Implementation and feasibility evaluation of a standard set of intensive care unit discharge criteria in a prospective before-after observational pilot study

Abstract

Background

Identifying ICU patients stable enough for discharge to lower care levels is usually subjective. We previously derived a standardized discharge criteria set including patient-specific aspects and organizational capabilities. This study evaluated its implementation feasibility, clinical utility for objective discharge readiness assessment, and descriptive patient flow and safety related outcome parameters in routine care.

Methods

A prospective two-phase before-after observational pilot study was conducted in two German ICUs in 2022. During baseline, discharge-readiness workflows were observed. After criteria-set implementation and pre-testing, criteria completion, user feedback, patient-flow, acuity, and capacity parameters were collected. Due to small cohorts, only descriptive statistics were reported.

Results

Eleven mainly patient-specific and some organization-specific criteria reached > 80% completion at discharge decision and were defined as a core set. User feedback supported extending this core set with organization-related criteria according to clinical setting, patient groups, pathways, and receiving-unit capabilities. Discharge readiness assessment should involve all relevant stakeholders to align care needs with available capabilities.

Conclusions

A core set of eleven criteria indicated ICU discharge readiness in clinical routine. Organization-specific criteria can support adaptation to different patient subgroups and pathways. Workflow integration may support timely and safe transfers, but patient flow and safety effects require validation in larger studies. Holistic discharge readiness assessment requires coordinated pathway-management processes along the care continuum. Future implementation should prioritize reporting automation, intuitive visualization, and workflow-integrated documentation, ideally supported by a patient data management system.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12913-026-15656-y.

Background

Proactive management of acute patient flow is essential to allocate limited intensive care resources to those who would benefit most [1]. For the optimal use of the Intensive Care Unit (ICU), adequate admission and discharge policies are required to improve equal and timely access and avoid discharges that might result in readmission or increased morbidity and mortality [1–3]. Objective decision making on stepping care levels up or down should consider patient individual factors, organizational abilities, and constraints [2]. A standardized assessment of ICU patients’ discharge readiness with the help of common criteria is not widely established in daily clinical practice, although it has been repeatedly recommended [2–9]. Various studies have previously been conducted to develop and validate patient discharge readiness indicators [5, 10–14] or tools to facilitate intensive care transition, like protocols, handover-forms and liaison nurses [3, 15–19]. Still, none of them has taken the holistic perspective of an individual patient and an organizational discharge readiness assessment in a standardized format to be used in daily clinical practice.

In a previous Delphi study, a set of 28 ICU discharge criteria was developed to improve the identification of stable patients eligible for transition to lower care levels, taking into account the characteristics of the receiving unit [4]. The set of criteria included patient-specific aspects (patient status, interventions, medications, diagnosis, prognosis, and treatment limitations), system-specific criteria (nursing workload criteria), and institutional factors (available technical infrastructure, skill sets, safety measures, protocols, and processes). Successful implementation of discharge criteria could potentially improve patient transfers and, thus allow new patient admissions while decreasing readmission rates and improving the overall quality of care transitions [5–10]. We designed a before-after observational pilot study implementing this previously consented set of discharge criteria to assess possible implementation methods in different discharge workflows, criteria adherence in the post-implementation phase, descriptive patient flow and safety-related outcome parameters, and overall clinical utility.

Methods

The set of 28 ICU discharge criteria, named Delphi-set hereafter, covering patient-specific and organ system related criteria as well as several organization-related aspects are detailed out in Additional file 1 (AF 1), Table S1. All criteria are binary scored (fit or not fit for ICU discharge) and vital parameters have target ranges and value calculation rules defined. Also, most of the criteria are flagged as “mandatory to be met.” In addition, applicable exceptions in specific patient groups (e.g. palliative care patients) are defined [4]. In this prospective two-phase observational trial, the Delphi-set was implemented at two adult ICUs in Germany. One was a 16-bed medical intensive care unit (MICU) in a university hospital (1585 beds), admitting ± 1150 patients annually. The second was an 11-bed interdisciplinary ICU (IICU) in a teaching hospital (449 beds), admitting ± 1000, mostly surgical, patients annually. All intensive medical diagnostic and therapeutic procedures, excluding extracorporeal membrane oxygenation, were available in both settings. The lower care discharge pathways differed between the two settings and were therefore described for contextual interpretation. At the MICU, patients could be transferred to a cardiac intermediate care unit with selected advanced care options, including oxygen therapy and continuous pulse oximetry, to cardiology or other general ward environments, or to external facilities such as long-term-care-institutions. At the IICU, the main lower care destinations were surgical floor and cardiology ward environments. The study included all ICU patients (> 16 years), discharged alive to a lower care setting within the observation periods. In the MICU, short-stay ICU patients (< 4 h and admitted for selected planned procedures) were excluded as they were not part of the unit’s general admission and discharge process.

The two-phase study consisted of a baseline analysis and a criteria implementation phase, where the difference in pre- and post-implementation discharge practice was analyzed. At the MICU, both phases ran exactly 4 weeks each (spring & summer 2022). At the IICU, both phases ran each over a full month during summer of 2022. A trained research student per study site performed on-site process observations, criteria adherence documentation, and data collection of selected key performance indicators (KPI).

