In the last decade, the laparoscopic approach for distal pancreatectomy (DP) has been increasingly adopted worldwide showing benefits in terms of reduced intraoperative bleeding, improved pain control, shorter hospital stay, and comparable postoperative morbidity to the traditional open technique [1]. The first multicenter randomized controlled trial (i.e., LEOPARD trial) comparing minimally invasive to open DP has recently been completed in the Netherlands, reporting improved clinical recovery in the minimally invasive surgery group [2]. The LEOPARD study found that despite a similar postoperative complication rate, patients treated with minimally invasive DP had a shorter time to functional recovery and improved quality of life in the first month after surgery. However, despite an early return of function and hospital discharge, 1 out of 4 patients in the trial experienced a hospital readmission within 90 days after surgery, which may eliminate the early benefits of minimally invasive surgery.

Readmissions represent an important metric for quality of surgical care, especially in the context of enhanced recovery pathways where length of hospital stay is shorter, and it has significant financial implications on the healthcare system [3, 4]. Furthermore, readmissions should be considered as a key patient-centered outcome hindering patient recovery and negatively affecting return to preoperative function. While a recent multi-institutional benchmarking study set the cut-off for readmissions after pancreaticoduodenectomy to < 21% [5], only a few retrospective studies have focused on rehospitalization following distal pancreatectomy [6]. A large two-institution trial reported a 14% overall readmission rate, with the most common reasons for readmission being the presence of an intra-abdominal abscess and other infectious complications [7]. The study showed that older age and postoperative complications during primary admission were the strongest predictors for post-discharge hospitalization. However, this study focused on a 10-year span between 2004 and 2014 with < 20% DPs performed laparoscopically, and no enhanced recovery pathway implemented resulting in a more conservative postoperative management and longer hospital stays compared to current perioperative care policies. Thus, a new study focused on readmissions after DP in a contemporary perioperative scenario is warranted.

The main objective of the present study is to evaluate the extent to which laparoscopic surgery impacts on postoperative readmissions and post-discharge emergency department (ED) visits following distal pancreatectomy. Secondarily, aims of this research include identification of perioperative variables associated with hospital readmission and post-discharge ED visits and verification of the impact of minimally invasive surgery on patient functional recovery.

Methods

Study design

This is a retrospective review of an institutional database including patients undergoing elective pancreatic resection in a high-volume center for pancreatic surgery [8]. This research was conducted and reported following the Strengthening the Reporting of Observational Studies in Epidemiology Statement (STROBE) guidelines [9]. This study was performed in accordance with the Declaration of Helsinki and all patients signed an informed consent before surgery including authorization to collect clinical data for research purposes. Since the study was retrospective, an approval by our institutional review board was waived.

All adult patients who underwent elective DP between January 2011 and December 2018 in the Division of Pancreatic Surgery at San Raffaele Scientific Institute in Milan, Italy were screened for inclusion in this retrospective cohort study.

Perioperative care

Since October 2010, all patients undergoing DP were treated within a dedicated enhanced recovery pathway as detailed in a previous publication [10]. All operations were carried out by a dedicated team of surgeons with a specific training in pancreatic surgery including minimally invasive DP [11]. The choice of minimally invasive versus open technique was at the surgeon’s discretion. As a general rule, patients with benign lesions or neuroendocrine tumors were approached laparoscopically, while pancreatic cancer cases were discussed on an individual basis according to tumor location and vessel proximity. Splenectomy and standard lymphadenectomy were routinely performed in cancer patients. A spleen preserving procedure with preservation of splenic vessels was attempted only in non-neoplastic cases at surgeon’s discretion. Pancreatic stump transection was performed with a stapler in laparoscopic procedures, whereas in open procedures, the pancreas was cut using the scalpel followed by selective suturing of the pancreatic duct when visible. One flat Penrose drain (12 mm; Redax®) was placed in proximity of the pancreatic stump. Drain fluid amylases were routinely drawn on POD3 and POD5. In the case of non-sinister fluid and drain amylase value less than threefold the normal range, removal of the pancreatic surgical drain was suggested but remained at the operating surgeon’s discretion.

