Predicting long-term risk of reoperations following abdominal and pelvic surgery: a nationwide retrospective cohort study.

Highlights One in seven patients has a reoperation in the 5 years after abdominal surgery. Important risk factors for reoperation are mesh, infection, and colorectal surgery. Individual risk for reoperation can be determined based on the prediction model. The prediction model can improve patient information and the informed consent procedure. In contemporary surgical practice, approximately half of all elective abdominal procedures are reoperations 1 – 3 . Reoperations have an inherent increased risk of complications, related to anatomical changes and the formation of adhesions in particular. These risks apply to patients undergoing emergency surgery for obstruction related to adhesions, as well as patients undergoing (elective) reoperation for indications unrelated to adhesions 1 , 4 . After previous laparotomy up to 90% of patients undergoing a reoperation require adhesiolysis to gain access to the operative area 5 . Following a laparoscopy, the overall incidence of adhesions is 60 and 44% of patients have adhesions at the previous operative site 6 . Adhesiolysis at reoperation increases the operative time by a median of 18 min and is associated with 6–10% risk of iatrogenic bowel injury 1 , 7 , 8 . Adhesiolysis is associated with a higher risk of infectious complications even if no bowel injury occurred 9 . Incremental hospital costs associated with adhesiolysis are estimated to be $4500 higher per operation compared with similar procedures without the need for adhesiolysis 1 . Identifying patients at high risk for future reoperations is considered difficult because most reoperations are not related to the previous procedure. Therefore, surgeons are often not directly confronted with the reoperations and subsequent adhesion-related complications. Moreover, risk factors for reoperations are not yet fully elucidated. A national survey demonstrated surgeons and gynecologists underestimate risk factors for adhesion formation that impact the need for adhesiolysis during reoperation 10 – 12 . Surgeons often fail to inform patients about the future risk of reoperation and other adhesion-related complications during informed consent 10 , 11 . Understanding the risk factors for future reoperation results in a more detailed informed consent and could guide the application of adhesion prevention strategies such as the use of adhesion barriers. The aim of this study was to construct an evidence-based prediction model based on the risk of future reoperation using data from the nationwide SCAR update study cohort. Models were fitted for the overall risk of reoperation, a reoperation in the same surgical area, and a reoperation in the same surgical area after more than 30 days. Most reoperations within 30 days are performed because of short-term complications of surgery such as anastomotic leakage or abscesses. These risks are often mentioned as such in informed consent. Patients reoperated after more than 30 days in the same surgical area are also the most likely to benefit from adhesion prevention strategies.

M.K.T.: methodology, formal analysis, validation, writing – original draft – review and editing, visualization, project administration. P.K.: conceptualization, visualization, data curation, writing – review and editing, funding acquisition. A.J.: visualization, data curation, formal analysis. T.D.G.: formal analysis, writing – review and editing. M.W.J.S., E.A.G., M.C.P., N.D.B., and H.v.G.: conceptualization, supervision, writing – review and editing. R.P.G.t.B.: conceptualization, methodology, supervision, funding acquisition, writing – review and editing.