The study was designed as a feasibility evaluation of implementing the Delphi set in routine clinical workflows, with clinical utility assessed through criteria completion, structured user feedback, and descriptive patient flow and safety-related outcome parameters.

During the baseline analysis, the current assessment of discharge readiness was observed by shadowing the ICU team. These observations were graphically mapped. The teams were still unaware of the criteria in the Delphi-set. The research students blindly counted how often the to-be-implemented discharge criteria were already applied in the daily routine. Congruence was scored. This was to assess to what extent the specific unit was already using the elements of the Delphi-set. Further, different outcomes of the discharge readiness assessment were defined: The patient could have been identified as being a “discharge option” (def.: the patient is getting well enough to be discharged to a lower level of care within the next 24 h but still benefits from intensive care or needs some therapeutic interventions to be completed, concrete discharge decision still pending) or the team could have directly taken a “discharge decision” (def.: patient is well enough to be discharged to the targeted receiving unit and transfer can be organized). When patients still required ICU-specific treatment, they were labelled as “therapy to be continued”. Among these patients, some could have been additionally defined as a “triage option” when capacities became constraint (def.: when admission of a new patient is indicated and current capacities are fully occupied, this patient can be discharged immediately as either the patient is the most stable among the current group of ICU patients or the patient’s outcome will be improved the least from further intensive care treatment). The care teams defined the patient’s status as a discharge or triage option, discharge decision or therapy ongoing as per their existing workflows and in the same way during both phases.

For the implementation phase, two ways of criteria implementation were developed for the Delphi-set. The MICU integrated these into their patient data management system (PDMS, Philips IntelliSpace Critical Care and Anesthesia (ICCA) (AF 1, Figure S2). Criteria completion was documented during the morning rounds (7 to 9 am). At the IICU, where documentation was mainly paper-based, a Microsoft Excel-based tool was developed to capture and visualize the criteria per individual patient throughout the stay (AF 1, Figure S2). Here, criteria completion was documented for each patient twice daily (6 am and 2 pm). At both sites, implementation of the Delphi-set was pre-tested and adjusted where needed based on user feedback before it was applied in the clinical routine. As an example of adjustments made, for criteria that were not applicable for a particular patient, day, or period, a “not applicable”-selection option was included next to the binary “yes” or “no” options. This criterion could then be ignored to simplify the workflow and included again when the criterion became relevant in the individual context. The Delphi-set was applied to all admitted patients from the start of the study period and to all new admissions from the time of admission. Where information was missing during the rounds, the research students supplemented and/or revised the criteria with the attending physicians and nurses during the day shifts. Criteria adherence was not mandatory for this observational trial and patient status definition was still independent from the criteria completion rate during the implementation phase. For the evaluation of adherence to the Delphi-set, criteria completion per criterion and per patient was counted at the different time points during the discharge process. Criteria could be either completed, not completed, or not applicable. In the MICU, process-wise, patients were not always defined as being a discharge option before a discharge decision. When criteria were not applicable at the point of discharge option or discharge decision, they were counted together with the completed criteria for completion rate calculation (def.: completion rate per process step and per criterion = number of patient cases with criterion completed + number of patient cases with criterion not applicable per process step/ number of all assessed patient cases per process step). Further, it was counted how many patients were discharged with all criteria completed (ACC) and how many patients were discharged with not all criteria completed (NACC). In the latter group, it was analyzed which criteria were not completed and how often. Lastly it was counted, how many patients were not discharged although all discharge criteria were completed.

KPIs of patient flow and outcome, resource use [1, 11, 12, 13, 14], patient acuity (SAPS II [15], SOFA [16]), and nursing workload scores (TISS10 [17], INPULS [18]) were measured daily throughout both phases (AF 1, Table S2), whereof INPULS was only measured in the MICU. For the time of discharge it was differentiated between urgent discharges (def.: when a patient has a “therapy ongoing” or “discharge option”-status during morning round and the patient has been discharged on that same day), in-time discharges (def.: time of actual discharge – time of discharge decision ≤ 2 h.) and delayed discharges (def.: time of actual discharge - time of discharge decision > 2 h.) [1, 19]. The > 2-hour threshold was adopted from the Institute for Healthcare Improvement patient flow framework, which describes a 2-hour ICU boarding time target as an operational patient flow metric. In this study, the threshold was used as a pragmatic operational benchmark to distinguish routine transfer logistics from clinically relevant same-shift discharge delays that may affect ICU capacity utilization and downstream patient flow. Further, the patient flow volume was quantified by the number of transfers to each receiving unit to compare the main receiving units during both phases with the aim to identify potential differences in criteria completion for different patient pathways. For readmissions, it was differentiated between readmissions ≤ 48 h post ICU discharge and > 48 h post ICU discharge, as only the earlier readmissions show stronger relations with the quality of the discharge decision and process [12, 13]. Unit occupancy rates (def.: number of beds used / number of beds operational and staffed) > 85% in the discharging or receiving unit were defined as strained, guided by previous studies [14, 20, 21, 22, 23]. The gathered KPI data was used to investigate selected cases of readmission, urgent or delayed discharges, or post discharge complications for KPI related reasons. Criterion completion in ≥ 80% of all patient cases at the point of discharge decision and a low ≤ 48 h.-readmission rate (< 5%, guided by previous studies [12, 24, 25, 26, 27]) without readmission reasons related to the discharge criterion were considered as confirming clinical utility of the criterion for a standard discharge readiness assessment.

