Nursing leadership and performance metrics in artificial intelligence integrated gastrointestinal surgical care: A narrative review

Abstract

Artificial intelligence (AI) is increasingly being integrated into gastrointestinal and complex surgical care, yet its practical impact on nursing work remains insufficiently defined. This narrative review examines how nursing leadership can support the safe use of AI across perioperative pathways and how its effects can be evaluated through measurable nursing-sensitive outcomes. Literature was identified through PubMed, Scopus, and Web of Science, with emphasis on peer-reviewed studies, systematic reviews, guidelines, consensus statements, and implementation frameworks relevant to gastrointestinal surgery, perioperative nursing, clinical decision support, and AI. The review focuses on preoperative assessment, intraoperative coordination, postoperative surveillance, and discharge planning in high-acuity surgical settings. Particular attention is given to workflow efficiency, response times, early recognition of complications, risk-stratified care, documentation quality, interdisciplinary communication, cognitive workload, and patient safety. The proposed framework is intended as a practical, non-validated structure for future evaluation rather than as an established measurement instrument. Its use will require prospective validation and adaptation to local staffing, digital infrastructure, resources, and clinical workflows. AI should therefore be judged not only by predictive performance, but by whether it helps nurses recognize risk earlier, communicate concerns clearly, and deliver safer perioperative care.

Key Words: Artificial intelligence; Gastrointestinal surgery; Nursing leadership; Perioperative nursing; Clinical decision support; Performance metrics

Core Tip: Artificial intelligence (AI) is increasingly entering gastrointestinal and complex surgical care, but its effect on nursing practice has not been clearly defined. This review argues that AI implementation should be evaluated not only by algorithmic performance, but also by its influence on nursing workflow, escalation, documentation, communication, early complication recognition, and patient safety. A structured set of nursing-sensitive performance metrics may help institutions assess whether AI tools genuinely improve perioperative care delivery in complex surgical settings.

INTRODUCTION

Artificial intelligence (AI) is entering gastrointestinal and complex surgical care at points where clinical information is dense, unstable, and difficult to interpret in real time. AI has been described as both promising and potentially disruptive in surgery[1], particularly when it begins to influence clinical decision-making rather than remaining a research tool[2]. In major abdominal surgery, clinical risk rarely appears as one abnormal value. A patient recovering from pancreaticoduodenectomy may initially look stable while gradually developing low-grade tachycardia, increasing drain output, delayed gastric emptying, borderline urine output, and rising inflammatory markers. After low anterior resection, an anastomotic leak may first appear as vague abdominal discomfort, increasing analgesic requirements, altered bowel or stoma output, reluctance to mobilize, or a subtle change in vital signs before overt sepsis becomes obvious. In these situations, prediction alone is not enough. What matters is whether risk information helps the team recognize a changing trajectory earlier and respond in time.

The clinical value of AI depends not only on model performance, but also on whether the technology can be safely used within real clinical systems[3]. This is particularly important because implementation failures often arise not from the algorithm alone, but from poor integration into workflow, unclear accountability, limited interpretability, or lack of clinician trust[4]. In gastrointestinal surgery, where postoperative deterioration may develop across several hours or days, machine learning models may improve prediction of postoperative outcomes[5], and AI-based approaches have been evaluated for major abdominal surgical complications[6]. The more important bedside question, however, is whether such prediction leads to earlier nursing reassessment, clearer escalation, and safer intervention.

Current discussions of AI in surgery still tend to emphasize model discrimination, calibration, predictive accuracy, or area under the receiver operating characteristic curve[5,6]. These measures are necessary, but they do not answer what the team should do differently when the model identifies risk. A postoperative alert has limited value if it appears in the electronic record without clarifying who should review the patient, how quickly reassessment should occur, which findings should be checked, and when escalation to the surgical or intensive care team is required.

This gap is particularly relevant to nursing. Nurses are often the professionals who notice that a patient is no longer following the expected recovery pathway. They assess pain, respiratory effort, wound appearance, drain characteristics, urine output, fluid balance, bowel function, stoma output, mobility, oral intake, mental status, and the patient’s own expression of concern. Early warning systems show that structured observation can support recognition of clinical deterioration[7-9], but AI introduces an additional layer of risk prediction that must still be interpreted through bedside assessment.

For this reason, AI implementation in gastrointestinal and complex surgical care should not be framed only as a technical or physician-centered development. It is also a nursing leadership issue. A risk signal after colorectal surgery does not improve safety by itself. It becomes useful when the nurse reassesses the patient, connects the alert with the bedside picture, communicates concern clearly, and documents the response. Similarly, enhanced recovery pathways in colorectal, pancreatic, liver, and esophageal surgery already depend on multidisciplinary coordination, patient education, early mobilization, nutrition, and structured monitoring[10-13]; AI should strengthen these processes rather than sit separately from them.

Nursing leadership is therefore central to the safe translation of AI outputs into clinical practice. Nurse leaders are positioned to decide how alerts enter the workflow, who is responsible for first response, how escalation is documented, and whether staff have the training and authority to act. This approach is consistent with sociotechnical models of health information technology, in which digital tools succeed or fail through their interaction with people, workflow, organizational culture, and clinical context[14-16]. In surgical care, this means that an AI model should not be evaluated only as software, but as part of a living clinical system.

This narrative review examines how nursing leadership can be redefined within AI-integrated gastrointestinal and complex surgical pathways. It focuses on the practical implications of AI for preoperative assessment, intraoperative coordination, postoperative surveillance, and discharge planning. Particular attention is given to high-acuity gastrointestinal, colorectal, hepatobiliary, pancreatic, and complex abdominal procedures, where small delays in recognition or escalation may have major consequences. The review also proposes a conceptual and practice-oriented framework of measurable nursing performance domains intended to support future evaluation of AI implementation in clinically meaningful terms. Accordingly, the purpose of this review is not to compare individual AI algorithms but to examine how AI implementation influences nursing leadership and how these changes may be evaluated through measurable nursing-sensitive outcomes.

SCOPE OF THE REVIEW

This article is a narrative review of AI-related changes in gastrointestinal and complex surgical care, with emphasis on nursing leadership and measurable performance. The discussion focuses on four areas most likely to influence nursing responsibilities: Predictive analytics, imaging-derived risk assessment, AI-enabled monitoring, and clinical decision support. The review does not rank individual models or perform a meta-analysis. Instead, it translates current evidence and implementation concerns into nursing-sensitive performance domains.

The review gives particular attention to patients undergoing major gastrointestinal, colorectal, hepatobiliary, pancreatic, esophageal, and complex abdominal procedures. These pathways are clinically demanding because they combine high complication risk with the need for coordinated perioperative care. AI tools introduced into these pathways must therefore be judged not only by predictive performance, but also by their ability to support real decisions at the bedside.

The reporting and implementation discussion is informed by emerging standards for AI evaluation. CONSORT-AI and SPIRIT-AI address reporting of clinical trials and protocols involving AI interventions[17,18], DECIDE-AI focuses on early-stage clinical evaluation of AI decision-support systems[19], and TRIPOD+AI, PROBAST, and PROBAST+AI provide guidance relevant to prediction models and risk-of-bias assessment[20-22]. These frameworks are relevant because nursing implementation depends on more than whether an algorithm performs well in retrospective validation. The literature informing this narrative review was identified through searches of PubMed, Scopus, and Web of Science. The primary literature search covered publications from January 2018 to May 2026, with the final search conducted in May 2026. Earlier publications were retained selectively when they represented foundational clinical definitions, established assessment tools, or seminal evidence directly relevant to the conceptual and clinical context of the review. Search concepts combined terms related to AI and machine learning with perioperative nursing, gastrointestinal and complex surgery, surgical data science, clinical decision support, patient deterioration, and implementation science. Because the purpose was to develop a clinically oriented narrative synthesis rather than a systematic review, studies were selected according to their relevance to the review questions rather than through a formal systematic screening process. Priority was given to peer-reviewed systematic and scoping reviews, international guidelines and consensus statements, AI reporting and evaluation frameworks, and clinically applicable studies relevant to gastrointestinal surgery, perioperative care, nursing practice, and implementation. Evidence directly addressing AI-supported nursing practice in gastrointestinal surgical settings was considered particularly relevant; where such evidence was limited, literature from adjacent fields was used to inform the conceptual discussion. No formal meta-analysis or systematic risk-of-bias assessment was undertaken.

