- Email: [email protected]
P–POSSUM scoring system for mortality prediction in general neurosurgery
Journal of Clinical Neuroscience 17 (2010) 567–570
Contents lists available at ScienceDirect
Journal of Clinical Neuroscience journal homepage: www.elsevier.com/locate/jocn
Clinical Study
P–POSSUM scoring system for mortality prediction in general neurosurgery W. Chen a,*, J.W.H. Fong a, C.R.P. Lind a,b, N.W. Knuckey a,b a b
West Australian Neurosurgical Service, Sir Charles Gairdner Hospital, 1st Floor, G Block, Hospital Avenue, Nedlands, Western Australia 6009, Australia Centre of Neuromuscular and Neurological Disorders, University of Western Australia, Perth, Australia
a r t i c l e
i n f o
Article history: Received 23 August 2009 Accepted 13 September 2009
Keywords: Audit Mortality rate Neurosurgery Operative variables Physiological variables Prediction Risk
a b s t r a c t The Physiological and Operative Severity Score for enUmeration of Mortality and morbidity (POSSUM) scoring systems have been designed for comparative audit and have been well validated in general and vascular surgery. The Portsmouth predictor equation (P–POSSUM) was highly predictive of mortality in a study of elective craniotomies for neurosurgery but has yet to be validated in spinal, peripheral nerve or acute cranial neurosurgery. The West Australian Categorisation of Operative Severity (WA classification) was created for all neurosurgical procedures. Case notes and laboratory results of 531 consecutive patients undergoing neurosurgery were reviewed retrospectively. All POSSUM variables were collected and the POSSUM and P–POSSUM mortality equations were applied. The observed mortality rate was 4.52% and the WA P–POSSUM predicted mortality rate was 4.58% (p > 0.951). The WA P–POSSUM rate was more predictive than either the WA POSSUM rate (10.9%, p < 0.0001) or the previously proposed elective craniotomy P–POSSUM classification (5.8%, p < 0.198). We concluded that the P–POSSUM model with WA classification has the potential to be used in mortality audits for general neurosurgery. By quantifying preoperative risk, P–POSSUM might provide a useful denominator to observed death rates for meaningful comparison of individual neurosurgeons and between departments. Ó 2009 Published by Elsevier Ltd.
1. Introduction Clinical governance and surgical accountability are becoming increasingly important as the general public and also health officials demand greater disclosure. Surgical audit of morbidity and mortality is performed in most units. The problem is that direct comparison between individual surgeons and different departments is confounded by variation in case mix. The Physiological and Operative Severity Score for enUmeration of Mortality and morbidity (POSSUM) was developed by Copeland et al. in 1991 with the purpose of adjusting for patients’ risk factors and hence allowing a comparative audit.1 The POSSUM model uses a logistic regression analysis of 12 physiological and six operative variables to predict mortality (Table 1). The main criticism of POSSUM was over-prediction of death in low-risk surgery.2–4 The equation was revised in 1996 by Whiteley et al. using a linear rather than logistic regression analysis of the same physiological and operative factors.5 This equation became known as the Portsmouth predictor equation (P–POSSUM). The POSSUM scoring systems are well validated as surgical audit tools in other specialities including general, vascular and colorectal surgery.2,3,6–8 POSSUM systems compared favourably with other risk-scoring systems such as APACHE and Surgical risk
* Corresponding author. Tel.: +61 08 346 2865. E-mail address: [email protected] (W. Chen). 0967-5868/$ - see front matter Ó 2009 Published by Elsevier Ltd. doi:10.1016/j.jocn.2009.09.020
score.9,10 The POSSUM methodology has been developed to improve its predictive accuracy in different specialities.11,12 No comprehensive POSSUM scoring systems for neurosurgery have been published. The only previous study of the POSSUM for neurosurgery was limited to elective craniotomies.13 Although P– POSSUM was highly predictive in that setting, most neurosurgical units also perform spinal and acute neurosurgical operations. All of these procedures were included in the current study.
2. Methods The West Australian Neurosurgical Service provides the only comprehensive public service to the 2.3 million population of the State of Western Australia. Services are provided at Sir Charles Gairdner Hospital and Royal Perth Hospital, both of which are Level One trauma centres. All patients over the age of 16 undergoing neurosurgery between 1 July and 31 October 2007 were reviewed from our comprehensive database. POSSUM/P–POSSUM physiologic and operative parameters (Table 1) were collected from the preoperative nursing check records, medical notes, and computerised and filed laboratory results. All physiologic and operative parameters were available for all patients except for those who underwent peripheral nerve day surgery under local anaesthesia. In these patients, laboratory results were assumed to be within the normal limits.
