Introduction

Pancreaticoduodenectomy for pancreas tumours and periampullary tumours is considered high-risk surgery and is associated with high morbidity (30–70%) and a mortality of 1–5% in specialized centres.1, 2 Centralization of pancreas surgery and advances in surgical techniques resulted in more patients being operated for advanced-staged tumours.3, 4 Patients with more comorbidity receiving pre-operative chemotherapy and/or vascular reconstructions in advanced disease, need more complex perioperative care. Currently this is facilitated by multiple guidelines and medical and nursing protocols. This complexity demands an overall multidisciplinary approach and clear communication.

Different departments are involved in the treatment during the patients’ journey through the surgical ward, operation theatre, post-anaesthesia care unit (PACU) and intensive care unit (ICU). However, large differences in the actual use of these protocols are present between the different units and medical and nursing staff members.5, 6 Moreover, while multidisciplinary teamwork for these patients is essential, the development and implementation of a clinical pathway (CP) involve many aspects of the total patient care and should therefore be multidisciplinary by doctors and nurses as well.

A CP may facilitate the care for this group of high-risk surgery patients by unifying different protocols into one multidisciplinary protocol for all units during the hospital stay of the patients. This may result in an increased protocol adherence, less morbidity and improved outcome. Key elements of a CP are guidelines, evidence-based clinical protocols and best practice rules, together with a coordinated sequence of activities of the multidisciplinary team.7 Registration, monitoring and evaluation of adherences, variances and outcomes are part of a CP and can be part of a process-driven pathway.8 A multidisciplinary CP has therefore many evaluation moments and scheduled actions. To keep the patient on the ‘pathway’, the CP mandates a registered response of the nurse or doctor if results are outside the range of the prescribed boundaries.

Many CPs have been developed for high volume with low-risk and with average-risk health care procedures in order to reduce complications.912 The post-operative phase of the patient spent in the ICU or PACU, however, is a seldom part of a CP.13 A CP including the PACU/ICU stay mandates an hour-to-hour care plan during the post-operative stay in the ICU/PACU.14 Many standardized care plans related to a pancreaticoduodenectomy have been published, focussing on the use of an enhanced recovery program after surgery (ERAS) with elements like early mobilization, early enteral feeding, pain treatment and reduction of iv fluid administrations to shorten the length of hospital stay.1519 In these care plans, a reduction of hospital length of stay (LOS), morbidity or mortality was not always observed. Crucially, the ICU period of these patients was not integrated in these protocols.

The aim of this study was first to determine the feasibility to develop and implement a multidisciplinary CP including a variance report for all pancreaticoduodenectomy patients during their entire hospital stay and second to determine if the use of this CP is associated with an improvement of patient’s morbidity and outcome.

Methods

Setting and Patients

The Radboud University Medical Center in Nijmegen is a 1000-bed university hospital, including a 32-bed closed-format ICU, a 5-bed PACU and a 30-bed gastrointestinal (GI) oncology surgical ward. An anaesthesiologist with a resident are supervising the PACU. The ICU is supervised by the intensivists, with intensivists-in-training and residents. They all work in close relation with the surgical team. On the surgical ward, nurses, physician assistants and young residents are caring for patients undergoing a pancreaticoduodenectomy, under daily supervision of the senior GI-oncology medical staff. Since the centralization in 2012 of pancreas surgery in the Netherlands, approximately 80 pancreas operations (60 malignant cases) are operated annually in the Radboudumc. As a result, the logistics and perioperative care of our pancreatic surgical program needed reflection and rescheduling.

Development of the CP

The development of the multidisciplinary CP for pancreaticoduodenectomy was a multistep procedure with the use of lessons learned from the development and implementation of the cardiac and oesophageal CPs, previously developed in Radboudumc, and started in 2013.

The first step was redefining and searching for evidence underneath the surgical, anaesthesiology and ICU protocols in the perioperative period. This was a multidisciplinary procedure, undertaken by the physician assistants, senior nurses, ‘key’ nurses and medical staff.2027 Instead of a traditional ‘day-to-day-care’ plan for the surgical ward, an ‘hour-to-hour’ care plan had to be developed, including the PACU and ICU care. It was important to identify potential barriers and facilitators in these settings, in order to tailor the implementation strategy.2831 An evidence-based implementation strategy according to Grol was used.32 Second, a unique variance report (‘Radboud variance report’; Appendix 1) had to be incorporated and developed together with the CP.33 This Radboud model of variance report enables nurses, physician assistants and young residents to execute predefined actions in accordance with and within the preset boundaries of a variance protocol, without having to wait for approval of the responsible physician first (Dutch law and order for health care professionals BWBR0006251 chapter IV, article 35).

