P-POSSUM Calculator: Surgical Mortality and Morbidity Risk

The Portsmouth Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity. Grade 12 physiological and 6 operative factors on the 1, 2, 4, 8 scale, and the published Prytherch equations convert the two totals into a predicted mortality and a predicted morbidity percentage. Built for surgical audit and education, not for deciding who should have an operation.

What P-POSSUM is

Every operation carries risk, but the risk is not the same for every patient or every procedure. A fit 40-year-old having a small elective operation and a frail 80-year-old having emergency abdominal surgery at midnight are not comparable cases, and any honest comparison of surgical outcomes has to account for that difference. Without risk adjustment, a surgeon who takes on the hardest cases looks worse than one who avoids them, which is exactly the wrong incentive. P-POSSUM exists to make the comparison fair.

The story starts with POSSUM, the Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity, published by Copeland, Jones and Walters in the British Journal of Surgery in 1991 (1991;78:355-360). They proposed 18 factors, 12 describing the patient's physiological state and 6 describing the operative insult, each graded on an exponential 1, 2, 4, 8 scale. The two summed scores were fed into logistic regression equations that predicted the probability of death and the probability of a postoperative complication. It was a genuine advance: one number for how sick the patient is, one number for how big the operation is, and a principled way to combine them.

But POSSUM had a flaw that showed up as it was used more widely. Its mortality equation overpredicted death, by more than twofold across all patients and by more than sevenfold in the lowest-risk group, those with a predicted risk of 5 percent or less. In other words, it was most wrong exactly where most patients sit. Prytherch and colleagues at Portsmouth set out to fix this. They collected POSSUM data prospectively on 10,000 general surgical episodes between August 1993 and November 1995, fitted a corrected mortality equation on the first 2,500 cases, and then validated it prospectively on the remaining 7,500, arranged chronologically in five groups of 1,500. The corrected equation matched observed mortality closely. That equation, published in the British Journal of Surgery in 1998 (1998;85:1217-1220), is P-POSSUM: the Portsmouth modification of POSSUM. The 18 factors and their grades are unchanged; only the mortality equation is new. The morbidity equation is the original POSSUM one, which Prytherch and colleagues left alone.

The 18 factors and the 1, 2, 4, 8 scale

The grading scale is deliberately exponential rather than linear. A factor graded 8 contributes eight times as much to its subtotal as a factor graded 1, so severe derangement of a single system moves the total far more than mild derangement of several. This matches clinical reality: a potassium of 6.2 or a systolic pressure of 80 matters more than a handful of borderline values. Note that not every factor uses the full scale. Age, white cell count and ECG have no grade 8, and multiple procedures and mode of surgery have neither grade 8 nor the intermediate steps their neighbours use: more than two procedures scores 4, and immediate emergency surgery scores 4. The table below gives every factor and every grade boundary, exactly as published.

Two details in the table deserve emphasis because they are easy to get wrong. First, the cardiac and respiratory factors each combine two rows, the clinical signs and the chest X-ray, and the published rule is to take the worse of the two. A patient with no cardiac failure but frank cardiomegaly on the film scores 8 for that factor, not 1. Second, the operative maximum is 40, not 48: it is 8 for operative severity, 4 for multiple procedures, 8 for blood loss, 8 for peritoneal soiling, 8 for malignancy and 4 for mode of surgery. The physiological maximum is 96, twelve factors at grade 8 each.

The two equations

Once the two totals are known, the equations convert them into probabilities. The mortality equation, from Prytherch and colleagues, is:

ln(R / (1 - R)) = -9.065 + 0.1692 x physiological score + 0.1550 x operative score

The morbidity equation, unchanged from the original POSSUM, is:

ln(R / (1 - R)) = -5.91 + 0.16 x physiological score + 0.19 x operative score

Here R is the predicted probability, and the left-hand side is the log-odds, also called the logit. To turn the linear predictor, call it x, into a probability, use the logistic function: R = 1 / (1 + e^(-x)). The calculator on this page does that conversion for you and shows the two percentages.

