Enhanced recovery after surgery in colorectal surgery: Do average benefits mask heterogeneous effects in patients underrepresented in trials?

Abstract

For elective colorectal surgery, enhanced recovery protocols are now standard practice, with convincing trial evidence showing fewer complications and shorter stays. What we know far less about is whether that same evidence reliably applies to the patients who were routinely excluded from those trials, older frail adults, those with severe obesity, and people carrying multiple chronic conditions. Three observations from the available data give reason for caution. First, baseline physiological vulnerability predicts complications irrespective of protocol fidelity, a finding that suggests the benefit of compliance is not uniform across patient groups. Second, patients least likely to meet trial eligibility complete fewer elements of the protocol; whether such shortfalls reflect sound clinical judgement or inadvertent undertreatment has not been established. Third, socioeconomic disadvantage and ethnic minority background are linked to wider adherence gaps and poorer outcomes, even where institutional compliance appears high. These signals converge on a single gap in measurement. Existing audit systems document whether a protocol step was carried out, but not why it was omitted. A framework that fails to account for why protocol steps are omitted cannot distinguish clinical adaptation from neglect. Building that distinction into perioperative audit may matter as much for the patients at the margins of evidence as achieving higher compliance scores does for the average patient.

Key Words: Colorectal surgery; Enhanced recovery after surgery; Protocol compliance; Patient selection; High-risk patients; Clinical reasoning; Audit; Healthcare disparities; Observer variation; Quality improvement

Core Tip: Enhanced recovery after surgery audits capture whether a protocol element was actually carried out, not why it was omitted. The reasons for omitting a step are not all the same: Some deviations reflect clinical caution, others simple oversight. That distinction matters most when the patient in question does not resemble the trial populations on which the protocol was originally based. Audit systems that record the rationale for non-adherence, rather than merely its occurrence, have the potential to improve perioperative care, although this hypothesis requires prospective validation.

INTRODUCTION

Enhanced recovery after surgery (ERAS) has been shown to reduce complications and shorten hospital stay after elective colorectal resection. ERAS traces back to Kehlet and Wilmore’s insight that surgical recovery is not a fixed postoperative trajectory but a physiological process amenable to intervention[1]. From that insight emerged a multimodal program intended to blunt the surgical stress response and make recovery more predictable[2]. Meta-analyses of randomized trials have documented consistent reductions in length of stay of 2 to 3 days, alongside a 30%-40% lower risk of overall morbidity[3-6]. Registry data add that compliance and outcomes appear linked in a dose-response fashion: Roughly every 10% gain in adherence corresponds to about a 5% drop in complication odds[7,8]. In most high-income settings ERAS is now standard care.

It is less clear whether patients outside trial populations benefit to the same degree. The trials that established ERAS drew on tightly selected cohorts, often excluding the very old, the severely obese, and those with heavy comorbidity[9,10]. Some exclusions are clinically unavoidable; randomizing a patient with decompensated heart failure to a fluid-restriction protocol would raise legitimate safety concerns. The problem is no longer the original exclusions, which were justified, but the fact that today’s protocols, barely altered, are being applied to the very patients those trials left out. Whether adherence still predicts benefit in these groups remains an open question.

Recent work offers mixed signals on this question. Studies restricted to elderly patients undergoing major abdominal surgery suggest that ERAS confers benefit in this group, yet compliance reporting is highly variable and complication rates remain higher than in younger cohorts[11]. Patients with inflammatory bowel disease, advanced age, or American Society of Anesthesiologists (ASA) status III-IV complete fewer protocol elements, and the reasons for these departures are poorly characterized[12]. Computational analyses of registry data have identified baseline physiological vulnerability, frailty, ASA class, body mass index, and age as independent predictors of complications, with effects that persist after accounting for protocol adherence[13,14]. Socioeconomic disadvantage and minority ethnicity have been associated with wider adherence gaps and worse outcomes, even in centers reporting high institutional compliance rates[15-17]. None of these observations challenges the value of ERAS in the populations for which it was designed. This suggests that the distribution of benefit varies across patient subgroups, and that aggregate compliance scores alone do not capture that variation.

The ERAS Society published its first consensus guideline for colorectal surgery in 2005, with subsequent updates in 2019 and 2025[18-20]. The latest update acknowledged, for the first time in explicit terms, that protocol elements may require modification in light of individual patient physiology. Stratified recommendations for frail patients, those with obesity, and those with cardiac or renal comorbidity now appear in the guidelines, a recognition that standardized pathways cannot be applied identically to all comers[20]. Yet between the guideline and the audit system lies a gap that has received almost no critical scrutiny. Current ERAS audit platforms record whether a protocol element was delivered, but not why it was not delivered[21-23]. The audit system records whether a deviation occurred, but not the clinical reasoning behind it—whether it reflected a defensible clinical judgment or an inadvertent omission.

This is not a minor point of measurement. If clinically indicated deviations are systematically more common in high-risk patients, as the limited available evidence suggests may be the case, then aggregate compliance scores will be biased in ways that current data cannot quantify[12,24]. The bedside question that audit frameworks still cannot answer is whether a protocol deviation represents thoughtful care or an avoidable failure. Programs geared entirely towards compliance, with no parallel mechanism to record justified adaptations, may inadvertently undermine the clinical discretion that matters most for vulnerable patients.

Here we bring together work from surgery, cognitive psychology, and safety science to focus on one overlooked distinction: Whether a protocol deviation represents a thoughtful adjustment or a simple failure to deliver. Let me be clear, we are not questioning the average benefit of ERAS. Our argument is this: A system that records why a step was omitted, not just that it was, might be just as important for perioperative quality as another percentage point of compliance. We ask three questions. First, what does the evidence actually tell us, and what does it miss, about how adherence affects outcomes in patients who would never have made it into the trials? Second, why could the pursuit of compliance scores end up distracting us from better care? Third, what would a practical, workable audit system that truly separates adaptation from omission look like?

We use non-adherence and non-compliance interchangeably for any protocol element not delivered as specified, regardless of the reason. We use deviation as a neutral term covering any departure, whether justified or not. We reserve omission for departures without a documented clinical rationale. By adaptation we mean deliberate departures made for clinical reasons in response to individual patient physiology or circumstances. We recognized that the boundary between adaptation and omission is often unclear in practice, and that very ambiguity is the central problem we are trying to get at.

Before proceeding, it is important to note the nature of the evidence we have assembled. We distinguish three levels of claim. Established findings are supported by multiple independent studies with consistent results, for example, the association between higher ERAS compliance and better outcomes in unselected populations, and the observation that high-risk patients complete fewer protocol elements. Interpretations are our own inferences from the available data, for example, the suggestion that aggregate compliance scores may underestimate the compliance-outcome association in high-risk subgroups. Untested hypotheses are proposals not yet empirically evaluated, for example, the proposition that adding a deviation rationale field to audit platforms would improve outcomes.