During the study, structured user feedback was captured for future clinical implementation, like criteria order and grouping, phrasing, technical implementation and visualization of the criteria. Descriptive statistics were reported. Microsoft Excel was used for analysis and graphical illustration of outcomes. Further, R language and environment for statistical computing (version 4.3.2) was used for further analysis and graphical illustration. Variables are presented as mean ± standard deviation (SD) and median with interquartile range (median [IQR]) for skewed data. Categorical variables are presented as absolute numbers and frequencies. ICU and hospital mortality rates are presented as relative frequencies with 95% confidence intervals, calculated using normal approximation.

Ethical and GDPR conformity declarations were obtained from both hospitals’ Institutional Review Boards (IRB; see declarations section). Based on the data anonymization efforts, the IRBs confirmed that there was no need for patients’ informed consent for participation in line with data privacy conformity to GDPR articles 5, 6, 9, 89 (11.01.2022 / 07.04.2022).

Results

Inclusion process, study cohort, and unit characteristics

Figure 1 illustrates the inclusion process to derive the final study cohort.

Fig. 1.

Study cohort inclusion flow chart

Table 1 describes the units’ characteristics and the included patient sub-cohorts per hospital as well as the total cohort. Blinded discharge criteria congruence was measured in 117 patient stays in total during the baseline phase. Discharge criteria application adherence was measured in 130 patient stays in total during the implementation phase. Periods of ICU capacity strain were measured during both phases, with mean (SD) ICU capacity utilization rates of 84.7% (14.6%) and 85.4% (11.8%) per day (baseline phase and implementation phase, total cohort). To transparently report the clinical outcome parameters collected during the before-after observation periods, ICU length of stay and readmission rates are presented descriptively in Table 1. In the total cohort, median ICU length of stay was 32.6 h [18.3, 60.7] during baseline and 29.4 h [18.9, 92.9] during implementation. Readmissions within 48 h remained low in both periods, with 3/117 patients (2.6%) during baseline and 4/130 patients (3.1%) during implementation. Readmissions occurring after 48 h were observed in 4/117 patients (3.4%) during baseline and 4/130 patients (3.1%) during implementation. These values are reported descriptively because the study was not powered to test outcome differences between periods. Site-specific receiving-unit patterns are summarized at the end of the Results section to provide contextual information without foregrounding a formal comparison between units. Further detail is reported in the supplementary material, so that main Results remain focused on total-cohort before-after implementation findings.

Table 1.

Unit characteristics, patient cohort, and descriptive clinical outcome parameters

Unit characteristics	MICU		IICU		Total cohort
	Baseline phase	Implementation phase	Baseline phase	Implementation phase	Baseline phase	Implementation phase
Total ICU patient group before exclusion criteria applied	146	186	64	73	210	259
ICU beds staffed	16 (16 days)/ 14 (12 days)	14	11	11	27 (16 days)/ 25 (12 days)	25
ICU capacity utilization rate, % mean (SD) median [IQR]	92.6 (8.2) 93.8 [87.5, 93.8]	82.4 (13.4) 78.6 [76.8, 92.8]	77.1 (15.4) 72.7 [63.6, 85.9]	88.3 (9.5) 88.8 [80.5, 90.9]	84.7 (14.6) 87.5 [72.7, 93.8]	85.4 (11.8) 85.7 [78.6, 92.3]
ICU mortality, % [95%-CI]	8.9 [3.9, 13.9]	12.9 [7.8, 18.0]	10.9 [2.5, 19.4]	4.1 [0.00, 9.4]	9.5 [5.3, 13.7]	10.4 [6.5, 14.3]
Hospital mortality, % [95%-CI]	1.6 [1.0, 2.1]	2.2 [1.6, 2.8]	2.3 [1.5, 3.1]	1.9 [1.2, 2.6]	1.9 [1.4, 2.3]	2.1 [1.6, 2.5]
Nurse : Patient ratio ICU, mean median [IQR]	1:1.9 1:2 [1:2, 1:2]	1:1.9 1:2 [1:2, 1:2]	1:1.5 1:1.5 [1:1.3, 1:1.7]	1:1.7 1:1.8 [1:1.4, 1:1.9]	1:1.7 1:2 [1:1.4, 1:2]	1:1.8 1:2 [1:1.7, 1:2]
Patient cohort (inclusion criteria: all adult patients (> 16 yrs), discharged alive during study periods from the two ICUs to lower care setting)
Included patient stays – n	61	60	56	70	117	130
Age, yrs. mean (SD) median [IQR]	65.6 (17.8) 69.0 [57.0, 78.0]	63.3 (17.3) 68.0 [52.8, 77.0]	65.9 (18.2) 68.5 [58.8, 79.0]	64.7 (17.8) 68.5 [54.5, 79.0]	65.7 (17.9) 69.0 [58.0, 79.0]	64.0 (17.5) 68.5 [53.3, 78.0]
Male sex, n (%)	35 (57.4%)	38 (63.3%)	32 (57.1%)	40 (57.1%)	67 (57.3%)	78 (60.0%)
ICU LOS, h mean (SD) median [IQR]	98.2 (205.5) 37.0 [16.0, 86.0]	68.9 (90.7) 25.0 [17.6, 75.6]	81.3 (187.9) 25.2 [19.5, 49.7]	80.6 (109.8) 30.0 [19.7, 101.8]	90.1 (196.6) 32.6 [18.3, 60.7]	75.2 (101.2) 29.4 [18.9, 92.9]
Readmissions, n (%) within 48 h > 48 h	3 (4.9%) 3 (4.9%)	2 (3.3%) 1 (1.7%)	0 (0%) 1 (1.8%)	3 (4.3%) 2 (2.9%)	3 (2.6%) 4 (3.4%)	4 (3.1%) 4 (3.1%)