Until 2012, a surgical pancreas matrix for (peri)operative care was used at the surgical ward. The historical control group was treated according to this matrix including the surgical medical and nursing protocols without the variance report. In the PACU and ICU, these patients were treated according to different PACU and ICU protocols. This pancreas matrix was used as backbone for further multidisciplinary development of the CP. As part of the development and implementation strategy, a small group of key nurses responsible for other CPs reflected on the concepts of the pancreas CP and variance report as part of a Prepare-Act-Reflect (PAR) Cycle.

The pancreas CP had to be a continuum from admission to discharge from the hospital. Essential elements included restrictive intra-operative fluid use, strict pain control, early mobilization, early drain and tube removal and early enteral feeding. Post-operatively, early warning scores (EWS) are measured at least once during every 8-h shift or more frequent, whenever indicated by the nurses, with strict directives for action by nurses according to the variance report.34

Patients with a malnutrition universal screening tool (MUST) score above 2 need an active feeding intervention according to the quality system of health care in the Netherlands. We decided that patients with a MUST above 2 should start with total parenteral nutrition (TPN) within 24 h after surgery. Publications on calorie deficit and enteral feeding or TPN after surgery in ICU patients often do not take into account malnutrition and MUST score >2. Our protocol prescribes that if the gastric tube can be removed, the patients need to start with oral/enteral feeding, and TPN needs to stop as soon as the oral intake of the patients is above 1000 kcal.22, 23, 3537 TPN should be started on day 3 if patients had a MUST score of 1 and enteral feeding had not been started on day 3. All patients with a gastroparesis without signs of sepsis or ileus on day 7 will be given a naso-jejunal tube by the gastroenterologist through the gastrojejunostomy and start enteral feeding.38 In contrast to ERAS-based protocols, deviations from the CP had to lead into prompt actions according to the variance report.

Implementation of the CP

After informative meetings for medical and nursing staffs, including reflections on the positive aspects of previous CPs, bedside training started on the surgical ward and PACU/ICU in 2014. Implementation of the pancreas CP would introduce an essential change in daily practice for most nurses, physician assistants and medical staff. The first step in teaching was getting acquainted to the CP vision that would result to one continuous multidisciplinary protocol.32 In nursing and medical staff meetings, updates of the project were discussed, and feedback was welcomed by the CP developers. During this teaching period, especially new PACU-specific aspects arose for the pancreas CP, including new variance report criteria, and as an interactive process of PAR cycles, these criteria were incorporated in the pancreas CP during the development. In this try-out period, feedback was asked and given every 4 weeks during the multidisciplinary team meetings of the project. After 4 months of teaching and try-out period, it was concluded that it was feasible and safe to use the pancreas CP with the Radboud model variance report for patients during their entire clinical stay, including the PACU/ICU. With the completion of this implementation step, the pancreas CP was considered being implemented and our study on the use of the CP and variance report for all pancreaticoduodenectomy patients started on the first of September 2014, 18 months after the start of the development of the CP, including many PAR cycles. Patients treated for other pancreas procedures than pancreaticoduodenectomy were considered candidates to have the benefits of the pancreas CP during their stay in PACU/ICU and ward, but were not included in this study. Protocol adherence was measured per pathway action. We considered protocol adherence if a deviation from the CP resulted in the correct action, according to the CP, or if no action was needed and no action was started. No protocol adherence was defined as wrong actions or no actions if actions were needed. Deviations from the CP had to be described in the variance report or patient record.

Design

This is a pre-post design study. After the implementation of the pancreas CP, patients treated according to the CP were compared with a historical control group of patients treated with standard perioperative care for pancreaticoduodenectomy according to the original pancreas matrix and monodisciplinary protocols and operated on between 2009 and 2012.

End Points

Primary endpoint was to determine the feasibility and safety, including incidence of post-operative complications, according to Clavien-Dindo classification, of the use the CP. Secondary endpoints were in length of stay (LOS) in-hospital, post-operative fluid balance, gastroparesis, protocol adherence to mobilization, drain removal, radiologic and surgical re-interventions, ICU readmission, hospital readmission and mortality rate.

