resource efficiency in Korean critical care.

KEY MESSAGES

SUPPLEMENTARY MATERIALS

Article information Continued