LITERATURE SEARCH AND SYNTHESIS

MEDLINE, EMBASE, and PsycINFO were searched from January 2000 to July 2026. Search terms combined Medical Subject Headings and free-text terms for the following concepts: (“colorectal surgery” OR “colonic surgery” OR “rectal surgery”) AND (“enhanced recovery” OR “ERAS” OR “fast-track”) AND (“compliance” OR “adherence” OR “protocol deviation”) AND (“frailty” OR “obesity” OR “comorbidity” OR “socioeconomic” OR “disparity” OR “high-risk”). The full search strategy for MEDLINE is provided in Supplementary Table 1, and the detailed study-level quality assessment is provided in Supplementary Table 2. We did not impose language restrictions at the search stage, although the final included studies were limited to English-language publications for practical reasons.

Two reviewers (Ding YB and Yu H) independently screened titles and abstracts, and then full texts against pre-specified eligibility criteria. Studies were eligible if they reported original data on: (1) ERAS protocol compliance in elective colorectal surgery; (2) Outcomes stratified by patient subgroup (age, comorbidity, body mass index, frailty, or socioeconomic/demographic factors); or (3) Reasons for protocol non-adherence. We excluded case reports, editorials, commentaries, and studies with fewer than 50 patients. Disagreements were resolved by consensus with a third reviewer (Zhan XL).

The search yielded 1847 unique records. After title/abstract screening, 214 full texts were assessed; 103 met inclusion criteria and are cited in this review. A PRISMA-style flow diagram is provided in Supplementary Figure 1. The reference lists of included articles and relevant systematic reviews were hand-searched for additional studies.

Because the evidence base spans clinical epidemiology, cognitive psychology, and organizational safety, each with its own methodological traditions, a narrative synthesis approach was adopted, following established guidance[25,26]. The reporting structure was informed by the PRISMA statement for systematic reviews, adapted for a narrative synthesis[27]. We chose a narrative synthesis because the heterogeneity of study designs, populations, and outcome measures precluded a meaningful quantitative meta-analysis. The Newcastle-Ottawa Scale was used for quality appraisal of cohort studies; the Cochrane Risk of Bias tool was applied to randomized trials where applicable. The results of these quality assessments are summarized in Table 1. Evidence from cognitive psychology and safety science is explicitly labeled as indirect and hypothesis-generating. This is therefore a narrative synthesis, not a quantitative meta-analysis; its contribution lies in conceptual integration across disparate bodies of evidence, rather than in the calculation of a summary effect size. The conceptual structure of the review is summarized in Figure 1.

Figure 1 The conceptual structure of this review. ERAS: Enhanced recovery after surgery.

Table 1 Summary of included study characteristics and quality assessment.

Study type	Number of studies	Quality assessment (NOS/RoB)	Key findings relevant to this review
Randomized controlled trials	8	Low-to-moderate RoB	ERAS reduces LOS by 2-3 days, complications by 30%-40%
Prospective multicenter cohorts	12	NOS 7-9 (high quality)	Dose-response: 10% compliance increment → approximately 5% complication reduction
Single-center retrospective cohorts	23	NOS 5-7 (moderate quality)	High-risk subgroups have lower compliance and worse outcomes
Registry analyses	7	NOS 7-8 (high quality)	Socioeconomic gradients in compliance and outcomes
Qualitative studies	4	CASP qualitative checklist	Barriers to adherence: Clinician judgement, patient factors, system constraints
Cognitive psychology/safety science	8	N/A (indirect evidence)	Automation bias, Goodhart’s Law - hypothesis-generating only

CASP: Critical Appraisal Skills Program; ERAS: Enhanced recovery after surgery; LOS: Length of stay; N/A: Not applicable; NOS: Newcastle-Ottawa Scale; RoB: Risk of bias.

WHAT THE EVIDENCE DOES AND DOES NOT SAY

The established association

In unselected colorectal populations, higher ERAS protocol adherence is consistently associated with better postoperative outcomes. A landmark multicenter cohort study showed that each 10% increment in adherence reduced the odds of complications by roughly 5%. That finding has since been replicated in several independent cohorts, and meta-analyses confirm a 30%-40% reduction in overall morbidity and a 2- to 3-day shorter hospital stay. A landmark prospective multicenter cohort analysis by Gustafsson et al[7] demonstrated a graded dose-response relationship: Each 10% increment in ERAS protocol adherence was associated with roughly a 5% reduction in the odds of postoperative complications. That finding was subsequently replicated in a systematic review by Ahmad et al[8], who further showed that postoperative elements, rather than preoperative or intraoperative ones, drove the association with optimal recovery. Meta-analyses of randomized trials have confirmed reductions in hospital stay of 2 to 3 days, alongside a 30%-40% lower risk of overall morbidity, when ERAS pathways are compared with traditional care[3-6,28].

Large-scale registry studies have introduced an additional layer of nuance. The Italian iCral2 multicenter prospective study confirmed that institutional adherence above 75% independently predicted faster functional recovery[29], and a pooled analysis of iCral2 and iCral3 data demonstrated lower rates of major morbidity and mortality with high compliance[30]. Longitudinal data from a Swedish center linked high adherence to improved 5-year survival after colorectal cancer surgery[31]. A difference-in-differences analysis found that ERAS implementation was associated with decreased organ-level complication rates, further strengthening the case for protocolized care[32]. Compliance with individual ERAS elements has been shown to vary considerably in routine practice, with early studies documenting adherence rates ranging from 40% to 90% across centers[33]. Taken together, these data leave little doubt that, in the populations studied, higher compliance is associated with better outcomes.

The populations studied, however, represent only a subset of those who undergo colorectal surgery in routine practice. The randomized trials that generated the evidence for ERAS systematically excluded individuals at the extremes of age, body mass index, and comorbidity burden[9,10]. The large registry studies, though more inclusive, still drew on centers with well-established ERAS programs and comparatively high baseline compliance, contexts where the most physiologically vulnerable patients would already have been managed with a degree of individualization. So the real question is whether that association holds equally for every patient now managed under these protocols, or whether it varies across subgroups.

SIGNALS FROM THE MARGINS

Physiological extremes

A number of recent studies have explored ERAS outcomes in subgroups that sit at the boundaries of trial eligibility. A systematic review and meta-analysis restricted to elderly patients undergoing major abdominal surgery reported that ERAS reduced hospital stay by a mean of 3.3 days and complications by roughly 40%. Yet compliance reporting varied widely, from 42% to 90% across studies, and complication rates in older cohorts remained higher than in younger ones, even within ERAS-managed pathways[11]. An Italian single-center prospective observational study of 656 patients found that patients with inflammatory bowel disease, advanced age, or ASA status III-IV completed fewer protocol elements[12]. Computational analyses of registry data have added a different angle: A retrospective cohort study by Ren et al[13] (649 patients) used logistic regression to identify predictors of postoperative complications within 30 days; preoperative anemia and old age were independent risk factors. A separate European analysis, using data from 2841 patients across 22 centers, arrived at a similar conclusion, patient-level risk factors carried predictive weight that was not erased by high compliance[14]. Studies in patients with class III obesity have highlighted physiological considerations that place standardized protocol elements in tension with individual physiology: Delayed gastric emptying affecting carbohydrate loading, weight-based fluid restriction creating relative hypovolemia, and obesity-related restrictive pulmonary physiology attenuating the respiratory benefits of upright positioning[34-36]. A recent cohort study of high-risk patients undergoing colon cancer surgery found that achieving 70% or higher compliance reduced the risk of prolonged stay by 86% and complications by 70%, leading the authors to propose that poor outcomes in high-risk patients reflect barriers to protocol completion rather than inherent physiological limitation[37].