Accordingly, evidence-based conclusions in this review refer to findings reported in the underlying literature, whereas the nursing-sensitive performance framework represents an author-proposed conceptual synthesis derived from those findings and from relevant adjacent evidence. The framework itself should not be interpreted as an empirically validated consequence of AI implementation in gastrointestinal surgical nursing.

Across the gastrointestinal AI literature considered in this review, model development and internal validation were substantially more common than external validation, calibration assessment, or evaluation after clinical implementation. These dimensions were therefore considered separately when interpreting the evidence. Discriminative performance alone was not taken as evidence of clinical effectiveness or applicability, and applicability was interpreted in relation to the population, surgical context, validation setting, and evidence of integration into routine care. Where calibration, external validation, implementation, or patient-level clinical benefit was not reported in the source literature, this absence was retained as an evidence limitation rather than inferred from model performance. Table 1 provides a structured narrative comparison of these characteristics for selected studies most relevant to the review.

Table 1 Evidence synthesis of selected artificial intelligence studies relevant to gastrointestinal surgical nursing.

Ref.	Setting/population	AI task	Validation and model performance	Implementation/clinical effectiveness	Applicability to nursing	Evidence relationship
Wang et al[5]	Systematic review and meta-analysis; 38 studies of gastrointestinal surgery	Prediction of postoperative outcomes	143 ML and 62 logistic-regression models; ML showed higher discrimination than LR (ΔAUC 0.07, 95%CI: 0.04-0.09); 44% of studies were at low risk of bias. Calibration and external validation were inconsistently reported across the underlying literature	Practical implementation and patient-level clinical effectiveness were not established	Relevant to GI surgical risk stratification and surveillance, but nursing-specific implementation and outcomes were not evaluated	Indirect
Bihorac et al[25]	Development/validation study; 51457 adults undergoing major inpatient surgery at a single center	Prediction of eight major postoperative complications and mortality	Repeated internal validation; complication AUCs 0.82-0.94 and mortality AUCs 0.77-0.83. Calibration was assessed. No external validation in this study	Designed for real-time perioperative use, but feasibility in clinical workflow required further prospective testing; clinical effectiveness was not demonstrated in this study	Potentially relevant to risk-based monitoring and escalation, but not GI-specific and nursing outcomes were not evaluated	Indirect
Bektaş et al[27]	Systematic review; 31 studies of colorectal surgery	Prediction of surgical outcomes using ML	Most studies were retrospective; external validation occurred in 4 studies (13%) and calibration was reported for 3 models (10%)	Routine clinical implementation and improvement in patient outcomes were not established	Directly relevant to colorectal surgical risk prediction; applicability to nursing workflow remains indirect because nursing-specific outcomes were not studied	Indirect
Bektaş et al[29]	Systematic review; 52 studies of hepatopancreaticobiliary surgery	Prediction of disease course, postoperative complications and other surgical outcomes	47/52 (90%) studies were retrospective and 5/52 (10%) prospective; external validation was generally absent	Authors identified external validation as necessary before clinical introduction; routine clinical effectiveness was not established	Relevant to procedure-specific risk assessment, but nursing implementation and nursing-sensitive outcomes were not evaluated	Indirect
Soria-Utrilla et al[34]	Retrospective observational study; 586 patients with colorectal cancer assessed before surgery	AI-assisted CT body-composition assessment	SMI discriminated low FFMI with AUC 0.82; sex-specific diagnostic cut-offs were derived within the study population. External validation was not reported	Diagnostic/morphofunctional assessment was evaluated; effects on downstream workflow or patient outcomes were not tested	May identify nutritional and functional vulnerability relevant to preoperative assessment and referral; direct nursing effects were not evaluated	Indirect
Cho et al[70]	Multihospital implementation study; 4 hospitals with different electronic record systems	One-hour fall-risk prediction with AI-powered decision support and tailored nursing recommendations	Locally optimized models were implemented across four sites; AUROC 0.8051-0.9581	Successfully integrated into different EMR systems. Nursing interventions increased, but fall rates did not significantly decrease	Provides direct evidence that AI can be incorporated into nursing workflow and outcomes tracking, but the setting was not gastrointestinal surgery	Direct nursing implementation evidence; non-GI setting

AI: Artificial intelligence; AUC: Area under the curve; AUROC: Area under the receiver operating characteristic curve; CT: Computed tomography; EMR: Electronic medical record; FFMI: Fat-free mass index; GI: Gastrointestinal; LR: Logistic regression; ML: Machine learning; SMI: Skeletal muscle index.

Selected studies illustrating the relationship between gastrointestinal surgical AI evidence, clinical implementation, and nursing-relevant outcomes are summarized in Table 1.

AI IN GASTROINTESTINAL AND COMPLEX SURGICAL CARE: WHAT CHANGES AT THE BEDSIDE?

AI applications in gastrointestinal and complex surgical care are diverse. Some tools predict postoperative complications, others support imaging interpretation, monitor physiological deterioration, or guide decision support. Technically, these tools differ. Clinically, they raise the same question: How does the information change the care delivered to the patient? Machine learning in medicine has been widely discussed as a method for extracting clinically useful patterns from complex data[23], while big data approaches in health care have expanded the volume and granularity of information available to clinical teams[24]. In surgical pathways, the challenge is to make these outputs useful without weakening clinical judgment (Table 2).

Table 2 Artificial intelligence applications across perioperative phases and nursing implications in gastrointestinal and complex surgical care.

Perioperative phase	AI-supported function	Surgical context	Nursing implications	Measurable nursing-sensitive indicators
Preoperative assessment	Risk stratification using demographic, laboratory, imaging, nutritional, and comorbidity data[5,25-30]	Pancreaticoduodenectomy; major hepatectomy; colorectal cancer surgery; multivisceral resection	Identification of patients requiring intensified education, nutritional support, prehabilitation, early stoma counselling, or planned higher-acuity postoperative monitoring	Completion of risk-based education; timely nutrition or physiotherapy referral; documentation of individualized monitoring plan; predefined escalation criteria
Preoperative imaging review	AI-assisted detection of sarcopenia, visceral obesity, frailty-related imaging markers, or anatomical complexity[34,35]	Hepatopancreatobiliary and oncological gastrointestinal surgery	Translation of imaging-derived risk into practical nursing care, including pressure-injury prevention, mobilization planning, respiratory preparation, and discharge support	Documentation of frailty or nutritional risk; early mobilization plan; referral to allied health professionals; pressure-injury risk reassessment
Operating room and immediate recovery	Prediction of hemodynamic instability, bleeding risk, respiratory compromise, or need for higher-acuity care[25,30,35]	Esophagectomy; major liver resection; complex abdominal surgery	Structured handover from operating room or recovery unit to ward or intensive care, with clear monitoring priorities	Completeness of AI-informed handover; time from recovery discharge to first ward reassessment; documentation of key intraoperative concerns
Early postoperative period	Detection of abnormal physiological patterns and early deterioration[5-9,25,30]	Anastomotic leak after colorectal surgery; pancreatic fistula; postoperative bleeding; sepsis; acute kidney injury	Prompt bedside reassessment, review of drain output, wound status, fluid balance, pain pattern, urine output, oxygen requirement, and escalation to surgical team	Time from alert to nursing reassessment; time from reassessment to medical escalation; proportion of alerts with documented response; missed or delayed escalation events
Postoperative ward care	Support for risk-stratified enhanced recovery pathways[10-13,47-49]	Colorectal surgery; pancreatic surgery; hepatobiliary surgery	Adjustment of monitoring frequency, mobilization targets, pain assessment, nutritional support, and discharge readiness according to risk	Adherence to individualized observation schedule; completion of ERAS-related nursing interventions; documented reasons for pathway deviation
Discharge planning	Prediction of readmission, delayed recovery, dehydration, infection, or poor self-management[5,25,27,30]	New ileostomy; complex wounds; drains at discharge; frail older patients; cancer surgery	Targeted patient education, confirmation of understanding, community nursing referral, stoma nurse involvement, and early follow-up planning	Completeness of discharge education; documentation of warning signs; referral to community or stoma services; early post-discharge contact; avoidable readmission