SD= Standard Deviation; IQR= Inter quartile range; CI= Confidence Interval; LOS = Length of stay

Workflow assessment

None of the sites had written ICU discharge guidelines prior to the study. At both sites, discharge decisions were taken by the clinicians in the morning shift and later confirmed by the nurses in charge during the rounds. Decisions were documented on the whiteboard and in the PDMS (only used in MICU). For patients with “discharge option” or “triage option”, reevaluations of their status took place during the day. Detailed baseline observations of the different discharge readiness assessment workflows are illustrated in AF 1, Figure S3 with additional process and decision-making description. In both units, planned discharges were mainly facilitated in the morning (AF 1, Figure S4). Discharge option or triage patients were discharged later during the day in case of stabilization and fitness for discharge or increased ICU capacity strain. Patient handovers to the general wards were done on a clinician-to-clinician and nurse-to-nurse level. In contrast to the baseline phase, during the implementation phase the team was aware of the discharge criteria and could refer to them for guidance via the available tools. However, due to the pilot and observational character of the study, criteria consideration was intended though strict adherence to them was not mandatory for discharge decisions. The remaining discharge readiness assessment workflow was unchanged. The research students documented the criteria discussed during the rounds in the respective tools (AF 1, Figures S1 and S2).

Criteria adherence during the baseline phase

For the baseline phase, the comparison of the applied criteria against the to-be-implemented criteria is illustrated in AF 1, Figure S5, showing how far or close the units were from the new discharge assessment concept. Additional details of the comparison are provided there. Adherence to criteria was observed in 87 patients. Only criteria expressed verbally by the clinician or discussed with the team could be scored by the research students; criteria assessed silently during clinical inspection could therefore not be quantified, although they may have contributed to clinical decision-making.

Criteria adherence and completion rates during the implementation phase

In the implementation phase, 80 patients (62% of all discharged patients) had all Delphi-set criteria completed (ACC) at the time of discharge decision. In 44 out of 91 patients (48%) with data available at discharge option, all discharge criteria had already been met at the moment the patient became a discharge option (AF 1, Figure S6). Fifty-seven patients (44%) were discharged with NACC (AF 1, Table S3), whereof one was a palliative care patient, and two patients had triage options. At time of discharge, ICU capacities were strained in 70% of cases (Table 1; AF 1, Figure S7).

Focusing the analysis on completion rates per criterion at the time of discharge decision, eleven criteria out of the set of 28 could be considered as a core set as they met the defined thresholds for clinical utility at both sites (AF 1, Table S4, column A) and had no related readmission reasons at overall low readmission rates. A completion trend analysis looking for early discharge readiness indicators also manifested these criteria, as nine out of the eleven criteria already scored ≥ 80% completion rates at the time of discharge option. Further, there were no criteria that scored > 80% completion at discharge option and then underscored 80% completion at discharge decision (Fig. 2). Figure 2 displays the implementation-phase completion rates of the eleven core criteria at the time points discharge option and discharge decision, reported on a total cohort basis. More granular site-level visualizations remain available in AF 1, Figure S6 for transparency, but are not intended to support a formal comparison between ICU types.

Fig. 2.

Completion rates of the eleven core-set criteria in the total cohort at discharge option and discharge decision. Initial criteria numbering is kept as per the Delphi-set

Six criteria were not applicable in > 80% of patient stays at both study sites (AF 1, Table S4, column B), whereof five were conditional questions (if), asking for a specific condition (e.g. long-term tracheostomy, percutaneous transient pacemaker, renal replacement therapy) related to care needs in the following lower care setting. The majority of these six criteria did not change their not applicable status between discharge option and discharge decision (AF 1, Figure S6). A high not applicable rate indicated that there were few patients with these care needs in both study groups. In addition, five of these six criteria already had high not-applied rates > 80% in the baseline analysis (indicated by * in AF 1, Table S4, column B) and therefore seem to be of limited relevance for routine discharge decisions in the studied settings.

