The applicability of POSSUM and P-POSSUM scores as predictors of morbidity and mortality in colorectal surgery

Objective: to apply the POSSUM and P-POSSUM scores as a tool to predict morbidity and mortality in colorectal surgery. Methods: we conducted a prospective cohort study of 551 patients submitted to colorectal surgery in a colorectal surgery tertiary referral hospital in Brazil. We grouped patients into pre-established risk categories for comparison between expected and observed morbidity and mortality rates by the POSSUM and P-POSSUM scores. Results: in the POSSUM morbidity analysis, the overall expected morbidity was significantly higher than that observed (39.2% vs. 15.6%). The same occurred with patients grouped in categories II (28.9% x 10.5) and III (64.6% x 24.5%). In category I, the expected and observed morbidities were similar (13.7% x 9.1%). Regarding the evaluation of mortality, it was statistically higher than that observed in category III patients and in the total number of patients (11.3% vs. 5.6%). In categories I and II, we observed the same pattern of category III, but without statistical significance. When evaluating mortality by the P-POSSUM score, the overall expected and observed mortality was similar (5.8% x 5.6%). Of the 31 patients who died, 20.2% underwent emergency procedures and sepsis was the main cause of death. Conclusion: the P-POSSUM score was an accurate tool to predict mortality and could be safely used in this population profile, unlike the POSSUM score.


INTRODUCTION
P eriodic auditing of a surgical service is essential to perform a critical evaluation and to gain quality.For this purpose, scores are used to predict postoperative morbidity and mortality.The result of this audit allows for better individual risk prediction, therapeutic planning and resource allocation, and comparison among populations of different geographic areas, with lower risk of failure [1][2][3][4] .
Several predictors of morbidity and mortality are available (ASA, APACHE, SAPS II).However, the POSSUM (Physiologic and Operative Severity Score for the Study of Mortality and Morbidity) has been commonly recommended as appropriate for surgical practice 2,[4][5][6] .The POSSUM model, first described in 1991 by Copeland et al. 4 , originally used 62 variables, 48 physiological and 14 surgical.After multivariate analysis techniques, these numbers were reduced to 12 physiological variables and six surgical.The system seeks to predict morbidity and mortality in the first 30 postoperative days and allows comparing the results within the institution over time or performing a cross-sectional comparative analysis with other institutions 2,4 .In the evaluation of morbidity, predetermined postoperative complications are considered, subdivided into infectious (subcutaneous infection, abdominal abscess, anastomotic fistula, pneumonia and sepsis) and non-infectious (cardiac events and pulmonary thromboembolism) 4 .This method of outcome evaluation was applied to a large number of patients, but it was observed that the system overestimated mortality, especially for low risk patients.Therefore, the P-POSSUM score (Portsmouth Physiologic and Operative Severity Score for the Study of Mortality and Morbidity) was developed, which, despite using the same variables, is able to reduce the overestimation calculated by POSSUM 3,7 .The P-POSSUM is calculated by adding a regression equation to the POSSUM calculation.

Original Article A B S T R A C T
Objective: to apply the POSSUM and P-POSSUM scores as a tool to predict morbidity and mortality in colorectal surgery.Methods: we conducted a prospective cohort study of 551 patients submitted to colorectal surgery in a colorectal surgery tertiary referral hospital in Brazil.We grouped patients into pre-established risk categories for comparison between expected and observed morbidity and mortality rates by the POSSUM and P-POSSUM scores.Results: in the POSSUM morbidity analysis, the overall expected morbidity was significantly higher than that observed (39.2% vs. 15.6%).The same occurred with patients grouped in categories II (28.9% x 10.5) and III (64.6% x 24.5%).In category I, the expected and observed morbidities were similar (13.7% x 9.1%).Regarding the evaluation of mortality, it was statistically higher than that observed in category III patients and in the total number of patients (11.3% vs. 5  After the surgical procedure, the patient's POSSUM and P-POSSUM were calculated through a risk calculator using the mathematical equation of the POSSUM system, created in Microsoft Excel format.In the service database, we entered the demographic data, the data referring to the surgical procedure and the respective POSSUM and P-POSSUM values. For analysis of morbidity, we divided the 551 patients into three morbidity categories estimated by POSSUM, validated in previous studies 8 .For the analysis of mortality by POSSUM and P-POSSUM, we grouped the patients into three and four categories of estimated mortality, respectively (Table 1).

RESULTS
With respect to procedures' modality, 83% (457) of the patients were submitted to elective procedures and 17% to emergency ones.Two hundred and fifty-seven (46.6%) patients underwent laparoscopic procedures, with a conversion rate of 6.6% (Table 2).In the POSSUM morbidity analysis, the total expected morbidity was 39.2%, 13.7% being in category I, 28.9% in category II and 64.6% in category III.Table 3 shows that the observed morbidity was lower than the  When evaluating mortality by the P-POSSUM, the expected general mortality was not different from the observed one (5.8%vs 5.6%, p>0.05).Table 5 shows that in the separate analysis of each of the four separately, there was no difference between the expected and the observed mortality.