AI: Artificial intelligence; ERAS: Enhanced recovery after surgery.

Preoperative risk stratification

Preoperative assessment in complex gastrointestinal surgery already requires judgment across multiple domains: Age, frailty, malnutrition, comorbid disease, previous chemotherapy, inflammatory status, liver function, renal reserve, body composition, and expected operative complexity. MySurgeryRisk showed how machine learning can be used to estimate major complications and death after surgery[25], while the ACS NSQIP Surgical Risk Calculator remains an established example of structured surgical risk estimation[26]. In gastrointestinal surgery, machine learning models have increasingly been studied for predicting postoperative outcomes after colorectal surgery[27], upper gastrointestinal cancer surgery[28], hepatopancreatobiliary surgery[29], and broader surgical complication prediction[30].

The nursing relevance of this information depends on whether it changes preparation. A patient with pancreatic cancer, weight loss, biliary obstruction, low albumin, and reduced functional capacity should not simply be labelled “high risk”. That classification should trigger practical nursing actions: Nutritional referral, clearer education about drains and delayed gastric emptying, assessment of family support, early mobilization planning, and a visible postoperative monitoring plan. This is consistent with enhanced recovery after surgery (ERAS) principles, where preoperative education, nutrition, mobilization, and multidisciplinary planning are central elements of perioperative care[10-13].

In colorectal surgery, predicted risk of prolonged recovery may support earlier stoma nurse involvement, more detailed discharge planning, and closer review of hydration and bowel function after surgery. Here, nursing leadership ensures that AI-generated risk information becomes part of preoperative counselling, multidisciplinary planning, and handover. The quality of implementation can be measured through completion of risk-based education, timely referral to nutrition or physiotherapy, documentation of individualized monitoring plans, and predefined escalation criteria for high-risk patients.

Imaging-derived risk and diagnostic support

AI is already prominent in gastroenterology through imaging and endoscopic applications. Reviews of AI in gastroenterology describe its use in lesion detection, image interpretation, and risk classification[31]. Computer-aided detection during colonoscopy has shown potential to improve colorectal neoplasia detection[32], while AI systems have also been evaluated for histology prediction and characterization of colorectal lesions[33]. Although these tools are often physician-facing, their outputs may still influence nursing responsibilities when they alter procedural planning, patient counselling, postoperative surveillance, or discharge needs.

In complex surgical care, imaging-derived risk may be relevant when it identifies frailty, sarcopenia, visceral obesity, tumor burden, anatomical complexity, or features suggesting higher postoperative risk[34]. A patient identified as nutritionally vulnerable or functionally frail should not simply carry an imaging label. The information should influence nursing care planning: Pressure-injury prevention, respiratory preparation, mobilization support, dietetic referral, family education, and discharge planning. Nursing leadership is needed to ensure that imaging-derived risk becomes an operational care plan rather than a radiology finding that never reaches the ward.

Intraoperative coordination and transition to postoperative care

The intraoperative and immediate postoperative period is a critical point for information transfer. Surgical data science has emphasized that operating room data, procedural context, and perioperative signals can be used to support surgical decision-making and quality improvement[35]. In practice, however, AI-supported outputs are useful only if they are communicated in a way that informs postoperative nursing priorities.

A patient undergoing major hepatectomy may leave theatre with concerns about blood loss, fluid shifts, coagulation status, lactate trends, and risk of liver dysfunction. A patient after esophagectomy may require careful attention to respiratory status, pain control, chest drainage, and early mobilization. Enhanced recovery after surgery recommendations for liver surgery and esophagectomy emphasize structured perioperative management and careful postoperative coordination[12,13]. If AI-supported risk information exists but is not included in handover, the receiving nurses may not know which changes should trigger concern.

Nursing leadership can strengthen this transition by standardizing how AI-derived risk information is included in perioperative handover. The receiving team does not need to know every technical detail of the model. It does need to know what the risk means for observation frequency, fluid balance monitoring, oxygen requirements, drain assessment, pain review, mobilization, and escalation. In this sense, AI-informed handover should work as a clinical translation tool, not as a technical report.

Postoperative surveillance and early complication recognition

Postoperative surveillance is where AI may most directly affect nursing practice. Surgical complications are commonly classified using standardized systems such as the Clavien-Dindo classification[36]. In gastrointestinal and hepatopancreatobiliary surgery, formal definitions exist for pancreatic fistula[37], delayed gastric emptying[38], post-pancreatectomy hemorrhage[39], post-hepatectomy liver failure[40], bile leakage[41], and anastomotic leakage after rectal surgery[42]. These definitions are important for research and reporting, but bedside recognition often begins earlier and less neatly.

A patient after colorectal resection may have pain, ileus, mild fever, and fatigue as part of expected recovery, but similar features may precede intra-abdominal sepsis. Anastomotic leakage is often recognized later than the first clinical signal, and delayed recognition remains a major concern after intestinal surgery[43]. A patient after pancreatic surgery may have delayed gastric emptying and drain output that require careful interpretation over several days. AI-enabled monitoring may help by drawing attention to combinations of signs that are easy to underestimate when viewed separately: Rising heart rate, increasing oxygen requirement, reduced urine output, persistent pain, abnormal laboratory trends, and delayed mobilization.

The alert, however, is only the beginning of the clinical process. The nurse must still reassess the patient, decide whether the signal fits the bedside picture, communicate concern, and document the response. Failure-to-rescue research shows that outcomes after surgical complications depend not only on whether complications occur, but also on whether teams recognize and manage them effectively[44,45]. This is why AI surveillance should be evaluated by response time, escalation quality, and clinical action, not only by prediction accuracy (Figure 1).

Figure 1 From artificial intelligence signal to nursing action: Author-proposed conceptual escalation pathway. The figure illustrates a non-validated framework linking artificial intelligence generated risk signals to bedside nursing reassessment, clinical interpretation, structured escalation, action, documentation, and outcome evaluation. It is intended as a conceptual model for future implementation and evaluation rather than as clinical guidance or a validated escalation protocol. AI: Artificial intelligence.

Discharge readiness and post-discharge risk

AI may also influence discharge planning by identifying patients at risk of readmission, poor functional recovery, infection, dehydration, inadequate pain control, or failure to manage a stoma, wound, or drain at home. In gastrointestinal surgery, discharge is not simply the moment when the patient is medically stable. It is the point at which responsibility shifts to the patient, family, community services, and outpatient teams.