For this study, the data from the SCAR update study were used 13 . Data for this study were retrieved from the Scottish Medical Record Linkage Database of the Scottish National Health Service (NHS Scotland). The data from the NHS Scotland includes all data from inhabitants of Scotland, with no opt-out. All patients (including minors) who underwent initial open or laparoscopic abdominal or pelvic surgery between June 2009 and June 2011 were included both elective and emergency operations. Patients with a history of previous abdominal or pelvic surgery were not selected. Abdominal and pelvic operations were identified by the Office of Population Censuses and Surveys Classification of Interventions and Procedures version 4 (OPCS-4) codes. Readmissions were defined by OPCS-4 codes combined with the International Classification of Disease, 10th Revision (ICD-10) diagnostic codes. All pelvic and abdominal reoperations during 5 years of follow-up were reviewed for this study. The most frequent indications for reoperations are described. More detailed information on this cohort can be found in a previous publication 13 . The variables incorporated in the prediction model were extracted from the OPCS-4 and ICD-10 codes. Demographics comprised age and sex and surgical variables were surgical approach (laparoscopic/open), surgical area of operation (abdominal wall, small pelvis, right upper quadrant, right lower quadrant, left upper quadrant or left lower quadrant), type of surgical procedure (liver, retroperitoneal, urological, upper gastrointestinal, gall bladder, appendix, gynecological uterus-sparing, gynecological with hysterectomy, rectum, colon or other), presence of inflammatory bowel disease (IBD), malignancy as an indication for initial surgery, previous radiotherapy, intra-abdominal infection, presence of a registered endometriosis diagnosis and mesh placement at initial surgery. Relevant ICD-10 codes for intra-abdominal infection have been listed in Appendix Table B (Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ). For any single surgical procedure more than one surgical area could be scored, for example, right hemicolectomy was considered as the right lower quadrant and right upper quadrant. Nomograms were built based on prediction models that were fitted using Cox regression analysis. A prediction model was built for the overall risk of reoperation 2 and 5 years after initial surgery, for reoperations in the same anatomical area, and finally for reoperations after 30 days in the same anatomical area. The multivariable prognostic model carried out in the Regression Modeling Strategies package in R (version 3.5.1) was used. To fit the prediction model we started with univariable Cox regression analysis, analyzing the difference between demographics, presence of IBD or malignancy, operated organ domain, and operation type. The incidence of repeat surgery and hazard ratios (HRs) with 95% CIs was calculated. Statistically significant differences were defined as a P -value of less than 0.05. All significant predictors for reoperations in the univariable analysis were included in the multivariable analysis. Using backward selection, variables were removed from the model in case of a P -value greater than 0.20. The c -statistic was used to quantify accuracy in terms of predictive discrimination. A c -statistics 0.65–0.70 were considered moderately predictive, 0.70–0.79 good prediction, 0.80–0.89 excellent prediction, and more than 0.9 outstanding 14 . To prevent overestimation with fitting on a single cohort, the validity of the created prediction model was evaluated by internal cross‐validation using a bootstrap procedure. The bootstrap procedure is the recommended method for internal validation of the predictive accuracy of prediction models 15 . From the bootstrap resampling procedure the adjusted c -statistic was calculated. The difference in the compared c -statistics is described as optimism. In case of optimism below 0.01, the original model was considered accurate and consequently retained. Furthermore, subgroup analysis was performed to verify the consistency of the subgroups. The subgroups are benign versus malignant indications for initial surgery, open versus laparoscopic initial surgery, and reoperation before and after 30 days after the initial surgery. To predict the risk of reoperation in the individual patient the nomogram was used by filling in the independent risk factors which corresponded to a number of points on the scale. The aggregated points of the risk factors translated to the prediction of the 2-year and 5-year risk of reoperation. The study is registered on Clinicaltrials.gov number NCT05516420 16 . The research proposal was approved by the board of the Electronic Data Research and Innovation Service of Public Health Scotland (PHS). This study is conducted and reported in line with the STROCSS criteria 17 (Supplemental Digital Content 1, http://links.lww.com/JS9/A322 ).

