Ethical statement

The study did not aim to change or modify existing clinical or laboratory practices. Steps were taken to ensure data collection was anonymous. The data collected was not used for making clinical decisions.

This study was reviewed and cleared by the Institutional ethics committee prior to the study as it was carried out as a thesis project for post-graduate degree certification.

Study design and setting

The study was a prospective observational cohort study carried out at the Department of Surgery, University College of Medical Sciences, and GTB Hospital, Delhi between December 2018 and March 2020. The hospital is a 1800-bedded multi-specialty tertiary care facility with facilities for acute and emergency surgery and a dedicated surgical ICU. Approximately 600–800 surgeries for gastrointestinal emergencies are carried out per year with the bulk being made up of secondary peritonitis followed by bowel obstruction.

The intention behind the study was to evaluate three different risk scoring systems in current use by surgical teams at the institution and affiliated hospitals for the purposes of triage, resource allocation, and prognostication.

The study has been written in conformation with STROBE Guidelines for cohort studies.

Inclusion and exclusion criteria

All consecutive patients of non-traumatic secondary peritonitis undergoing emergency laparotomy by three different general surgical units in this period were included in the study. Secondary Peritonitis was defined as an intra-abdominal infection that extended beyond the organ of origin and caused either a localized or diffuse inflammation of the peritoneum with soiling of the peritoneal cavity.

The exclusions were as follows:

Patients with traumatic perforations due to blunt or penetrating trauma.

Postoperative peritonitis due to leaks who had undergone index surgery elsewhere.

Those who could not be taken up for surgery either due to lack of consent or preoperative death were excluded from the study.

Outcome measures

The primary outcome measure was postoperative mortality (either in a hospital or within 90 days of the procedure if discharged) or survival.

Treatment

Every patient followed the same standard pathway using the Surviving Sepsis Guidelines (Rhodes et al. 2016). Clinical and biochemical assessment was carried out to determine and classify the presence of sepsis, septic shock, and organ dysfunction according to internationally accepted criteria (Sartelli et al. 2014; Sartelli et al. 2012; Rhodes et al. 2017). After confirmation of diagnosis and adequate resuscitation, patients were taken up for exploratory laparotomy after pre-anesthetic assessment by the anesthesia team on duty. The procedure performed was decided by the operating surgeon, either the consultant on duty or senior resident after a discussion with the consultants.

Data collection

The clinical findings were recorded from hospital preoperative notes, operative notes, anesthetic charts, and postoperative ward notes. After the initial registry, patients were followed till the end of their stay in the hospital (discharge or mortality). For patients who were discharged to home, follow-up visits occurred at 7 days, 28 days, and 3 months.

The data collected was of the following types:

1.

Preoperative data including demographic data, co-morbid history, examination findings, laboratory investigations, and radiological findings.

2.

Intraoperative findings, i.e., degree of contamination, etiology of perforation, source of contamination, intraoperative blood loss, method of abdomen closure, and need for blood transfusion.

3.

Postoperative course including the need for ICU stay, course of disease, and any postoperative complications.

The final etiology was defined by intraoperative findings, histopathological, and microbiological examination.

Postoperative mortality was defined as intrahospital death or death within 90 days of the index procedure.

To reduce bias, all consecutive patients with secondary peritonitis who underwent laparotomy were included in the study. All the data points required for the calculation of scores were collected from patient records and verified by two different investigators.

Statistical analysis and scoring systems

The scoring systems to be evaluated were chosen after taking a survey of multiple surgeons at our institution and affiliated hospitals about the risk scoring systems being used commonly to prognosticate and triage patients. The scoring systems used were p-POSSUM (Portsmouth Modification to Physiological and Operative Severity Score for Enumeration of Mortality) (Prytherch et al. 1998), MPI (Mannheim Peritonitis Index) (Linder et al. 1987), and Jabalpur Peritonitis Index (JPI) (Mishra et al. 2003).

Using the patient data and variables, risk scores for every patient, under each of the three systems to be assessed were calculated. Receiver operator characteristic (ROC) curves were constructed for sensitivity analysis for each of the 3 risk-scoring systems. These ROCs are used to determine diagnostic performance and compare the three scores based on the area under the curve (AUC) (Soreide 2009). The receiver operator characteristic curve was also used to define a cutoff score, using the Youden Index (Safari et al. 2016) beyond which patients were considered to be high risk (Safari et al. 2016). Based on the stratification of patients into high- or low-risk populations and the mortality rate in these, sensitivity, specificity, positive predictive value, and negative predictive value were calculated.

After cutoff scores were calculated using ROC, further calibration of scores was done using chi-square test for observed to expected mortality rates (Oliver et.al 2015) to ensure the applicability of results.

p value of < 0.05 was considered significant.

Scoring systems

Both p-POSSUM and MPI are commonly used systems that have been reported to have high accuracy based on the area under the curve (AUROC) in receiver operator characteristic curves (González-Pérez et al. 2019; Neary et al. 2007; Scott et al. 2014) with AUROC greater than 80% indicating good diagnostic ability (Safari et al. 2016; Hanczar et al. 2010). The Jabalpur peritonitis index is easy to use with few components and perhaps more suited to Indian populations as the original patient cohort was based in India. Due to its simplicity and low number of variables, it is also commonly used in low-resource settings where extensive preoperative workup may not always be feasible.

p-POSSUM

p-POSSUM, standing for Portsmouth Modification to Physiological and Operative Severity Score for the Enumeration of Mortality was devised by Prytherch et al. (1998). The system uses a 12-factor physiological score for patient condition prior to surgery and a 6-factor operative severity score, both of which were derived from earlier observations on 1372 patients (Copeland et al. 1991). The physiological and operative scores are used to give a predicted percentage risk of mortality for a patient by calculating via the p-POSSUM equation as follows:

$$\mathrm[\mathrm/ (1-\mathrm)]\hspace=-9.065+(0.1692\hspace\times \hspace\mathrm)\hspace+\hspace(0.1550\hspace\times \hspace\mathrm)$$

where R is the predicted risk of mortality.

Mannheim Peritonitis Index (MPI)

Based on the clinical observations and risk factors from 1243 patients of purulent peritonitis, Linder et al. (1987) devised the Mannheim Peritonitis Index for predicting mortality in patients of perforation peritonitis. A total of 8 factors are included in the scoring system covering demographic, physiological, and disease-specific factors. The total score possible is 47. In the original study, with a cutoff value of > 26, MPI helped in identifying patients at increased risk of mortality with good sensitivity ( 84%), specificity (79%), and overall accuracy (81%).

Jabalpur Peritonitis Index

Mishra et al. (2003) devised the Jabalpur peritonitis index for perforation peritonitis as a simplified system for use in resource-poor situations where extensive preoperative investigations may not be available. One hundred forty patients were studied prospectively, and multiple regression analysis was employed to identify 6 factors which had a high association with mortality. Using 9 as a cutoff value, beyond which 50% mortality was observed, the authors determined the system to have a sensitivity of 87% and specificity of 85%.