For nurses, an AI-generated readmission risk score is useful only if it improves the quality of preparation. A patient discharged after colorectal surgery with a new ileostomy may need more than standard instructions if there is risk of dehydration, poor self-management, or limited support at home. A patient recovering from pancreatic surgery may require clear guidance on drain care, oral intake, warning signs, and when to contact the surgical team. Surgical safety evidence, including checklist-based approaches, has shown that structured team processes can reduce harm when they are embedded into routine care[46]. AI-supported discharge planning should follow the same principle: It should strengthen communication and preparedness, not merely add another risk score.

Nursing-sensitive indicators in this phase may include completeness of discharge education, documentation of warning signs, confirmation of patient understanding, timely referral to community nursing or stoma services, early post-discharge contact, and avoidable readmission. These measures help determine whether AI-supported discharge planning improves continuity of care rather than simply adding another prediction to the record.

REDEFINING NURSING LEADERSHIP IN AI-SUPPORTED SURGICAL PATHWAYS

The integration of AI into complex surgical care changes the expectations placed on nursing leadership. Traditional perioperative nursing leadership includes staffing, supervision, protocol adherence, patient safety, education, and coordination of care. These responsibilities remain essential. What changes is that nurse leaders must now manage the space between digital prediction and clinical response.

Clinical decision-support systems are more likely to influence care when their recommendations are timely, actionable, and integrated into workflow[47-49]. In gastrointestinal surgery, this principle is not abstract. A risk alert after pancreaticoduodenectomy may suggest increased probability of postoperative sepsis or clinically relevant pancreatic fistula. The formal definition of postoperative pancreatic fistula is useful for classification[37], but it does not tell the bedside nurse how to respond to the early combination of fever, cloudy drain fluid, delayed gastric emptying, rising inflammatory markers, and increasing fatigue. If the alert remains isolated from nursing assessment, it becomes another notification. If it is linked to a clear escalation pathway, it may support earlier recognition and more consistent response.

Nursing leadership therefore becomes a form of clinical translation. Nurse leaders do not need to become data scientists, but they need to understand how AI outputs enter the workflow, where they may fail, and how they should be interpreted alongside bedside observation. In practice, a model may identify that a patient is moving away from an expected recovery pattern, but the nurse still determines whether the patient is pale, anxious, breathless, reluctant to mobilize, or reporting a different quality of pain.

From protocol supervision to risk-responsive coordination

AI-supported pathways require nurse leaders to move beyond passive protocol supervision. A conventional protocol may define routine vital-sign frequency, drain monitoring, mobilization targets, pain assessment, nutrition steps, and discharge criteria. Enhanced recovery pathways provide a strong basis for this structure[10-13]. In risk-responsive care, however, these actions may need to change according to the patient’s evolving risk profile.

A patient after low anterior resection who is progressing well may follow a standard recovery pathway. Another patient after the same procedure, with increasing pain, persistent tachycardia, reduced oral intake, and rising inflammatory markers, may require closer observation before a clear diagnosis is made. If an AI system identifies this patient as deviating from expected recovery, nursing leadership must ensure that the ward response is not improvised. Monitoring frequency, escalation thresholds, communication with the surgical team, and documentation of concern should already be defined.

This approach is particularly important in hepatopancreatobiliary surgery. After major hepatectomy, changes in lactate, coagulation, bilirubin, urine output, mental status, and hemodynamic stability may suggest early liver dysfunction or systemic deterioration. Post-hepatectomy liver failure and bile leakage have established definitions[40,41], but the ward response depends on whether early bedside changes are recognized and escalated before deterioration becomes severe.

Accountability and escalation

AI-supported care raises questions of accountability. When a model generates a warning, responsibility for the next step must be clear. If an alert is visible to several professionals, each may assume someone else has acted. If it is sent only to nurses without escalation authority, nurses may carry responsibility without the necessary support.

In a busy surgical ward, these ambiguities matter. A nurse caring for several postoperative patients may receive an alert for a patient after colorectal surgery whose heart rate has been slowly increasing overnight. The patient is afebrile but reports increasing abdominal discomfort and has reduced urine output. Without a clear pathway, the nurse may document the observations and continue routine monitoring. With a defined AI-supported escalation process, the same situation may trigger immediate reassessment, repeat vital signs, review of fluid balance, structured communication to the surgical resident, and earlier laboratory or imaging review.

Nurse leaders can reduce ambiguity by defining alert categories, response targets, escalation steps, and documentation expectations. High-priority indicators should relate to time-sensitive safety concerns requiring prompt bedside reassessment and communication with the responsible surgical team, whereas moderate-priority indicators may trigger intensified monitoring or earlier review. Lower-priority indicators may be used primarily for workflow monitoring, documentation quality, or service evaluation. Universal numerical thresholds or fixed weighting of these indicators cannot currently be recommended, because appropriate targets are likely to vary according to surgical procedure, patient acuity, staffing, institutional resources, and local workflow. Each institution should therefore define and prospectively evaluate its own clinically appropriate thresholds and priority hierarchy before routine use.

Protecting clinical judgment

One of the most important leadership tasks is to protect clinical judgment. AI systems may appear objective because they produce numerical scores or risk categories, but their outputs are shaped by training data, included variables, missing data, local documentation practices, and the clinical population in which they are used. Alarm fatigue and repeated electronic warnings are already recognized safety problems in monitoring and clinical decision support[50-53], and AI may worsen these problems if alerts are poorly prioritized or insufficiently explained.

Nurses must therefore be encouraged to question, verify, and contextualize AI outputs. A patient after esophagectomy with increasing oxygen requirement and reduced cough effectiveness may need urgent attention even if a general deterioration score has not crossed a threshold. A patient with postoperative tachycardia may trigger an alert, but the cause may be pain, hypovolemia, bleeding, anxiety, pulmonary embolism, sepsis, or atrial arrhythmia. The alert identifies concern; it does not replace assessment.

Leadership culture is crucial. If nurses feel that AI overrides their bedside judgment, they may hesitate to escalate concerns that are not digitally supported. If alerts are frequent and poorly explained, staff may ignore them. Nurse leaders should therefore promote a culture in which AI is treated as an additional source of information, similar to a laboratory trend or early warning score, rather than as a final decision.

Staff preparation and digital confidence

The success of AI-supported care depends heavily on staff preparation. Nurses need practical education that goes beyond a short technical demonstration. They need to understand what the tool is intended to do, which patient groups it applies to, what type of data it uses, how alerts are displayed, what actions are expected, and what limitations are known. Nursing-focused reviews have emphasized that AI adoption may affect nursing roles, decision-making, documentation, workload, and professional accountability[54-60].

Training should be connected to real surgical scenarios. It is more useful for nurses to discuss how an alert should be handled in a patient after pancreatic surgery with rising drain amylase and new fever than to receive a general explanation of predictive analytics. Similarly, training around readmission prediction should be linked to discharge planning for patients with new stomas, poor oral intake, complex wounds, or limited family support.

Digital confidence does not mean blind trust. It means that nurses understand the system well enough to use it safely, question it appropriately, and document their response. Nurse leaders can support this confidence through simulation, case-based teaching, bedside coaching, and feedback after real alerts.

