Article Text
PDF
Abstract
Objectives This study aims to determine hospital variation and intensive care unit characteristics associated with failure to rescue after abdominal surgery in Norway.
Design A nationwide retrospective observational study.
Setting All 52 hospitals in Norway performing elective and acute abdominal surgery.
Participants All 598 736 patients undergoing emergency and elective abdominal surgery from 2011 to 2021.
Primary outcome measure Primary outcome was failure to rescue within 30 days (FTR30), defined as in-hospital or out-of-hospital death within 30 days of a surgical patient who developed at least one complication within 30 days of the surgery (FTR30). Other outcome variables were surgical complications and hospital FTR30 variation. Statistical analysis was conducted separately for general surgery and abdominal surgery.
Results The 30-day postoperative complication rate was 30.7 (183 560 of 598 736 surgeries). Of general surgical complications (n=25 775), circulatory collapse (n=6127, 23%), cardiac arrhythmia (n=5646, 21%) and surgical infections (n=4334, 16 %) were most common and 1507 (5.8 %) patients were reoperated within 30 days. One thousand seven hundred and forty patients had FTR30 (6.7 %). The severity of complications was strongly associated with FTR30. In multivariate analysis of general surgery, adjusted for patient characteristics, only the year of surgery was associated with FTR30, with an estimated linear trend of −0.31 percentage units per year (95% CI (−0.48 to –0.15)). The driving distance from local hospitals to the nearest referral intensive care unit was not associated with FTR30. Over the last decade, FTR30 rates have varied significantly among similar hospitals.
Conclusions Hospital factors cannot explain Norwegian hospitals’ significant FTR variance when adjusting for patient characteristics. The national FTR30 measure has dropped around 30% without a corresponding fall in surgical complications. No association was seen between rural hospital location and FTR30. Policy-makers must address microsystem issues causing high FTR30 in hospitals.
- gastrointestinal tumours
- health services administration & management
- quality in health care
Data availability statement
Data may be obtained from a third party and are not publicly available. Data used in this study cannot be made publicly available due to the conditions of the data-sharing agreement.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
The strength of the study is the completeness of the data, covering all of Norway’s abdominal surgery for 10 years.
The ability to link two national population registries using a personal personal identification number code is unique.
The reduction of FTR30 by approximately one-third may be linked to the national implementation of the WHO Surgical Safety Checklist and rapid response teams.
Data validity and reliability must be considered.
Introduction
Failure to rescue (FTR) is defined as postoperative death following a surgical complication, and the FTR rate is an internationally accepted quality indicator.1–7 FTR reflects the ability of a hospital to provide appropriate care to surgical patients after a complication. Notably, the principles of FTR go beyond recognising postoperative mortality to understanding the sentinel events that precede it.8 Macrosystem causes of FTR are related to hospital characteristics like hospital bed size, decreasing nurse–patient ratio, university hospitals and small intensive care units (ICUs).9 10 Microsystem causes of FTR are related to dysfunctional communication, teamwork and safety culture.11
FTR is a significant driver of variations in postoperative mortality.1 We have shown that Norwegian hospitals varied in readmission rate, 30-day mortality and length of stay.12 Specific procedures are associated with high-risk and high-volume FTR trajectories, indicating national variability in FTR.13 Nevertheless, knowledge gaps exist. Most FTR evidence comes from the USA, where healthcare has a different structure than in Europe.5 6 14 15 Organisation of care is a significant predictor of FTR, and a higher number of healthcare vendors increase fragmentation of care.9 16 Thus, differences in healthcare organisation may impact FTR, implying more research from a European hospital setting.16 Others acknowledge this concern, and the UK Perioperative Quality Improvement Programme studies risk-adjusted complications and death following major surgery.17 Finally, a rural hospital localisation, such as many Norwegian hospitals, could affect the ICU’s capacity to manage critical postoperative problems that need transfer to a specialised referral ICU.18–20 The existence of these knowledge gaps necessitates more investigation.
Thus, the primary objective of this study was to assess at a national level hospital and ICU characteristics associated with variations in FTR after abdominal surgery. Second, we wanted to explore temporal trends of FTR and to identify hospital outliers significant for future quality improvement initiatives.
Methods
This study is consistent with the Strengthening the Reporting of Observational Studies in Epidemiology statement.