Patients were discharged home after achieving the following criteria: tolerance of solid food intake with no need for intravenous fluids, return of gastrointestinal function (i.e., passage of stool), adequate pain control with oral analgesia, independent mobilization (or restored at the preoperative level), and no clinical signs of untreated medical issues. If there were no clinical signs of infection, patients with pancreatic fistula were discharged home with a drain and followed up with scheduled office visits. Patients were given an email address they could contact at all times to communicate any clinical issue to the on-call member of the surgical staff.

Data collection

Research fellows entered individual patient data in the Division of Pancreatic Surgery database at time of patient admission, during hospital stay, and after hospital discharge up to 90 days postoperatively. The database features preoperative patient characteristics, surgical, and postoperative data as well as follow-up information. Preoperative variables included demographics, comorbidities, lab testing, and diagnosis. Intraoperative data included type of surgery, duration of surgery, blood loss, and need for intraoperative blood transfusion. Postoperative data included functional recovery measures, medical and surgical complications, and post-discharge outcomes. For the purpose of this study, a clinical researcher reviewed all patients’ hospital medical records and follow-up visits up to 90 days after surgery.

Definition of outcomes

Primary outcome of the study was hospital readmission, defined as any hospitalization, including those in other institutions, occurring after discharge within 90 days after surgery.

Secondary outcomes included ED visits recorded up to 90 days after surgery and time to functional recovery. Functional recovery was defined as the achievement of the following criteria: recovery of gastrointestinal function (i.e., passage of stool); adequate pain control with only oral medications; sufficient caloric intake in the absence of intravenous fluid administration; and restored mobility. Functional recovery outcomes were assessed daily by physicians during morning clinical rounds and noted in a specific clinical chart.

Additionally, postoperative complications and length of hospital stay (LOS) were assessed. Postoperative pancreatic fistula (POPF), delayed gastric emptying (DGE), and post-pancreatectomy hemorrhage (PPH) were classified and graded using the International Study Group on Pancreatic Surgery (ISGPS) definitions [12,13,14]. Grades B and C POPF were considered as clinically relevant. Surgical site infections (SSIs) were classified as superficial incisional, deep incisional, or organ–space according to the definition by the Center for Disease Control and Prevention (CDC) [15]. Complication severity was classified using the Clavien–Dindo classification [16]. Major complications were defined as Clavien–Dindo grade III or higher. The comprehensive complication index (CCI) was then generated for each patient [17]. This is a validated measure summarizing the complete spectrum of complications occurred and their severity in a single score ranging from 0 to 100.

Statistical analysis

Normality of continuous variables was assessed by visual inspection of frequency histograms and with Kolmogorov–Smirnov test. Descriptive data are displayed as mean (standard deviation), median [25th percentile–75th percentile], and number of patients (percentage). Categorical variables were compared by the chi-square test or Fisher’s exact test, as appropriate. Continuous variables were compared by the Student’s t test if normally distributed, or non-parametric Mann–Whitney U test.

Univariate logistic regression analysis was performed to evaluate variables correlated with hospital readmission and ED visits. To identify independent predictors of readmission, a multivariate analysis was conducted. No variable pre-selection according to univariate analysis was performed. All clinically relevant variables for the outcomes were initially entered in the multivariate regression models and a backward stepwise elimination approach was used to determine factors significantly associated with the outcome interest. Only variables that exhibited a p < 0.10 were retained in the final model and collinear variables were dropped out in order to eliminate confounding factors.

A univariate and multivariate linear regression was used to analyze factors affecting functional recovery expressed as log-transformed time to functional recovery. Backward stepwise elimination was used to determine factors significantly associated with this outcome, and variables significant at p < 0.10 were retained in the final multivariable model.

The type of surgery variable was analyzed as per protocol (i.e., patient converted from laparoscopy to open surgery were considered in the analysis as open surgery patients). To confirm our findings, a sensitivity analysis with an intention-to-treat approach was also performed for the main study outcomes.

p values < 0.05 were considered to be statistically significant in the final model. All statistical tests were two-sided. Analyses were performed with STATA version 14.2 software (StataCorp, College Station, TX, USA).