MEASURABLE PERFORMANCE DOMAINS FOR AI-INTEGRATED SURGICAL NURSING

AI implementation should be evaluated through outcomes that reflect real care delivery. Algorithmic performance remains important, but it is not sufficient. Reporting and evaluation frameworks for AI emphasize that clinical effectiveness, workflow integration, transparency, and implementation context should be described alongside model performance[17-22]. For nursing leadership, the central question is whether AI changes what happens to the patient in a measurable and beneficial way (Table 3). Because no formal evidence-grading methodology was applied in this narrative review, the proposed domains should not be interpreted as carrying graded levels of evidentiary strength. For future validation studies, individual indicators should be operationalized prospectively by specifying the numerator and denominator where applicable, measurement timing, data source, intended use, and relevant confounders within the local clinical setting.

Table 3 Nursing-sensitive performance domains for evaluating artificial intelligence integrated gastrointestinal and complex surgical care.

Performance domain	Definition	Clinical relevance	Suggested indicators
Workflow efficiency[14-16,47-49]	The extent to which AI supports prioritization without adding unnecessary digital burden	Surgical nurses often manage multiple unstable patients, drains, wounds, medications, mobilization targets, and discharge tasks during the same shift	Time spent retrieving risk information; duplicate documentation steps; task completion rates; staff-reported usability; time from alert visibility to review
Response-time optimization[44,45,47-49]	The interval between AI-generated concern and meaningful clinical action	Delays in response may worsen outcomes in sepsis, anastomotic leak, bleeding, respiratory compromise, or organ dysfunction	Time from alert to bedside reassessment; time from nursing concern to surgical review; time from review to intervention; proportion of high-priority alerts acted on within target time
Early complication recognition[5-9,25-30,44,45]	Earlier identification of deterioration before crisis-level presentation	Complications after colorectal, pancreatic, and hepatobiliary surgery often develop through subtle trends rather than one obvious sign	Time from first abnormal trend to escalation; proportion of complications preceded by documented nursing concern; unplanned ICU transfers; failure-to-rescue events
Adherence to risk-stratified pathways[10-13]	Alignment of nursing care with the patient’s predicted and evolving risk profile	High-risk patients may require intensified monitoring, earlier allied health input, or deviation from standard enhanced recovery pathways	Completion of individualized observation plans; adherence to escalation thresholds; timely nutrition, physiotherapy, wound, or stoma referral; documented pathway deviations
Documentation accuracy[17-22]	Completeness and consistency of clinically relevant nursing documentation following AI-supported alerts or risk classifications	AI systems and clinical teams both depend on accurate documentation of observations, actions, and reasoning	Proportion of AI-supported alerts with documented nursing reassessment and response; completeness of predefined clinical fields relevant to the identified risk; documentation of escalation and subsequent action; missing or incomplete required entries, assessed through structured record audit
Interdisciplinary communication[47-49]	Timely and documented transfer of clinically relevant AI-derived risk information between nurses and other professionals involved in the patient’s care	AI-derived risk information is clinically useful only if it reaches the right professionals in time	Proportion of relevant handovers containing AI-derived risk information; proportion of escalations with documented acknowledgment by the receiving team; time from nursing escalation to team acknowledgment; documentation of agreed actions or follow-up
Cognitive workload and alert burden[50-53]	The effect of AI tools on mental workload, interruption, and alarm fatigue	Excessive or poorly explained alerts may reduce trust and delay response to clinically important signals	Number of alerts per shift; proportion of actionable alerts; staff-rated alert usefulness; false or non-actionable alert frequency; missed alerts
Patient safety outcomes[44-46]	Patient-level outcomes potentially influenced by AI-supported nursing surveillance and coordination	The final aim of AI implementation is safer care, not simply more prediction	Delayed deterioration events; postoperative sepsis; unplanned ICU admission; readmission; medication errors; prolonged length of stay; patient-reported discharge preparedness

The performance domains presented in this table represent a conceptual and practice-oriented framework proposed by the authors to support future implementation and evaluation of artificial intelligence integrated surgical nursing. They should not be interpreted as a formally validated measurement instrument. AI: Artificial intelligence; ICU: Intensive care unit.

The proposed performance domains should be regarded as a conceptual implementation framework intended to support future evaluation of AI-integrated nursing practice rather than as a validated measurement instrument. Prospective validation will be required before these indicators can be adopted as standardized performance measures. The following domains are proposed as nursing-sensitive indicators for AI-integrated gastrointestinal and complex surgical care. They complement traditional surgical outcomes by focusing on the operational processes through which patient safety is achieved. Because direct empirical evidence on AI-integrated nursing practice in gastrointestinal surgery remains limited, some of these domains are informed by evidence from adjacent fields and should be interpreted as testable implementation concepts rather than demonstrated causal effects of AI on nursing practice or patient outcomes.

Workflow efficiency

Workflow efficiency is often discussed as a technical advantage of AI, yet in surgical nursing it is experienced very differently. During a single shift, nurses move continuously between medication administration, drain assessment, wound care, pain management, mobilization, documentation, communication with physicians and families, and preparation for discharge. In such an environment, another digital system is valuable only if it genuinely helps nurses identify priorities instead of adding another layer of interruption. Human factors and sociotechnical research have repeatedly shown that technologies introduced without considering everyday clinical practice are unlikely to improve either efficiency or patient safety[14-16].

Consider a morning shift on a gastrointestinal surgical ward. Several patients require routine postoperative review, but one patient recovering from a major hepatectomy begins to show subtle physiological deterioration, while another patient following colorectal surgery appears to be at increased risk of early readmission. If AI highlights these patients before deterioration becomes clinically obvious, the nurse can reorganize priorities accordingly. The benefit lies less in the alert itself than in enabling earlier clinical attention where it is most needed.

Workflow efficiency should not be judged by the presence of AI within the electronic record but by whether it simplifies clinical work. Useful evaluation measures include the time required to retrieve clinically relevant information, unnecessary duplication of documentation, completion of risk-based nursing interventions, the interval between alert generation and clinical review, and nurses’ own assessment of whether the technology supports rather than complicates patient care.

Response-time optimization

In gastrointestinal and complex surgical care, the value of AI is measured less by the moment an alert appears than by what happens immediately afterwards. Postoperative deterioration often develops gradually, and opportunities for intervention may be lost if abnormal findings are not assessed promptly. From a nursing perspective, the critical interval is therefore the period between the first indication of increased risk and the clinical actions that follow. Studies of failure to rescue consistently demonstrate that patient outcomes are strongly influenced by how rapidly deterioration is recognised, communicated, and managed rather than by the complication itself[44,45].

A patient recovering from low anterior resection illustrates this challenge well. Progressive tachycardia, increasing abdominal discomfort, reduced urine output, and reluctance to mobilize may initially appear as isolated postoperative findings. Considered together, however, they may represent the earliest manifestations of an anastomotic leak or evolving intra-abdominal sepsis. An AI system may recognise this pattern before it becomes clinically obvious, but patient safety still depends on how quickly the nurse reassesses the patient, informs the surgical team, initiates further investigation, and documents the response.

Evaluation should focus on the entire clinical pathway rather than on the alert itself. Measuring the interval from AI notification to bedside reassessment, from nursing reassessment to communication of clinical concern to the responsible surgical team, from that communication to medical review, and ultimately to therapeutic intervention provides a more meaningful assessment of implementation than reporting prediction accuracy alone. These indicators are particularly relevant in high-risk complications such as anastomotic leakage, pancreatic fistula, postoperative haemorrhage, and sepsis, where even relatively short delays may influence clinical outcomes.