Each criterion was completed in > 10% of patient stays at both sites. Differences in completion or not applicable rates between sites are presented descriptively in Table S4, column C. These values should be interpreted as contextual information reflecting implementation with differing patient pathways and receiving-unit capabilities rather than as evidence of systematic differences between the ICU types.

Urgent discharges, delayed discharges, and readmissions

Urgent discharges (def. in AF 1, Table S2) in face of strained unit capacities happened in both phases (13% of all patient-stays during baseline, 11% in implementation phase). Criteria fulfillment among urgently discharged patients ranged from ACC (n = 2) to NACC (n = 12), with up to six criteria not completed. Here, no impact of the criteria implementation could be analyzed in context with guiding patient selection for urgent discharges.

Delayed discharges (def. in AF 1, Table S2 and Figure S7) occurred at both sites and in both phases (total: 68 (58%)/ 84 (64.6%) of all patient stays (baseline / implementation phase)) with a total of accumulated hours of discharge delay of 792.5 h. during the baseline and 628.4 h. in the implementation phase. The major reasons were no bed availability at the receiving units and internal and external organizational difficulties. 48 (57%) of all patients with discharge delays (n = 84) in the implementation phase had ACC. Reasons for delayed discharges did not differ much between NACC and ACC groups, as it was mainly capacity constraints at the receiving units that could be resolved within the same shift. 45 (94%) of the ACC group were then discharged within 24 h. post criteria completion. For patients with NACC, but a discharge decision taken and a delayed physical discharge (36 (43%)), the top three not-completed criteria were stable blood oxygenation (no. 3); If patient needs continued monitoring at receiving unit, are required technology/staff capabilities in place? (no. 14) and Do urine output, electrolyte level, and renal function allow patient discharge? (no. 19). Here, all three criteria relate to available technical, monitoring or care capacities at the receiving unit and could hinder timely discharge if not in place. A general analysis on delay windows and underlying reasons for all delayed discharges is shown in AF 1, Figure S7. We could observe that discharge delays went up in occurrence but down in the total amount of hours in the implementation phase. However, we could not conclude that criteria implementation and adherence had an impact on that, as the main causes of discharge delays were constraints in the organizational readiness and capacity of receiving units rather than individual patient discharge readiness. This was shown also by the main not completed criteria and the discharge delay root cause analysis in AF 1, Figures S6 and S7.

Readmissions within 48 h occurred in both phases at low rates (Table 1). Because event numbers were very small and discharge decisions occurred under varying criteria completion status and capacity utilization rates at ICUs and receiving wards, these values should be interpreted descriptively. The before-after clinical outcome parameters did not show a clear signal suggesting improved readmission rates after implementation. ICU length of stay was slightly lower by median values in the total cohort during the implementation phase, but variability was wide and the observation periods were short. Therefore, these findings should not be interpreted as evidence of an effect of criteria implementation on ICU length of stay or readmission rates.

Nurse workload and acuity scores at discharge decision

For patients discharged with NACC, their nursing workload and acuity scores at discharge were compared to the scores’ medians of the patients discharged with ACC (AF 1, Table S5 and Figure S8). Due to manual score calculations and lack of reporting in some cases, scores were only available for a subset of the total cohort (INPULS 29%; TISS 53%; SAPS II 61%; SOFA 90%). Because of incomplete score availability and varying data quality, these data are reported descriptively only and were not used to infer associations between nursing workload, acuity, criteria completion status, or discharge readiness.

User feedback analysis from the implementation phase for future improvements of the criteria set

AF 1, Table S6 provides collected user feedback on a per criterion level. In general, the perception was that a list of 28 criteria is too long for daily clinical utilization. In the implementation phase, we already introduced the option of marking criteria as not applicable over the course of the ICU stay or at least until they become applicable. That reduced the daily effort of checking these criteria. With the result of six criteria with not applicable status labels in > 80% of the patient discharge decisions, these criteria could be options for deletion from the general criteria set, although no. 20 as one of them could be rephrased as proposed in AF 1, Table S6. In general, all criteria should be phrased shorter, especially for application in a PDMS as the number of letters per criterion entry is a limiting factor. With that, an abbreviations glossary was recommended. Localization (language/ processes/ order of criteria/ integration in PDMS workflow) was partially challenging and should be taken into consideration for further development or validation. Not all patients were discharged to general ward environments. Therefore, receiving unit would be a more general term than general ward which some criteria refer to. The criteria set should be customizable per type of receiving unit that differs from a general ward setting (e.g., CPU, intermediate care units, external organizations, etc.). User feedback supported a selection of a criteria core set, which is the primary result of this study (AF 1, Table S4, column A). Further, high priority criteria (with mandatory completion required) could be identified, e.g., for night discharges. Other criteria could be added based on the receiving unit’s set-up. When further criteria are added, criteria interdependencies (e.g., if criteria 7 does not apply, 7a (conditional criteria with if) will not be displayed) should be considered and implemented. Criteria on lab parameters, nutrition, anticoagulation, and neurology were missed, but no proposals specified. Having had the completed and not applicable statuses automatically copied from day-to-day assessment until it was changed, was perceived as utile and was recommended for future utilization. A potential future automated and rule-based assessment of criteria covering vital parameters has been seen skeptically because of the artefact burden. Rather, an integrated cockpit-like view on needed vital parameters, lab values, etc. next to the criteria list was recommended.