Results

Three hundred seventy-six patients who underwent distal pancreatectomy were included in the study. Overall, 244 (64.8%) patients initially underwent a laparoscopic approach, 25 of them (10.2%) required conversion to open surgery. Thus, 157 (41.8%) patients were included in the open group and 219 (58.2%) patients in the laparoscopic group. Preoperative and intraoperative characteristics are summarized in Table 1. Around one-third of patients (n = 127, 33.7%) included in the study was 70 years or older, and about half (n = 180, 47.9%) was overweight (BMI > 25 kg/m2). Main indications for surgery were pancreatic ductal adenocarcinoma, neuroendocrine tumors, and benign cystic neoplasms. Patients in the open group were older, mostly due to a greater proportion of pancreatic cancer patients, whereas neuroendocrine tumors and benign cystic neoplasms, which occur at a younger age, were more frequently approached laparoscopically. Multivisceral and vascular resections were significantly higher in patients undergoing open surgery, translating in a higher operative time and blood loss compared to the laparoscopic group.

Table 1 Perioperative characteristics of patients undergoing open and laparoscopic distal pancreatectomy
Full size table

Postoperative outcomes and complications during primary hospital stay and at 90 days after surgery are reported in Table 2. The median time to functional recovery after DP was 5 days (IQR 4–6 days), while median LOS was 3 days longer (8 days, IQR 6–10). One hundred sixty-eight patients (44.7%) were discharged with an abdominal drain in place. At discharge, 90 (23.9%) patients had developed a clinically relevant pancreatic fistula. At 90 days after surgery, following the 2016 ISGPS definition, clinically relevant pancreatic fistula was diagnosed in 174 patients (46.3%). Interestingly, SSIs were the complication with the largest increase after discharge (72% increase, from 39 to 67 patients) mostly due to an increase in organ–space SSIs.

Table 2 Postoperative clinical outcomes and complications during primary hospital stay and at 90 days after surgery in all patients
Full size table

Table 3 reports postoperative outcomes at 90 days after surgery in the laparoscopic and open group. No difference was found between groups for postoperative surgical complications, while cardiac and respiratory complications were lower in the laparoscopic group. Median complication severity as measured by CCI was 8.7 [0–22.6] in the laparoscopic group, which is equivalent to a single Clavien–Dindo grade 1 complication, versus 20.9 [0–24.2] in the open group, which is a single Clavien–Dindo grade 2 complication (p = 0.214). Median LOS was one day shorter in the laparoscopic [7 (6–10) days] group compared to open [8 (7–10) days] (p = 0.004).

Table 3 Postoperative clinical outcomes and complications at 90 days after surgery in patients undergoing open versus distal pancreatectomy
Full size table

Table 4 includes information on post-discharge outcomes. ED visits within 90 postoperative days occurred in 62 (16.5%) patients, 41 (18.7%) who underwent successful laparoscopic resection, and 21 (13.4%) an open DP (p = 0.162). This finding was confirmed when an intention-to-treat analysis was performed (n = 44, 18.0% in laparoscopic vs. n = 18, 13.6% in open DP patient; p = 0.253). The most common symptoms leading the patients to the ED were fever and abdominal pain. Hospital readmission at 90 days was observed in 46 (12.2%) patients of those observed in the ED, 31 (14.2%) who had completed a laparoscopic resection, and 15 (9.6%) an open DP (p = 0.179). No difference was found also when an intention-to-treat analysis was performed (n = 33, 13.5% in laparoscopic vs. n = 13, 9.8% in open DP patients; p = 0.282). The most common indication for readmission was the presence of an infected abdominal collection (n = 37, 80.4%), which was treated with antibiotic therapy and in selected cases with either percutaneous or endoscopic drainage. Of these 37 patients, 31 had a diagnosis of POPF during index admission. Less commonly, PPH secondary to POPF determined readmission (n = 6, 13%). The median time of readmission was around 2 weeks after hospital discharge and 25 days after the initial surgery. No significant difference in post-discharge outcomes was found between patients in the open and laparoscopic group.

