Introduction

Enhanced recovery after surgery (ERAS) protocols are multimodal perioperative care pathways developed for several surgical procedures to achieve early recovery after surgery by preserving preoperative organ function and reducing physical stress responses. The key elements of the ERAS protocols include preoperative counselling and nutritional assessment, avoidance of preoperative fasting and bowel preparation, standardised opioid-sparing analgesic and anaesthetic regimens, and early refeeding and mobilisation. Such multimodal stress-minimising approach has been shown to reduce the rates of morbidity, improve recovery and shorten length of stay (LOS) after major colorectal surgery.1 Despite the strong background theory and the available evidence supporting the potential improvements in colorectal cancer surgery,1–4 the ERAS protocols still pose a challenge to traditional surgical doctrine.

After the fourth updated version of the ERAS protocol in colorectal cancer,1 only three selected hospitals, particularly open to change, have adopted this approach in routine care in Piemonte, an Italian region of around 4.2 million inhabitants.5 The Piemonte region has a large network of medium-small public-funded hospitals treating colorectal cancer, referring to the Piemonte e Valle d’Aosta Cancer Network (http://www.reteoncologica.it/). Initiatives to improve the quality of cancer care within the network were mainly based on the adaptation and implementation of regional clinical guidelines (https://next.cpo.it/it/publications) and their monitoring with administrative data.6 The slow, spontaneous diffusion of the ERAS protocol within the hospital network has had a limited impact on the overall quality of perioperative care at the regional level and may have increased heterogeneity between centres and inequalities between patients. In addition to the usual barriers to implementing new organisational models, the limited evidence from properly designed randomised trials2 7 8 and the lack of structured local audit and feedback (A&F) strategies may have limited the dissemination of the ERAS protocol.9

The ERAS Protocol Implementation in Piemonte Region for Colorectal Cancer Surgery (ERAS Colon-Rectum Piemonte study)10 was conducted to promote a systematic adoption of the protocol throughout the entire regional hospital network, with the goal of enhancing quality of care through an equitable and pragmatic approach. The study, registered with ClinicalTrials.gov (NCT04037787), is part of a larger project that evaluates the effectiveness of A&F interventions in different settings and health services (EASY-NET), a network project funded by the Italian Ministry of Health and the participating regions.

The aim of this study was to assess the impact of introducing an ERAS protocol, supported by an A&F intervention,11 on LOS and other clinical outcomes in a large population undergoing elective surgery for colorectal cancer, using a stepped wedge, cluster randomised trial (SW-RCT) design12 and involving the entire hospital regional network.

Methods

The article was written according to the Consolidated Standards of Reporting Trials (CONSORT) statement for SW-CRT13–15 and the Reporting on ERAS Compliance, Outcomes and Elements Research (RECOvER) checklist16 (online supplemental annex). The study protocol has been previously published.10

Design and participants

The ERAS Colon-Rectum Piemonte study is a pragmatic SW-RCT conducted among patients with colorectal cancer at 29 general surgery departments located in public hospitals in the Piemonte region.

Given all the available evidence on the ERAS protocol, with a favourable balance of benefits and risks (for both patients and staff), the SW-CRT was considered an appropriate design to allow sufficient time for training all the regional hospital teams, to send them periodic feedback while the study was ongoing and to achieve roll-out of ERAS across the entire hospital network at the end of the trial.12

All regional general surgery units that performed at least 30 elective surgical procedures for colorectal cancer in 2018 were invited to participate. The surgical units formed the study clusters, which sequentially adopted the ERAS protocol, by group of units.

All consecutive patients with colorectal cancer scheduled for elective surgery between 1 September 2019 and 31 May 2021 were included, with very few exclusions (eg, emergency admissions and patients with an American Society of Anesthesiologists (ASA) score of 5). All participants provided written informed consent and data were collected by the hospital staff on paper case report forms.

The recruitment period, originally set at 15 months, was later extended to 21 months to compensate for the negative impact of the COVID-19 pandemic on hospital activity. The third step was completed in August 2021 instead of May 2021 to allow for more time to organise the web-based training (previously scheduled in person) and to ask centres to start implementing ERAS during a remission of the COVID-19 pandemic in September 2021. The last step was extended due to concerns that the COVID-19 pandemic would be a barrier to the centres’ ability to change clinical practice and improve quality of care.

During the first 3 months of the study (baseline), standard treatment was continued in all groups. Thereafter, the four groups started adopting the ERAS protocol every 3 months, until all groups had a ‘control’ period of standard treatment and an ‘experimental’ period of application of the ERAS protocol, as described in online supplemental figure S1.