Statistics

Continuous variables were described as median and interquartile range [IQR] and tested with the Mann-Whitney U test. Differences in dichotomous variables were analyzed using the chi-squared test. Due to the exploratory nature of this study, and to increase the sensitivity to detect differences between groups, no correction for multiple testing was performed. With our convenience sample size of 95 patients in the CP group and 52 patients in the control group, our study had 80% power to demonstrate a 7% absolute reduction of post-operative complications. All statistical analyses were performed using SPSS version 20.01 for Windows (IBM, SPSS statistics, Chicago, IL, USA).

Results

Development Results of the CP

Nurses, physiotherapists, dieticians and medical staff specialized in pancreas surgery contributed to the development of the pancreas CP and the variance report. This resulted in a set-up of clear and safe boundaries in taking clinical treatment decisions and an upscaling system to consultation with a key nurse or senior staff members, if actions according to the variance report did not seem right.

First, the pancreas CP for medical and nursing decisions was written according to existing evidence-based protocols, best practices and guidelines. Finally, a multidisciplinary variance report was incorporated (Appendix Table 4: summary of the differences between CP and control surgery and Appendices 2 and 3: variance report).

For the analysis of the developmental process, we evaluated barriers and facilitators for protocol adherence. For this, interviews and questionnaires were used, focussing on possible barriers and facilitators for protocol adherence to the new CP. An important facilitator was the motivation of nursing and medical staff to ask for guidance and training in the use of this protocol. The most important barrier was that using the protocols was experienced as a time consuming processes of getting acquainted with the system, resulting in feelings of loss of autonomy for doctors and nurses. Key nurses together with medical leadership were essential for awareness, feedback and motivation during development, implementation and the use of the CP.

Implementation Results of the CP

First, the medical aspects of the CP were implemented on the ward followed by the nursing aspects. Because of the lack of experience with CPs, the care providers working on the PACU received more time for training and bedside teaching and started later with implementation. Key nurses at the surgical ward gave guidance and were partner for the key nurses of the PACU.

Evaluation after the implementation process was performed every 2 months during the first 6 months and after this period whenever needed. These evaluations resulted mostly in questions or new ideas for a change in the CP from the units or when less compliance was observed. The variance report was an important tool for evaluating compliance. When compliance of one of the CP domains was below 80%, feedback was given by the key nurse or surgeon through focussed teaching sessions for nurses and residents.

After a period of 18 months, the pancreas CP was implemented and evaluation of protocol adherence was 80% for PACU/ICU periods and 60% for the surgical ward. The latter was mainly influenced by a low compliance to drain removal (<50%). According to the pancreas CP, drain removal was allowed if amylase level in the drain was below 500 U/l and volume below 200 ml/day. Deviations turned out to be primarily a system problem of postponing drain removal during weekends. After recognition of this system problem, an active policy started and protocol adherence on this item improved to above 80%.

Following the implementation, in September 2014, the outcome study of the pancreas CP was started (Fig. 1 implementation flowchart).

Fig. 1
figure 1

Implementation of pancreas CP and study flowchart

Clinical Outcomes

Between September 2014 and September 2016, in total, 95 elective consecutive pancreaticoduodenectomy patients were treated within the pancreas CP. Semi-acute pancreaticoduodenectomies (for bleeding tumours) and other types of resections (e.g. total pancreatic resections or pancreaticoduodenectomies with resection of a secondary colorectal tumour) were no part of the study. A cohort of 52 consecutive elective pancreaticoduodenectomy patients treated before the CP implementation period between 2009 and 2012 was identified as historical control group. Their perioperative treatment had been according to the underlying matrix protocol that was used as base for the development of the CP. Three surgeons in the pre-CP period operated on the pancreaticoduodenectomy patients. Results between these surgeons did not differ, and perioperative care was regulated by protocols. These surgeons were also responsible for pancreas surgery in the CP period.

Baseline characteristics between the two groups were not significantly different, apart from a higher number of CP patients receiving portal vein resection or celiac trunk/superior mesenteric artery (SMA) vessel exploration (Table 1).

Table 1 Baseline characteristics of pancreas CP and control groups of pancreaticoduodenectomy

Intra-operative Data

The median intra-operative amount of fluids administered was 3900 ml [IQR 3000–4600] in the CP patients versus 5200 ml [IQR 4000–6000] in the control group (p < 0.001). Post-operative fluid balance and fluid balance on day 1 post-operative were also significantly lower in the CP group versus the control group (p < 0.001; Table 2). Although more portal vein resections and celiac trunk and explorations along the SMA were performed, blood loss was less in the CP patients: 755 ml [IQR 500–1100] versus 1303 ml [IQR 656–2402] (p < 0.001, Table 2).