Patient population and data source
We conducted the study by combining two national registries:
The Norwegian Patient Registry (NPR) contains patient administrative data from all Norwegian hospitals from 2010 to 2021. The data contain information related to emergency or elective admission, primary and secondary diagnosis codes according to the Norwegian version of International Classification of Diseases 10th revision(ICD-10), surgical and medical procedures, age, sex, Charlson Comorbidity Index, date and time of ward admission and discharge, and procedures for all department stays. We coded the surgical procedures according to the Norwegian version of the Nordic Medico-Statistical Committee Classification of Surgical Procedure (NCSP-N). Data on abdominal surgical procedures and complications were extracted (online supplemental tables 1–3).
Supplemental material
The National Population Register contains data for all Norwegian citizens, that is, addresses and movement, spouses, divorces and next of kin, children, and information related to death (date of death, place, examination at death).
The National Population Register was linked to the NPR, using every citizen’s personal identification number (PIN). Procedures with a missing PIN, admission type, or vital status or a recorded date of death more than 24 hours before department admission were excluded. Procedures on patients aged 18 years or older whose hospital episode spanned at least from one date to the next were included (ie, day-case surgery was not included). Consecutive hospital stays within 8 hours were grouped into hospital episodes. Both inpatient and outpatient care were included. NPR data from 2010 were only used for the computation of the Charlson index. Due to errors in identifying some hospitals in NPR for 2016, this year was excluded from the analyses.
The surgeries were either elective or emergency gastrointestinal, gynaecological, urological or vascular procedures, denoted as surgery class in the following (online supplemental table 1). The admission years 2011–2021, except 2016, were used for descriptive analyses. To allow for sufficiently long patient histories prior to surgery, only the years 2013–2021 were included in the statistical multivariate modelling. Date of death was unavailable for the last quarter of 2021. Accordingly, this period was excluded from the analysis datasets.
Two datasets were extracted for the statistical analyses (online supplemental figure 1. Flow chart): abdominal surgeries and general surgeries with a complication. This distinction was made to avoid confounders caused by patient and surgical heterogeneity between rural, regional and university hospitals. For instance, advanced cancer surgeries are centralised to university hospitals, whereas general surgical procedures are performed at most Norwegian hospitals.
Thus, the data sets were as follows:
Dataset I: General surgical procedures with a corresponding complication: Hernia repair (JAB10, JAB20, JAB30, JAB11) Laparotomy (JAH00, JAH01), Appendectomy (JEA00, JEA0), Right colectomy (JFB30, JFB31) and Cholecystectomy (JKA11).
Dataset II: Abdominal surgical procedures with a corresponding complication.
Dataset I and the corresponding statistical analyses are reported in the main manuscript, whereas the analyses from dataset II are reported in online supplemental file.
Dependent and predictor variables
Patient demographics, such as age, sex and Charlson Comorbidity index, were gathered from NPR data. The comorbidities used to calculate the Charlson Comorbidity Index were those included in the Quan et al’s ICD-10 version of the Charlson Comorbidity Index.21 22 In the multivariate analysis, the Charlson index was truncated at a value of 12. The Charlson Comorbidity Index was determined from previous admissions 3 years before, but not including the current episode of care. The number of hospital episodes in the year before the operation was also calculated.21
The annual number of ICU admissions (ie, number of annual ICU stays in days) and SAPS II score (Simplified Acute Physiology Score) were extracted from the National Intensive Care Registry for 2020 and 2021 (https://helse-bergen.no/norsk-intensivregister-nir). The Rural ICU Index was defined as the driving distance from local and regional hospitals to the nearest referral ICU and was calculated using Google Maps. According to the organisational hospital structure, hospitals were divided into four groups: day-care hospitals (D, outpatient surgery), local— rural hospitals (L), regional hospitals (R, main referral hospital in the healthcare trust) and university hospitals (U). (online supplemental tables 4 and 5).
Complications were classified according to Clavien-Dindo scales I–IV, which grades the severity of the complication.23 Postoperative complications were defined in two ways: (1) from diagnosis codes, as shown in online supplemental table 2, identified in the same hospital episode as the surgery or within 30 days after the index surgery and (2) from second surgeries, defined as any reoperation within 30 days of the index surgery, defined by the NCSP-N codes JWW96, JWC00, JWF00, JWA00, JWE00, JWB00, JWC01, JWE01, JWE02, JWF01, JWW97 and JWW98 (online supplemental table 3). It was impossible to determine whether a diagnosis code signified a condition present before surgery or arising after.
Our final study populations included all surgical procedures with at least one reported complication or reoperation (online supplemental figure 1). If a patient had more than one complication, the complication with the highest Clavien-Dindo score was chosen.