Randomisation

Surgical units were stratified by volume of colorectal procedures performed in 2018 and then randomly divided into four groups with a similar number of procedures. The randomisation procedure was carried out by the clinical epidemiology unit after the surgical units had been anonymised. The allocation was concealed to centres until 2 months before the start of the experimental period to allow sufficient lead time to train the local ERAS team and organise the trial activities.

Due to the nature of the intervention, it was not possible to blind participants, carers and researchers to group allocation.

Procedure

Each centre identified an ‘ERAS team’, including at least a surgeon, an anaesthetist, a nurse and a dietitian, to support local implementation of the protocol and to serve as a reference for data collection. These local teams received a 1-day interactive course on the principles of ERAS as well as organisational aspects and practical experience. All four editions of the training were delivered by a team of experts with consolidated experience in teaching and working with the ERAS protocol, with the last two editions delivered online due to COVID-19 restrictions.

The ERAS protocol to be implemented at the regional level was adapted from the ERAS Society guidelines for colorectal surgery.1 17 The ERAS items are described in online supplemental table S1, together with related indicators and discharge criteria.

A newsletter was sent to all local ERAS teams every 2 months to maintain commitment and motivation for the overall project and to share information on the progress of the study. To monitor the completeness of study enrolment at each centre, a graph was constantly updated on the study website comparing the number of patients actually enrolled with the expected number in the same calendar period of the previous year. According to the A&F intervention, once the experimental period had started, the hospital teams were given access to a feedback section of the study website to assess their progress in implementing the protocol, so that critical issues could be immediately identified and corrective actions addressed. The feedback indicators were also discussed with each group of centres in meetings, mostly online, a few months after the introduction of ERAS, together with the experts previously involved in the training and the study coordination/data management team.

Details on the intervention are reported in the Template for Intervention Description and Replication checklist18 in the online supplemental materials.

Outcomes

The primary endpoint was LOS, which was calculated after excluding outliers (LOS >94th percentile).

The secondary endpoints were the percentages of LOS outliers; of postoperative complications, defined according to the Clavien-Dindo classification19; and of admissions to the intensive care unit, transfusions and reinterventions during the postoperative hospital stay. Other clinical outcomes assessed within 30 days of discharge were any readmission to the emergency department (ED), readmission to hospital and reintervention.

Postoperative complications were analysed as the presence of at least one complication, total and major complications (Clavien-Dindo III–IV) or death. All the outcomes before discharge were collected with the case report form. Outcomes at 30 days after discharge were collected from regional administrative data.

The quality of postoperative recovery was measured with the validated Italian version of the Quality of Recovery-15 (QoR-15) questionnaire,20 21 filled in by patients approximately 48 hours after surgery. The QoR-15 is an instrument based on 15 items with a scale of 0–10 and a Visual Analogue Scale (VAS) for well-being, where 0 indicates the worst health status and 10 the best. Other secondary outcomes, including patients’ and professionals’ interviews, and an analysis of healthcare costs will be presented in another article.

Difference in compliance with ERAS items between the two study periods was measured overall, by phase of care and by single items.

Statistical analysis

Considering available literature and local data, the study was planned with a statistical power of 0.98 and with an alpha error of 0.05 (two-sided) to detect a reduction of 1 day of mean LOS without outliers (from 9.0 to 8.0, SD=3.7). Details on the sample size calculation were extensively reported in the study protocol publication10 and are summarised in the online supplemental materials.

Compliance with ERAS

Compliance with ERAS items during the two study periods was measured as the mean percentage of adherence with a list of indicators (online supplemental table S1), overall and for groups of items classified by phase of care (preoperative, intraoperative and postoperative). Multilevel linear models were used to estimate the difference in average compliance levels between the two study periods, overall and by phase of care, adjusting for patient characteristics and time period, and considering surgical units as random effects. Patient covariates included in the models were sex, age, Charlson Comorbidity Index (0 or 1), body mass index (BMI) class (<18.5, 18.5–24.9, 25–29.9, ≥30) and ASA score (1–2 or 3–4). The same set of covariates was used for adjusting Poisson models used to estimate the difference in compliance between the two study periods for each individual ERAS indicator (dichotomous).

Primary endpoint

LOS without outliers was described as median, mean and SD. The difference between the two study periods was estimated using a multilevel linear model adjusted for patient characteristics and time period and accounting for surgical units as random effects. In addition to the covariates included in the analysis of compliance, the cancer site (colon, rectum), the creation of a stoma and the type of surgical access, classified as open or minimally invasive (laparoscopic or robot-assisted), were also included.