Table 4 Postoperative ED visits and hospital readmission features in patients undergoing open versus distal pancreatectomy
Full size table

Factors influencing readmissions and ED visits

The univariate and multivariate analyses of factors associated with 90-day hospital readmission are reported in Table 5. At univariate logistic regression, laparoscopic surgery had no impact on readmissions. Occurrence of postoperative complications including pancreatic fistula, PPH, infections, and being discharged with an abdominal drain in place were associated with readmission at univariate analysis. Conversely, a short hospital stay (i.e., discharge within POD 7) demonstrated a protective effect on readmissions at univariate analysis. At multivariate analysis, a low preoperative physical status (i.e., ASA score 3 or more) (OR 2.30, 95% CI 1.16–4.56, p = 0.017), clinically relevant POPF (OR 6.77, 95% CI 2.89–15.88, p < 0.001), and PPH (OR 3.94, 95% CI 1.18–13.13, p = 0.025) were the only independent factors significantly associated with hospital readmission within 90 days after DP. When an intention-to-treat analysis was performed for type of surgery, the laparoscopic approach still had no association with hospital readmission at univariate (OR 1.45, 95% CI 0.73–2.86, p = 0.284) and multivariate linear regression.

Table 5 Univariate and multivariate logistic regression analysis for predictors of 90-day hospital readmission
Full size table

Table 6 shows logistic regression analysis for factors influencing ED visits within 90 days postoperatively. At multivariate analysis occurrence of an infectious complication during primary hospital stay (OR 2.18, 95% CI 1.11–4.28, p = 0.024) and being discharged home with an abdominal drain (OR 5.76, 95% CI 2.98–11.11, p < 0.001) were significantly associated with an access to the ED after discharge. At intention-to-treat analysis, the laparoscopic approach had no impact on ED visits at univariate (OR 1.41, 95% CI 0.78–2.55, p = 0.255) and multivariate linear regression.

Table 6 Univariate and multivariate logistic regression analysis for predictors of 90-day emergency department visits
Full size table

Factors influencing functional recovery

Univariate and multivariate linear regression analysis on perioperative variables associated with time to functional recovery is reported in Table 7. At multivariate analysis, successful laparoscopic surgery was the only factor shortening time to functional recovery with an estimated reduction by 13% (95% CI − 19 to − 6%; p < 0.001). Older age, occurrence of delayed gastric emptying, organ–space SSI, and reoperation significantly delayed time to functional recovery by 9% (95% CI 2–15%; p = 0.012), 30% (95% CI 11–49%; p = 0.001), 31% (95% CI 19–42%; p < 0.001), and 62% (95% CI 44–79%; p < 0.001), respectively. When an intention-to-treat analysis was performed for type of surgery, the laparoscopic approach remained significant at multivariate linear regression (− 12%; 95% CI − 19 to 6%; p < 0.001).

Table 7 Linear regression analysis for perioperative predictors of postoperative time to functional recovery
Full size table

Linear regression analysis of variables associated to time to reach an adequate oral intake was reported on Supplementary material 1. At multivariate analysis, successful laparoscopic surgery reduced by 19% time to achievement of adequate oral intake. Conversely, delayed gastric emptying, organ–space SSI, and reoperation significantly delayed postoperative time to adequate oral intake by 48%, 37%, and 70%, respectively.

Supplementary Table 2 shows the perioperative factors associated to postoperative time to adequate pain control with oral analgesia. At multivariate analysis, laparoscopic surgery reduced by 12% time to achievement of adequate pain control. organ–space SSI and reoperation significantly delayed postoperative time to adequate pain control by 25% and 19%, respectively.

Discussion

In modern perioperative care settings where patients’ postoperative hospital stay is progressively being shortened, hospital readmissions represent a very relevant outcome, with implications both on patient recovery and healthcare costs [3]. However, this aspect has been poorly studied in patients undergoing DP with a minimally invasive approach. In the present series, including 376 consecutive patients who underwent DP, mostly with a laparoscopic approach, in a single high-volume center with an established enhanced recovery pathway, 12.2% patients required hospital readmission and 16.5% at least one ED visit after discharge. Minimally invasive surgery did not have a significant impact on hospital readmissions nor ED visits compared to the open approach, while it reduced patient time to reach functional recovery.