Early complication recognition

Recognizing postoperative complications before they become clinically obvious remains one of the greatest challenges in gastrointestinal surgery and one of the areas in which AI may offer the greatest practical benefit for nurses. Many serious complications evolve gradually rather than presenting as a single dramatic event. Patients often deteriorate through a series of subtle physiological and clinical changes that, when viewed individually, appear unremarkable but together signal that recovery is no longer following the expected course. Although machine-learning models have demonstrated promising results in predicting complications after abdominal and gastrointestinal surgery[5,6,25-30], their true clinical value lies in helping healthcare teams recognize deterioration early enough to influence patient outcomes.

In everyday nursing practice, postoperative deterioration may emerge through a changing pattern of clinical findings rather than a single decisive abnormality. Experienced nurses may recognize that recovery is deviating from its expected course before a definitive complication is established. AI may support this judgement by identifying combinations or trends that warrant closer assessment, while the interpretation of those signals remains grounded in bedside evaluation.

Evaluating AI solely by its predictive performance provides only a partial picture of its clinical usefulness. More informative measures include whether deterioration was recognized earlier than would otherwise have been expected, whether nursing concern was documented before the complication became evident, how rapidly escalation occurred after the first abnormal trend, and whether timely intervention prevented progression to critical illness. Indicators such as unplanned intensive care admission, rescue interventions following delayed deterioration, and documentation of early warning signs may therefore provide a more meaningful assessment of AI-supported postoperative surveillance than prediction accuracy alone.

Adherence to risk-stratified care pathways

Risk stratification has value only when it changes the care that patients actually receive. ERAS pathways already provide structured recommendations for colorectal, pancreatic, liver, and esophageal surgery, yet clinical recovery rarely follows a single standardized trajectory. Frailty, malnutrition, extensive blood loss, organ dysfunction, and unexpected postoperative events frequently require individualized management despite adherence to established protocols[10-13].

AI may help identify these patients before clinical deterioration becomes obvious, allowing perioperative care to be adapted while recovery is still on course. Rather than replacing ERAS pathways, risk prediction can support more flexible implementation. A patient considered to be at increased postoperative risk may benefit from closer physiological monitoring, earlier dietetic assessment, intensified mobilization support, more frequent pain evaluation, or additional discharge planning. In contrast, patients progressing as expected may continue through standard recovery pathways without unnecessary intervention.

The responsibility for translating predicted risk into consistent clinical practice largely falls to nursing leadership. Classifying a patient as “high risk” has little practical value unless that designation is reflected in daily care. Observation frequency, escalation criteria, nutritional support, mobilization targets, wound assessment, and discharge preparation should all reflect the patient’s evolving clinical profile rather than a fixed protocol alone.

Implementation can be evaluated using indicators that demonstrate whether risk-adapted care actually occurred. Completion of individualized observation plans, timely referral to physiotherapy, nutrition, wound-care or stoma specialists, adherence to predefined escalation thresholds, and documented justification for pathway modification provide measurable evidence that AI-supported risk assessment has influenced patient management rather than remaining an isolated prediction.

Documentation accuracy and clinical reasoning

Clinical documentation is often viewed as an administrative responsibility, yet in AI-supported surgical care it becomes an essential component of clinical decision-making. Prediction models rely on accurate data, but clinicians rely just as heavily on nursing documentation to understand how a patient’s condition has evolved, what assessments have already been performed, which concerns prompted escalation, and how the multidisciplinary team responded. Contemporary AI reporting frameworks place considerable emphasis on data quality and transparency because unreliable documentation ultimately limits the clinical usefulness of any predictive model[17-22].

In gastrointestinal surgery, the clinical value of documentation lies in its ability to show change over time. Small shifts in drain output, oral intake, mobility, bowel function, pain, or fluid balance may be difficult to interpret in isolation, but become more meaningful when recorded consistently across successive assessments.

The purpose of AI should therefore not be to generate more documentation but to improve its clinical relevance. Instead of encouraging additional administrative tasks, AI-supported systems should guide nurses toward recording the observations most likely to influence subsequent clinical decisions. Respiratory findings, abdominal examination, wound appearance, drain characteristics, urine output, pain progression, mental status, and the actions taken after an alert provide information that supports both patient care and future interpretation of clinical deterioration.

Ultimately, documentation serves as more than a legal or administrative record. It provides the narrative of the patient’s recovery and allows clinicians to reconstruct how deterioration developed, how concerns were recognized, and whether opportunities for earlier intervention were identified. In this way, high-quality nursing documentation strengthens both clinical judgement and the performance of AI-supported care pathways.

Interdisciplinary communication

Successful implementation of AI depends as much on communication as on predictive performance. Risk information has little clinical value if it remains confined to a dashboard or electronic record instead of reaching the professionals responsible for patient care. In gastrointestinal and complex surgical pathways, safe decision-making relies on continuous collaboration between surgeons, anesthesiologists, nurses, intensivists, physiotherapists, dietitians, pharmacists, wound-care specialists, stoma nurses, and, in many cases, oncology or interventional radiology teams. Previous studies evaluating clinical decision-support systems have consistently shown that digital recommendations influence practice only when they are delivered at the right time, to the right professionals, and within existing clinical workflows[47-49].

Communication becomes particularly important when postoperative recovery begins to diverge from the expected pathway. Imagine a patient after major abdominal surgery whose physiological trends suggest an increasing risk of respiratory deterioration. The usefulness of the prediction does not depend on the numerical risk score itself but on whether the information prompts coordinated action. Nurses may increase respiratory assessment, physiotherapists may prioritize mobilization and airway clearance, physicians may reconsider analgesia or request additional investigations, and the multidisciplinary team may modify the monitoring plan before significant deterioration develops.

A similar situation arises during discharge planning. When AI identifies a patient with a newly created ileostomy as being at increased risk of dehydration or readmission, the prediction should influence the entire discharge process rather than simply appear in the electronic record. Nursing education, stoma care review, dietary advice, community referral, and follow-up arrangements all become part of the same coordinated response.

Assessing communication therefore requires more than documenting whether information was transferred. More meaningful indicators include the completeness of structured handovers, evidence that concerns were communicated and acknowledged, the interval between escalation and multidisciplinary response, and whether relevant risk information remained visible throughout the patient's journey from admission to discharge.

Cognitive workload and alert burden

Nurses working in gastrointestinal and complex surgical units rarely manage a single source of clinical information at a time. During an ordinary shift they are expected to respond to physiological monitoring, medication administration, laboratory results, early warning scores, family concerns, electronic documentation, patient requests, and unexpected clinical deterioration, often while coordinating care for several high-acuity patients simultaneously. Introducing AI into this environment should reduce cognitive burden rather than compete for attention.

The challenge is not the presence of alerts themselves but the way they are presented. When notifications are frequent, poorly prioritized, or provide little clinical context, they become increasingly difficult to distinguish from routine electronic interruptions. Alarm fatigue has long been recognised as a patient-safety concern in critical care and clinical decision-support systems, and similar problems may arise if AI generates a large number of alerts that rarely require meaningful action[50-53].

Maintaining clinicians' confidence therefore depends on alert quality rather than alert quantity. Nurses are more likely to engage with AI-supported recommendations when notifications are clearly prioritised, accompanied by clinically understandable explanations, and linked to actions that can realistically be performed within routine workflow. Conversely, repeated non-actionable alerts may gradually reduce trust in the system, increasing the likelihood that clinically important warnings will eventually be overlooked.

Other unintended effects should also be considered. Repeated reliance on algorithmic recommendations may contribute to automation bias, particularly when staff defer to an AI output despite conflicting bedside findings. Conversely, poorly calibrated systems may encourage unnecessary reassessment or escalation, increasing workload without improving care. Surveillance bias may also arise when patients identified as high risk receive more intensive observation and consequently have more abnormalities documented than lower-risk patients. These effects are important when evaluating implementation because additional monitoring, documentation, and escalation may divert nursing time from other patients and clinical priorities.