Overall, 72 270 patients underwent their initial abdominal or pelvic surgery during the study period; reoperations were performed in 10 467 (14.5%) patients in the 5 years after the initial surgery (Fig. 1 ). In 9497 (13.1%) patients reoperation was performed in the same anatomical area. Reoperations were performed during a mean follow-up of 70.2 months (SD=22.2, range: 0–95.1). Table 1 displays the baseline characteristics. During the study period, 21 519 (29.7%) patients underwent a laparoscopic procedure. All other operations (50 751, 70.2%) were open procedures. In 16.0% of procedures (11 560 of 72 270) malignancy was the indication for the initial surgery. Reoperations per surgical area. Baseline characteristics of the study population Data are n , n (%), or mean [SD (range)]. GI, gastrointestinal; HR, hazard ratio; IBD, inflammatory bowel disease. There was wide variation in the indications for reoperation. In Table A (Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ), the 10 most common indications for reoperation after laparotomy or laparoscopy are presented. These 10 indications were related to 60% of reoperations after open initial surgery and 91% of reoperations after laparoscopic procedures. Most reoperations were unrelated to the initial surgery. Some reoperations were related but unplanned, such as the correction of abdominal wall hernia and adhesiolysis for obstruction. Only 18.5% of these common reoperations could be identified as a possibly planned reoperation (i.e. reoperations that could be anticipated at the initial procedure such as ileostomy closure). Reoperations for disease recurrence were relatively uncommon and not included in the list of common procedures for reoperation. Almost all variables significantly impacted the risk for reoperation overall in the univariable (Table C, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ) and multivariable analyses (Table 2 ) and were entered into the prediction model (Fig. 2 ). Only endometriosis was excluded from the multivariable analysis through backward selection. Risk factors for reoperation overall in multivariable analysis GI, gastrointestinal; HR, hazard ratio; IBD, inflammatory bowel disease. Nomogram to calculate the risk for reoperation overall. By drawing a vertical line from each variable to the point axis on top and summing the individual points for all variables, the total score is calculated. From the total score axis perpendicular to the bottom, the linear predictor and the 2-year and 5-year risk for reoperation are determined. An example of a calculation using the nomogram (Fig. 2 ); male patient, 50 years old (23 points), laparoscopic (0 points) right hemicolectomy (colon 71 points) due to malignancy (14 points) as initial surgery (taken the highest surgical area, seen right hemicolectomy is performed in right upper quadrant (55 points) and right lower quadrant (16 points). Summed up a total of 179 points, translating to a 14% risk for reoperation in 2 years postoperative and an 18% 5-year risk. GI, gastrointestinal; IBD, inflammatory bowel disease. Placement of mesh at initial surgery had the highest impact on the risk of reoperation overall (35.2%, HR=7.06, 95% CI: 6.60–7.55). Other relevant risk factors were operations involving the rectum (30.2%, HR=5.40, 95% CI: 4.19–6.96) or colon (18.7%, HR=3.39, 95% CI: 2.63–4.36) and intra-abdominal infection (17.6%, HR=2.24, 95% CI: 2.10–2.50). Following open procedures, more patients (15.0%) underwent reoperation (HR=1.30, 95% CI: 1.23–1.39) compared with laparoscopic procedures (8.6%). Reoperations were performed more often in women (13.4%) compared with men (12.3%, HR=1.15, 95% CI: 1.10–1.21). The results were robust in subgroup analysis, comparing benign with malignant indications for initial surgery (Tables D, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 and E, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ) and open with laparoscopic approach (Tables F, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 and G, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ). The only noticeable difference between subgroups was the impact of sex in the subgroups of malignant surgery. The risk for reoperation was increased in men with malignancy as an indication for initial surgery. In all other subgroups, women had a higher risk for reoperation. In the subgroup of laparoscopic initial operations, intra-abdominal infection was excluded as a risk factor for reoperation from the multivariable analysis through backward selection. The nomogram for reoperation overall (Fig. 2 ) had good prognostic accuracy with a c -statistic of 0.72. The accuracy of the model remained the same after internal validation through bootstrapping ( c -statistic=0.72). Due to optimism in the model of 0.001, the original model could be retained. Univariable analysis was performed on all variables (Table H, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ). Through backward selection endometriosis and intra-abdominal infection were excluded from multivariable analysis (Table 3 ). Consequently, the remaining variables were entered into the prediction model. Multivariable analysis of risk factors for reoperation in the same surgical area GI, gastrointestinal; HR, hazard ratio; IBD, inflammatory bowel disease. The type of surgical procedure had the greatest impact on the risk for reoperation in the same surgical area with the highest reoperation rate after rectal surgery (26.0%, HR=9.48, 95% CI: 6.79–13.22), followed by uterus-sparing gynecological procedures (14.5%, HR=4.73, 95% CI: 3.37–6.64) and colonic surgery 14.1% (HR=5.44, 95% CI: 3.90–7.59). After open surgery 5773 of 50 751 (11.4%) patients underwent reoperation (HR=1.42, 95% CI: 1.32–1.54) compared with 966 of 21 519 (4.5%) after laparoscopic procedures. When mesh was placed during the initial surgery, reoperation was higher (14.4%, HR=2.48, 95% CI: 2.25–2.74) than after surgery without mesh placement (9.0%). The subgroup of malignant indications for initial surgery showed the intra-abdominal infection was retained in the multivariable analysis and IBD was excluded through backward selection in the multivariable analysis for the risk of reoperation in the same area (Tables I, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 and J, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ). The male sex had an increased risk in the malignant subgroup analysis compared with the female sex in the main analysis. In the laparoscopic subgroup, sex was excluded for multivariable analysis through backwards selection. All other variables were maintained in the multivariable subgroup analysis of the open versus laparoscopic approach (Tables K, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 and L, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ). However, endometriosis and radiotherapy did not impact the risk of reoperation in the same area in the open approach subgroup significantly. The predictive accuracy of the nomogram for reoperation in the same surgical area was good ( c -statistic=0.72) and remained the same after internal validation through bootstrapping ( c -statistic=0.72). Consequently, the optimism of less than 0.001 did not require adjustment of the original model. After univariable analysis on all variables (Table M, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ), intra-abdominal infection and endometriosis were excluded from multivariable analysis (Table N, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ) through backward selection. Operations involving the rectum and colon were performed in 25.7% (HR=10.67, 95% CI: 7.48–15.23) and 13.6% (HR=5.92, 95% CI: 4.15–8.44), respectively, imposing the greatest impact on the risk for reoperation after 30 days in the same surgical area. After surgery with the use of mesh, 13.7% of patients underwent reoperation (HR=3.01, 95% CI: 2.71–3.34). In the subgroup analysis of benign versus malignant indications for initial surgery, endometriosis posed no significant impact. Intra-abdominal infection was only significant in the malign indication subgroup (Tables O, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 and P, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ). In the laparoscopic subgroup sex was excluded for multivariable analysis through backward selection, in the open approach subgroup endometriosis and radiotherapy had no significant impact on the risk of reoperation in the same area after 30 days (Tables Q, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 and R, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ). For the subgroup of reoperation within 30 days after initial surgery, IBD, intra-abdominal infection, radiotherapy in history, and mesh placement had no significant impact on the risk of reoperation in the univariable analysis (Table S, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ) and therefore were not incorporated in the multivariable analysis (Table T, Supplemental Digital Content 2, http://links.lww.com/JS9/A323 ). The nomogram based on the multivariable analysis for predicting reoperation later than 30 days is seen in Figure 3 . The predictive accuracy for this nomogram was good ( c -statistic=0.73) and internal validation through bootstrapping demonstrated comparable accuracy ( c -statistic=0.73). With an optimism of 0.001, the original nomogram was retained. : Nomogram to calculate the risk for reoperation later than 30 days in the same area. By drawing a vertical line from each variable to the point axis on top and summing the individual points for all variables, the total score is calculated. From the total score axis perpendicular to the bottom, the linear predictor and the 2-year and 5-year risk for reoperation are determined. GI, gastrointestinal; IBD, inflammatory bowel disease.