568
W. Chen et al. / Journal of Clinical Neuroscience 17 (2010) 567–570
Table 1 Parameters used in POSSUM scoring system Physiological parameters
Operative parameters
Age Cardiac Signs Respiratory history Systolic blood pressure Pulse Glasgow coma scale Haemoglobin White cell count Urea Sodium Potassium Electrocardiogram
Operative severity Operative urgency Multiple procedures Total blood loss Peritoneal soiling Presence of malignancy Mode of surgery
ln½R=ð1 RÞ ¼ 7:04 þ 0:13 physiological score
POSSUM = Physiological and Operative Severity Score for enUmeration of Mortality and morbidity.
Using the severity score of general surgery as a guide, we assigned comparable severity scores for neurosurgical operations: The West Australian Categorisation of Operative Severity (WA classification) (Table 2). In POSSUM and P–POSSUM, aortic, abdominoperineal resective, pancreatic resective, liver resective or oesophagogastrectomy procedures were classified ‘‘Major plus”. All other laparatomies were classified as ‘‘Major”. By the same principle, we placed aneurysm clipping, arteriovenous malformation resective, skull base resective and intramedullary spinal surgeries in the ‘‘Major plus” category. Transoral and instrumented spine, as well as all other craniotomy including burr hole surgery with brain penetration, were classified as ‘‘Major”. Burr holes for extracerebral pathology or strain-gauge intracranial pressure monitoring, decompressive spine surgery, spinal hardware removal and replacement of extracerebral neurostimulator components were classified ‘‘Moderate” risk. Ulnar and carpal tunnel decompressive surgery were categorised as ‘‘Minor” severity. Although this work was not done during the study period, we suggest that reconstructive and brachial plexus peripheral nerve surgery should be categorised as ‘‘Major” or ‘‘Moderate” severity. The WA classification contrasts with that of Ramesh et al., who classified all elective craniotomies as ‘‘Major plus”.13 POSSUM includes a classification of urgency of surgical procedures. These are: (a) elective; (b) emergency to theatre within 2 hours of decision to operate; (c) emergency to theatre more than 2 hours after the decision to operate; and (d) surgery within 24 hours of admission. For purposes of analysis, we also dichoto-
Table 2 WA Classification of neurosurgical operations Minor
Major
Carpal tunnel Ulnar neurolysis
Craniectomy Craniotomy Cranioplasty Deep brain stimulation External ventricular drain insertion Shunt procedure Spinal fusion Stereotactic biopsy Transoral procedure Other instrumented spinal surgery
Moderate
Major +
Burr holes for subdural haemorrhage Depressed skull fracture Diskectomy Intracranial pressure monitor insertion Laminectomy Ommaya reservoir insertion Spinal hardware removal
mised this scheme into: (i) elective – category (a); and (ii) acute – categories (b), (c), and (d). The clinical endpoint in this study was the same as most recent POSSUM publications:5,13 inpatient mortality. Inpatient deaths were prospectively recorded in the departmental audit database. The POSSUM (Eq. 1) and P–POSSUM (Eq. 2) calculations were each performed using the WA classification, the Ramesh et al. operative score for the craniotomies13 and the WA classification scores for all the non-cranial procedures. R represents the predicted mortality score.
Aneurysms Arteriovenous malformation Intramedullary tumour Skull base resective
þ 0:16 operative score
ð1Þ
ln½R=ð1 RÞ ¼ 9:37 þ 0:19 physiological score þ 0:15 operative score
ð2Þ
In a neurosurgical audit, operations are commonly divided into three different groups: cranial, spinal, and peripheral nerve. 2.1. Statistics Observed and expected mortality rates were compared. Student’s t-test was used to calculate the test of fit between predicted and observed rates. The receiver–operator characteristics (ROC) curve was used to assess the neatness of fit between the observed and predicted mortality rates using P–POSSUM.14 Alpha was defined as 0.05. 3. Results A total of 531 operations (365 cranial, 154 spinal and 12 others) were performed between July and October 2007. Of these, 371 operations were elective and 160 were acute. The mean age was 54.6 years (range: 16–95 years) and the ratio of males to females was 3:2. In this present series, the overall observed mortality rate was 4.52% (95% confidence interval [CI]: 2.75%, 6.29) (Fig. 1). The P– POSSUM calculated using the WA classification predicted a mortality rate of 4.58% (p = 0.952). The P–POSSUM calculated with the Ramesh et al. classification for craniotomies together with WA classification for other neurosurgery predicted a mortality rate of 5.12% (p = 0.197). The predicted mortality using POSSUM with the WA classification was 10.9%, or with the Ramesh et al. classification was 13.5% (p < 0.0001 for each). The predictive accuracy of the WA P–POSSUM equation was assessed using a ROC curve. The analysis of the WA P–POSSUM resulted in a mortality prediction (area under the curve) of 0.780 (78%) (Fig. 2).