Table 2 Intra-operative results of pancreas CP and control groups of pancreaticoduodenectomy

Post-operative Data

Adherence of pain and hemodynamic interventions according to the variance report was 100% at the PACU/ICU, and a step-up approach regarding pain control was adequately used according to CP protocol. Hemodynamic interventions in accordance with the variance report were not needed and not started in 17% of the CP patients, and 57% of the CP patients needed an extra hemodynamic intervention which was subsequently started according to the CP protocol. In total, 26% of the patients were treated with vasopressors on arrival in the PACU/ICU, which could be reduced during their stay. Significantly more CP patients were swing mobilized within 24 h compared with the control group, respectively, 83 versus 19%, p = 0.001. Especially poor pain control and patients’ feelings of weakness, early after the operation, were recorded as reasons not to start swing or mobilization at the surgical ward. Trigger for complications was the EWS; in 32% of the patients in the CP group, the EWS was above 3. Interventions on a high EWS were adequate and according to the variance report >95% of the patients.

Considering clinical outcome, major complications according to the Clavien-Dindo classification grade 3 or more occurred less frequently (13 vs 27%, p = 0.02) in the CP group, compared to the control group.39 One patient had a Clavien-Dindo 4b complication as a result of pancreatic leakage complicated by sepsis with EWS >6 on day 7 and hemorrhagic bleeding on day 14 in the CP group. This complication was successfully treated by radiologic coiling of the gastroduodenal artery and splenic artery.

Less patients suffered from gastroparesis grades B and C in the CP group compared to the control group, 9 versus 62%, p < 0.001, as were radiologic interventions: 11 versus 27%, p = 0.04. In the control group, the gastric tube was not removed when production was reduced but was left in place and blocked and could be removed if after measurement of retention after 8 and 16 h, it was less than 100 ml per 8 h. Pancreatic leakage and chylus leakage, readmission to ICU and readmission to hospital did not significantly differ between the CP group and control group. Median times to drain removal were also not influenced. The mortality rate was low and not different between groups (Table 3).

Table 3 Post-operative data of pancreas CP and control groups of pancreaticoduodenectomy

Discussion

This study illustrates that development of a CP for pancreaticoduodenectomy is an iterative multidisciplinary process, starting with a dynamic protocol with improvements through PAR cycle evaluation and change moments. Implementation of the pancreas CP in all units involved in the entire (peri-) operative process (OR, PACU/ICU/surgical ward) took 18 months. Process evaluation of the prospective CP group revealed that protocol adherence was successfully achieved in >80% for most of the criteria throughout the clinical stay. Comparison of both cohort groups on main clinical outcomes showed that major complications according to the Clavien-Dindo classification grade 3 or more and hospital LOS in the CP group were significantly lower compared to the control group. In addition, implementation of the CP was associated with a reduction of gastroparesis, an improved post-operative fluid balance, and patients in the CP group were more likely to receive early mobilization and adequate actions on EWS above 3. These data illustrate that implementation of a CP in this specific group of patients is feasible, safe and likely to be beneficial for the patient.

Analyzing reasons not to follow the variance report was part of this study. Human factors were often reasons for deviation from the report, for example, insecurity of young professionals on decisions leading to postponing gastric tube removal. The prevention of gastroparesis is part of a very active PAR cycle in the CP. Nurses, young doctors and patients want to prevent discomfort for the awake patient while repositioning the tube, even if early removal is according to protocol. The action was a team reflection on the discomfort of a needless gastric tube for too long and, as a result, delay in starting early oral nutrition and well-being.

Postponing early mobilization because of patients’ pain or weakness did occur. In all situations, the iterative process of repeated and specific education was important to explain the reasons behind the CP and guidance.

Considering the diverse landscape of CPs and surgical care plans, it is difficult to compare the different studies. In studies, related to implementation of CPs, not all hospital wards involved in the clinical process (like PACU/ICU) were included, which negatively influences the continuous care process for the patient. Also different treatment regimes make reliable comparison and evaluation of different CPs difficult. Regarding the available studies, we found only studies not covering the whole clinical stay, excluding parts of the post-operative period. In these studies usually some specific aspects like ERAS, drain and gastric tube removal were addressed.18 A standardized care plan for pancreaticoduodenectomy patients was retrospectively studied in another study focussing on predictors of LOS in-hospital.15 Specific ERAS pathways, without PACU/ICU periods involved, focussed on in-hospital LOS, outcome mortality and morbidity. While these were unchanged, measurement of protocol adherence was not part of the study.16 Braga et al. evaluated the compliance to the enhanced recovery protocol and concluded that patients with low compliance had a higher incidence of complications.40

Our results are in pursuance of previous studies that showed that a CP or standardized care plan for pancreaticoduodenectomy patients resulted in an earlier start of solid enteral feeding and a shorter hospital LOS and less readmissions. Importantly, protocol adherence to predefined targets has not been part of these studies as was analysis of the reasons not following the protocol and its association to outcome.