An unrestricted research grant was provided by Nordic Pharma for this investigator initiated research.

The study is registered on Clinicaltrials.gov number NCT05516420 .

Masja K. Toneman and Richard P.G. Ten Broek.

Using a large national database, nomograms were constructed to predict the risk for reoperation in individual patients. The discriminative value of the prediction model for the risk of reoperation overall was considered moderate, whereas the prediction models for reoperation in the same surgical area were considered good. The risk factors with the highest impact on reoperations in all the models were mesh placement at initial surgery and operations involving the colon or rectum. Furthermore, an IBD diagnosis, younger age, malignancy, open surgical approach, and female sex increased the risk for reoperation. Epidemiological findings of this cohort demonstrated that approximately one in seven patients will undergo abdominal or pelvic reoperation within 5 years after initial abdominal surgery. In multivariable analysis, the rate of reoperation overall was one in five after laparoscopic surgery compared with almost one in three after open surgery. The rate of reoperation in the same surgical area was 30% lower after laparoscopic surgery compared with open procedures. Our study addresses an important knowledge gap in predicting the risk of abdominal reoperation in the general surgical population. Previous research on predicting the risk of reoperations are mainly based on short-term reoperations for complications of the initial operation 18 . However, the literature shows that the majority of reoperations are unplanned, and half of the reoperations are unrelated to the initial operation 2 . Also in our cohort, only one in five of reoperations could be identified as a possibly planned reoperation. These possibly planned reoperations were included in our study because we aimed to predict the overall incidence of reoperation for the individual patient. Moreover, it is difficult to ascertain if reoperation is planned or not based on the available administrative codes. Our nomograms provide an overall prediction for the risk of reoperation, the estimation of the risk for reoperation might be further tailored based on foreseeable and planned reoperation. In a recent survey among Dutch surgeons, only one in seven surgeons acknowledged the increased risk of adhesion-related complications in IBD patients. In our cohort, reoperation risk was highest for patients with IBD. As many as three out of 10 patients with IBD underwent reoperation within 5 years and one out of four patients had a reoperation in the same surgical area. This correlates with previous literature on IBD patients, reporting reoperation in almost 20% of patients within 30 months and 25% within 5 years after the first resection 19 – 21 . In a previous single-center cohort of general surgery, some risk factors for reoperation were confirmed, including female sex, younger age, and an open approach. In that study, a higher rate of reoperation was recorded (27%), possibly due to the relatively higher number of extensive operations performed in a tertiary referral setting, the low number of laparoscopies, and the inclusion of patients with several previous operations 2 . In females, the higher reoperation risk might be partly attributable to fertility-related operations. Notably, uterus-sparing gynecological operations had a higher impact on the reoperation risk than procedures with hysterectomy. The increased risk of reoperation following oncological surgery is explained by future operations for local recurrence or metastases and procedures involving (reversal of) the ileostomy or colostomy 2 . For younger patients, the increased risk for reoperation might be attributable to the fact that younger patients are more often deemed fit for surgery should a new condition be diagnosed for which surgery is indicated, whereas in older and frail patients surgery for relative indications might be more frequently avoided 2 . A major strength of this study is the building of the nomogram based on the national database (Scottish Medical Linkage Database, managed by the NHS), with validated data and low migration numbers allowing a population-based study in a broad surgical cohort 22 . The NHS database consists of all Scottish patients undergoing the first abdominal surgery during the study period, without the possibility of opt-out. The broad population on which the model was fitted also comes with some limitations, as some surgical indications and procedures might come with specific inherent risks that are not fully captured by multivariable analysis. However, we deliberately performed our analyses on a broad surgical cohort, because the risk of reoperation and subsequent adhesiolysis is relevant to the entire abdominal surgical population and our aim was to develop a model which is suitable for the majority of patients. The epidemiology of reoperations for a number of specific conditions such as IBD or malignancy have been described previously 19 , 21 , 23 . Another limitation is the lack of an external population for validation of the model. However, this cohort consists of nationwide unselected patients representative of the daily practice of surgery departments in a high-income country with high data density due to centralized reporting. Internal cross-validation using bootstrapping was performed, which demonstrated good internal validation with no need for correction of all original prediction models. Excluding patients with previous abdominal surgery potentially limits the generalizability of the model. Patients undergoing reoperation are part of daily clinical practice, as more than half