These observations tell us that patients with greater physiological vulnerability have worse outcomes and complete fewer protocol elements. But they cannot tell us why. Two interpretations are possible. The first is causal: High-risk patients do worse because they receive less of the protocol. The second is contextual: The protocol was developed in a narrower trial population, so some elements may not suit certain high-risk patients, and clinical teams, perhaps appropriately, withhold them. The current data cannot distinguish between these interpretations, because the datasets underpinning these analyses did not record why individual protocol elements were omitted.

One thing worth bearing in mind: The link between low compliance and poor outcomes might not be what it looks like. It is easy to assume that low adherence causes complications, but complications themselves can make it impossible to follow the protocol. A patient with a leak, ileus, or pneumonia cannot mobilize early, cannot eat, cannot have the catheter out. In that case, the low compliance is a consequence, not a cause. Most registry data do not track whether a protocol element was delivered before or after a complication occurred, nor do they account for the fact that the timing itself can distort the association. This matters most for postoperative steps, which clinicians naturally adjust as a patient’s condition evolves, far more than pre- or intraoperative elements. The sensible way forward is to analyze compliance separately by phase—preoperative, intraoperative, and postoperative—and to be explicit about timing. The data we have do not allow that. That represents a gap in the literature, not a limitation of our review.

Ceiling effects

The adherence-outcome relationship also appears to flatten at the upper end. Several studies suggest that once compliance reaches about 75%, further increments add little additional benefit. The Italian ERCOLE cohort of 1138 patients found that functional recovery time was shorter with adherence of 75% or higher, but no significant difference existed between the 75%-99% and complete adherence groups[38]. A Belgian retrospective cohort by Hartman et al[39] (284 patients, 2007-2013) found that greater adherence to the fast-track protocol was associated with fewer postoperative complications and shorter hospital stay. A Swiss retrospective analysis similarly noted that higher composite compliance reduced complications, but no individual ERAS element independently predicted outcome after adjustment[24]. Let us be clear: We are not saying that high compliance does not matter. We are saying that once you pass a certain point, say, around 75%, the extra benefit of chasing the remaining 20% is small. And that raises a practical question: Would we do more good by ensuring that the essential elements actually reach the patients who stand to gain the most, rather than by obsessing over moving the center’s compliance score from 85% to 95%?

Socioeconomic and demographic disparities

Physiology is only part of the story. Socioeconomic disadvantage and ethnicity also matter, patients from these groups have lower adherence and worse outcomes, even when they are treated in centers that otherwise report high compliance. A single-center study of 1519 elective colorectal procedures in the United States found that low socioeconomic status was independently associated with higher odds of infectious complications (odds ratio 2.46) and all in-hospital complications (odds ratio 1.50)[15]. An English cohort of 1835 colorectal cancer resections reported a similar gradient: Patients from the most deprived quintile had significantly higher overall complications and longer postoperative stay compared with the most affluent quintile[40]. Racial disparities have been documented as well. Black patients in a United States academic center had lower rates of preoperative counselling, carbohydrate loading, and early mobilization, even after adjustment for socioeconomic status and comorbidity[16]. A matched cohort study from a Veterans Affairs hospital found that ERAS eliminated racial differences in length of stay[41], but more recent data from civilian academic centers have shown persistent disparities in both process measures and outcomes[15,16]. Studies restricted to patients with inflammatory bowel disease have similarly found that racial disparities in postoperative complication rates persisted both before and after ERAS implementation[42]. The evidence comes predominantly from the United States, and its whether it translates to health systems with universal coverage is uncertain. The implication is clear: Standardized pathways will not deliver equitable outcomes unless they are designed with patients’ social circumstances in mind.

THE MEASUREMENT BLIND SPOT

What current audits capture, and what they miss

Across all the signals described above runs a common methodological thread: The measurement of compliance itself. Contemporary ERAS audit systems record whether a protocol element was performed[21-23]. The system reduces everything to a single percentage. A hemodynamic instability, a patient’s refusal, or a prescribing oversight by the junior team, they all feed into the same bucket: Non-adherent. The dataset never asks why.

A single-center study by Roulin et al[24] provides the closest available look at what lies beneath the aggregate score. In a detailed audit of elective colorectal ERAS cases, the investigators classified each instance of non-compliance as either clinically justified or not. The majority of deviations were judged to be medically indicated, initiated by the treating physician rather than representing oversight. That study is small, single-center, and its classification relied on retrospective chart review by a local team, limitations that prevent broad generalization. What the clinical record does contain, though, is enough to tell a justified deviation from an unjustified one, even if the audit system is not currently set up to capture that distinction.

Why this distinction matters

The conflation of justified adaptation and inadvertent non-adherence has consequences that extend beyond the statistical. In research, if clinically indicated deviations are more common in high-risk patients, a pattern of which the available evidence offers suggestive but inconclusive support, then aggregate compliance scores may underestimate the association between protocol adherence and good outcome in the subgroups for whom the question of benefit is most urgent. The direction and magnitude of any such bias cannot be determined from current data, and could in principle vary across settings and patient groups.

This is not just an academic distinction. When an audit system lumps every deviation together, whether for good reason or bad, it effectively says to clinicians: A deviation is a black mark, full stop. There is evidence, albeit indirect, that this message is felt. A qualitative study of barriers and facilitators to ERAS adherence in the United Kingdom found that junior surgeons and nurses reported feeling obliged to follow protocol elements even when they had reservations about their suitability for specific patients[43]. A review by Roulin and Demartines[44] discussed the principles and implementation challenges of ERAS in gastrointestinal surgery. These data do not establish that ERAS auditing causes clinicians to override their judgement. The conditions for such pressure are clearly present. But a measurement system that cannot distinguish good judgement from error has no way of knowing whether that pressure is actually exerting an effect.

This is not a gap in theory, it is a gap in practice. It is why the question we posed earlier, whether high-risk patients do worse because they get too little of the protocol or because the protocol asks too much of them, has not been answered. The data to answer it are generated every day in every colorectal unit. They are simply not being captured.

WHY COMPLIANCE SCORES CAN MISLEAD: THE MEASUREMENT PROBLEM

The case for stratified audit

Current audit systems record what was delivered but not why something was withheld. That is not a problem unique to colorectal ERAS, as other areas of medicine have faced the same issue and have already built workable solutions. Computerized provider order entry systems, for instance, routinely allow clinicians to record a reason when they override a drug safety alert, contraindication, patient intolerance, or intentional deviation based on clinical judgement, and audit logs distinguish between these categories[45]. The World Health Organization surgical safety checklist, now used in operating theaters worldwide, explicitly permits items to be marked as not applicable rather than forcing a binary completed/not completed classification[46]. Even sepsis bundles, often criticized for their rigidity, allow clinicians to note when an element is contraindicated, because not every patient fits the mold[47]. What these systems share is that they track not just whether a step was completed but why it was skipped.