Secondary endpoints

The percentages of outliers for LOS, postoperative complications (total, surgical and medical), incidence of transfusions, access to ICU after surgery, 30-day mortality after surgery, and ED admissions, readmissions and reinterventions were all analysed as dichotomous variables using random-effects logistic regression models, with the same set of covariates included in the model for the LOS analysis, except for BMI, cancer site and ASA score.

For the QoR assessment, only questionnaires with all 15 items completed within 1–4 days after surgery were included. The mean total scores and the mean scores for the physical and psychological subscales were reported. The effect of ERAS on the QoR scales and well-being VAS was estimated using multilevel linear models, with centres serving as random effects and adjusting for the same set of covariates used for the other secondary outcomes.

Subgroup and sensitivity analyses

For the primary endpoint, planned subgroup analyses were conducted by patient characteristics (sex, age, education, tumour location, Charlson Comorbidity Index, ASA score, surgical approach) and by centre characteristics (level of compliance with the ERAS protocol during the control period, increase in compliance with the ERAS protocol after its adoption, completeness of enrolment, volume of interventions). Enrolment completeness was assessed using the regional hospital discharge record database. To assess the achievement curve, we also analysed the change in LOS according to the time (in quarters) elapsed since the introduction of ERAS.

To account for the learning phase in each centre, the impact of the intervention on LOS was also analysed excluding the first month of each implementation period of the ERAS protocol.

Subgroup analyses were also conducted for compliance, using the same analytical approach as for LOS, to identify facilitators and barriers to the intervention. Finally, the association between the level of compliance with the ERAS protocol (10% increase) and the clinical outcomes was analysed, overall and by study period. To control for a possible reverse-causation effect between compliance with the protocol (especially for postoperative items) and patient-level outcomes, LOS was analysed with centre mean compliance with the protocol as a fixed effect.

Statistical analyses were performed using SAS V.9.4.

Results

Of the 36 public general surgery facilities treating patients with colorectal cancer in the Piemonte region in 2018, 3 were excluded because they had already implemented the ERAS protocol, 3 because they had a low case load and 1 declined to participate. Six centres had a case load slightly below 30 cases in 2018, but they asked to be included in light of an expected increase in activity during the study period. The final number of participating centres was therefore 29.

As described in the flow chart of the study (figure 1), 2626 patients were included (a more detailed flow chart can be found in online supplemental figure S2).

Figure 1

ERAS Colon-Rectum Piemonte study flow. ERAS, enhanced recovery after surgery.

The personal and clinical characteristics of the participants are shown in table 1. Overall, the mean age was 72 years (SD 10.9), 43% were women, about 70% of cases had colon cancer and 50% had a Charlson Comorbidity Index ≥1. No evident unbalances were observed between the two study periods.

Table 1

Personal and clinical characteristics of the participants

Table 2 shows the impact of the implementation of ERAS on the outcomes of the study.

Table 2

Effect of ERAS implementation on study outcomes

Primary endpoint

The mean LOS without outliers (defined as LOS >20 days, corresponding to the 94th percentile of the LOS distribution) was 8.5 days during the control period (SD 3.9) and 7.5 (3.5) during the experimental period, with a raw reduction of 1 day (table 2). After excluding six patients with missing data on the covariates, the estimated adjusted difference between the two periods was a reduction of −0.58 days (95% CI −1.07, −0.09; p=0.021).

The planned subgroup analyses, depicted in figure 2, revealed only moderate differences in LOS reduction by patients’ and centres’ characteristics. A tendency towards larger effects on LOS reduction was observed in centres that had a lower compliance with ERAS at baseline (−0.78 days; 95% CI −1.31, −0.26; p value for interaction=0.056) and in centres with ≥80% completeness of enrolment (−0.68 days; 95% CI −1.18, −0.18; p value for interaction=0.058). For minimally invasive surgery, where the initial LOS was already at the target level of 8 days, the adoption of ERAS had a smaller impact on LOS (−0.49 days; 95% CI −1.00, 0.02), while for open surgery, with an initial LOS of 10 days, the reduction was more pronounced (−1.03 days; 95% CI −1.83, −0.22; p value for interaction=0.162).

Figure 2

Estimated difference in LOS (primary outcome) between the two study periods and related subgroups analyses: patients’ and structure characteristics. ASA, American Society of Anesthesiologists; ERAS, enhanced recovery after surgery; LOS, length of stay.* Interaction p value.

Online supplemental figure S3 describes the change in LOS according to time elapsed since the introduction of ERAS (in quarters), showing a stable effect over time.

After excluding data collected in the first month of the roll-out period of the ERAS protocol, the results did not change (LOS reduction −0.61; 95% CI −1.15, −0.07).