Unit-specific observations were retained only as contextual information and are not part of a formal comparison between ICU types. Receiving-unit pathways differed between sites: patients from the MICU were most frequently transferred to cardiology-related destinations, including an intermediate care unit where selected advanced care options such as oxygen therapy and continuous pulse oximetry monitoring were available, whereas patients from the IICU were most commonly transferred to surgical floor and cardiology settings. These pathway differences may help readers understand why some criteria were completed, not completed, or not applicable at different frequencies. However, the study objective focuses on before-after implementation aspects of the Delphi-set in the total cohort. The study was not designed or powered to compare ICU types or to attribute observed differences to units, patient case mix, implementation method, or receiving-unit structure. More detailed site-level findings are therefore provided in the supplementary material only for transparency.

Discussion

The preceding Delphi process [4] defined 28 ICU discharge criteria that should close the gap of a widely applicable assessment tool for ICU patients that considers the patient’s and organization’s discharge readiness related aspects. As a feasibility evaluation, the present pilot study implemented these criteria in different clinical settings and focused on three aspects: (1) criteria application adherence; (2) criteria completion in context with descriptive patient-flow, outcome, and resource-utilization parameters; and (3) implementation in different clinical workflows and user feedback on clinical utility.

The criteria set

The criteria adherence analysis of the implementation phase revealed eleven criteria that were met in > 80% of all discharge decisions (Fig. 2; Table S4, column A). Potentially serving also as an early indicator of discharge readiness, all of them were already met in > 70% of patients being a discharge option. We would recommend using these eleven criteria as a general set of core criteria for further validation efforts in different settings in daily clinical practice. Criteria with the largest differences between sites (Table S4, column C) may be useful as optional additions when local patient pathways or receiving-unit capabilities require a broader assessment. However, these descriptive site-level differences should not be interpreted as evidence of systematic unit-level effects. With the aim of providing a set of discharge criteria that consider patient as well as organization specific discharge readiness factors, we could conclude that in the core set (column A), eight criteria assess patient specific factors, and three criteria also consider the capabilities of the receiving unit. A pragmatic two-step approach may therefore be considered, with a general core set for initial patient-stability screening and additional organization-specific criteria selected according to the intended receiving unit. The latter criteria should be ideally agreed on together with the receiving units’ team to avoid overestimating their capacity. For those six criteria that were not applicable in > 80% of the patient stays in both sites (Table S4, column B), we would recommend adding them only exceptionally to the core set in settings where they would potentially reach far higher application rates. In such cases, their relevance to assess discharge readiness could be tested again.

Impact on patient flow-, outcome and resource utilization

Regarding patient flow, both early and delayed discharges were equally present. Most of the urgent discharges of patients without all criteria completed were triage decisions due to ICU capacity constraints. Some patients were transferred to receiving units with monitoring or care capabilities beyond a standard general ward environment. This may explain why selected patients could be discharged despite NACC, but the present study does not allow conclusions on whether specific receiving-unit structures directly influenced discharge safety or outcomes. A lack of staffed beds in the receiving units and internal and external organizational difficulties were the major underlying reasons for patients with discharge decisions, but delayed discharges. The proportionally high occurrence of discharge delays mirrors the findings of other studies in this field [7, 28–31]. Because this study used a pragmatic > 2-hour operational definition, discharge-delay frequencies should be interpreted as indicators of short-term workflow and capacity constraints rather than directly compared with studies using longer thresholds, such as > 6 h or discharge after midnight. Wrong allocations and capacity mismatch in the acute care environment is evident [1, 3, 14, 32–34], which was substantiated by our results. Accumulating the discharge delay hours for both study periods to 1428 h., underlines suboptimal capacity utilization, especially in context of generally high ICU capacity rates. For 59.5 patient days over the entire study period, one bed could have been allocated to a more acute patient. Most of the patients’ discharge delays could be solved within the first 24 delay hours. For readmissions, we assume that the low readmission rates could be related to the presence of significant discharge delays, as patients may have been stabilized and monitored longer in the ICU and reached the GW in a better state. Also, the 44 patients that had ACC at the time of discharge option and no discharge decision taken until the next round of assessment, would support this finding. For the observed readmission cases, ICU capacity rates differed between 78,6% and 90%, GW capacities were strained at discharge decision, and the acuity score values of the readmitted patients were mainly above the average score values at discharge, assuming patient deterioration happened at the lower care level. From this study, we could not conclude that criteria completion rates had an impact on guiding patient selection for urgent discharges or could shorten discharge delay, as cohorts were rather small, adherence was not mandatory, and the main root cause of the urgent and delayed discharge laid in constraints of capacity and process organization rather than in individual patient’s discharge readiness.