Applying this principle to ERAS audit would require no new infrastructure. The ERAS Interactive Audit System already supports customizable data fields for center-specific quality metrics[21]; the Korean ERAS Research and Audit System offers real-time dashboards with continuous data-completeness monitoring[22]. Adding a single categorical field that records, for each omitted protocol element, one of four reasons: Clinical contraindication, patient refusal or intolerance, system constraint, or no reason documented—would build on existing platforms rather than replace them. The importance of designing measurement systems that capture context rather than mere completion has been articulated in frameworks developed for patient safety monitoring[48], and the distinction between process compliance and thoughtful deviation resonates with the Safety-II perspective, which emphasizes understanding why things go right under variable conditions rather than simply counting when they go wrong[49]. A three- to five-center pilot could test this within 6 months. It would need to show whether the extra data entry time is acceptable, whether deviation classifications are reliable across raters, and how deviation patterns vary across patient subgroups.

The other problem with aggregate compliance scores is that they treat every protocol element equally. Failing to provide carbohydrate loading is recorded as a failure in exactly the same way as failing to mobilize a patient on day 1, even though the two could hardly be more different.

Carbohydrate loading has moderate evidence and few real contraindications; early mobilization, however, depends on pain, stability, and the absence of delirium or lines. Weighted scores have been proposed but have not gained widespread adoption. Patient factors make the problem even more complex; an element that is genuinely not right for one patient is not equivalent to a step that was simply overlooked. Our proposed stratified audit would allow us, for the first time, to see whether some elements are more often withheld for good reason, and whether compliance patterns are driven by a few specific elements or distributed uniformly.

What aggregate compliance scores cannot tell us is whether protocol deviations in high-risk patients are mostly clinically appropriate or potentially avoidable, and that is precisely what stratified audit could clarify. If frail patients who develop complications are overrepresented among those whose deviations have no documented rationale, it suggests they are missing elements they could have tolerated, a gap that targeted improvement might address. If, by contrast, complications occur despite high rates of justified adaptation, then the protocol itself, rather than merely its delivery, may need adjustment for that group. The current data cannot tell us which of these is true. Stratified audit would not be about dictating care; it would make clinical behavior visible enough to study, and, ultimately, to improve.

Concerns about the unintended consequences of performance measurement in healthcare have been documented across multiple settings[50,51]. Quality improvement programs can produce strikingly different results when implemented in different institutional contexts, even when the protocol itself is identical, because local culture, leadership, and team dynamics shape how standardized tools are used in practice[52]. When measures become targets, they may cease to be good measures, a dynamic first described in economics as Goodhart’s Law and subsequently observed in healthcare contexts ranging from emergency department waiting times to sepsis bundle reporting[47,53-55]. Whether comparable effects are at play in colorectal ERAS programs is unknown, but an audit system that distinguishes justified adaptation from inadvertent omission would, at the very least, make such effects detectable. Any such system must also navigate the tension between “no blame” cultures that encourage open reporting and the need for accountability when deviations are not clinically justified[56].

The equity dimension

The measurement gap also cuts across equity. Studies consistently show that socioeconomically disadvantaged patients and certain minority groups have lower ERAS adherence and worse outcomes, even in centers with strong overall compliance[15,16,40]. Why? We do not know. It could be that clinicians are sensibly holding back. Or it could be that barriers like literacy, language, transport, or food insecurity are getting in the way. These are not just different explanations, they call for different responses. If low adherence in disadvantaged patients mostly reflects comorbidity that legitimately rules out certain steps, then lower compliance is not necessarily a sign of poor care. If, on the other hand, these patients are systematically missing elements like preoperative education, carbohydrate loading, or early mobilization, not because of their physiology, but because of poor communication, language gaps, or discharge plans that assume too much about home support, then our audit system, by counting only what was done, is hiding a problem we could fix. A single-center study from the United States found that Black patients had lower rates of documented preoperative counselling and carbohydrate loading after adjustment for socioeconomic status and comorbidity[16], a pattern consistent with differential delivery rather than differential physiology. These factors work through different pathways, and treating them as one risks missing what each actually needs.

Socioeconomic disadvantage, measured by income, occupation, or area-level deprivation, affects access to preoperative resources (carbohydrate loading, education), the ability to take time off work for surgery and follow-up, and the availability of home support after discharge. The English cohort study by Godbole et al[40] documented a clear gradient: Patients from the most deprived quintile had significantly higher complications and longer stay than those from the most affluent quintile. This gradient is likely to reflect material barriers rather than physiological differences.

Race and ethnicity, in the United States context, Black and Hispanic patients, have been associated with lower rates of documented counseling and protocol completion even after adjustment for socioeconomic status. The mechanisms here are more complex, encompassing both healthcare system factors (differential referral, access to specialist care) and interpersonal factors (communication, trust, implicit bias). The finding from Wahl et al[41] that ERAS eliminated racial differences in length of stay in a Veterans Affairs setting, while Leeds et al[16] found persistent disparities in a civilian academic center, suggests that the effect of race on ERAS outcomes is moderated by the healthcare context.

Language barriers affect a patient’s ability to understand preoperative instructions, engage with education, and communicate symptoms postoperatively. These barriers are distinct from socioeconomic status and race; a patient may be affluent and from a racial minority or non-minority group yet still face language barriers that compromise protocol completion.

Health literacy, the capacity to obtain, process, and understand basic health information, determines whether a patient can follow a carbohydrate-loading schedule, recognize early signs of complications, or adhere to mobilization targets. Theiss et al[57] found that limited health literacy was associated with higher complication rates after colorectal surgery, an effect likely independent of income, race, or language.

Structural barriers, including transport difficulties, food insecurity, and unstable housing, affect the patient’s ability to attend appointments, obtain prescribed medications or supplements, and recover in a supportive environment. These factors are often correlated with socioeconomic disadvantage but are not synonymous with it; a middle-income patient with limited transport options may face the same practical barrier as a low-income patient.

The implication for audit is that these different mechanisms require different responses. A documentation field that records “patient inability” as a single category, without distinguishing whether the inability was due to language, health literacy, transport, or personal circumstances, would replicate the same conflation we have criticized in existing audit systems. The deviation rationale field we propose would need to be sufficiently granular to distinguish these mechanisms, or, at minimum, to flag that such distinctions are important.

Social vulnerability also exerts an influence on those elements of the pathway that lie beyond the hospital setting. Preoperative carbohydrate loading requires that the patient obtain and consume a specific drink 2 hours before surgery; for a patient who relies on public transport, has limited health literacy, or faces food insecurity, this seemingly simple step may be insurmountable[57,58]. Post-discharge follow-up requires telephone access, a stable address, and the ability to take time off work or from caregiving. When patients who lack these are recorded as non-adherent, the audit entry treats a social barrier as a clinical failure and provides no indication of what the system could do differently—for example, providing carbohydrate drinks free of charge at the pre-assessment clinic or scheduling follow-up calls outside normal working hours.