DISCUSSION
Postoperative morbidity and mortality rates are objective measures of outcomes that can be used to modify behavior and assess the quality of care.Previous studies have evidenced the important role of the POSSUM system as a predictor of postoperative morbidity and mortality 5,9,10 .A systematic review of the literature that evaluated POSSUM and its variants in patients undergoing surgery for colorectal cancer showed that the POSSUM score was able to predict morbidity in a reliable way and the P-POSSUM was the most accurate predictor of mortality, even when compared with CR-POSSUM 1 .In their prospective study, Chatterjee et al. 10  showing no mortality predictors specific to each disease.
In the present study, the POSSUM score was able to predict morbidity and mortality with accuracy only in patients with low risk of complications and mortality.
In patients with a high risk of morbidity and mortality, there was an overestimation, making the tool flawed and inadequate to evaluate this population profile.On the other hand, P-POSSUM was able to predict morbidity and mortality in a reliable manner when performed by category or general mortality, showing to be a safe and accurate index.With regard to deaths, the rates observed are in line with the data in the literature.Of the thirtyone deaths, the majority were secondary to emergency procedures and in patients who had high morbidity.The implementation of a results evaluation system to analyze the mortality rate as the P-POSSUM in our service was of great value, as it allowed risk stratification.Thus, it was possible to evaluate the results in a timely manner, clearly identifying the situations in need of resources to improve quality.In addition, it allowed the implementation of corrective measures in specific groups aimed at improving results, with targeted interventions.
The POSSUM and P-POSSUM scores were developed based on several surgical procedures.Although studies have shown the value of the POSSUM score in colorectal cancer surgery 1,12 , Tekkis et al. 13 developed the CR-POSSUM (Colorectal Possum) variation, which uses fewer parameters, facilitating calculation and decreasing variations.The CR-POSSUM evaluates the physiological variables age, cardiovascular system, systolic pressure, pulse rate, hemoglobin and urea, and the surgical variables peritoneal contamination, malignancy status and surgery size.A study comparing POSSUM and CR-POSSUM as predictors of mortality in 120 patients who underwent surgical resection for colorectal cancer showed that P-POSSUM, despite a non-significant overestimation of mortality in 25%, may be used to predict mortality.
However, when compared to CR-POSSUM, this variant was more accurate 14 .Two other prospective studies evaluated the use of the POSSUM score to predict morbidity and mortality in colorectal surgeries in 304 and 899 patients 15,16 .
In both studies, only CR-POSSUM was accurate in predicting mortality, while POSSUM overestimated morbidity and mortality.P-POSSUM overestimated mortality in the first study and underestimated the risk of dying after surgery in the other 16 .Corroborating with these findings, Constantinides et al. 17 , when evaluating the POSSUM score in 324 patients with complicated diverticulitis, found that CR-POSSUM was able to predict the results more reliable than P-POSSUM.
Our study has as limitations the selected sample.
Since it is a high complexity reference hospital, it deals with patients with more complex clinical conditions and, consequently, they are expected to have a greater potential for complications and associated deaths.
P-POSSUM proved to be an accurate tool for this high-risk population and it is not possible to safely extrapolate its use to low-risk patients.Data collection by more than one service surgeon was another limiting factor in the study, as it allows subjective data used in the calculation of the score, such as estimated blood loss, to be interpreted differently, leading to variations in the final score result.
The analysis of the population in a stratified manner, by age and by disease, although previously described without a statistical difference in diseasespecific morbidity and mortality 12 , could allow a thorough evaluation, especially with regard to morbidity, enabling the implementation of targeted interventions.
The evaluation of the P-POSSUM in a population of low risk is necessary to validate it as suitable for use in a generalized way in the patients submitted to colorectal surgeries in the institution.Regarding morbidity, further studies are needed to validate a score that accurately predicts morbidity in both low and high-risk patients.As seen previously seen [14][15][16] , the use of CR-POSSUM may be more adequate to the profile of the population studied, but other studies are necessary to validate its adoption.
The data presented show that the P-POSSUM score was able to accurately predict mortality, but the population evaluated was composed of high-risk patients.
This score can be safely used in this specific population, allowing for improvements in the postoperative routines and in performing internal audit with regard to mortality.

METHODSA
prospective study, conducted by the Coloproctology Clinic of the Hospital Felício Rocho-BH-MG -Brazil, followed Brazilian governmental standards for human research and was approved by the institution's ethics and research committee (CEP), under number 43647815.1.0000.5125.There were 551 patients undergoing colorectal surgeries enrolled between January 2011 and June 2014.The medical staff filled the protocols for clinical follow-up, specific for each disease, including data on postoperative morbidity and mortality, and later stored in the database maintained by the Coloproctology Clinic.
evaluated 50 patients with perforating peritonitis.The POSSUM score was a good indicator of postoperative outcome, but the limitations of the study were the sample size and the reliability evaluation of the use of the score in highrisk patients and emergency procedures, preventing its extrapolation to the low-risk population.Oomen et al. 11 compared the different POSSUM scores in 241 patients submitted to resections for sigmoid tumor or diverticular disease, and did not observe any difference in mortality,

Table 1 .
Categories of morbidity and mortality.
The overall mortality rate was 5.62%.Of the patients who died, 19 did after emergency surgery, and 12, after elective procedures, a mortality rate of 20.2% and 2.62%, respectively.Sepsis was the main cause of death in 67.7% of the patients.