Fig. 1. Comparison of observed mortality (shaded area; 95% confidence interval, 2.75–6.29%) to predicted mortality using two classification systems of operative severity (Ramesh et al.13 and The West Australian Categorisation of Operative Severity [WA classification]) and two scoring systems for mortality and morbidity (The Physiological and Operative Severity Score for enUmeration of Mortality and morbidity [POSSUM1] and the Portsmouth predictor equation [P–POSSUM5]).
W. Chen et al. / Journal of Clinical Neuroscience 17 (2010) 567–570
569
No further analysis of POSSUM is presented because of the gross over-prediction of overall mortality. 3.1. Acute versus elective neurosurgery The observed mortality rate for elective patients undergoing any form of neurosurgery was 1.62% (95% CI: 0.33%, 2.90%) (Fig. 3a): P–POSSUM with WA classification predicted mortality rates of 2.41% (p = 0.253); and with Ramesh et al. operative scores P–POSSUM predicted a mortality rate of 3.00% (p = 0.050). The observed mortality rate for acute operations was 11.25% (95% CI: 6.34%, 16.2%) (Fig. 3b): the P–POSSUM with WA classification predicted mortality rates of 12.35%, and P–POSSUM with Ramesh et al. predicted a 9.61% mortality rate (p > 0.500 for both). 3.2. Acute versus elective cranial neurosurgery
Fig. 2. Receiver operating characteristics graphs for: the Portsmouth Predictor Equation–Physiological and Operative Severity Score for enUmeration of Mortality and morbidity (P–POSSUM) scoring system (solid line); and The West Australian Categorisation of Operative Severity (WA classification) (WA P–POSSUM) (dashed line). (Mortality rate is predicted by the area under the WA P–POSSUM curve – 0.780 [78%]).
The observed mortality rate for elective cranial neurosurgery was 2.73% (95% CI%: 0.57%, 4.88%) (Fig. 3c). The P–POSSUM predicted mortality rates with the WA classification was 2.87% (p = 0.903), and with Ramesh et al. was 3.84% (p = 0.363). The observed mortality rate for acute cranial neurosurgery was 12.41% (95% CI: 7.03%, 17.80%) (Fig. 3d). The P–POSSUM predicted mortality rate with the WA classification was 9.73% (p = 0.363), and with Ramesh et al was 12.71% (p = 0.908).
Fig. 3. Comparison of observed mortality to predicted mortality for patients with (a) elective operations; (b) acute operations; (c) elective cranial procedures; (d) acute cranial procedures; (e) all cranial procedures; and (f) spinal procedures. The classification systems of operative severity were: the Portsmouth Predictor Equation– Physiological and Operative Severity Score for enUmeration of Mortality and morbidity (P–POSSUM5); and the West Australian Categorisation of Operative Severity (WA classification) (WA P–POSSUM). The shaded area of each graph represents the observed mortality rate and the 95% confidence interval.
570
W. Chen et al. / Journal of Clinical Neuroscience 17 (2010) 567–570
Table 3 Comparison of predicated mortality with observed spinal mortality of one, two and three Sensitivity analysis t-Test with spinal mortality = 1 WA P–POSSUM
0.031
t-Test with spinal mortality = 2 WA P–POSSUM
0.280
t-Test with spinal mortality = 3 WA P–POSSUM
0.695
P–POSSUM = Portsmouth predictor equation of the Physiological and Operative Severity Score for enUmeration of Mortality and morbidity, WA = West Australian (WA) Categorisation of Operative Severity.
quired to place the mortality rate within the model’s predicted range. Chance may account for there being no deaths, in the current spinal sample size of 154 operations. Spinal neurosurgery entails an overall lower risk of mortality than cranial neurosurgery. As in P–POSSUM applications in other specialties, the model over-predicts deaths in this relatively low risk surgery.4,18 The mathematical basis of the POSSUM equations does not allow a predicted mortality of zero. The minimum possible mortality rate as scored by P–POSSUM is 0.22% when all parameters are at the lowest risk. The P–POSSUM is a promising mortality audit tool in neurosurgery. Further use of the P–POSSUM with the WA classification would be useful to validate subgroup comparisons. Application of this system in other centres will demonstrate the generalisability of the model.