Broader frameworks for understanding health disparities are relevant here. Health inequalities are shaped by the conditions in which people are born, grow, live, work, and age—the social determinants of health—and these structural factors operate through pathways that clinical protocols alone cannot address[59,60].

Implicit bias among healthcare professionals has been documented across multiple settings, including surgery, and its potential influence on clinical decision-making is increasingly recognized[61]. Language barriers have been independently associated with poorer health outcomes and lower adherence to medical recommendations[62]. Racial disparities in surgical care and outcomes have been comprehensively reviewed, with contributing factors identified at patient, provider, and systemic levels[63]. The National Institutes of Health and American College of Surgeons have jointly called for a coordinated research agenda to address surgical disparities, emphasizing that measurement is a prerequisite for improvement[64]. An audit system that recorded non-delivery reasons would tell us something we cannot currently see: Who is missing what, and why. If a particular group has a disproportionate number of omissions related to systems or personal barriers, that points to what the system can do differently, not to a failure of the clinical team.

Recent work has begun to turn its attention to equity-focused adaptations to ERAS protocols. Structured approaches incorporating social determinants screening, tailored patient education, and proactive post-discharge case management have been proposed, although the evidence base remains limited to expert consensus and small-scale quality improvement projects[65]. Audit redesign alone is not the answer; it never was. But unless we change what we measure, we cannot tell whether a patient failed to receive a step because they could not manage it, or because they should never have been asked to do it in the first place. That distinction is not academic, it is essential to delivering equitable care.

We acknowledge that the evidence on racial and socioeconomic disparities reviewed here is predominantly from the United States. This reflects the distribution of published research, not a judgment that these patterns are unique to the United States. Whether similar disparities exist in health systems with universal coverage, such as the National Health Service in the United Kingdom or the public hospital networks in East Asia, remains less certain. The English cohort study by Godbole et al[40] demonstrated socioeconomic gradients in outcomes under a universal coverage model, suggesting that the phenomenon is not confined to systems with private insurance and residential segregation. Across universal coverage systems, the mechanisms may differ: Disparities may operate more through health literacy, culturally appropriate care, and access to primary care than through insurance or out-of-pocket costs. The absence of comparable data from East Asia, Latin America, and the Middle East is an important omission in the published literature.

The global dimension adds a further layer. ERAS implementation in low- and middle income countries faces structural constraints that make protocol completion a fundamentally different challenge than in well-resourced settings[66-68]. The global surgery community has articulated the scale of unmet surgical need and the infrastructure gaps that constrain perioperative quality improvement in resource limited environments[69]. Whether stratified audit, which presupposes a baseline level of data infrastructure and personnel, is feasible or prioritized in such contexts is uncertain. Yet even in settings where full protocol compliance is unrealistic, documenting the reasons for non-completion—whether resource unavailability, supply chain interruptions, or workforce shortage—could serve an advocacy function by making visible the gap between what guidelines recommend and what local conditions permit[70].

Theoretical perspectives: A brief note

It is worth drawing on two frameworks from outside surgery: One from economics and one from psychology. Both describe how the pursuit of a metric can, under certain conditions, begin to subvert the very thing it is meant to measure. Neither has been validated in an ERAS context, and neither should be read as proof that harm is actually happening, but both offer useful ways of thinking about what might be at stake.

The first is Goodhart’s Law, an economic principle stating that when a measure becomes a target, it ceases to be a good measure. In healthcare, this dynamic has been documented: Emergency department waiting-time targets in the English National Health Service incentivized admissions before the 4-hour deadline, including some that were unnecessary[53].

Public reporting of sepsis bundle compliance in the United States has been associated with overdiagnosis and overtreatment[47,54]. A systematic review of healthcare providers’ attitudes toward pay-for-performance found that most providers had serious concerns about unintended consequences[55]. The notion that medicine has entered an “era of metrics,” in which measures of quality become ends in themselves, has been articulated as a caution against unreflective pursuit of compliance scores[71]. Whether comparable effects operate in colorectal ERAS programs is unknown. But the conditions for these dynamics are real: Benchmarking, score pressure, and an audit system that never asks why. Whether any harm has resulted is an open question, and one that current audit data are not designed to address.

The second framework, drawn from cognitive psychology, is automation bias: The tendency, documented in simulation studies, to follow standardized recommendations even when contradictory information is clinically relevant[72,73]. The concept originates in human factors research, where it describes the observation that operators who rely extensively on automated systems may show measurable deterioration in the capacity to detect and respond to non-routine events[74]. High-reliability organizations address this vulnerability not by eliminating protocols but by cultivating a capacity for mindful deviation when circumstances demand it[75]. In medicine, a systematic review found that clinically correct decisions were abandoned in favor of incorrect decision-support advice in a measurable proportion of test cases[76]. Experiments with medical students and residents have shown that protocol-based training can improve performance on cases that match the protocol while impairing reasoning on cases that deviate from it[77,78]. A systematic review of cognitive biases in surgical settings found consistent evidence that heuristics influence surgical decision-making, although the review did not specifically address longitudinal effects of protocolized practice[79].

The broader literature on clinical reasoning distinguishes between intuitive and analytical cognitive processes, as described by the dual-process theory, and has identified conditions under which intuitive expert judgement is trustworthy and other in which analytical override is warranted[80,81].

Strategies for reducing cognitive bias, including debiasing techniques, have been proposed, although their effectiveness in real-world clinical settings remains under investigation[82].

These findings come from controlled laboratory settings with simplified tasks and homogeneous participants; whether they apply to experienced colorectal surgeons managing complex patients has not been tested. We include them not to suggest that such a problem exists, but to highlight that similar cognitive trade-offs have been documented in other clinical domains and that an audit system that ignores clinical reasoning will also miss whether such trade-offs are occurring.

The gap between theory and evidence remains wide, and we do not attempt to close it here. What unites these two frameworks is a structural insight: Measurement systems, by rewarding some behaviors and ignoring others, reshape clinical incentives, however subtly. A stratified audit tool would not eliminate this dynamic, but it would make its consequences observable and therefore correctable.

A RESEARCH AGENDA THAT RESPECTS CLINICAL REALITY

At this point, the evidence we have reviewed does not justify a change in clinical practice. The signals, persistent complications in high-risk subgroups despite moderate compliance, ceiling effects at high adherence levels, socioeconomic gradients, and the common measurement blind spot, are suggestive but not conclusive. They are a reason to investigate, not to act. What follows is a research agenda organized around three sequential priorities, each feasible with existing infrastructure and each building on the previous one.

First priority: Adding deviation rationale to existing audit platforms

The most immediately practicable proposal is also the most straightforward: To embed a field for recording the reason for each protocol deviation within the ERAS Interactive Audit System or equivalent national platforms. The technical requirement is modest: A single categorical variable added to the data-entry screen for each protocol element, with four options—clinically contraindicated, patient refusal or intolerance, system constraint, and no reason documented.