Evaluation should therefore extend beyond counting alerts. More informative measures include the proportion of alerts that result in clinical action, the frequency of false or non-actionable notifications, staff perceptions of usefulness, missed high-priority alerts, and the overall effect of AI on nurses’ mental workload during routine perioperative care. These indicators provide a clearer understanding of whether AI is supporting clinical decision-making or unintentionally contributing to cognitive overload.

Patient safety outcomes

Ultimately, AI-supported nursing performance should be linked to patient safety. In gastrointestinal and complex surgical care, relevant outcomes include failure to rescue, unplanned intensive care admission, postoperative sepsis, delayed recognition of bleeding or leak, surgical site infection, medication errors, readmission, prolonged length of stay, and patient-reported preparedness for discharge. Surgical safety research has shown that outcomes depend on both prevention of complications and timely rescue when complications occur[44-46].

Patient safety outcomes should be interpreted cautiously because they are influenced by case complexity, baseline patient risk, staffing, surgeon and team experience, anesthetic care, institutional resources, and other aspects of perioperative management. Improvements in these outcomes therefore cannot be attributed specifically to AI-supported nursing practice without comparative study designs and appropriate adjustment for confounding. Nursing-sensitive process measures may provide a more proximal indication of whether implementation has changed care delivery, but they should not be interpreted as evidence of a causal effect on patient outcomes. Future prospective studies should evaluate process and patient outcomes separately and account for relevant patient-, procedure-, and institution-level factors.

PRACTICAL FRAMEWORK FOR NURSING LEADERSHIP

A practical framework for nursing leadership in AI-integrated surgical care should not be viewed as a sequence of isolated managerial tasks. Rather, it represents a continuous process in which preparation, clinical translation, monitoring, and evaluation influence one another throughout the patient’s perioperative journey.

Preparation begins long before AI tools become part of routine clinical practice. Successful implementation depends on ensuring that nurses understand not only how a system functions, but also why it is being introduced, what clinical problem it is expected to address, and where its limitations lie. Scenario-based education, predefined escalation pathways, and clear expectations regarding the role of AI within existing workflows are likely to improve staff confidence during implementation. Current nursing literature consistently highlights professional preparedness and accountability as fundamental prerequisites for safe AI adoption, particularly in environments where rapid clinical decisions are required[54-60].

Translation connects AI-generated information with a defined clinical response. Nurse leaders should ensure that risk information is linked to appropriate observation, reassessment, communication, and escalation rather than remaining an isolated entry in the electronic record. The required response will vary according to the procedure, patient condition, and nature of the identified risk, but it should be clinically interpretable and clearly integrated into existing nursing workflows.

Monitoring requires continuous observation of how AI functions within everyday clinical practice rather than how it performs under ideal research conditions. Nurse leaders should regularly review whether alerts reach the appropriate professionals, whether recommended actions are completed within acceptable timeframes, whether documentation remains clinically meaningful, and whether staff perceive the technology as genuinely useful during routine care. Viewing implementation through a sociotechnical perspective is particularly valuable because technology, clinical workflow, organizational culture, and human behavior continuously interact during patient care[14-16].

Evaluation represents the final step, but also the starting point for further improvement. Introducing AI into surgical pathways should not be regarded as success in itself. Meaningful implementation should be demonstrated through measurable improvements in workflow efficiency, communication, documentation quality, response times, recognition of clinical deterioration, discharge preparation, and patient outcomes whenever these can reasonably be assessed. Continuous evaluation allows institutions to refine AI-supported pathways while preserving professional accountability and maintaining patient safety (Figure 2).

Figure 2 Nursing leadership framework for artificial intelligence integrated gastrointestinal surgical care: Author-proposed conceptual model. The figure presents a non-validated framework in which nursing leadership supports staff preparation, translation of artificial intelligence outputs into bedside actions, implementation monitoring, and evaluation of measurable performance outcomes. The framework requires prospective evaluation and local adaptation before implementation in clinical practice. AI: Artificial intelligence.

In this framework, nursing leadership is not a passive recipient of technological change. It is an active force that shapes how AI enters clinical practice. Without nursing leadership, AI may remain a physician-facing or administrative tool. With nursing leadership, it can become part of a safer, more coordinated, and more accountable surgical care pathway.

IMPLEMENTATION CHALLENGES AND ETHICAL CONSIDERATIONS

Implementation of AI in gastrointestinal and complex surgical care must address clinical, ethical, and organizational risks. Responsible machine learning in health care requires attention to safety, bias, generalizability, monitoring, and accountability[61]. Ethical analyses of AI implementation have also emphasized that clinical teams must understand who is responsible when algorithmic recommendations influence patient care[62]. Interpretability and explainability are likewise important for ensuring that AI-supported recommendations can be understood and appropriately contextualized by healthcare professionals[63-65]. These concerns are directly relevant to nursing, because nurses may be expected to act on AI-generated alerts during routine postoperative care. Experience from early clinical implementation suggests that successful AI adoption depends on considerably more than predictive performance. Hospitals with limited digital infrastructure, fragmented clinical workflows, insufficient interoperability between information systems, or inadequate staff training may experience little measurable benefit despite technically accurate algorithms. Staffing shortages, organizational culture, resistance to workflow change, and limited institutional resources may further influence implementation success. These observations underline that introducing AI into clinical practice is primarily an organizational transformation rather than a purely technological intervention[14-16,47-49,61] (Table 4).

Table 4 Implementation risks and nursing leadership safeguards in artificial intelligence-supported surgical pathways.

Implementation risk	How it may appear in practice	Potential consequence	Nursing leadership safeguard
Alert fatigue	Frequent low-priority alerts during already busy ward shifts	Important signals may be ignored or delayed	Define alert priority levels; monitor alert volume; remove non-actionable alerts; collect staff feedback
Unclear accountability	Alert visible to several professionals, but no one is clearly responsible for first response	Delay in reassessment or escalation	Assign first-response responsibility; define response time; document ownership of action
Over-reliance on AI	Staff wait for digital confirmation before escalating bedside concern	Clinical deterioration may be underestimated when no alert is generated	Reinforce that nursing concern overrides absence of alert; include this principle in training
Under-use of AI	Alerts are ignored because they are poorly understood or not integrated into workflow	Technology has little real-world clinical effect	Provide scenario-based training; integrate alerts into handover and ward routines
Poor interpretability	Nurses receive a risk score without explanation of relevant contributing factors	Low trust, inconsistent response, or inappropriate escalation	Require clinically understandable alert displays; include contributing variables where possible
Documentation burden	AI adds new forms or fields without reducing existing duplication	Increased workload and incomplete records	Align documentation templates with nursing workflow; remove duplicate fields; audit documentation quality
Inequitable performance	Model performs less well in frail, older, complex, or underrepresented patient groups	Risk may be underestimated or overestimated in vulnerable patients	Monitor subgroup performance; encourage clinical override; review unexpected deterioration events
Weak handover integration	Risk information remains in the electronic record but is not communicated during transitions	Receiving nurses may not know what to monitor	Add AI-derived risk status to structured perioperative handover
Lack of staff confidence	Nurses are unsure whether they are expected to trust, question, or act on alerts	Variable adoption and inconsistent practice	Use case-based education, simulation, and feedback after real alerts
Unmeasured implementation	Institution installs AI tool but does not evaluate nursing workflow or patient safety impact	No evidence that care delivery improved	Use a nursing-sensitive audit dashboard before and after implementation

AI: Artificial intelligence.