3.3. Cranial versus spinal neurosurgery References The observed mortality rate for all cranial operations was 6.58% (95% CI: 4.03%, 9.13%) (Fig. 3e). P–POSSUM predicted mortality rates were 5.59% with the WA classification and 7.38% with Ramesh et al. (p > 0.475 for both). There were no deaths in the spinal surgery group. The P–POSSUM predicted spine surgery mortality with the WA classification was 2.43% (Fig. 3f). As there was no death in the spinal group, we proceeded to perform a sensitivity analysis (Table 3) to assess the impact of up to three deaths occurring. The analysis suggests that at two deaths, there would be no significant difference between the observed and predicted mortality rate. We did not analyse peripheral nerve surgery as an entity as the number of such operations performed during this period was too small to allow meaningful analysis. 4. Discussion Comparative audit is only meaningful when it takes into account both clinical and operative risk factors. Comparison of raw mortality data without consideration of these risk factors may penalise surgeons and practices carrying out high-risk surgery. The POSSUM scoring systems have been shown to be valuable audit tools as they correct the case mix between surgeons15 and hospitals.3,16 This study shows that both P–POSSUM models are predictive of the overall mortality in a general neurosurgical service. However, POSSUM consistently overestimated the mortality rates in all groups of patients. This is consistent with the findings in other surgical populations.2,4,5,13,17 We introduce the WA classification of operative severity which enables the neurosurgeon to classify not only elective craniotomies,13 but also spinal and peripheral nerve neurosurgery and all acute neurosurgical operations. P–POSSUM using the WA system is highly predictive of overall mortality. In addition, we provide further evidence following Ramesh et al. of the generalisability of P–POSSUM in neurosurgery.13 In the study of our spinal surgery patients, there were no deaths and this represents a significant difference from the model’s prediction. The sensitivity analysis suggests that two deaths were re-
1. Copeland GP, Jones D, Walters M. POSSUM: a scoring system for surgical audit. Br J Surg 1991;78:355–60. 2. Wiejesinghe LD, Mahmood T, Scott DJ, et al. Comparison of POSSUM and the Portsmouth predictor equation for predicting death following vascular surgery. Br J Surg 1998;85:209–12. 3. Tekkis PP, Kessaris N, Kocher HM, et al. Evaluation of POSSUM and P-POSSUM scoring systems in patients undergoing colorectal surgery. Br J Surg 2003;90:340–5. 4. Griffiths H, Cuddihy P, Davis S, et al. Risk adjusted comparative audit. Is POSSUM applicable to head and neck surgery? Clin Otolaryngol 2002;27:517–20. 5. Whiteley MS, Prytherch DR, Higgins B, et al. An evaluation of the POSSUM surgical scoring system. Br J Surg 1996;83:812–5. 6. Prytherch DR, Whiteley MS, Higgins B, et al. POSSUM and Portsmouth POSSUM for predicting mortality. Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity. Br J Surg 1998;85:1217–20. 7. Prytherch DR, Ridler BM, Beard JD, et al. A model for national outcome audit in vascular surgery. Eur J Vasc Endovasc Surg 2001;21:477–83. 8. Neary WD, Heather BP, Earnshaw JJ. The Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (POSSUM). Br J Surg 2003;90:157–65. 9. Jones DR, Copeland GP, de Cossart L. Comparison of POSSUM with APACHE II for prediction of outcome from a surgical high-dependency unit. Br J Surg 1992;79:1293–6. 10. Brooks MJ, Sutton R, Sarin S. Comparison of surgical risk score, POSSUM and pPOSSUM in higher-risk surgical patients. Br J Surg 2005;92:1288–92. 11. Tekkis PP, Prytherch DR, Kocher HM, et al. Development of a dedicated riskadjustment scoring system for colorectal surgery (colorectal POSSUM). Br J Surg 2004;91:1174–82. 12. Tekkis PP, McCulloch P, Poloniecki JD, et al. Risk-adjusted prediction of operative mortality in oesophagogastric surgery with O-POSSUM. Br J Surg 2004;91:288–95. 13. Ramesh VJ, Umamaheswara RG, Guha A, et al. Evaluation of POSSUM and PPOSSUM scoring systems for predicting the mortality in elective neurosurgical patients. Br J Neurosurg 2008;22:275–8. 14. Eng J. ROC analysis: web-based calculator for ROC curves. Baltimore: Johns Hopkins University.; [updated 17.05.06; cited 13.01.09]. 15. Copeland GP, Sagar P, Brennan J, et al. Risk-adjusted analysis of surgeon performance: a 1-year study. Br J Surg 1995;82:408–11. 16. Sagar PM, Hartley MN, Mancey-Jones B, et al. Comparative audit of colorectal resection with the POSSUM scoring system. Br J Surg 1994;81:1492–4. 17. Midwinter MJ, Tytherleigh M, Ashley S. Estimation of mortality and morbidity risk in vascular surgery using POSSUM and the Portsmouth predictor equation. Br J Surg 1999;86:471–4. 18. Kuhan G, Abidia AF, Wijesinghe LD, et al. POSSUM and P-POSSUM overpredict mortality for carotid endarterectomy. Eur J Vasc Endovasc Surg 2002;23:209–11.