It helps to see the arithmetic once. Take the lowest possible scores: physiological 12 and operative 6. The mortality linear predictor is -9.065 + 0.1692 x 12 + 0.1550 x 6, which is -9.065 + 2.0304 + 0.9300 = -6.1046. Then R = 1 / (1 + e^(6.1046)) = 0.00223, or about 0.22 percent. That is the published minimum predicted mortality for P-POSSUM, and it is a useful sanity check: any correct implementation must produce 0.22 percent for the minimum scores. The morbidity linear predictor for the same scores is -5.91 + 0.16 x 12 + 0.19 x 6 = -2.85, giving R = 1 / (1 + e^(2.85)) = 0.0547, or about 5.47 percent. The chart below shows how the predicted mortality climbs as the physiological score rises, at three fixed operative scores, with every point computed from the published equation.

Worked examples

Three worked examples show how the grades, the totals and the equations fit together. Each has been computed by hand and can be checked against the calculator above.

Example one, low risk. A 55-year-old woman is booked for a minor elective procedure. She has no cardiac or respiratory history, blood pressure 120, pulse 72, Glasgow Coma Score 15, haemoglobin 14.0, white cell count 7.0, urea 5.0, sodium 140, potassium 4.2, and a normal ECG. Every physiological factor grades 1, so the physiological score is 12. The operation is minor and elective, one procedure, blood loss under 100 ml, no peritoneal soiling, no malignancy: every operative factor grades 1, so the operative score is 6. Predicted mortality is 0.22 percent and predicted morbidity is 5.47 percent, the published minimums.

Example two, moderate risk. A 68-year-old man needs a major bowel resection for a primary tumour, done electively. He takes antihypertensive therapy (cardiac grade 2), has dyspnoea on exertion (respiratory grade 2), blood pressure 145 (grade 2), pulse 92 (grade 2), Glasgow Coma Score 15 (grade 1), haemoglobin 12.0 (grade 2), white cell count 12.0 (grade 2), urea 8.0 (grade 2), sodium 138 (grade 1), potassium 4.0 (grade 1), and a normal ECG (grade 1). The physiological grades sum to 22. The operation is major (4), a single procedure (1), blood loss 300 ml (2), minor serous soiling (2), primary malignancy only (2), elective (1): the operative score is 12. The mortality linear predictor is -9.065 + 0.1692 x 22 + 0.1550 x 12 = -3.4826, giving a predicted mortality of 2.98 percent. The morbidity linear predictor is -5.91 + 0.16 x 22 + 0.19 x 12 = -0.11, giving a predicted morbidity of 47.25 percent. Notice how much higher the morbidity figure is: that gap is expected, and the FAQ section explains why.

Example three, high risk. A 78-year-old man needs an immediate emergency laparotomy. He is 78 (age grade 4), has a raised jugular venous pressure (cardiac 8), dyspnoea at rest (respiratory 8), blood pressure 85 (grade 8), pulse 125 (grade 8), Glasgow Coma Score 10 (grade 4), haemoglobin 9.5 (grade 8), white cell count 22.0 (grade 4), urea 16.0 (grade 8), sodium 128 (grade 4), potassium 6.1 (grade 8), and atrial fibrillation with a rapid rate, which counts as an abnormal rhythm (ECG grade 4). The physiological score is 76. The operation is major plus (8), involves two procedures (2), blood loss 1200 ml (8), free bowel content in the peritoneum (8), nodal metastases (4), done as an immediate emergency within 2 hours (4): the operative score is 34. The mortality linear predictor is 9.0642, giving a predicted mortality of 99.99 percent, and the morbidity linear predictor is 12.71, giving a predicted morbidity effectively at the ceiling. This is the kind of case the exponential scale was built for: several grade-8 factors compound into a prediction near certainty.

Reading the result honestly

P-POSSUM deserves its reputation, but the reputation needs the caveats that come with it. First, it is a comparative audit tool. Its purpose is to let a unit, a surgeon or a time period be compared against risk-adjusted expectations: observed deaths versus the deaths the case mix would predict. That is a different job from telling an individual patient their personal chance of dying, and the score was never validated for bedside decision-making. A percentage on a screen must not decide whether someone is offered surgery.

Second, even the corrected equation is least reliable where most patients are. Prytherch and colleagues showed that the original POSSUM overpredicted mortality most severely in the lowest-risk patients, and although P-POSSUM fixed the overall fit, later validation studies have repeatedly found some overprediction persisting in low-risk and elective populations. Treat a predicted mortality of, say, 1 or 2 percent as an approximate figure with wide uncertainty, not as a precise personal forecast. The morbidity equation, inherited unchanged from POSSUM, has the same tendency: predicted complication rates in the teens should be read as rough estimates.