Comparing our study to these studies, a similar effect on reduction of complications, hospital LOS, readmissions, gastroparesis, time to enteral feeding and time to mobilization was found. Our present study also illustrates that it is feasible to implement a CP that covers the entire clinical admission, applying different targets of the various involved units (e.g. focus on hemodynamic and respiratory vital parameters at the PACU/ICU, versus focus on EWS and ERAS criteria at the surgical ward). Nurses were also able to start adequate therapy in accordance with the variance rapport when EWS deviated from the target. Moreover, new to the other studies is that this study, via the variance report method, exposed the barriers and facilitators of CP adherence. In addition, these two monthly formal meetings to evaluate variance report deviations and their barriers and facilitators enabled us to discriminate the difference of loss of compliance to a protocol due to complicated discourse of operations, versus loss of professional adherence to the CP protocol.

The current study has several limitations. Most importantly, this is a single-centre pre-post-intervention study. The intensity and duration to develop the CP, as well as the implementation process, limit the feasibility of using other study designs. In addition, the historical group was not formally matched, which, together with the fact that no randomization was carried out, induces a higher risk of confounding factors. No relevant differences in patient characteristics between the different study periods were observed. However, the case load per surgeon increased, which could be considered as a possible confounding factor. We considered the development of a CP as the most appropriate intervention to re-schedule the process. Prospective complication registration was part of the daily supervised perioperative care as well as the discharge procedure in both groups. Moreover the prospective database on outcome and complications of the control group (2009–2012) served as a document to identify barriers and facilitators for building the CP. Furthermore, no relevant changes in other procedures, staffing levels, technical infrastructure or other major changes that could influence patient management occurred, and during the whole study period, there were no changes in interventions that are known to influence morbidity or mortality in the ICU such as strict glucose regulation, early goal-directed therapy, use of corticosteroids, prone positioning and low tidal volume ventilation. Second, no a priori power calculation was carried out, implying that the risk for a type 1 or 2 error has not been overcome. Using our convenience sample, we did calculate that our study has 80% power to demonstrate a 7% change in complication rate, while we observed that the complication rate halved. Nevertheless, the sample size of the study and the discussed design issues should make us aware of the possible overestimation of the outcome differences. In contrast, this does not necessarily apply for the process analysis part. As no comparison of the CP group was made to the control group, the conclusions of the process analysis merely indicate that CP development, implementation and high level of adherence to such a CP, throughout all units involved in the perioperative process, are feasible within a relative short period and up to a high standard.

Lessons Learned

This study shows us, in line with the implementation of our cardiac surgery CP and oesophageal surgery CP,41 that it is feasible to develop and implement a CP for pancreaticoduodenectomy procedures for all involved units like the PACU/ICU and surgical ward through the entire clinical perioperative period. In all units, the CP targets need to be aligned and the use of a variance report discriminates complication-related to failure of professional adherence. Implementation is an iterative process that takes time to become comfortable in use for all involved units. Key nurses together with medical leadership were essential for awareness, feedback and motivation during development, implementation and the use of the CP.

Future Perspectives

In order to overcome the methodological drawbacks of this study and to validate the CP methods, a multicenter stepped-wedged cluster randomized controlled trial would be ideal. However, due to the complexity of the implementation and intervention with barrier and facilitator analysis in different hospitals and units, interpretation of the results will be difficult. Exploring the validity of similar CPs is in line with the need for quality assurance of standardized treatment regimes with high protocol adherences.

For the near future, continuous monitoring, wearables and electronic medical data recording with pop-up facilities warning medical and nursing staff for deviations from the CP will likely be of help in building more complex pathways. Possibly, patients with high comorbidity will be able to follow their personalized clinical pathway (pCP) with the help of dedicated staff.

Conclusion

The use of the CP was associated with a reduction of perioperative morbidity. Essential new tools include a variance report analysis, scheduled barrier and facilitator analyses and the iterative PAR cycle protocol development, performed by a multidisciplinary team. Development, implementation and use of a CP throughout the hospital stay for patients undergoing pancreaticoduodenectomy are a multistep procedure in which we showed that this is feasible and safe.