Contents lists available at ScienceDirect
Journal of Clinical Neuroscience journal homepage: www.elsevier.com/locate/jocn
Clinical Study
P–POSSUM scoring system for mortality prediction in general neurosurgery W. Chen a,*, J.W.H. Fong a, C.R.P. Lind a,b, N.W. Knuckey a,b a b
West Australian Neurosurgical Service, Sir Charles Gairdner Hospital, 1st Floor, G Block, Hospital Avenue, Nedlands, Western Australia 6009, Australia Centre of Neuromuscular and Neurological Disorders, University of Western Australia, Perth, Australia
a r t i c l e
i n f o
Article history: Received 23 August 2009 Accepted 13 September 2009
Keywords: Audit Mortality rate Neurosurgery Operative variables Physiological variables Prediction Risk
a b s t r a c t The Physiological and Operative Severity Score for enUmeration of Mortality and morbidity (POSSUM) scoring systems have been designed for comparative audit and have been well validated in general and vascular surgery. The Portsmouth predictor equation (P–POSSUM) was highly predictive of mortality in a study of elective craniotomies for neurosurgery but has yet to be validated in spinal, peripheral nerve or acute cranial neurosurgery. The West Australian Categorisation of Operative Severity (WA classification) was created for all neurosurgical procedures. Case notes and laboratory results of 531 consecutive patients undergoing neurosurgery were reviewed retrospectively. All POSSUM variables were collected and the POSSUM and P–POSSUM mortality equations were applied. The observed mortality rate was 4.52% and the WA P–POSSUM predicted mortality rate was 4.58% (p > 0.951). The WA P–POSSUM rate was more predictive than either the WA POSSUM rate (10.9%, p < 0.0001) or the previously proposed elective craniotomy P–POSSUM classification (5.8%, p < 0.198). We concluded that the P–POSSUM model with WA classification has the potential to be used in mortality audits for general neurosurgery. By quantifying preoperative risk, P–POSSUM might provide a useful denominator to observed death rates for meaningful comparison of individual neurosurgeons and between departments. Ó 2009 Published by Elsevier Ltd.
1. Introduction Clinical governance and surgical accountability are becoming increasingly important as the general public and also health officials demand greater disclosure. Surgical audit of morbidity and mortality is performed in most units. The problem is that direct comparison between individual surgeons and different departments is confounded by variation in case mix. The Physiological and Operative Severity Score for enUmeration of Mortality and morbidity (POSSUM) was developed by Copeland et al. in 1991 with the purpose of adjusting for patients’ risk factors and hence allowing a comparative audit.1 The POSSUM model uses a logistic regression analysis of 12 physiological and six operative variables to predict mortality (Table 1). The main criticism of POSSUM was over-prediction of death in low-risk surgery.2–4 The equation was revised in 1996 by Whiteley et al. using a linear rather than logistic regression analysis of the same physiological and operative factors.5 This equation became known as the Portsmouth predictor equation (P–POSSUM). The POSSUM scoring systems are well validated as surgical audit tools in other specialities including general, vascular and colorectal surgery.2,3,6–8 POSSUM systems compared favourably with other risk-scoring systems such as APACHE and Surgical risk
* Corresponding author. Tel.: +61 08 346 2865. E-mail address: [email protected] (W. Chen). 0967-5868/$ - see front matter Ó 2009 Published by Elsevier Ltd. doi:10.1016/j.jocn.2009.09.020
score.9,10 The POSSUM methodology has been developed to improve its predictive accuracy in different specialities.11,12 No comprehensive POSSUM scoring systems for neurosurgery have been published. The only previous study of the POSSUM for neurosurgery was limited to elective craniotomies.13 Although P– POSSUM was highly predictive in that setting, most neurosurgical units also perform spinal and acute neurosurgical operations. All of these procedures were included in the current study.