Outcome variables
The primary outcome was the 30-day mortality rate after an emergency or elective surgical procedure, referred to as FTR30. FTR30 was defined as death inside or outside the hospital, irrespective of cause, in patients with at least one complication within 30 days of the surgical intervention. Originally, FTR was defined as death during hospitalisation of surgical patients who developed at least one complication during the hospital stay. We used a revised definition of FTR as 30-day mortality to avoid the bias of in-hospital mortality due to variation in length of stay.24
Statistical analyses
In the tables comparing hospital types, χ2 tests were used to test for homogeneity. To adjust for multiple comparisons, the false discovery rate (FDR) was computed.25 26 The analysis was performed in two stages: first, a patient-level model was developed to adjust the FTR30 rates for individual risk factors. Second, a model was built to evaluate the association of hospital characteristics with the risk- adjusted FTR30 rates.
Risk adjustment model
Using stepwise logistic regression with the Bayesian information criterion (BIC), models for predicting individual risk of FTR30 were built. The candidate explanatory variables were patient age, sex, Charlson Comorbidity Index, the number of previous admissions 1 year before index admission, Clavien-Dindo grade of complication, surgery class, whether the admission was elective or emergency, hospital characteristics and admission year. The latter two were always kept in the models. Two-way interactions between patient variables were also included as candidate variables. Both square root and spline functions of the Charlson index were entered. This model neglects the between-hospital and between-year variation and will, therefore, not give valid inferences. Therefore, the final model was modified to include indicator variables for each hospital-year combination instead of hospital characteristics and year, and re-estimated. For each combination of hospital and year, the mean predicted probability of FTR30, based on all the patients in the dataset, was computed. The procedure was done separately for datasets I and II. For each predictor variable, ORs with 95% CIs were reported.
Hospital characteristics model
The variables ICU admissions, SAPS II score and The Rural ICU Index were grouped in quintiles for the final analysis. This was necessary to avoid a singular model, as these variables are highly interdependent. Given the dependencies and the small sample size of the data set, they were entered as numerical variables (scores 1–5) in the model. The hospital categories R (regional) and U (university) were merged, and denoted RU. A linear regression model was fitted to the hospital-year data, with adjusted FTR30 rate as the dependent variable, a linear term in year and the hospital characteristics: hospital type, quintiles of ICU admissions, SAPS II score and The Rural ICU Index, as explanatory variables. Stepwise regression, with the BIC criterion, was used to search for any two-way interactions, while keeping year and the hospital characteristics in the model. Estimated linear regression coefficients were reported for all explanatory variables, with 95% CIs. To describe the variation between hospitals and identify outliers, FTR30 rates were averaged over year per hospital and displayed graphically.
All data preprocessing and statistical analyses were performed in R V.3.5.1 and V.4.0.2 (The R Foundation, Vienna, Austria).
Results
Demographics
Data from 52 hospitals in the time 2011–2021 (except 2016) were included in the descriptive analyses. This consisted of 422 126 patients, undergoing 598 736 abdominal surgical procedures (online supplemental figure 1).
In general surgery, there was a significant difference between hospital types in the proportion of emergency cases (FDR=0.0006). Most patients were admitted as emergency cases at local (n=6726, 53.5%) and university hospitals (n=5260, 55.2%). University hospitals had younger patients (>18–60 years, n=3506, 36.8%), compared with regional hospitals (>18–60 years, n=1019, 28.4%, FDR=0.0006). Similarly, university hospitals had the highest proportion of patients with a Charlson Comorbidity Index>1 (n=4337, 45.5%, FDR=0.0006) (table 1). A similar demographic profile existed for abdominal surgeries (online supplemental table 6).
Descriptive statistics general surgery
The ICUs were classified according to three different characteristics, that is, annual ICU volume, SAPS II score and the Rural ICU Index. The highest annual ICU volume was 914 patient admitted more than 24 hours, and the lowest was 50 admittances more than 24 hours. The maximum median SAPS II score was 48, whereas the lowest was 26. The highest Rural ICU index was 787 km (online supplemental table 5).
Complications and reoperations
Of the 598 736 surgeries, there were 183 560 (30.7%) surgical procedures with a postoperative complication or reoperation (online supplemental figure 1).
