15 year longitudinal study of mortality in a general neurosurgical practice.

David R. Sandeman

Consultant Neurosurgeon

North Bristol NHS Trust

Abstract

This paper analyses in patient mortality rates in a prospectively acquired database of all admissions under one consultant neurosurgeon over a 15 year period from April 1999 to March 2014. Out of 6006 admissions, there were 163 deaths (2.7%). Surgical mortality was 3.2% (129/4255 cases). In 105 of the surgical patients, the aim of surgery was recorded preoperatively as being ‘to save life’ (80%). 14 patients had surgery for pathology that was not life threatening (11%) - four had burr hole surgery for a chronic subdural haematoma, one died after surgical treatment for subarachnoid haemorrhage. Seven died from complications of tumour surgery and two following epilepsy surgery. A further ten died after endovascular treatment of an aneurysm (9%). The average annual mortality was 10.8 cases (SD +/- 3.06). 147 patients were admitted as an emergency (90%). Only five patients were admitted from a waiting list to be classified as elective retrospectively. The paper helps to define that vital element of neurosurgery practice, which involves the management of patients with life threatening pathology. Elective mortality rates may be a truer representation of individual surgical competence but these do not reflect overall neurosurgical practice.

Introduction

It is quite understandable that society should demand high standards of its surgeons. However surgical competence is a difficult thing to define and to measure objectively in a way that allows valid comparison between units and individuals. Recently crude mortality rates have been published for individual surgeons (1). For this data to be a valid as a comparator of surgical competence in surgical specialties, like neurosurgery, where so much activity is directed towards reversing the effects of life threatening pathology, it needs to be set against a background of a recognised baseline mortality, something which has yet to be defined. This paper is presented to try to address this.

Materials and Methods

Since August 1998, this author has collected data on his practice prospectively on Excel spreadsheets. Information on all admissions under his name has been collected on a daily basis. The dataset on each patient has included demographic data, data on clinical presentation, referral source, surgery performed, surgical complications, length of stay, surgical complications and outcome. The data has therefore included data on death in hospital, readmission and second operations, the main benchmarks for surgical competence. This paper focuses on the mortality data.

In addition, data on the aim of surgery and initial outcome from surgery has also been collected prospectively. This has allowed some assessment of the indications for and expected outcome from surgery. This, in turn, has allowed some assessment of the number of patients who died following surgery aimed, unsuccessfully, at reversing the life threatening effects of the presenting pathology. Referral source was also recorded allowing identification of those patients referred via outpatients, the standard definition of an ‘elective’ case in most series.

The dataset refers to all in patient deaths that occurred over a 15 year period from April 1999 to March 2014. The categories for each datafield collected of each patient were as follows:

The type of admission was defined as:

  • ‘Emergency’ – admitted on ‘take’,
  • ‘Urgent’ – admitted within 1 month of referral, irrespective of referral source
  • ‘Routine’ – admitted from an out patient generated waiting list (elective).

Referral source was coded as:

  • Emergency referral
  • Non ‘take’ – urgent referral
  • Consultant to consultant referral
  • General outpatients
  • Specialist outpatients – oncology / epilepsy etc
  • Planned readmission

The first three categories represent emergency referrals, the latter admissions booked through outpatient / waiting list process

The aim of surgery at each admission was defined in three ways:

  • To ‘save life’,
  • To improve the presenting condition,
  • To obtain a diagnosis.

Outcome at discharge was defined simply as:

  • Condition ‘improved’,
  • Condition ‘unchanged’
  • Condition ‘worse’
  • Dead.

The dataset does not include information on outcome after discharge so that 30 day mortality data is incomplete in this series. Nevertheless a comparison with the departmental Hospital Episodes Statistical (HES) data released for 2012/13 has been made.

The author has had a full time NHS contract throughout this study period, which has included a full share of the regional on call rota, emergency admissions accounting for 50% of admissions. His elective practice evolved through three phases during the study period from a traditional general practice, through one focussing on neuro-oncology, epilepsy and pain surgery to one where the main focus has been epilepsy surgery. In broad terms there was a 60:40 split between cranial and spinal surgery. The dataset includes private patients.

Results

In the study period, there were 6006 admissions, 4255 of which had surgery (71%). 2632 were cranial procedures (62%). 1555 were spinal procedures (37%) and 32 peripheral nerve operations (1%). There were 2872 emergency admissions (50%), 1731 urgent admissions (30%) and 1202 routine admissions (20%). Table 1 shows annual data for admissions and deaths in the study period. The total number of deaths during this period was 163 (2.7% of all admissions). The average annual mortality was 10.8 with a standard deviation of 3.06. The number of surgical deaths was 129 (2.4% of all admissions, 3.2% of the surgical cases). Deaths from cranial pathology numbered 153 (94%) and from spinal pathology 10 (6%) – Figure 1

57 of the patients that died did so as a result of trauma (51 cranial, 6 spinal) which represented 35% of overalldeaths. were recorded as secondary to trauma, 51 due to cranial trauma and 6 attributed to spinal trauma (35% of all deaths), 74 were due to spontaneous intracranial haemorrhage (45%), about half due to Subarachnoid haemorrhage alone. 21 patients with tumours died, deaths occurred in tumour patients,17 cranial and 4 spinal(13%). Six deaths were due to infection, five cranial and one spinal (4%). 3 occurred as a result of end stage complications of congenital problems – spina bifida and two Arnold Chiari malformation patients and two occurred in patients undergoing surgery for epilepsy.