2. Methods The West Australian Neurosurgical Service provides the only comprehensive public service to the 2.3 million population of the State of Western Australia. Services are provided at Sir Charles Gairdner Hospital and Royal Perth Hospital, both of which are Level One trauma centres. All patients over the age of 16 undergoing neurosurgery between 1 July and 31 October 2007 were reviewed from our comprehensive database. POSSUM/P–POSSUM physiologic and operative parameters (Table 1) were collected from the preoperative nursing check records, medical notes, and computerised and filed laboratory results. All physiologic and operative parameters were available for all patients except for those who underwent peripheral nerve day surgery under local anaesthesia. In these patients, laboratory results were assumed to be within the normal limits.
568
W. Chen et al. / Journal of Clinical Neuroscience 17 (2010) 567–570
Table 1 Parameters used in POSSUM scoring system Physiological parameters
Operative parameters
Age Cardiac Signs Respiratory history Systolic blood pressure Pulse Glasgow coma scale Haemoglobin White cell count Urea Sodium Potassium Electrocardiogram
Operative severity Operative urgency Multiple procedures Total blood loss Peritoneal soiling Presence of malignancy Mode of surgery
ln½R=ð1 RÞ ¼ 7:04 þ 0:13 physiological score
POSSUM = Physiological and Operative Severity Score for enUmeration of Mortality and morbidity.
Using the severity score of general surgery as a guide, we assigned comparable severity scores for neurosurgical operations: The West Australian Categorisation of Operative Severity (WA classification) (Table 2). In POSSUM and P–POSSUM, aortic, abdominoperineal resective, pancreatic resective, liver resective or oesophagogastrectomy procedures were classified ‘‘Major plus”. All other laparatomies were classified as ‘‘Major”. By the same principle, we placed aneurysm clipping, arteriovenous malformation resective, skull base resective and intramedullary spinal surgeries in the ‘‘Major plus” category. Transoral and instrumented spine, as well as all other craniotomy including burr hole surgery with brain penetration, were classified as ‘‘Major”. Burr holes for extracerebral pathology or strain-gauge intracranial pressure monitoring, decompressive spine surgery, spinal hardware removal and replacement of extracerebral neurostimulator components were classified ‘‘Moderate” risk. Ulnar and carpal tunnel decompressive surgery were categorised as ‘‘Minor” severity. Although this work was not done during the study period, we suggest that reconstructive and brachial plexus peripheral nerve surgery should be categorised as ‘‘Major” or ‘‘Moderate” severity. The WA classification contrasts with that of Ramesh et al., who classified all elective craniotomies as ‘‘Major plus”.13 POSSUM includes a classification of urgency of surgical procedures. These are: (a) elective; (b) emergency to theatre within 2 hours of decision to operate; (c) emergency to theatre more than 2 hours after the decision to operate; and (d) surgery within 24 hours of admission. For purposes of analysis, we also dichoto-
Table 2 WA Classification of neurosurgical operations Minor
Major
Carpal tunnel Ulnar neurolysis
Craniectomy Craniotomy Cranioplasty Deep brain stimulation External ventricular drain insertion Shunt procedure Spinal fusion Stereotactic biopsy Transoral procedure Other instrumented spinal surgery
Moderate
Major +
Burr holes for subdural haemorrhage Depressed skull fracture Diskectomy Intracranial pressure monitor insertion Laminectomy Ommaya reservoir insertion Spinal hardware removal
mised this scheme into: (i) elective – category (a); and (ii) acute – categories (b), (c), and (d). The clinical endpoint in this study was the same as most recent POSSUM publications:5,13 inpatient mortality. Inpatient deaths were prospectively recorded in the departmental audit database. The POSSUM (Eq. 1) and P–POSSUM (Eq. 2) calculations were each performed using the WA classification, the Ramesh et al. operative score for the craniotomies13 and the WA classification scores for all the non-cranial procedures. R represents the predicted mortality score.
Aneurysms Arteriovenous malformation Intramedullary tumour Skull base resective
þ 0:16 operative score
ð1Þ
ln½R=ð1 RÞ ¼ 9:37 þ 0:19 physiological score þ 0:15 operative score
ð2Þ
In a neurosurgical audit, operations are commonly divided into three different groups: cranial, spinal, and peripheral nerve. 2.1. Statistics Observed and expected mortality rates were compared. Student’s t-test was used to calculate the test of fit between predicted and observed rates. The receiver–operator characteristics (ROC) curve was used to assess the neatness of fit between the observed and predicted mortality rates using P–POSSUM.14 Alpha was defined as 0.05. 3. Results A total of 531 operations (365 cranial, 154 spinal and 12 others) were performed between July and October 2007. Of these, 371 operations were elective and 160 were acute. The mean age was 54.6 years (range: 16–95 years) and the ratio of males to females was 3:2. In this present series, the overall observed mortality rate was 4.52% (95% confidence interval [CI]: 2.75%, 6.29) (Fig. 1). The P– POSSUM calculated using the WA classification predicted a mortality rate of 4.58% (p = 0.952). The P–POSSUM calculated with the Ramesh et al. classification for craniotomies together with WA classification for other neurosurgery predicted a mortality rate of 5.12% (p = 0.197). The predicted mortality using POSSUM with the WA classification was 10.9%, or with the Ramesh et al. classification was 13.5% (p < 0.0001 for each). The predictive accuracy of the WA P–POSSUM equation was assessed using a ROC curve. The analysis of the WA P–POSSUM resulted in a mortality prediction (area under the curve) of 0.780 (78%) (Fig. 2).