The classification we propose is deliberately simple and should be understood as a starting point rather than a definitive solution. The four categories are not mutually exclusive in practice. A patient who refuses early mobilization may do so because of uncontrolled pain (a clinical factor) or because of anxiety and inadequate explanation (a communication factor). A system constraint, such as a shortage of nursing staff, may make early mobilization infeasible, but the decision to code it as system constraint rather than no reason documented depends on the clinician’s willingness to record it as such. These overlaps and ambiguities are unavoidable in the early stages of developing a new audit dimension. We propose the four-category framework as a testable prototype, not as a final taxonomy. Refinement through Delphi consensus, pilot testing, and prospective validation will be essential before any such system can be recommended for routine use.

The classification itself would require standardization. A Delphi consensus process involving high-volume ERAS centers could develop operational definitions for each category, specifying, for instance, what constitutes a clinical contraindication to early mobilization vs what should be recorded as patient intolerance. Delphi methodology has been widely used in surgical quality improvement to generate consensus where direct evidence is absent, and its application to deviation classification would follow established protocols[83,84].

The feasibility of such a system would then require testing in a multicenter pilot study. Three to five centers could implement the stratified audit tool over 6 months, generating data on three outcomes: The additional time required for data entry per patient, the inter-rater reliability of deviation classification between centers, and the distribution of deviation types across patient risk strata. Inter-rater reliability would be assessed using kappa statistics, with a target of at least 0.70 based on precedent from surgical deviation classification studies[85]. A previous single-center audit demonstrated that retrospective classification of ERAS deviations is feasible and that the majority of departures from protocol were clinically indicated[24]; what is now required is prospective, multicenter validation.

The analytical value of such a system would be tested by comparing outcomes between patients whose deviations were recorded as clinically indicated and those for whom no reason was documented. If the former group fares no worse than the latter, in terms of complications, length of stay, or readmission, the implication is that clinical teams are already making appropriate judgements about which patients can safely depart from which elements, and that current compliance scores, by treating all deviations as equivalent, are misclassifying good care as a defect. If, conversely, the no reason documented group has systematically worse outcomes, a quality gap is identified that targeted improvement efforts might close. As an illustrative estimate of the sample size that would be required for such a comparison: Assuming a deviation rate of 15%-20%, consistent with published data[24], a multicenter cohort of approximately 1500 patients would provide 80% power to detect a 10% absolute difference in complication rates between deviation categories at α = 0.05, accounting for center-level clustering with an estimated intraclass correlation coefficient of 0.02[86]. These figures are intended to demonstrate feasibility, not to constitute a formal power calculation for a specific study; any future trial would require its own sample-size estimation based on site-specific parameters.

The implementation costs of modifying existing audit platforms should also be considered; based on published estimates of ERAS program costs, these costs are likely to be modest relative to the potential gains from more targeted quality improvement[87].

A secondary effect of the intervention—the possibility that the audit tool itself changing clinical behavior—also merits attention. The act of documenting a reason for each omission may make clinicians more reflective about departures from protocol. Qualitative evidence suggests that junior surgeons and nurses already feel pressure to follow protocol elements even when they have reservations[43], and providing a structured field for recording clinical contraindications may legitimize appropriate deviation, counteracting any perceived pressure towards rigid compliance. We could test this during the pilot. If the proportion of deviations coded as clinically justified rises, without any worsening of outcomes, that would be a sign that the audit tool is supporting, not constraining, clinical judgement.

Second priority: Understanding the reasons behind low compliance in high-risk groups

Stratified audit data, once available, would permit hypotheses to be advanced about why particular patient groups complete fewer elements of the protocol. Testing those hypotheses requires methods that go beyond what a registry can provide. A sequential explanatory mixed-methods design, in which quantitative analysis of deviation patterns is followed by qualitative exploration of the mechanisms underlying those patterns, offers a rigorous and established approach[88].

The quantitative component would use the stratified audit dataset to identify which patient phenotypes and protocol elements are most closely linked to particular categories of deviation. If, for example, frail patients are disproportionately represented in the clinical contraindication category for early mobilization, while socioeconomically disadvantaged patients are over-represented in the system constraint or patient inability categories for carbohydrate loading and preoperative education, the data themselves point towards different interventions, protocol modification in the first case, service reconfiguration in the second. Multilevel logistic regression models, with patients nested within centers, would account for the hierarchical structure of the data and permit estimation of center-level variability in deviation patterns[89].

The qualitative component would explore what the audit fields cannot capture. Semi structured interviews with purposively sampled patients from high-risk subgroups—including those who are frail or severely obese, have limited health literacy, or come from socioeconomically disadvantaged backgrounds—would probe the lived experience of attempting to complete ERAS elements.

Interviews with clinicians, surgeons, anesthetists, ward nurses, ERAS coordinators, would examine the decision-making behind protocol departures. A patient who declines early mobilization may do so not because of laziness or non-compliance, as a bare audit entry might imply, but because of uncontrolled pain, unrecognized anxiety, or a culturally mediated belief that rest is essential to recovery. A surgeon who withholds goal-directed fluid therapy from a patient with borderline renal function may have made a sound clinical decision but may not have documented it in terms that an audit field can easily capture. Qualitative research in perioperative settings has demonstrated feasibility: Studies of barriers and facilitators to ERAS adherence have used focus groups and interviews to uncover reasons for non-completion that are invisible in registry data[43,90]. Thematic analysis, following an established framework such as Braun and Clarke’s[91] reflexive approach, would generate a taxonomy of deviation reasons grounded in both patient experience and clinical reasoning.

This qualitative work is unlikely to yield a single definitive answer, and that is not its goal. But it would give us a more nuanced sense of why adherence scores are lower in some groups, and whether those lower scores reflect appropriate adaptation, remediable barriers, or some mix of the two. That insight could then inform more precise audit classifications, sharper quality improvement strategies, and perhaps even tailored modifications to specific protocol elements for specific patient phenotypes.

Third priority: Testing whether audit redesign changes clinical behavior

The first two priorities establish feasibility and generate hypotheses. The third tests whether acting on those hypotheses improves care. A stepped-wedge cluster randomized trial would evaluate the introduction of stratified audit with feedback, a system in which centers receive regular reports not only of their compliance rates but of the distribution of deviation reasons across patient subgroups, against standard compliance-only audit.

The stepped-wedge design, in which all centers begin in the control condition and transition to the intervention at randomly assigned time points, is well suited to health service interventions of this kind[92]. Each center acts as its own control, reducing confounding by institutional culture and case-mix, and the sequential roll-out eliminates the ethical concern of withholding a potentially beneficial audit tool from any center permanently. Randomization would be stratified by center volume and baseline compliance, given the known association between institutional experience and ERAS outcomes[8]. The primary outcome would be the proportion of deviations for which a clinical reason was documented, serving as a direct process measure of the intervention. Secondary outcomes would comprise risk-adjusted complication rates, the proportion of deviations coded as having no documented reason, expected to fall should the intervention encourage more reflective practice, and, crucially, whether any rise in documented clinical contraindications is accompanied by a shift in outcomes for high-risk subgroups.