Emergency admissions accounted for 147 deaths (90%), whereas only 15 patients (9%) died after urgent admission. One routine admission, a patient with epilepsy died. 147 of patients were admitted as emergencies (90%) of admissions, 15 were urgent admissions (9%). Only one, an epilepsy patient, was admitted as a routine case (1%). Five patients were admitted via outpatients and therefore could be defined as elective admissions (3%). In those undergoing surgery, the stated preoperative aim of surgery was recorded as ‘to save life’ in 105 patients (80%), to improve their presenting condition in 20 cases (16%) of cases and to establish a diagnosis in 4 cases (3%).

Examining the detail of the different categories of admission gives some insight into the cause of death.

Cranial Trauma. Of the 51 patients dying from cranial trauma, 39 died from an acute head injury and 12 from a chronic subdural haematoma. 18 had a primary brain injury either in isolation or combined with multiple trauma (46%). In 12 of these, no neurosurgical procedure was performed (67%), 6 had a burr hole procedure for intracranial pressure monitoring or insertion of an external ventricular drain (33%). Nine patients presented with an intracranial haematoma in addition to a primary brain injury, all of which underwent emergency surgery to try and reverse a rapidly deteriorating condition (23%). A further 15 presented deteriorating from a surgical haematoma (Acute subdural haematoma -11, Extradural haematoma - 3, intracerebral haematoma – 1) without obvious other brain injury (38%). The average age in the acute cranial trauma group was 59 years with a range of 3 – 94 years).

13 patients died following surgery for chronic subdural haematoma (24%). In this group the average age was 78 years (range 31 – 94 years). 9 patients in this group presented inextremis. One wasn’t operated on. Eight died following surgery. In four patients the aim of surgery was recorded as being to improve their neurological deficit rather than to save their life.

In this analysis, only 4/51 patients presented in a condition that was not immediately life threatening (8%). None were admitted via outpatients

Spontaneous Intracranial Haemorrhage. 74 patients are included in this group, 33 presented with a coma producing subarachnoid haemorrhage (45%). A further 26 patients presented in coma with an intracranial haematoma (35%) – 18 with intracerebral haemorrhage, 3 with intraventricular haemorrhage and 2 with a spontaneous acute subdural haemorrhage. Of these 59 poor grade patients, 11 patients were not operated on (19%). 48 patients had surgery to stabilise their condition - in 34 patients the primary procedure was a burrhole for either external ventricular drainage (30 patients) or intracranial pressure monitoring (4 patients). 12 had a craniotomy, 5 to evacuate haematomas alone, 5 to evacuate an intracerebral haematoma and clip the underlying aneurysm and 2 who had haematoma evacuation combined with coiling. Three patients had coiling alone. 14 patients presented with non coma producing haemorrhage. Three died without surgery, one from a rebleed and two from stroke. One died from vasospasm after craniotomy, Ten died after coiling, including one elective coiling.

In this analysis, the majority of surgery was performed in poor grade patients to try and reverse a life threatening presentation. Only 11 procedures were done on stable patients and only one of these underwent neurosurgery, it being the unit policy to attribute all coiling deaths to the admitting surgeon rather than the endovascular neuroradiologist. (In the author’s view,if crude mortality is to be used as a benchmark for surgical competence then it would be appropriate to classify these cases as non surgical). None were admitted via outpatients.

Cranial Tumours.Of the 17 patients that died following admission with a tumour, Two were not operated upon, one presenting with fixed pupils and one from a pulmonary embolus prior to surgery. 6 had surgery which failed to reverse rapidly progressing raised intracranial pressure (two burr hole procedures and four craniotomies). Seven died from complications of surgery, three from postoperative haematoma, two from pulmonary embolus and two from complications of surgery for recurrent glioblastoma - hydrocephalus and wound breakdown

In this analysis 7/17 patients (41%) died following operations that were not aimed at saving life. Three patients were admitted via outpatients.

Infection.All five deaths in this group were in patients undergoing life saving surgery to try and stabilise their deterioration, from meningitis in three cases, a cerebral abscess in one and a subdural empyema in another. All were admitted as emergencies.