of abdominal operation performed in current practice are reoperations 1 – 3 , 7 . The reasons for including only patients undergoing their first operation are both methodological and clinical. Methodologically, the risk of reoperation cannot be attributed to characteristics relating to the most recent operation if patients undergoing reoperation are included in the model. Clinically, patients undergoing their first operation seem most relevant as these patients are often unaware of the relatively high risk of future reoperation. Second, utilizing adhesion prevention strategies is conceptually most efficacious when applied directly at the first incidence of peritoneal trauma 4 , 24 . A specific limitation of excluding patients with prior surgical history is that more patients with malignancy at a later age and a history of previous surgery at a younger age might have been excluded. However, the number of older patients and patients with malignancy was still high. The effect of age and malignancy was also consistent throughout the subgroup analyses. Therefore, we consider the model appropriate also for elderly and oncological patients. For our assessments, we had to rely on the OPCS and ICD-10 codes available for each readmission. This type of data is not granular enough to score variables such as surgical experience, and prescribed medication or to differentiate accurately between elective and emergency surgery. Moreover, ideally, a model would be fitted on the outcome ‘reoperation requiring adhesiolysis.’ In clinical practice, the OPCS-10 and ICD-10 codes for adhesiolysis are often used only if adhesiolysis was the primary indication for surgery (e.g. in case of adhesive small bowel obstruction). Furthermore, adhesiolysis is often poorly documented even in operative reports 25 . Also in the SCAR update database, only a very small proportion of reoperations were coded as an adhesiolysis procedure. For this reason, nomograms were built for reoperation in the same surgical area in which the need for adhesiolysis is reportedly high even following laparoscopic surgery 6 . The corresponding nomogram had a good predictive ability. Our prediction model enables the identification of patients at increased risk of reoperations in support of clinical decision-making and informed consent. Although some risk factors are intuitively related to a higher risk of reoperation, surveys among surgeons, gynecologists, and general practitioners have demonstrated that the incidence of reoperation and the impact of risk factors for adhesion formation are often underestimated and poorly mentioned in the informed consent procedure 10 , 12 , 26 . If the adhesion-related risk is mentioned, it is often in patients with an increased risk of small bowel obstruction. The high incidence of abdominal reoperations indicates that these risks should also be discussed. In case of a planned reoperation, the use of adhesion barriers might be discussed. In these patients with an expected reoperation, adhesion barriers have been demonstrated to reduce operative time and potentially the risk of iatrogenic injury 27 . Accurate and relevant patient information is essential in the informed consent and shared decision-making process 28 . Many efforts have been made in the past decades to improve patient information regarding surgical procedures and the development of decision aids in particular 29 . Most patients are now well informed about the success rate and short-term complications of surgery. Information about long-term complications related to adhesions, such as reoperation risk but also small bowel obstruction, and chronic pain is often inadequate 10 , 12 . Strategies to minimize the extent of adhesion formation include minimally invasive surgery and the use of adhesion barriers. Laparoscopy has been shown to reduce the extent and incidence of adhesion formation to the previous incision site to 40% and any adhesions to 60%, while after open surgery the incidence is up to 80 and 90%, respectively 6 . Despite the decrease of adhesions and adhesion-related complications, the burden after laparoscopy remains high 13 . Adhesion barriers could further reduce adhesion formation and have been demonstrated to reduce operative time in two-stage procedures. Nevertheless, the search for an optimal barrier continues because of some important practical limitations of currently available barriers 27 , 30 . Recent surveys in the UK and The Netherlands have demonstrated that the use of adhesion barriers is rare 10 , 31 .

Not commissioned, externally peer-reviewed.

The constructed prediction model provides a prognosis of the individual risk for reoperation. Overall, in our cohort, one in seven patients had a reoperation within 5 years after abdominal surgery. The highest impact on the risk of reoperation is provided by the placement of mesh at initial surgery followed by the type of surgical procedure, with the highest risk after initial colorectal operations. Furthermore, previous radiotherapy, an IBD diagnosis, surgical area, malignancy as an indication for initial surgery, open approach, younger age, and female sex increased the risk for reoperation in all prediction models.

The authors declare that they have no financial conflict of interest with regard to the content of this report.

The data that support the findings of this study are available from the Scottish National Health Service (NHS). Restrictions apply to the availability of these data, which were used under license for this study. Data derived from the analysis are available from the authors with the permission of the Scottish NHS. The used repository was the Safe Haven ( https://shs.epcc.ed.ac.uk/2fa/scotnsh.html ) from the NHS.