The sample size required depends on the expected effect size, which is unknown. A trial of sixteen centers, each contributing 150 patients before and after the transition, would provide 80% power to detect a 5% absolute reduction in the no reason documented deviation rate, from a baseline of approximately 30% to 25%, at α = 0.05, assuming an intracluster correlation coefficient of 0.02[86]. This is a realistic target for a multicenter ERAS collaborative. The trial would not, by itself, answer whether stratified audit improves clinical outcomes; that would require a larger and longer study. It would, however, establish whether audit redesign changes what is recorded, a necessary precursor to any claim about whether it changes what is done. A process evaluation embedded within the trial, using qualitative methods to explore how centers use the stratified audit data, would provide essential context for interpreting the quantitative findings, following Medical Research Council guidance on process evaluation of complex interventions[93].

A critical caveat to this proposed trial, and to the broader research agenda, is that improved documentation is not synonymous with improved care. A reduction in the proportion of deviations recorded as “no reason documented” may reflect more complete recording rather than better clinical reasoning. Conversely, clinicians who become more aware of the audit system may become more likely to document a clinical rationale for deviations that previously went unrecorded, without any change in their actual decision-making. This “recording effect” is a real risk in any audit intervention, and it can only be addressed through validation against independent clinical adjudication. Future studies should include a process evaluation in which recorded reasons are compared against case-note review by an external panel, blind to the audit data, to assess the accuracy of the recorded classifications. Until such validation is performed, improvements in documentation should be interpreted with caution.

Supplementary consideration

The proposals set out above are concerned with the measurement framework through which ERAS is delivered. They do not address the content of the protocols themselves. A parallel research effort, not detailed here, concerns the development of evidence-based modifications to individual ERAS elements for specific high-risk phenotypes: What fluid strategy for the patient with class III obesity, what analgesic regimen for the patient with chronic kidney disease, what mobilization target for the patient with severe frailty. The 2025 ERAS Society guidelines have begun this work[20], and computational phenotyping methods may accelerate it[13,14]. The argument of the present review is that such efforts will remain incomplete without a measurement system capable of distinguishing, at scale, whether a given modification was deliberate or accidental. The research agenda proposed here is not an alternative to stratified protocol design; rather, it constitutes the measurement infrastructure on which such design depends.

The measurement challenges we have described are likely to be amplified in emergency colorectal surgery, where patient physiology is more variable, preoperative optimization is limited or absent, and protocol deviations are more frequent. The 2025 ERAS Society guidelines explicitly note that evidence for emergency surgery is less robust than for elective procedures, and the boundary between “elective” and “emergency” is itself a spectrum rather than a binary distinction. We have chosen to focus this review on elective colorectal ERAS because the evidence base is larger and the measurement framework better established; however, the core problem, the conflation of adaptation and omission, is no less relevant, and perhaps more so, in the emergency setting. Whether the audit redesign we propose would be feasible or effective in emergency contexts remains an open question that future research should address.

LIMITATIONS

This review carries several constraints that should be stated plainly before its conclusions are weighed. First, it is a narrative review, not a systematic one. Although we conducted the literature search across multiple databases using predefined terms and supplemented by hand-searching of reference lists, the synthesis was qualitative and the selection of studies for detailed discussion involved author judgement. Selective reporting cannot be excluded. The choice of a narrative rather than a systematic approach was a pragmatic one: The evidence base cuts across clinical epidemiology, cognitive psychology, organizational safety, and health equity research, each with its own methodological traditions and standards of quality, and such breadth could not readily be accommodated by any single quantitative synthesis. The cost of this breadth is that different readers, applying different selection criteria, might arrive at different emphases[25]. Narrative reviews have been defended as uniquely suited to topics that cross disciplinary boundaries and require interpretive synthesis rather than effect-size aggregation, but their vulnerability to author bias is irreducible and should be acknowledged rather than minimized[94].

More fundamentally, the central thesis of this review rests on indirect evidence: Namely, that current ERAS audit systems, by conflating clinically justified adaptation with inadvertent non-adherence, may be misdirecting quality improvement. No prospective study has tested whether distinguishing deviation types improves outcomes. The observations assembled here, from registry analyses to single-center audits to qualitative interviews, are consistent with the thesis but do not confirm it. The distinction between hypothesis generation and hypothesis confirmation is particularly important in this context, because the measurement infrastructure required to test the central claim does not yet exist in routine ERAS audit. A review that identifies a measurement gap cannot, by definition, fill it with data that the measurement system was not designed to collect. Subgroup analyses, even when statistically significant, do not constitute proof of effect modification unless they are pre-specified and adequately powered; the signals reviewed here are almost entirely post hoc and should be interpreted accordingly[95].

The theoretical frameworks discussed, Goodhart’s Law and automation bias, originate in disciplines beyond surgery, and their applicability to colorectal ERAS settings remains entirely unexplored. The evidence for metric distortion in healthcare derives largely from emergency department targets and sepsis bundle reporting, settings that differ from elective surgical pathways in their incentive structures, clinical workflows, and patient populations. Evidence for automation bias, by contrast, comes chiefly from simulation studies with medical students and junior doctors, rather than from experienced surgical teams who have managed complex patients over years of protocolized practice. These frameworks are offered as heuristic tools for generating testable hypotheses, not as descriptions of documented phenomena in perioperative care. Their inclusion in this review reflects a deliberate choice to look beyond surgery for conceptual resources, but the gap between theory and surgical evidence is wide and should not be minimized. The history of cross-domain translation from aviation to medicine, while productive, also demonstrates that direct extrapolation without domain-specific validation can mislead[96]. The cognitive psychology literature on diagnostic reasoning has consistently shown that findings from controlled experiments do not translate straightforwardly to complex clinical environments[80].

Fourth, the health equity evidence reviewed here is drawn almost entirely from the United States. The association between socioeconomic disadvantage and worse ERAS outcomes has been replicated in an English cohort, but the mechanisms linking race, ethnicity, and protocol adherence have been studied predominantly in a healthcare system with specific features, private insurance, residential segregation, a particular history of medical discrimination, that limit generalizability. Whether similar patterns exist in the universal coverage systems of Europe, the public hospital networks of East Asia, or the resource-constrained settings of low- and middle-income countries is largely unknown. The global ERAS literature documents implementation gaps between high- and low-resource settings, but the intersection of socioeconomic disadvantage and protocol adherence within those settings has received far less attention.

Fifth, the literature search was restricted to English-language publications indexed in MEDLINE, EMBASE, and PsycINFO. Grey literature, conference abstracts, and non-English publications were excluded. The exclusion of non-English sources is a particular concern for the equity dimension of this review: Colorectal surgical research from Asia, Latin America, and the Middle East, where socioeconomic and demographic patterns differ from those of North America and Western Europe, may have been missed. Language bias in medical research is well documented: Trials with statistically significant results are more likely to be published in English-language journals, and exclusion of non-English studies can affect the direction and precision of pooled estimates[97]. The exclusion of grey literature introduces a further source of potential bias, as studies with null or negative findings are less likely to appear in peer-reviewed journals and more likely to remain in conference proceedings or institutional reports[98]. This review cannot quantify the magnitude of these biases, but it is important to acknowledge that their direction would be towards an overestimation of the strength of the signals described.