Secondary endpoints

No differences were observed in the frequency of outliers in LOS, complications, need for transfusion, access to intensive care in the postoperative period and in-hospital mortality (table 2).

The incidence of postoperative complications is described in online supplemental table S2 by study period. The occurrence of complications did not differ either overall (OR 1.22; 95% CI 0.89, 1.68; p=0.211) or by type (surgical, medical) or severity (Clavien-Dindo I–II, III–IV).

Outcomes at 30 days after discharge did not differ between the two study periods.

Patients included in the QoR (1762, 73.5%) and VAS (1850, 77.2%) analyses are described in online supplemental figure S4. The mean QoR score was 7.12 and 7.46 in the control and ERAS periods, with a small improvement of 0.24 points (95% CI 0.01, 0.47; p=0.039). This improvement was mainly due to a difference in the physical scale. There was also a small improvement between the two study periods in the VAS score for general well-being (0.31; 95% CI 0.05, 0.57; p=0.021).

Compliance with ERAS protocol

Overall, the level of compliance with the ERAS protocol changed from 52.4% during the control period to 67.3% during the experimental period, with an adjusted absolute difference of +13% (95% CI 11.4%, 14.7%; p=0.0001) (table 3A). Compliance was lowest for postoperative items (35.4% and 52.5% in the control and experimental periods). The largest absolute difference between the two periods was recorded for preoperative items (+18.2%; 95% CI 16.5%, 20.0%; p=0.0001). Intraoperative items also appeared to be frequently used during the control period (67.8%) and showed a small increase in compliance (+5.4%; 95% CI 2.4%, 8.3%; p<0.001). Table 3B shows the adjusted implementation effect (risk ratio) on compliance for each individual ERAS indicator, grouped by phase of care. Online supplemental figure S5 describes the change in compliance according to the time elapsed since the introduction of ERAS (in quarters). Compared with the baseline period, the increase in compliance was larger in the first two quarters after starting the experimental period (around 11%) but remained stable afterwards (around 8.5% in all the following quarters). The results of the subgroup analyses are shown in online supplemental figure S6. Patients’ characteristics did not impact the change in ERAS compliance, but low compliance with the ERAS items prior to initiation of the study, high completeness of patient inclusion in the study and high volume of surgical activity were relevant facilitating factors.

Table 3

Compliance with ERAS items in the two study periods and the adjusted effect: (A) difference in compliance (%) overall and by phase of care and (B) risk ratio on compliance for each single ERAS indicator

Online supplemental table S3 shows the impact of compliance with ERAS items overall and stratified by study period on study outcomes. A 10% increase in overall compliance was associated with a reduction in LOS (−0.65 days; 95% CI −0.76, −0.54) and in most clinical outcomes, with a stronger effect during the ERAS period, when an integrated implementation of all items was supported by the A&F approach. After adjustment for mean centre-level compliance, the estimated LOS reduction was −0.30 days (95% CI −0.50, −0.10) (data not shown).

DiscussionKey findings

In this large, pragmatic SW-RCT of patients surgically treated for colorectal cancer, implementation of the ERAS protocol supported by an A&F intervention across the network of regional hospitals in Piemonte reduced the mean LOS by 0.6 days during the experimental period compared with the control period. The subgroup analyses showed greater improvements in hospitals where the opportunity to improve compliance with the ERAS protocol was greater. The application of the ERAS programme did not lead to any meaningful impact on postoperative complications, either during hospitalisation or 30 days after discharge.

The A&F initiative supporting ERAS implementation determined a relevant change in clinical practice, with an absolute increase in compliance with the ERAS protocol of around 13%, an impact much larger than the average effect (median 4.3% improvement) estimated by a previous systematic review.22

Comparison with existing literature

The decrease in LOS observed in our study is consistent with, but smaller than, that estimated by a meta-analysis of randomised controlled trials for elective colorectal surgery, reporting a mean LOS of 5.8 and 8 days in the ERAS and control groups, respectively.2 Similar effects on LOS were reported by meta-analyses of randomised trials including colorectal surgery and other types of interventions.3 4 A reduction in median LOS associated with the adoption of the ERAS programme in colorectal cancer surgery was also reported by cohort studies conducted within hospital networks,23 24 despite a reduced impact when the results were adjusted for the background time trend reduction of LOS.25 Analyses of data from the international multicentre ERAS registry for elective colorectal cancer resection reported a median overall LOS of 6 days (IQR 4–8).26

The absence of a meaningful impact of our study on postoperative complications is in contrast to previous experiences,2–4 but in line with some recent systematic reviews of randomised controlled trials assessing specific ERAS items, as prehabilitation27 and early mobilisation interventions.28

According to a Cochrane systematic review of the efficacy of A&F22 in modifying healthcare behaviours, the global change in compliance observed in the present study can be considered a relevant result. Specific comparisons with previous experiences of ERAS implementation are not easy because it is not clear whether and how explicit A&F interventions were adopted.9 Consistent with the findings of Nelson et al,29 the greatest increase in compliance was for preoperative items (+18.2%), which included practices not commonly found outside ERAS, such as counselling, nutritional risk assessment and introduction of a carbohydrate load.

