Publications

15Feb2010

An Assessment of the POSSUM System in Orthopaedic Surgery

Mohamed K FRCS*, Copeland GP ChM**, Boot DA FRCS***,
Casserley HC FRCS***, Shackleford IM FRCS*** Sherry PG FRCS***,
Stewart GJ FRCS***

*Orthopaedic Registrar, Warrington Hospital
** Consultant General Surgeon, Warrington Hospital
*** Consultant Orthopaedic Surgeon, Warrington Hospital

Key words: Orthopaedic surgery, Surgical audit, POSSUM scoring system

First revision 11.12.2001

Abstract

The present article describes the development and validation of a scoring system for orthopaedic surgical audit. The present system is a minor modification of the established POSSUM scoring system widely used in general surgery. The orthopaedic POSSUM system developed yields predictions for mortality and morbidity which correlate well with the observed mortality and morbidity rates in a sample of 2326 orthopaedic patients operated upon in a twelve month period.

Introduction

The assessment of outcome after surgical endeavour is not a new science. As early as 1750 BC King Hammurabi of Babylon issued a number of decrees related to surgeons and their surgery. His most infamous of these codices was that if a surgeon operated on a free man and the patient became blind or worse still died then the surgeon should have his operating hand cut off. While to many surgeons world wide this codex may still seem to be in operation, a number of groups have attempted to devise more reliable and robust methods for assessing the outcome from surgical intervention13.

Clearly the use of raw mortality and morbidity data will produce differences in outcome when comparing differing units, but these apparent differences may be explained merely by variations in case mix and the type of surgery45. In general surgery methods have been devised to allow comparison between units by taking the physiological status of patients and the operative severity into account167.

In the general surgical setting the POSSUM1 and P-POSSUM8 systems have proved to be the most reliable and widely applicable of all the scoring systems so far devised, and they have both been found to be equally applicable in sub-specialities including vascular surgery, surgical gastroenterology and urology. Indeed the POSSUM system has been found to be applicable in many differing health care organisations world wide911.

A number of orthopaedic authors have drawn attention to the variability in outcome following orthopaedic surgery, in particular fractured neck of femur and have argued that this variability is more likely to be related to case mix than to hospital facilities or surgeons12. That mortality and morbidity can be high particularly with proximal femoral fractures is well recognised13. However more recent studies have cautioned that predicting the outcome of orthopaedic intervention in the elderly injured patient merely on the basis of the injury severity score can be hazardous, as host factors appear to be of greater importance14.

A system which is heavily weighted towards physiological status would appear, potentially, to be of benefit in assessing the outcome from orthopaedic surgical intervention. Clearly such a system should allow comparison based on the patient’s physiological status and an assessment of the magnitude of surgery and its timing.

During the period 1996-1998 we have utilised similar methodologies, to those described and used previously by ourselves in the development of the POSSUM system1, to design an orthopaedic operative severity score which would allow the general surgical logistic regression equation to be applied in an orthopaedic setting. In total 22 operative severity factors were assessed which were reduced by multivariate analysis to the minimum number necessary to produce an accurate estimate of mortality and morbidity. The resulting operative severity score was not surprising similar in many ways to the general surgical score although individual factors and weightings for these factors were dissimilar.

The present study attempts to validate the application of this new method of assessing the outcome after orthopaedic surgery.

Patients and Methods

All consecutive patients admitted to Warrington Hospital during a 12 month period in whom orthopaedic surgery was performed on a non day case basis were assessed using the new orthopaedic POSSUM system.

This includes a physiological assessment and an operative severity assessment. The physiological assessment included twelve variables each divided into four grades with an exponentially increasing score value (1,2,4 and 8) (see table 1). In the main all score variables were available for all patients, but where no score variable was available a score of 1 was allocated. The operative severity score included six score variables each divided into four grades with an exponentially increasing score value (see table 1). Definitions of operative magnitude are illustrated in table 2. Where a operation is not listed the most closely approximately operation grouping was chosen. The number of operations variable indicates the chronology of procedures occurring within thirty days of the preceding operation. All operative severity score variables were obtainable in all patients, although histological confirmation was necessary in some individuals to complete all score elements.