Unplanned hospital readmissions are an important metric as they are reflective of patient outcomes, health system performance, and care coordination [18]. Since they are mostly related to surgical complications, they are not a rare occurrence after a major operation like pancreatectomy. A recent meta-analysis reported readmission rates between 12 and 30% after pancreatic resection [6]. However, a minority of studies included DP patients, as most were focused on pancreaticoduodenectomy, where the open approach is prevalent, and severe postoperative morbidity higher. The recently published LEOPARD study, where an enhanced recovery pathway was used, and patients were followed-up prospectively for 90 days after DP, reported a high overall readmission rate around 27%. In the present research at San Raffaele hospital, with a similar perioperative care program and follow-up time, readmissions were lower (12.2%), similar to other retrospective studies available in literature [1, 7, 19]. The large difference with the LEOPARD study may be attributable to a different post-discharge management, but also to an underestimation in retrospective studies due to missing information, patients lost to follow-up, or reporting bias.

A small retrospective comparative study from Baker et al. reported a preliminary experience on laparoscopic distal pancreatectomy showing a significantly increased hospital readmission rate in patients treated with laparoscopy, eliminating the perceived benefit of a shorter LOS [20]. Conversely, in our study, the type of surgical approach had no impact on hospital readmission rates confirming results from the DIPLOMA trial, a previous multicenter propensity score matched study including more than 1200 distal pancreatic resections [19]. This may reflect the results from more experienced centers who have already completed a learning curve in laparoscopic DP, but also a standardization of postoperative care (i.e., same perioperative pathway regardless of surgical technique) with the use of predefined discharge criteria, reducing the risk of a premature discharge.

The main reason for readmission in the present series was the occurrence of an intra-abdominal abscess requiring antibiotic treatment and, in most cases, a drainage procedure (percutaneous or endoscopic). Less commonly, an episode of late post-pancreatectomy bleeding brought the patient back to the hospital. Our findings overlap previous research from Marchegiani et al. showing that intra-abdominal infected fluid collections are the most common reason for rehospitalization [7]. These results point out a significant difference with pancreaticoduodenectomy patients where many patients require readmission very early after discharge because of gastrointestinal symptoms secondary to DGE, dehydration, and failure to thrive [4, 6].

Both intra-abdominal abscess and late PPH after DP are usually secondary to pancreatic fistula and superinfection of the pancreatic fluid despite the presence of an abdominal drain [21]. In fact, occurrence of clinically relevant POPF during index admission was significantly associated with readmission at multivariate analysis showing the highest odds ratio. Multiple studies have demonstrated that early drain removal can be associated with a reduced risk of clinically relevant POPF and potentially with a reduced risk of abdominal fluid contamination [22]. However, there is still no clear evidence regarding the optimal cut-off value for drain fluid amylase on the first postoperative day for predicting clinically relevant fistula after DP [23]. In our series, drain fluid amylase were evaluated on POD 3 and POD5, and the drain was left in place if the value was greater than threefold the normal range, leading to more than 40% patients discharged home with an abdominal drain. This policy, along with a rigorous use of the 2016 ISGPS POPF definition (i.e., patients still carrying a drain at 3 weeks from surgery despite no symptoms fall into grade B POPF category), is possibly responsible for the higher rate of clinically relevant POPF compared to other studies. Surgical drain may also be omitted during DP, according to a recent multicenter RCT [24] where no difference in clinically relevant POPF rate was found between the drain and no drain policy. Nonetheless, drain placement appears to be associated with a reduced severity of POPF as reported in a 2019 multicenter international collaborative study [25]. For these reasons, a specific tool for POPF risk stratification in DP is urgently needed, as identifying patients at negligible risk for POPF could allow drainless surgeries with a potential decrease of intra-abdominal infections due to drains maintenance.

Patient age, preoperative physical status, and comorbidities have also been associated with an increased risk of readmission in previous studies [4, 7, 26]. In our research, the only preoperative factor significantly associated with hospital readmission was a poor patient physical status as measured by an ASA score 3 or greater, while age and specific concomitant disease had no impact. A low preoperative physical status identifies patients with frailty phenotype and more comorbidities, which may not cope with a complication managed at home (e.g., SSI treated with antibiotics). Actually, there is recent evidence that patient frailty as measured by specific scores correlates with an increased risk of readmission early after surgery [27, 28]. Scores such as the modified Frailty Index (mFI) and the Risk Analysis Index (RAI) could also be implemented into clinical practice to flag patients at higher risk for readmission.