Fig. 1. Comparison of observed mortality (shaded area; 95% confidence interval, 2.75–6.29%) to predicted mortality using two classification systems of operative severity (Ramesh et al.13 and The West Australian Categorisation of Operative Severity [WA classification]) and two scoring systems for mortality and morbidity (The Physiological and Operative Severity Score for enUmeration of Mortality and morbidity [POSSUM1] and the Portsmouth predictor equation [P–POSSUM5]).
W. Chen et al. / Journal of Clinical Neuroscience 17 (2010) 567–570
569
No further analysis of POSSUM is presented because of the gross over-prediction of overall mortality. 3.1. Acute versus elective neurosurgery The observed mortality rate for elective patients undergoing any form of neurosurgery was 1.62% (95% CI: 0.33%, 2.90%) (Fig. 3a): P–POSSUM with WA classification predicted mortality rates of 2.41% (p = 0.253); and with Ramesh et al. operative scores P–POSSUM predicted a mortality rate of 3.00% (p = 0.050). The observed mortality rate for acute operations was 11.25% (95% CI: 6.34%, 16.2%) (Fig. 3b): the P–POSSUM with WA classification predicted mortality rates of 12.35%, and P–POSSUM with Ramesh et al. predicted a 9.61% mortality rate (p > 0.500 for both). 3.2. Acute versus elective cranial neurosurgery
Fig. 2. Receiver operating characteristics graphs for: the Portsmouth Predictor Equation–Physiological and Operative Severity Score for enUmeration of Mortality and morbidity (P–POSSUM) scoring system (solid line); and The West Australian Categorisation of Operative Severity (WA classification) (WA P–POSSUM) (dashed line). (Mortality rate is predicted by the area under the WA P–POSSUM curve – 0.780 [78%]).
The observed mortality rate for elective cranial neurosurgery was 2.73% (95% CI%: 0.57%, 4.88%) (Fig. 3c). The P–POSSUM predicted mortality rates with the WA classification was 2.87% (p = 0.903), and with Ramesh et al. was 3.84% (p = 0.363). The observed mortality rate for acute cranial neurosurgery was 12.41% (95% CI: 7.03%, 17.80%) (Fig. 3d). The P–POSSUM predicted mortality rate with the WA classification was 9.73% (p = 0.363), and with Ramesh et al was 12.71% (p = 0.908).
Fig. 3. Comparison of observed mortality to predicted mortality for patients with (a) elective operations; (b) acute operations; (c) elective cranial procedures; (d) acute cranial procedures; (e) all cranial procedures; and (f) spinal procedures. The classification systems of operative severity were: the Portsmouth Predictor Equation– Physiological and Operative Severity Score for enUmeration of Mortality and morbidity (P–POSSUM5); and the West Australian Categorisation of Operative Severity (WA classification) (WA P–POSSUM). The shaded area of each graph represents the observed mortality rate and the 95% confidence interval.
570
W. Chen et al. / Journal of Clinical Neuroscience 17 (2010) 567–570
Table 3 Comparison of predicated mortality with observed spinal mortality of one, two and three Sensitivity analysis t-Test with spinal mortality = 1 WA P–POSSUM
0.031
t-Test with spinal mortality = 2 WA P–POSSUM
0.280
t-Test with spinal mortality = 3 WA P–POSSUM
0.695
P–POSSUM = Portsmouth predictor equation of the Physiological and Operative Severity Score for enUmeration of Mortality and morbidity, WA = West Australian (WA) Categorisation of Operative Severity.
quired to place the mortality rate within the model’s predicted range. Chance may account for there being no deaths, in the current spinal sample size of 154 operations. Spinal neurosurgery entails an overall lower risk of mortality than cranial neurosurgery. As in P–POSSUM applications in other specialties, the model over-predicts deaths in this relatively low risk surgery.4,18 The mathematical basis of the POSSUM equations does not allow a predicted mortality of zero. The minimum possible mortality rate as scored by P–POSSUM is 0.22% when all parameters are at the lowest risk. The P–POSSUM is a promising mortality audit tool in neurosurgery. Further use of the P–POSSUM with the WA classification would be useful to validate subgroup comparisons. Application of this system in other centres will demonstrate the generalisability of the model.