Outcome was assessed as 30 day morbidity and mortality. 30 day mortality and morbidity assessment was chosen to allow comparability with the general surgical system and in view of the Department of Health’s publication of 30 day mortality statistics. The presence of the following complications were recorded as morbidity:

  1. Infection- chest, urinary, wound, joint, bony, septicaemia, pyrexia of unknown origin, or deep collection
  2. Haemorrhage- deep or wound
  3. Other wound problems- seroma requiring drainage or dehiscence
  4. Thromboembolic complications- deep venous thrombosis, pulmonary embolus, myocardial infarction or cerebrovascular accident
  5. Cardiac- failure, hypotension, or abnormal rhythm
  6. Respiratory- failure, collapse, or pneumonia
  7. Renal- failure or retention
  8. Unanticipated dislocation of prosthetic implant

Exact definitions are as have been previously published1.

Mortality and morbidity predictions (R1 for mortality and R2 for morbidity) for individual patients were estimated using previously determined equations.

Loge R1/(1-R1) = -7.04 + (0.13 x physiological score) + (0.16 x operative severity score)

Loge R2/(1-R2) = -5.91 + (0.16 x physiological score) + (0.19 x operative severity score)

Results

During a twelve month period 2326 patients underwent orthopaedic surgery on a non-day case basis. 44% of these patients were operated on in the elective setting, the remainder, 56%, were operated on urgently or as emergencies.

The overall mortality rate was 2.2%, this being 0.2% for elective cases increasing to 3.8% for urgent and emergency surgery. The overall morbidity rate was 10.8%, being 4.4% for elective cases and 15.9% for patients undergoing urgent or emergency surgery. The complication profile is shown in table 3.

Using the POSSUM logistic regression equations yielded an overall predicted mortality rate of 53 patients (versus 51 observed) and a predicted morbidity rate of 254 patients (versus 252 observed). The risk spectrum for both mortality and morbidity are shown in table 4, and follow the expected exponential distribution.

The number of operations performed by each of the six surgeons contributing to the study is shown in table 5. As can be seen there are apparent differences in each surgeon’s overall mortality and morbidity rates.

However when corrected for case mix using the POSSUM system there would appear to be little difference between observed mortality and morbidity rates and those predicted by POSSUM (table 6).

The predictive accuracy of these equations was assessed by the determination of receiver operating characteristic curves (ROC curves), by determining classification matrices for different levels of predicted mortality and morbidity. The resultant ROC curves are illustrated in Figures 1 and 2, which show good correlation across the risk range.

Discussion

There has been increasing worldwide public and political interest in the assessment of quality of care with regard to surgical outcome. This is perhaps easier within the surgical specialities as death following surgery is a fairly obvious adverse outcome. This has led many lay and non-surgical clinicians to suggest that mortality rates may be a suitable indicator of surgical prowess. As is illustrated by King Hammurabi of Babylon who removed the hands of surgeons whose patients died this is not a new phenomenon. There is, however, increasing awareness that such ‘raw’ mortality data may be at best incorrect and at worst dangerous. In addition morbidity is often ignored. Death following surgery is a rare event as demonstrated in the present study and usually follows a number of antecedent complications. If improvements in care are to be made then reduction in complications may well proceed reduction in mortality.

The overall mortality and morbidity rates for the twelve month study period appear to be in keeping with previously published rates for non day case surgery in the district general hospital setting1,6. Indeed the risk profile of the study sample does show similarities to the general surgical distribution16 suggesting that the present study group is comparable to the average district general orthopaedic case mix. In the general surgical setting however there is a tendency to have a greater number of patients whose risk is greater than 70% for mortality1467.

The present study shows an excellent correlation between the overall observed mortality and morbidity rates and the predictions derived from the POSSUM logistic regression equations. The physiological variables assessed were those shown by logistic regression analysis to be the most important in predicting mortality and morbidity. Other additional variables were not found to independently improve the predictive ability of the logistic regression equation. We have not compared the POSSUM predictions with P-POSSUM as previous authors have clearly demonstrated that if the correct mathematical model is applied there are no significant differences in the predictive ability of either method15.

There would appear at first sight to be significant differences in outcome between the six surgeons studied. Mortality rates varied between 1.1% and 3.0%, and morbidity rates between 4.6% and 13.4%. Similar variations have been demonstrated in the general surgical setting and can be explained on the basis of case mix and operative severity. The orthopaedic POSSUM predictions in the present analysis would lend weight to the argument that a similar explanation may also account for apparently marked differences in surgical outcome in orthopaedic surgery. This is well demonstrated by the close correlation between observed and predicted rates for both mortality and morbidity in individual surgeons.