In dataset I, for general surgery performed at university hospitals, circulatory collapse was most frequent (n=2412, 25.3%) among those having a complication, whereas cardiac arrhythmia was most frequent at local hospitals (n=3082, 24.5%) (table 2). In general surgery, 1507 out of 25 775 (5.8%) patients were reoperated within 30 days (online supplemental table 7). Reoperations for deep abdominal abscesses were more frequent at the university hospitals (135 out of 601, 22.46%, FDR=0.057) (online supplemental table 7).
Thirty-day complications within general surgery
Failure to rescue
For general surgery, the overall FTR30 number was 1740 (1+809+285+645, table 1) and the FTR30 rate was higher at university hospitals (6.8%) compared with local hospitals (6.4%) with FDR of 0.002 (table 1). In 2011–2021, there was no decrease in postoperative complications graded as Clavien-Dindo>2 (online supplemental table 8). However, there was a considerable reduction in FTR30 the last 10 years (figure 1A,B and online supplemental table 8). Additionally, in the hospital-year multivariate analysis for general surgery, the year of surgery was significantly associated with an absolute reduction of FTR30 estimated at −0.31 percentage points per year (95% CI (−0.48 to –0.15)), corresponding to a relative FTR30 reduction of approximately one-third over the 10 years (table 3).
(A) FTR30 within general surgery Norway (%), risk adjusted for patient characteristics. The years, 2011 and 2016, are not included in the figure. The FTR30 was for general surgery in 2011 8.6% and in 2021 5.2%. (B) FTR30 within abdominal surgery Norway (%), risk adjusted for patient and characteristics. The years, 2011 and 2016, are not included in the figure. The FTR30 was for abdominal surgery in 2011 5.9% and in 2021 4.3%. FTR30, failure to rescue within 30 days.
Logistic and linear regression models for FTR30 within general surgery
Within general surgery, we found that the more severe complications (Clavien-Dindo 3 and 4) were highly associated with FTR30: the OR for Clavien-Dindo grade 4 with grade one as reference was 25.83 (95% CI (11.93 to 55.92)) on the patient level. Similarly, emergency surgery was highly associated with FTR30 (OR 4.42 with elective surgery as reference, (95% CI (3.84 to 5.08)) (table 3).
In the hospital year level multivariate model, after adjusting for patient characteristics, annual ICU volume, SAPS II score and Rural ICU Index were not associated with the FTR30 rate. For abdominal surgery, year of surgery and SAPS II score contributed significantly to the model, with an estimate of −0.154% (95% CI −0.229% to –0.078%) and 0.277% (95% CI 0.092% to 0.463%), respectively. Importantly, we found no association between local hospitals, the Rural ICU index and FTR30 (table 3, online supplemental table 9).
Lastly, we assessed FTR30 variation between Norwegian hospitals. There exists a significant FTR30 variation, where comparable hospitals have a considerable difference in rate of FTR30, when adjusted for patient characteristics (figure 2A,B, online supplemental figure 2). Some local and regional hospitals have a FTR30 rate above 10%, whereas comparable hospitals have a FTR30 rate around 4%. The between-hospital SD, after adjusting for patient and hospital characteristics was 1.49%.
(A) FTR30 (%) at Norwegian hospitals (general surgery), risk adjusted for patient characteristics. X-axis: % of FTR30. (B) FTR30 (%) at Norwegian hospitals (abdominal surgery), risk adjusted for patient characteristics. X-axis: % of FTR30. FTR30, failure to rescue within 30 days.
Discussion
In Norway, there has been an estimated relative decrease of 31% in FTR30 the last 10 years. However, it is worth noting that the rate of major postoperative complications (namely, those classified as Clavien-Dindo>2) has not shown a corresponding fall (figure 1A,B, online supplemental table 8).
There is a significant difference in FTR30 across similar Norwegian hospitals, especially for commonly performed surgical operations such as appendectomies, cholecystectomies, laparotomies and colectomies (defined as general surgery) (figure 2A,B). In the multivariate analyses, adjusting for patient characteristics, we found no signs of a higher FTR30 rate at local hospitals compared with regional and university hospitals. This finding indicates that the level of safety experienced by surgical patients in rural areas is comparable to that seen in metropolitan university hospitals.
Many organisational characteristics are linked to FTR, for instance, an association between fragmentation of care, which is characterised by readmission to a hospital different from the initial hospitalisation.9 18 27 28 The findings of our study suggest that the rural ICU index did not serve as a significant predictor of FTR30, implying that the distance required for transferring patients between university, regional and local hospitals in rural areas did not have a discernible influence on FTR30.