The before-after descriptive outcome parameters reported in Table 1 therefore provide transparency for readers but should be regarded as hypothesis-generating. They indicate that ICU length of stay and readmission rates can be feasibly collected alongside criteria completion data, while confirming that larger confirmatory studies are required to evaluate whether mandatory criteria adherence can influence timely discharge, readmissions, or other patient-flow and safety outcomes.

Nursing workload and acuity scores were explored descriptively in relation to criteria completion status, but interpretation was limited by incomplete score availability and varying data quality. Although some score patterns appeared similar between patients discharged with ACC and NACC, the present data do not allow conclusions on whether nursing workload or acuity scores are associated with criteria completion status or discharge readiness. Therefore, these findings should be regarded only as exploratory and should not be used to define complementary discharge criteria in the present study. Future studies with larger cohorts, automated and timely score calculation, and standardized score availability should evaluate whether nursing workload or acuity measures can add value to discharge readiness assessment.

Implementation aspects

The workflow observation during the baseline phase served well to adapt the implementation of the Delphi set to the different site needs (tooling with PDMS versus MS Excel, criteria order). A structured daily documentation of the discharge readiness assessment with the help of defined discharge criteria was new to the workflow at both sites. Because implementation formats differed between sites, potential effects of PDMS-based versus Excel-based documentation can only be considered as a hypothesis-generating implementation aspect and not as a conclusion supported by the present data. In general, our observations and user feedback further confirmed that concise and structured criteria verbalization, intuitive, workflow-integrated digital documentation, reporting automation and key data visualization, would support day-to-day progress comparison and decision making with the interdisciplinary care team and keep utilization effort limited in clinical routine. These findings align with previous research [3, 6, 9, 34–36]. Integration into a PDMS may facilitate this process but should be regarded as an implementation enabler rather than as a mandatory prerequisite. Automated calculations for vital parameter criteria fulfillment are not recommended at this stage because of potential artefacts and high tolerances in some patients, that would pass set thresholds. Further, clinical users favored having a shorter core criteria list with complementary criteria based on the hospital’s organizational framework conditions. Our proposal with a core and an extended criteria set would cater to this need. Although standardized processes and objective clinical decision support are appreciated to support care transitions, clinicians prefer to be able to overrule the suggestions and decide in a subjective manner if required by the situation.

These study results could guide the set-up of larger confirmatory studies and provide ideas for testing the impact of process changes regarding patient flow management. For any discharge process improvement, a detailed workflow analysis is needed to identify all stakeholders and involve them in a standardized way of data provision, constraints communication, discharge readiness assessment and patient handover. Care capacity overestimation and readmissions could be avoided when taking the present capacity rates of the receiving unit more into consideration, having a real dialog about the patients care needs, and using nursing-activities-score like TISS or INPULS as additional discharge decision support [3, 6]. Further, ICUs and receiving units should align and synchronize their discharge and admission hours and processes more along the patient pathway [3]. They should establish a mandatory and constant assessment and documentation of discharge readiness progress to help identify transfer candidates early and plan discharges more in advance [37]. Ideally, this should span even across discharges from the general ward to home or external facilities. Here, electronic progress documentation with visibility for related departments could further support organizational efficiency. For an implementation of a standardized discharge assessment process with defined criteria, it would be beneficial to also have consented and documented ICU admission criteria to compare the individual patient status against in a daily manner [38].

Limitations

Our study focused more on an in-depth analysis of criteria implementation within the context of processes, patient cases and framework conditions, rather than a broad overview. Therefore, the choice was to select only two study sites from the same health care region but in different settings and with different specializations. The local study cohorts with between 56 and 70 patients were rather small, and the study period was only 4 weeks or a month per phase and site. That limited collecting larger, more generalizable datasets, but had the advantage of complementing electronic data collection with observations and manual data collection from the shadowing activities, which brought additional valuable context. The involved data collection team with only two research students could only cover the morning shifts with limited observation possibilities on handovers from day to night shift. Complementary observation data coming from outside these periods relied on reported information. Data inconsistencies from manual care documentation, delayed electronic data entries, and missing data, including laboratory values and nursing workload or acuity scores, influenced data quality and limited interpretation. Because INPULS, TISS, SAPS II, and SOFA scores were available only for subsets of the cohort, no robust conclusions can be drawn regarding associations between nursing workload, acuity level, criteria completion status, or discharge readiness. These score-based findings were therefore retained only as descriptive exploratory data and should be validated in larger cohorts with automated and timely score calculation.

Further, the criteria set, as it is currently phrased, covers only discharges to general wards. In our study, the criteria were applied to all in scope patients, also those that were discharged to specialized wards and external long-term care facilities. This may have derived slightly broader user feedback on needs for further development for a wider scope of receiving units, which we saw as a benefit.

Interpretation of potential effects on ICU length of stay and readmission rates was further limited by the small cohorts, short study periods, and seasonal differences between observation phases. In addition, adherence to the discharge criteria was not mandatory, and discharge timing was strongly influenced by capacity and process constraints in the receiving units. Significant discharge delays may have contributed to the low early readmission rates because patients remained stabilized and monitored in the ICU longer than clinically necessary. Therefore, the observed clinical outcome parameters cannot be causally attributed to criteria completion or implementation and should be validated in larger studies with mandatory criteria adherence and broader assessment of downstream capacity-management processes.