Implementation also requires investment beyond software acquisition. Resources are needed for staff education, technical support, system maintenance, interoperability with existing electronic health records, governance, cybersecurity, and continuous evaluation of clinical performance. These organizational requirements should be considered when estimating the overall value of AI-supported surgical care[61].

Interpretability is essential. Nurses do not need full access to model architecture, but they do need clinically understandable outputs. A warning that states only “high risk” is less useful than one that identifies the relevant pattern: Rising heart rate, worsening oxygen requirement, low urine output, increasing inflammatory markers, or deviation from expected mobilization. Explainability in health care AI is increasingly recognized as a requirement for safe clinical use[63-65].

Bias and inequitable performance must also be considered. Algorithms may perform differently across populations, hospitals, and documentation systems. Evidence of bias in health care algorithms has shown that apparently objective systems may reproduce inequities when trained or implemented without adequate oversight[66]. In complex gastrointestinal surgery, this matters for frail patients, older adults, patients with cancer-related malnutrition, and patients whose social circumstances affect recovery after discharge.

Finally, patient-centered care must remain central. AI may classify patients into risk categories, but nurses often understand contextual factors that are not fully captured by data: Fear, poor health literacy, lack of home support, reluctance to report symptoms, cultural barriers, or family concerns. World Health Organization guidance on AI for health emphasizes that governance, transparency, safety, and human-centered use are essential for responsible implementation[67,68]. AI-supported care should therefore strengthen, not narrow, the human assessment of surgical recovery.

LIMITATIONS

This review has several limitations. As a narrative review, literature selection was guided by relevance to the review objectives rather than by a predefined systematic-review protocol, and some degree of selection and citation bias cannot be excluded. The evidence considered is heterogeneous in study design, clinical setting, patient population, and stage of AI development. More importantly, direct evidence evaluating AI-supported nursing practice specifically within gastrointestinal and complex surgical pathways remains limited. Several of the nursing implications and performance domains discussed in this review therefore draw partly on evidence from surgical risk prediction, clinical decision support, early-warning systems, implementation science, and broader nursing literature. These sources provide a useful basis for developing testable concepts, but they do not establish that AI-supported nursing interventions improve patient outcomes in gastrointestinal surgery. Finally, the proposed performance domains constitute a conceptual framework rather than a validated measurement instrument. Their feasibility, reliability, and clinical relevance require prospective evaluation before they can be recommended as standardized performance measures. The framework should not be assumed to transfer unchanged across surgical populations, institutions, or healthcare systems; its indicators, thresholds, and implementation priorities will need to be adapted to the specific procedure, patient population, staffing model, level of digital maturity, available infrastructure and resources, and local clinical workflow. Regional differences in workforce organization, access to digital technologies, and institutional capacity should also be considered when determining which measures are feasible and clinically meaningful.

FUTURE DIRECTIONS

Future research should move beyond demonstrating that AI can accurately predict surgical risk and begin examining whether these predictions translate into meaningful improvements in perioperative care. From a nursing perspective, the central questions are no longer limited to model performance. It is equally important to determine whether AI-supported alerts shorten the time to bedside reassessment, facilitate earlier communication with the surgical team, improve recognition of postoperative deterioration, and ultimately contribute to safer patient management. These implementation outcomes deserve the same level of attention as traditional measures such as discrimination and calibration because they determine whether AI genuinely changes clinical practice rather than simply producing additional digital information[17-22].

Future progress will also depend on involving nurses much earlier in the development and evaluation of AI systems. In many institutions, digital tools are introduced after technical development has already been completed, leaving frontline professionals to adapt their practice to technologies they had little opportunity to shape. Nurses are uniquely positioned to identify whether alerts arrive at clinically appropriate moments, whether the proposed actions are feasible within routine workflow, and whether additional documentation requirements improve or hinder patient care. Their practical experience should therefore inform the design, refinement, implementation, and evaluation of AI-supported perioperative systems.

Emerging digital approaches may also broaden the role of AI in nursing beyond individual risk prediction. Digital twins, which create dynamic virtual representations of patients, workflows, or healthcare processes, are increasingly being explored for clinical and operational decision support. Their potential relevance to nursing includes anticipating workload, testing workflow changes, and supporting more timely allocation of clinical resources. However, most healthcare digital-twin applications remain at an early stage, and direct evidence of benefit in surgical nursing practice is still limited[69]. Another important direction is the integration of AI with standardized nursing terminologies. Structuring nursing assessments, diagnoses, interventions, and outcomes in a consistent and computable form can improve semantic interoperability and make nursing-generated data more suitable for analytics and clinical decision support. Multihospital implementation work has already shown that standardized nursing terminologies can support the deployment of AI-based decision-support tools across different electronic record systems while allowing nursing processes and outcomes to be tracked more consistently[70].

Implementation studies should extend beyond algorithm validation and evaluate how AI influences everyday clinical work. Hospitals introducing AI into gastrointestinal and complex surgical pathways would benefit from collecting baseline data before implementation and comparing them with outcomes after deployment. Nursing-sensitive indicators such as escalation delays, documentation quality, missed clinical deterioration, unplanned intensive care admissions, discharge preparedness, and staff workload may provide a more realistic assessment of clinical value than predictive accuracy alone.

Finally, perioperative nursing education will inevitably evolve alongside digital transformation. Future nurses are unlikely to require expertise in algorithm development, but they will need sufficient digital literacy to interpret AI-supported recommendations, recognize situations in which bedside assessment and algorithmic predictions diverge, understand the limitations of predictive models, and appreciate the ethical responsibilities associated with AI-assisted decision-making. Preparing nurses for these evolving responsibilities will become an essential component of safe perioperative practice as AI becomes increasingly integrated into routine gastrointestinal and complex surgical care[17-22,54-60]. Prospective multicentre studies will be needed to validate these proposed performance domains and determine their reliability, feasibility, and clinical utility across different surgical settings[17-22].

CONCLUSION

AI is becoming part of everyday gastrointestinal and complex surgical practice, yet its value will ultimately be determined by what happens at the bedside rather than by algorithm performance alone. Predictive models may identify patients at increased risk, but safer care depends on whether those predictions are translated into timely assessment, appropriate escalation, effective communication, and coordinated multidisciplinary action. From this perspective, AI should be viewed as a clinical support tool that complements, rather than replaces, nursing expertise.

The findings presented in this review suggest that nursing leadership will play a decisive role in determining whether AI becomes a meaningful component of perioperative care or remains an isolated digital technology. Nurse leaders are responsible for integrating AI-supported information into everyday clinical workflows, establishing clear escalation pathways, supporting professional judgement, reducing unnecessary alert burden, and ensuring that digital recommendations are interpreted within the clinical context of each individual patient. These priorities are consistent with the broader direction of contemporary AI evaluation, which increasingly emphasizes implementation, transparency, and human oversight alongside technical performance[17-22].

Rather than focusing exclusively on predictive accuracy, this review proposes a conceptual framework for considering AI implementation through nursing-sensitive performance domains that reflect routine clinical practice. Drawing on the reviewed evidence and relevant adjacent literature, these author-proposed domains include workflow efficiency, response time, early recognition of postoperative deterioration, adherence to individualized care pathways, documentation quality, interdisciplinary communication, cognitive workload, and patient safety outcomes. They should be regarded as candidate measures for future evaluation rather than as established performance standards. The framework has not been derived through formal consensus or instrument-development methods and requires prospective validation before its indicators can be recommended for standardized use. Its immediate value lies in providing a clinically grounded structure for asking whether AI-supported changes in surgical care are reflected in nursing workflow, clinical response, and patient safety while preserving professional nursing judgement.