Substantial international variation is seen in the rates of surgical reoperation. Patients who return to the operating theatre in the postoperative period experience some of the worst preventable complications. Our study has observed a reoperation rate of 5.8% for general surgical procedures, which is higher than research conducted in the United Kingdom and New Zealand (online supplemental table 7).1 29
Norway introduced the nationwide Safe Surgery Saves Lives initiative in 2010 to enhance the patient safety culture (https://www.itryggehender24-7.no/). They implemented the WHO Surgical Safety Checklist as a standard of care in 2011, and ‘Rapid Response Teams’ were implemented at all Norwegian Hospitals in 2017.30–32 A significant decrease of around 30% in FTR30 has been shown in this study, notwithstanding the absence of any reduction in the incidence of serious complications (Clavien-Dindo>2) (online supplemental table 8). The potential impact of these nationwide quality improvement activities on the reduction of FTR should be acknowledged.
Communication, teamwork and safety culture are key domains contributing to the timely recognition and effective management of surgical complications.5 6 11 Towards this, an emerging literature shows that hospital resources, attitudes and behaviours (microsystem factors) are characteristics fundamental for a deeper understanding of the sentinel events associated with FTR. In our opinion, hospital differences in safety culture are an essential driver contributing to the significant Norwegian FTR30 variability.8 33 34
There exist limitations to this study; most importantly, this work does not necessarily imply causation, and our results must be interpreted with care. As previously indicated, there have been efforts to implement national quality-enhancing measures. However, the impact of these actions on the FTR30 rate is undecided. The generalisability might be limited, as few other countries have publicly financed healthcare such as National Health Service and the Scandinavian model.
Second, data validity and reliability must be considered. NPR has a high degree of data completeness when comparing diagnoses and surgical procedures to quality registries.35 According to recent research, the Global Trigger Tool (GTT) method resulted in a sensitivity of 76% and a specificity of 65% for the identification of postoperative complications. These findings highlight GTT’s potential to enhance the accuracy of complication detection.36 The authors of this study recognise the issue above as a potential limitation in identifying surgical complications. Nevertheless, our study revealed a complication rate of 30.6%, consistent with the data reported by Storesund (30.3%).36
In the case of multiple complications, we included only the most severe (according to Clavien-Dindo grade) in the analysis. Thus, the increased severity of multiple organ failure was not accounted for. A more detailed analysis would also necessitate consideration of the causal relationship between complications. The task would have posed challenges due to the limited availability of temporal data on complications in our data source.
Lastly, there is a potential risk of misclassification bias associated with the definition of ‘FTR’. Patients may be erroneously categorised as FTR when they present with a condition requiring a critical surgical intervention for survival. Nevertheless, one may argue that top-tier medical facilities exhibit streamlined protocols and collaborative efforts, decreasing death rates after complications. Further investigation is warranted, particularly regarding the definition and classification of FTR.
The strength of the study is the completeness of the data, covering all of Norway’s abdominal surgery for 10 years. The ability to link population registries using a personal PIN is unique, and the accurate information related to date of death, made it possible to provide an accurate overview of FTR at a national basis.
Data availability statement
Data may be obtained from a third party and are not publicly available. Data used in this study cannot be made publicly available due to the conditions of the data-sharing agreement.
Ethics statements
Patient consent for publication
Ethics approval
The Norwegian Directorate of Health and the Norwegian Data Protection Authority approved this retrospective observational study.The Norwegian Institute of Public Health has been given comprehensive permission by the Norwegian Directorate of Health to investigate the Norwegian Patient Registry (NPR) and the National Population Register data and is not requesting a reference number or ID for ethical clearance. Patient consent for publication is not required.
Acknowledgments
We would like to thank the Department of Research and Innovation at the University Hospital North Norway and the head of the department Dr Ingvild Pettersen, for administrative support. We would like to thank Anne Marie Grenersen for budget organising support. This work uses data provided by Norwegian patients, and collected by Norwegian health care workers, being part of the mandatory data reporting system within Norway.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @KnutAugestad
Contributors AKL: contributed to data analysis and interpretation—critical review of the final manuscript. KDS: contributed to data preparation and statistical analysis—critical review of the final manuscript. KMA: secured funding. Study conception contributed to data preparation, analysis and interpretation and wrote the draft manuscript. Critical review of the final manuscript. JH: study conception. Data preparation and statistical analysis, critical review of final manuscript. KMA is guarantor and accepts full responsibility for the work and the conduct of the study, had acess to the data, and controlled the descision to publish.
Funding This study was financed by a Helse Nord Health Care Trust research grant: HARM Score project HST1245-15.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.