Interpretation

Although the adjusted effect on LOS in our study (−0.6 days) was smaller than planned in the study protocol (−1 day, from 9 to 8), it should be noted that the mean LOS in the baseline period of the study was already reduced to 8.5 days, which left little room for further improvement. The reduction in LOS from 2018 to the end of 2019 may be partly the result of a long-term trend and partly an indirect effect of the centres’ involvement in writing the study protocol.

The decrease in LOS was achieved in the first period of implementation of the ERAS protocol, that is, up to 6 months, and thereafter LOS remained stable. This result seems to contradict the notion of a learning curve.30 31 However, a major influence on this result is probably the COVID-19 pandemic, which affected the organisation of the hospital and the implementation of the study. In the fourth quarter of the implementation of ERAS, corresponding to the peak of the pandemic in the Piemonte region (March–May and October–December 2020), no decrease in LOS was observed compared with baseline, likely due to organisational stress for COVID-19 management.

In general, the COVID-19 pandemic had a strong impact on the study’s ability to change clinical practice and improve the quality of care. As reported by the healthcare staff in the feedback meetings, it affected the full implementation of the ERAS protocol and weakened the adoption of the A&F approach in several ways. ERAS implementation requires close collaboration between different health professionals,1 with nurses and anaesthetists playing a key role, both of whom were heavily involved in pandemic management. In addition, ERAS benefits from strong patient32 33 and caregiver34 involvement, which was drastically reduced by social distancing measures. Similarly, A&F requires discussion of results in meetings with all stakeholders, but due to the pandemic, only one meeting was held per group of centres, and for two groups only via web conference. To limit the negative impact of COVID-19, the study duration was extended by three additional months in both the third and fifth steps. As this extension affected the middle and last period of the study, the variation in the final study sample size and the imbalance between the two groups were negligible.

Strengths and weaknesses

The most original features of our study, which to our knowledge is the largest ERAS randomised trial to date, are the cluster randomisation design and its pragmatic approach, the implementation of the programme within the entire regional hospital network, and the high level of engagement and involvement of most eligible patients. Participating centres also included those usually excluded from research projects and unlikely to adopt the ERAS and A&F approach on their own. The research framework allowed a full adoption of ERAS across the entire regional network, reducing heterogeneity in patient care and consequently inequalities.

The subgroup analyses on compliance with ERAS items found that some centre characteristics resulted as facilitating factors. Other factors perceived as facilitators of such positive results were the substantial methodological and organisational support, the leading role of referral centres experienced in delivering ERAS and the presence of the PeriOperative Italian Society offering specific expertise. In addition, the intervention has benefited from the presence of a regional oncology network and the strong commitment of the regional healthcare authority.

Our study differs from most previous studies, mainly monocentric and with small sample sizes, in which individual patients were randomised within the same ward, with a high risk of bias. This is because implementing the ERAS protocol requires cultural and organisational changes that cannot be achieved with an ‘on/off’ intervention at the patient level. In addition, the stepped wedge design allowed us to account for time trend effects, a bias that typically occurs in studies comparing outcomes between pre-implementation and postimplementation periods.25 29

As the ERAS protocol circulated within the regional hospital network as part of the ERAS Colon-Rectum Piemonte study protocol, group contamination cannot be excluded. The groups waiting to implement ERAS may have anticipated some changes during the standard period, reducing the potential difference between the two periods in terms of adherence to ERAS items and impact on clinical outcomes. The suboptimal level of compliance achieved during the ERAS period (67%), at least in part attributable to the COVID-19 pandemic, may have compromised the ability to achieve relevant effects on secondary outcomes.

In our opinion, the A&F approach was useful in engaging clinicians and centres and in overcoming resistance to the cultural and organisational changes required by the ERAS implementation, but a formal analysis of the data collected by surveying local ERAS teams was not included in the present paper.

Because the study was conducted in a region with a public health system, the results may have limited generalisability to other countries with different health systems. A final issue is the recognition that it is difficult to monitor detailed quality of care measures after the study is completed using only currently available data.