Third, the score has a constructive use beyond prediction. Several of its factors are modifiable before an elective operation: anaemia can be investigated and treated, potassium and sodium can be corrected, euvolaemia can be restored, and cardiac medication can be optimised. A physiological score of 22 that falls to 16 after two weeks of optimisation is a genuine reduction in predicted risk, and that is arguably the most useful thing the score does. It turns a vague sense that a patient is high risk into a list of named, gradable, correctable problems.

Finally, the score is only as good as its inputs. In the published validation work, missing variables were scored as 1, that is, treated as normal, which is a convention for analysis, not a clinical recommendation. In practice an unknown potassium should be measured, not assumed normal. Score what you have measured, and be suspicious of a reassuringly low total built on missing data.

Frequently asked questions

What does P-POSSUM predict?

P-POSSUM predicts two separate probabilities for a patient undergoing surgery: the predicted risk of death (mortality) and the predicted risk of a postoperative complication (morbidity), each expressed as a percentage. It does this by adding 12 physiological grades and 6 operative grades, each scored 1, 2, 4 or 8, and passing the two totals through published logistic regression equations. The lowest possible scores (physiological 12, operative 6) give a predicted mortality of about 0.22 percent and a predicted morbidity of about 5.5 percent.

How is the P-POSSUM score calculated?

Each of the 18 factors is graded 1, 2, 4 or 8 on an exponential scale, where 1 means normal and 8 means severely deranged. The 12 physiological grades are summed to give the physiological score (range 12 to 96) and the 6 operative grades are summed to give the operative score (range 6 to 40). The mortality equation is ln(R/(1-R)) = -9.065 + 0.1692 x physiological score + 0.1550 x operative score, and the morbidity equation is ln(R/(1-R)) = -5.91 + 0.16 x physiological score + 0.19 x operative score. Each linear predictor is converted to a probability with R = 1 / (1 + e^(-x)).

What is the difference between POSSUM and P-POSSUM?

POSSUM, published by Copeland, Jones and Walters in 1991, is the original 18-factor system with its own mortality and morbidity equations. P-POSSUM (Portsmouth POSSUM), published by Prytherch and colleagues in 1998, keeps the identical 18 factors and 1, 2, 4, 8 grading but replaces the mortality equation, because the original POSSUM mortality equation overpredicted death by more than twofold overall and by more than sevenfold in the lowest-risk patients. The P-POSSUM mortality equation was fitted on 10,000 general surgical episodes in Portsmouth and matched observed mortality closely in validation. The morbidity equation is unchanged from POSSUM.

Why is the predicted morbidity so much higher than the predicted mortality?

Because complications are far more common than death after surgery, and the two equations reflect that. The morbidity equation has a less negative intercept (-5.91 versus -9.065) and steeper coefficients, so for the same scores it produces a much larger probability. For the lowest possible scores the predicted morbidity is about 5.5 percent while the predicted mortality is about 0.22 percent. A high predicted morbidity with a low predicted mortality is the normal pattern, not an error.

Can P-POSSUM be used to decide whether an individual patient should have surgery?

No. P-POSSUM was developed as a comparative audit tool: it lets hospitals and surgeons compare observed outcomes against risk-adjusted expected outcomes, so that differences in case mix do not distort the comparison. It was not designed to make treatment decisions for individual patients, and the literature notes that it tends to overpredict risk in low-risk populations. The score can flag modifiable risk factors, such as anaemia or electrolyte disturbance, that are worth correcting before surgery, but the decision to operate rests on clinical judgement, not on a percentage.

Where do the P-POSSUM equations come from?

The mortality equation comes from Prytherch DR, Whiteley MS, Higgins B, Weaver PC, Prout WG and Powell SJH, "POSSUM and Portsmouth POSSUM for predicting mortality", British Journal of Surgery, 1998;85:1217-1220 (doi:10.1046/j.1365-2168.1998.00794.x). The authors applied the original POSSUM system to 10,000 general surgical episodes in Portsmouth, found that it overpredicted mortality, and fitted the corrected equation on the first 2,500 cases before validating it prospectively on the remaining 7,500. The 18 factors and their 1, 2, 4, 8 grades come from the original POSSUM paper by Copeland GP, Jones D and Walters M, British Journal of Surgery, 1991;78:355-360.