The ROC curves would suggest that the orthopaedic POSSUM score is equally applicable across the spectrum of surgical risk and that the individual patient predictions may be of benefit when assessing at clinical audit an individual patient in who death or complication occurs. Such a system is already in operation within general surgery and may also be of benefit in orthopaedic surgery. We have previously drawn attention to the benefits of audit in patients who survive but in who a predicted risk of death exceeds 50%, as these patients often have more to teach us than those that die. Further studies are at present in progress to quantify the benefits of preoperative optimisation in the high risk orthopaedic patient, as recently Wilson et al16 have demonstrated potentially significant improvements in overall care are possible by optimisation in this particular group.

Clearly no regression equation for risk assessment should remain static over time, although no significant change has been needed over the past ten years. However should dramatic changes occur in the future the equation could be easily updated without the need to alter the score variables and comparison over time would still be possible. Should this occur historically scored patients would be assessed with the present equation and new patients assessed with an updated equation.

The present study indicates that orthopaedic POSSUM can be used as an audit aid to assess the quality of orthopaedic care. It is possible a quality measure similar to that used in general surgery, the O/E ratio (the ratio of observed adverse events to predicted adverse events) could be used within orthopaedic surgery as a more reliable quality measure than simple mortality and morbidity rates. Using such a method a ratio of 1.00 indicates average care. A ratio greater than 1.00 worse care and ratio less than 1.00 better care. Further studies are in progress in other units to assess its more wide spread usage and application.

References

  1. Copeland GP, Jones D, Walters M. POSSUM: a scoring system for surgical audit. Br J Surg 1991;78:355-360
  2. Jones HJS, de Cossart L. Risk scoring in surgical patients. Br J Surg 1999;86:149-157
  3. Pillai SB, van Rij AM, Williams S, Thomson IA, Putterill MJ, Greig S. Complexity and risk-adjusted model for measuring surgical outcome. Br J Surg 1999;86:1567-1572
  4. Copeland GP, Jones D, Wilcox A, Harris PL. Comparative vascular audit using the POSSUM scoring system. Ann R Coll Surg Engl 1993;75:175-177
  5. Sagar PM, Hartley MN, MacFie J, Taylor BA, Copeland GP. Comparison of individual surgeon’s performance. Dis Colon Rectum 1996;38:654-658
  6. Copeland GP, Sagar P, Brennan J, Roberts G, Ward J, Cornford P et al. Risk adjusted analysis of surgeon performance. Br J Surg 1995;82:408-411
  7. Copeland GP. Assessing the surgeon: 10 years experience with the POSSUM system. J Clin Excell 2000;2:187-190
  8. Prytherch D, Whiteley MS, Higgins B, Weaver PC, Prout WG, Powell SJ. POSSUM and Portsmouth POSSUM for predicting mortality. Br J Surg 1998;85:1217-1220
  9. Brunelli A, Fianchini A, Xiume F, Gesuita R, Mattei A, Carle F. Evaluation of the POSSUM scoring system in lung surgery. Thorac Cardiovasc Surg 1998;46:141-146
  10. Gotohda N, Iwagaki H, Itano S. Can POSSUM, a scoring system for perioperative surgical risk, predict postoperative clinical course. Acta Medica Okayama 1998;52:325-329
  11. Tekkis PP, Kocher HM, Bentley AJ, Cullen PT, South LM, Trotter GA, Ellul JP. Operative mortality rates among surgeons: comparison of POSSUM and p-POSSUM scoring systems in gastrointestinal surgery. Dis Colon Rectum 2000;43:1528-1532
  12. Withey C, Morris R, Beech R, Backhouse A. Outcome following fractured neck of femur- variation in acute hospital care or case mix? J Public Health Med 1995;17:429-437
  13. Thomas M, Eastwood H. Re-evaluation of two simple prognostic scores of outcome after proximal femoral fractures. Injury 1996;27:111-115
  14. Van der Sluis CK, Timmer HW, Eisma WH, ten Duis HJ. Outcome in elderly injured patients: injury severity versus host factors. Injury 1997;28:588-592
  15. Wijesinghe LD, Mahmood T, Scott DLA, Berridge DC, Kent PJ, Kester RC. Comparison of POSSUM and the Portsmouth predictor equation for predicting death following vascular surgery. Br J Surg 1998;85:209-212
  16. Wilson J, Woods I, Fawcett J, Whall R, Dibb W, Morris C, McManus E. Reducing the risk of major elective surgery: randomised controlled trial of preoperative optimisation of oxygen delivery. BMJ 1999;318:1099-1103