Further research

Because the initial Delphi study reflected consensus across different European healthcare settings, future implementation research should test the criteria set in different healthcare systems and geographies. To increase robustness and usability of the criteria set, future studies should include flexible criteria adaptation, longer observation periods, larger cohorts, automated data reporting, intuitive visualization, and different patient groups and receiving-unit types. Workflow-integrated electronic documentation, including PDMS-based implementation where available, should be evaluated as one possible enabler of scalable routine use. Having access to larger cohorts with high quality electronically registered data would also support future development, e.g. the use of deep learning algorithms to support discharge readiness assessments.

Conclusion

Caregivers need objective decision criteria to assess discharge readiness for ICU patients that could be transferred to the next lower level of care when ICU capacities are strained. Decisions need to be based on the individual patients’ needs but also on the organizational capabilities across the care continuum. Catering to this need, we pilot-tested a consented set of 28 objective discharge criteria for its clinical utility in daily routine. Baseline analysis insights helped to integrate the criteria set into different clinical settings. In an implementation phase, criteria adherence in context with selected KPIs was evaluated against a definition of clinical utility. From the 28 criteria applied, a core set (11 criteria) incorporating mainly patient- and some organization-specific criteria was derived and evaluated as useful for broader clinical routine. This core set could serve well to identify stable ICU patients that could be discharged to a standard general floor environment. It could be scaled up with additional, more organization-specific criteria that consider different patient groups, patient pathways and receiving units’ framework conditions. A nursing activity score could be a suitable criterion addition to align on care needs with the receiving units. Discharge assessment workflow standardization including all stakeholders, reporting automation and visualization ideally in a PDMS, is key for constant use in daily clinical practice. However, having a sophisticated ICU discharge assessment tool in place will not significantly foster timelier discharges, when capacity constraints in the lower care levels are a major driver of ICU discharge delay. Thus, only standardized and aligned discharge processes along the care continuum supported by suitable tooling will improve patient flow, patient outcome and acute care capacity utilization.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ICU

Intensive care unit

MICU

Medical intensive care unit

GW

General ward

PDMS

Patient data management system

ICCA

Philips IntelliSpace Critical Care and Anesthesia

IICU

Interdisciplinary intensive care unit

AF

Additional file

KPI

Key performance indicator

ACC

All criteria completed

NACC

Not all criteria completed

SD

Standard deviation

ED

Emergency department

CPU

Chest pain unit

n.s.

Not specified

n.a.

Not applicable

LOS

Length of stay

Author contributions

M.H. set-up the study protocol and recruited the study sites. J.B., H.B. and M.W. reviewed the protocol and provided methodological guidance. R.R., S.S., J.-K.S. supported study realization and guided criteria implementation at the individual sites. S.N. and A.B, as research students on site, performed the workflow observation and collected, de-identified, structured, and analyzed all relevant data from the two study periods. A.S. performed data cleaning, descriptive data analysis, summary reports and graphical illustration. First results data preparation and analysis were done by M.H., A.S., A.B. and S.N. M.W., H.B., J.B., R.R., S.S. and J.-K.S. contributed through data review, results discussion, methodological guidance and agreement on the manuscript structure. M.H. prepared the manuscript which was carefully reviewed, edited and finally approved by all authors.

Funding

No additional funding. The corresponding author is an external PhD candidate at Erasmus University Rotterdam, salaried by Philips Medizin Systeme Böblingen GmbH, see details above.

Data availability

All referred tables and figures starting with AF in their name are provided in the additional file 1 (AF 1, online supplementary material). The raw data was collected, structured, de-identified and analyzed on each study site separately. Summary data was joined from both sites for further descriptive analysis. Underlying summary data for all tables and figures in the main manuscript and the online supplement section can be made available in an anonymous form from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The university hospital’s Ethics Committee (EC, Ethik-Kommission an der Medizinischen Fakultät der Eberhard-Karls-Universität und am Universitätsklinikum Tübingen) provided a waiver (doc. no. 960/2021BO2) due to the observatory study format that was targeted on process quality improvement in routine care. For the academic teaching hospital, the EC of Ärztekammer Nordrhein also provided a waiver (80/2022), pointing to the observational and quality-targeted nature of the study. For both sites, the IBRs declared that due to the data anonymization efforts, patients’ informed consent for participation was not necessary as per GDPR articles 5, 6, 9 and 89 and in line with GDPR data privacy conformity (11.01.2022 / 07.04.2022). The study was performed in accordance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

M.H. is an external PhD student at the Erasmus University of Rotterdam, The Netherlands, affiliated to the Department of Adult Intensive Care, Erasmus Medical Center, and salaried by Philips Medizin Systeme Böblingen GmbH. This study forms part of M.H.’s PhD research work, that is supervised by Professor Jan Bakker (supervisor) and Professor Hendrik Bracht (co-promoter). All other authors have no competing interests to disclose.