In the present series, patients were readmitted at a median time of 2 weeks after initial discharge and around 4 weeks after surgery, suggesting that a tight follow-up including office visits and frequent phone calls should be maintained for at least the first month after surgery. A cautious approach should be reserved to patients discharged with POPF and an abdominal drain in place, and patients with lower physical status as they have a higher risk of post-discharge complications.

A very unique feature of our study was the assessment of post-discharge ED visits, which have been rarely reported in retrospective studies. Three out of four patients who presented to the ED after discharge were readmitted, while one out of four was discharged home. This latter subgroup of patients represents a target for quality improvement, as they could have been managed in an outpatient setting or through the local community services or family physicians. The most relevant factor significantly associated with ED visits was home discharge with an abdominal drain, confirming the major impact that POPF and drain management have on patient recovery after surgery.

Another secondary objective of the present study was to evaluate the impact of minimally invasive surgery on time to patient functional recovery. The implementation in recent years of novel perioperative care pathways (i.e., enhanced recovery programs) has allowed a quicker postoperative rehabilitation with shorter hospital stay and considerable benefits on patient-centered outcomes also in the context of pancreatic surgery [29]. In the present series, where enhanced recovery was applied in all patients and median time to functional recovery was five postoperative days, a successful laparoscopic resection was the only significant factor reducing time to functional recovery after DP by 1 day. This finding is secondary to an earlier recovery of oral feeding and adequate pain control with oral medication after minimally invasive DP compared to open surgery. Conversely, older age, multivisceral resection, and postoperative complications, such as SSIs, significantly prolonged time to functional recovery. These results corroborate the primary outcome of the LEOPARD study [2] where laparoscopy reduced median time to functional recovery by 2 days compared to open DP. Despite the use of predefined discharge criteria, it should be noted that, in our study, median LOS was 2–3 days longer than time to functional recovery. LOS is invariably influenced by non-clinical factors unrelated to patient recovery status, such as healthcare system culture, surgeon preference, and patient expectations. Reducing the gap between achievement of functional recovery and home discharge represents another target for quality improvement.

Limitations and strengths

The main limitation of this study is its retrospective nature, which may carry selection bias, missing information bias, and potentially inaccurate data collection. Despite the routine postoperative follow-up in all patients with post-discharge office visits and phone calls up to 90 days after surgery, we cannot exclude that readmissions occurring in other hospitals may have been missed, thus underestimating the frequency of this outcome. Another shortcoming was the absence of patient reported outcome measures, such as health-related quality of life measures and more detailed information on level of functionality including physical, emotional, and social components after discharge. Additionally, the definition of functional recovery used in this trial was only based on clinical parameters during the early hospital phase of recovery, missing other domains that are more relevant to patients (e.g., emotional functions, return to activities of daily living, interpersonal relationships) [30]. Further studies focusing on patient-centered outcomes are needed to better understand the impact of specific surgical procedures on patients’ recovery.

Nevertheless, this study has different strengths including a large sample size consisting of almost 400 consecutive patients that were treated with a homogenous perioperative care pathway. Although readmissions can be considered a rare outcome, the large sample size allowed us to perform a detailed analysis including multivariate regressions with multiple predictors. Other strengths of the study include the use of validated definitions for postoperative morbidity including ISGPS definitions for post-pancreatectomy complications and CDC definitions of SSIs, which facilitates comparability with other series.

Conclusions

In this retrospective study including a consecutive series of patients undergoing distal pancreatectomy in a high-volume hospital for pancreatic surgery, hospital readmissions and emergency department visits were not influenced by the surgical approach. A low preoperative physical status, occurrence of postoperative pancreatic fistula, and hemorrhage during first admission were significantly associated with post-discharge readmission within 90 days. Our findings suggest that patients carrying these risk factors be monitored closely after discharge as they are at greater risk for further complications and hospital readmission.