Tables

Table 1: POSSUM sheet for physiological score

SCORE 1 2 4 8
Age (yrs) <60 61 - 70 >71
Cardiac signs Normal On Cardiac drugs or steroid Oedema WarfarinJVP
CXR Border Cardio Cardio megaly
Resp. signs Normal SOB Exertion SOB stairs SOB rest
CXR Mild COAD Mod COAD Any other change
SYSTOLIC 110 - 130 131 - 170 >171 <89
BP (mmHg) 100 - 109 90 - 99
Pulse (/min) 50 - 80 81 - 100 101 - 120 >121
40 - 49 <39
Coma Score 15 12 - 14 9 - 11 <8
Urea (mmol/l) <7.5 7.6 - 10 10.1 - 15 >15.1
Na (mmol/l) >136 131 - 135 126 - 130 <125
K (mmol/l) 3.5 - 5 3.2 - 3.4 2.9 - 3.1 <2.8
5.1 - 5.3 5.4 - 5.9 >6
Hb (g/100ml) 13 - 16 11.5 - 12.9 10 - 11.4 <9.9
16.1 - 17 17.1 - 18 >18.1
WCC (x1012/l) 4 - 10 10.1 - 20 >20.1
3.1 - 3.9 <3
ECG Normal AF (60 - 90) Any other change

 

Table 2: POSSUM sheet for operative severity score

SCORE 1 2 3 4
Op magnitude MinorInterMajorMajor +
Number of ops within 30 days 1 2>2
Blood loss per operation <100 101 - 500 501 - 999 >1000
Contamination No Incised wound ie stab Minor contam or necrotic tissue Major contam or necrotic tissue
Presence of malignancy No 1o Node metastases Distant metastases
Time of operation Elec. Emerg. resus poss <48hrs Emerg. immed <6hrs

 

Table 3:Number of patients experiencing complications postoperatively.

(A number of patients experienced multiple complications)

Complications Number
Haemorrhage 14
Infection
Chest 59
Urinary 34
Wound 73
Septicaemia 7
PUO 5
Respiratory Failure 17
Cardiac
Hypotension 29
Cardiac Failure 23
Arrhythmia 6
Thrombotic
Deep venous thrombosis 12
Pulmonary embolus 9
Myocardial infarction 17
Cerebrovascular infarction 7
Limb occlusion 2
Other vascular complications 4
Renal Failure 15
Urinary retention 32
Other wound problems 10
Prosthetic problems 7
Miscellaneous 26

 

Table 4: Risk spectrums for both morbidity and mortality.

Risk Band Mortality (number of patients) Morbidity (number of patients)
<10% 2086 1042
10-20% 140 600
20-30% 51 230
30-40% 23 124
40-50% 11 99
50-60% 3 72
60-70% 6 48
70-80% 3 49
80-90% 1 37
>90% 2 25

 

Table 5: The variability in surgeon workload, mortality and morbidity rates

Surgeon Number of Procedures Mortality Morbidity
1 495 3.0% 12.9%
2 474 1.1% 10.5%
3 372 2.4% 11.0%
4 216 1.8% 4.6%
5 426 2.1% 9.6%
6 343 2.6% 13.4%

 

Table 6: Comparisons between observed and predicted mortality and morbidity rates for individual surgeons during the study period

Surgeon Mortality Observed Mortality Predicted Morbidity Observed Morbidity Predicted
1 15 15 62 64
2 5 6 52 50
3 9 10 41 43
4 4 4 10 10
5 9 9 41 40
6 9 9 46 47

Legends

Fig 1. Receiver operating characteristic curve for mortality. A curve approaching the linear line indicates no predictive ability for the assessing system. The further from the linear line the better the predictive ability.

Fig 2. Receiver operating characteristic curve for morbidity.

  • 15 Feb, 2010
  • Claire Bale
  • 0 Comments

Share This Story

Categories

Comments

WordPress Appliance - Powered by TurnKey Linux