Sixth, the potential for publication bias extends beyond language and grey literature. The evidence base synthesized here consists of published, peer-reviewed studies. Negative or null findings on heterogeneity of ERAS benefit, unintended consequences of standardization, or disparities in protocol adherence are less likely to be submitted or accepted for publication[99]. The phenomenon of non-publication of studies with negative results has been documented across multiple fields of medicine, and the resulting distortion of the published evidence base is increasingly recognized as a threat to the validity of literature syntheses[100]. Absence of evidence for particular patterns should not be mistaken for evidence of their absence. Where gaps in the literature are identified in this review, they may signal either a genuine absence of the phenomenon in question or a failure of the publication record to capture it. The distinction is unknowable from published data alone.

Seventh, the field is moving rapidly. The 2025 ERAS Society guideline update, published after the initial drafting of this review, was incorporated into the analysis, but further evidence may have appeared between the search date and publication. Systematic reviews have been shown to lose currency within two to five years of publication, depending on the pace of research in the field[101]. The review should be read as a snapshot of the evidence available at the time of writing, not as a final adjudication.

Eighth, the authors are clinicians by training, with expertise spanning general surgery and ultrasound medicine. Two authors (Yu H and Zhan XL) are colorectal surgeons with direct clinical experience in perioperative care; the remaining author (Ding YB) is a diagnostic specialist whose ultrasound practice is relevant to colorectal surgical assessment. The interpretation of evidence from cognitive psychology, organizational behavior, and health equity research reflects an effort to engage with literatures outside the authors’ primary discipline, but the risk of misinterpretation or oversimplification cannot be eliminated. We have taken care to frame cross-disciplinary concepts as heuristic tools rather than as settled evidence, and we emphasize that our use of terms such as “automation bias” and “Goodhart’s Law” is intended to generate hypotheses, not to assert that these mechanisms have been demonstrated in perioperative care. The review has not been independently reviewed by a cognitive psychologist or a health equity researcher, and the absence of such interdisciplinary scrutiny is a genuine limitation. Researcher positionality, the influence of disciplinary training and clinical experience on the questions asked and the interpretations offered, is increasingly recognized as a source of bias in qualitative and mixed-methods research, and its relevance to narrative reviews that cross disciplinary boundaries is no less pertinent[102].

Ninth, the scope of this review is deliberately narrow. It addresses ERAS in elective colorectal surgery. Findings may not generalize to emergency colorectal procedures, to other surgical specialties in which ERAS protocols have been adopted, or to enhanced recovery pathways in non-surgical contexts. Within colorectal surgery, the review does not address the cost-effectiveness of ERAS, the comparative effectiveness of individual protocol elements, or the organizational resources required for implementation. The boundary between elective and emergency colorectal ERAS is itself an area of active investigation, and the measurement concerns raised here may apply with equal or greater force in the emergency setting, where patient physiology is more variable and protocol deviation more frequent[103].

Finally, the global dimension of the measurement problem discussed here has been addressed only in outline. ERAS implementation in low- and middle-income countries faces structural constraints, inadequate funding, insufficient personnel, supply chain interruptions, that make protocol completion a fundamentally different challenge than in well-resourced settings[66,67]. Whether stratified audit, which presupposes a baseline level of data infrastructure and personnel, is feasible or prioritized in such contexts is uncertain. The recommendations advanced in this review are intended primarily for health systems with established ERAS programs and the capacity for electronic audit; their relevance to other settings would require separate evaluation. A summary of the key limitations and their implications for the strength of the conclusions is provided in Table 2.

Table 2 Key limitations and their implications.

Limitation	Nature of the constraint	Implication for conclusions
Narrative review design	Non-systematic study selection; author judgement in synthesis	Different reviewers might emphasis different evidence
Indirect evidence for core thesis	No prospective study has tested whether distinguishing deviation types improves outcomes	Central claim is hypothesis-generating, not confirmatory
Post hoc nature of subgroup signals	Signals of heterogeneity derive from unplanned analyses	Cannot distinguish effect modification from confounding
Cross-domain theoretical frameworks	Goodhart’s Law and automation bias untested in ERAS settings	Theoretical discussion is heuristic, not evidentiary
Geographic concentration of equity evidence	Racial and socioeconomic disparities data predominantly from United States	Findings may not generalize to universal coverage health systems
English-language restriction	Grey literature and non-English publications excluded	Language bias and publication bias cannot be quantified
Single-discipline authorship	Authors are clinicians with expertise spanning surgery and ultrasound medicine	Risk of oversimplification in psychology and equity literatures
Temporal scope	Search conducted to July 2026	New evidence may have since emerged
Scope limited to elective colorectal surgery	Findings may not extend to emergency surgery or other specialties	Boundary of applicability is defined but untested
High-income country orientation of recommendations	Stratified audit presupposes electronic audit infrastructure	Applicability to low-resource settings is uncertain

ERAS: Enhanced recovery after surgery.

CONCLUSION

Enhanced recovery protocols have been shown to reduce length of stay and complication rates after elective colorectal surgery. That is a genuine and substantial achievement, backed by a level of evidence that few other perioperative interventions can rival. Nothing in the present review calls it into question. Our argument is narrower. The patients on everyday surgical lists are not the same as those who populated the ERAS trials. When these patients complete fewer protocol elements, as they consistently do, current audit systems cannot tell us why. Whether a patient has fluids restricted because of renal impairment, cannot mobilize because of severe frailty, or simply has carbohydrate loading forgotten, the record is the same: Non-adherent. As long as audit frameworks treat every departure as equivalent, the push for higher compliance scores may inadvertently divert improvement efforts away from the very patients who stand to gain least from the protocol.

Closing this gap does not mean abandoning the compliance paradigm; it means adding one more dimension to it. A deviation rationale field, clinical contraindication, patient intolerance, system constraint, or no reason documented, would, if integrated into existing digital audit platforms, reveal what is currently hidden: The clinical reasoning behind protocol departures. The infrastructure is already there. The ERAS Interactive Audit System and equivalent national platforms already capture compliance data in real time; adding a categorical field for each protocol element would be a technically modest change. Its real value lies in what it would enable: An audit process that distinguishes justified adaptation from omission, research that tests whether that distinction predicts outcomes, and quality improvement that actually targets the real issue.

We are not arguing for less measurement, we are arguing for better measurement. A metric that only asks was it done? Cannot tell us should it have been done? Patients on the margins of the evidence are not problems to be designed around. They are exactly why the protocol needs to evolve, and why the tools we use to track it must evolve with it. Let us be clear about what we are and are not proposing. The deviation rationale field, the stratified audit framework, and the stepped-wedge trial design are intended to generate hypotheses, not to guide practice. The measurement gap is real, and the proposed approaches are feasible, but whether addressing the gap improves outcomes remains to be tested. The real value of this review, as we see it, is not in supplying answers, but in sharpening the question: When a patient at the margins of trial evidence departs from the ERAS protocol, was that departure an act of clinical judgement or a missed opportunity? Until audit systems can tell us, we cannot know.

ACKNOWLEDGEMENTS

The authors thank the investigators of the international enhanced recovery after surgery registries and the clinical research groups whose published work provided the empirical foundation for this review.