3.3. Cranial versus spinal neurosurgery References The observed mortality rate for all cranial operations was 6.58% (95% CI: 4.03%, 9.13%) (Fig. 3e). P–POSSUM predicted mortality rates were 5.59% with the WA classification and 7.38% with Ramesh et al. (p > 0.475 for both). There were no deaths in the spinal surgery group. The P–POSSUM predicted spine surgery mortality with the WA classification was 2.43% (Fig. 3f). As there was no death in the spinal group, we proceeded to perform a sensitivity analysis (Table 3) to assess the impact of up to three deaths occurring. The analysis suggests that at two deaths, there would be no significant difference between the observed and predicted mortality rate. We did not analyse peripheral nerve surgery as an entity as the number of such operations performed during this period was too small to allow meaningful analysis. 4. Discussion Comparative audit is only meaningful when it takes into account both clinical and operative risk factors. Comparison of raw mortality data without consideration of these risk factors may penalise surgeons and practices carrying out high-risk surgery. The POSSUM scoring systems have been shown to be valuable audit tools as they correct the case mix between surgeons15 and hospitals.3,16 This study shows that both P–POSSUM models are predictive of the overall mortality in a general neurosurgical service. However, POSSUM consistently overestimated the mortality rates in all groups of patients. This is consistent with the findings in other surgical populations.2,4,5,13,17 We introduce the WA classification of operative severity which enables the neurosurgeon to classify not only elective craniotomies,13 but also spinal and peripheral nerve neurosurgery and all acute neurosurgical operations. P–POSSUM using the WA system is highly predictive of overall mortality. In addition, we provide further evidence following Ramesh et al. of the generalisability of P–POSSUM in neurosurgery.13 In the study of our spinal surgery patients, there were no deaths and this represents a significant difference from the model’s prediction. The sensitivity analysis suggests that two deaths were re-
1. Copeland GP, Jones D, Walters M. POSSUM: a scoring system for surgical audit. Br J Surg 1991;78:355–60. 2. Wiejesinghe LD, Mahmood T, Scott DJ, et al. Comparison of POSSUM and the Portsmouth predictor equation for predicting death following vascular surgery. Br J Surg 1998;85:209–12. 3. Tekkis PP, Kessaris N, Kocher HM, et al. Evaluation of POSSUM and P-POSSUM scoring systems in patients undergoing colorectal surgery. Br J Surg 2003;90:340–5. 4. Griffiths H, Cuddihy P, Davis S, et al. Risk adjusted comparative audit. Is POSSUM applicable to head and neck surgery? Clin Otolaryngol 2002;27:517–20. 5. Whiteley MS, Prytherch DR, Higgins B, et al. An evaluation of the POSSUM surgical scoring system. Br J Surg 1996;83:812–5. 6. Prytherch DR, Whiteley MS, Higgins B, et al. POSSUM and Portsmouth POSSUM for predicting mortality. Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity. Br J Surg 1998;85:1217–20. 7. Prytherch DR, Ridler BM, Beard JD, et al. A model for national outcome audit in vascular surgery. Eur J Vasc Endovasc Surg 2001;21:477–83. 8. Neary WD, Heather BP, Earnshaw JJ. The Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (POSSUM). Br J Surg 2003;90:157–65. 9. Jones DR, Copeland GP, de Cossart L. Comparison of POSSUM with APACHE II for prediction of outcome from a surgical high-dependency unit. Br J Surg 1992;79:1293–6. 10. Brooks MJ, Sutton R, Sarin S. Comparison of surgical risk score, POSSUM and pPOSSUM in higher-risk surgical patients. Br J Surg 2005;92:1288–92. 11. Tekkis PP, Prytherch DR, Kocher HM, et al. Development of a dedicated riskadjustment scoring system for colorectal surgery (colorectal POSSUM). Br J Surg 2004;91:1174–82. 12. Tekkis PP, McCulloch P, Poloniecki JD, et al. Risk-adjusted prediction of operative mortality in oesophagogastric surgery with O-POSSUM. Br J Surg 2004;91:288–95. 13. Ramesh VJ, Umamaheswara RG, Guha A, et al. Evaluation of POSSUM and PPOSSUM scoring systems for predicting the mortality in elective neurosurgical patients. Br J Neurosurg 2008;22:275–8. 14. Eng J. ROC analysis: web-based calculator for ROC curves. Baltimore: Johns Hopkins University.