Epilepsy.There were two surgical deaths on the epilepsy surgery program, including the one patient in this series admitted as a routine admission. One patient had an urgent frontal lobectomy went into status postoperatively. One patient who had a seizure induced cardiac arrest following drug reduction for stereo EEG. Both patients were admitted via outpatients

Spinal pathology. Of the 11 patients dying of spinal pathology, 6 patients died following cervical trauma, 2 without surgery and 4 following spinal fixation. One patient died from systemic effects of spinal infection and four with spinal malignancy, two of which did not undergo surgery. None were admitted via outpatients.

Deaths following non emergency surgery.In the series overall, 17 patients had surgery to improve their presenting condition four chronic subdural cases, one spontaneous haemorrhage case, six tumour cases, one epilepsy case, one congenital case and four spinal cases. Four had surgery to establish a diagnosis – three tumour patients and one epilepsy patient undergoing stereo EEG, a total of 21 deaths from pathology that was not immediately life threatening (0.7% of non emergency admissions). Five patients were admitted through outpatients, three tumour patients and two epilepsy patients (0.2% of non emergency operations)

Comparison with National Neurosurgical Audit Program (NNAP) preliminary data on unit mortality 2012/13 (2)

Table 2 compares the Hospital Episode Statistical (HES) data for 2012/13 for the Bristol neurosurgical unit with the longitudinal data from this study. Although the HES data presented here was preliminary and unvalidated, there is nevertheless good correlation between the individual longitudinal data and the annual departmental data for 2012/13, the overall percentage mortality being the same with a similar range of mortality between adult neurosurgeons as seen over the years in the author’s practice.

Discussion

This data is presented to help define what, historically, a normal baseline for mortality has been for a surgeon with a mixed emergency and elective practice and a fairly even spread between cranial and spinal cases. What is immediately apparent is that, unlike many other surgical disciplines, the nature of neurosurgery is such, that a ‘baseline mortality’ is an integral part of normal neurosurgical practice. In this series the average annual mortality was 11 deaths per year (2.7% of admissions) with a range from 6 to 16 (1.7 to 3.8%). The standard deviation was 3, allowing for an ‘acceptable’ range of mortality as defined by the NNAP from 0 to 17 (+/- 2 x SD). The author acknowledges that no attempt has been made to carry out any more sophisticated statistical analysis on the smaller sub groups

Attempts have been made to establish baseline mortality in other healthcare systems. Hammers et al looked at institutional mortality over a 12 months period using data generated from monthly mortality meetings (3). In their series there was a higher percentage of elective deaths (15%) suggesting that, in the USA, neurosurgery may not have quite the same role in managing life critical pathology as happens in the UK.

Historically the difficulty with assessing information like this was the lack of comparable data. The recent publication of mortality rates for all surgeons for the year 2012/13 generated from Hospital episode statistics (HES data) provides a comparable dataset (2). For the author the overall numbers were comparable but the detail was not. The national figures recorded seven deaths against his name as opposed to six in this series. (The national dataset included a patient that died within a month of admission but of a cause unrelated to that admission). However of the six patients recorded in the national database, only three correlated with this dataset. Three patients were wrongly attributed to the author by the national audit process and three were not attributed to him that should have been. These included a patient who died from a direct surgical complication that was wrongly attributed to a colleague. Although one can make the assumption that inaccuracies in locally collected, trust data will apply to all surgeons in a unit equally, it is less valid to assume that discrepancies are balanced out across the country. Accurate local data collection is essential if valid comparison between units is to be made in a national dataset.

Comparing the mortality rates between adult surgeons in North Bristol NHS trust in 2012 shows the same range of mortality as this longitudinal study, an average of 11 deaths per year with a range of 7 to 16. This suggests that an annual mortality rate between 2 and 4%, does represent ‘normal’ practice in a busy regional neurosurgical unit.

Given the spread of pathology that contributed to this mortality data it is interesting to speculate what factors influence mortality rate, Firstly definition of surgical mortality is crucial. Should this include cases that have not undergone surgery? Excluding these would reduce overall mortality by 20%. Admission policy for key conditions is also key, the most obvious of which is poor grade subarachnoid haemorrhage. A national policy not to transfer these patients for neurosurgical care would reduce neurosurgical mortality by a further third! Timing of surgery must also be relevant, although not easily defined in this data set. An internal audit of cases referred from outside the normal Bristol catchment area showed a greater than 75% mortality (unpublished data) which must reflect the delay in referral. Case mix is also relevant. Mortality following spinal surgery is an order of magnitude less than cranial (0.4% compared with 4.7%). It should be noted that this series does not include any deaths following elective spinal surgery. In this area mortality data may not be a very good benchmark of surgical competence – analysis of patients whose condition has been made worse by surgery would be more relevant, although harder to quantify in any comparable way.

So how does one use mortality as a bench mark for surgical competence in neurosurgery? One approach would be to assign only those deaths in patients presenting with ‘non lethal’ pathology who had surgery i.e. those patients whose indication for surgery was not to save life - 21 cases in this series (0.7% of non emergency admissions) However this figure comes from an assessment of indication for surgery done prospectively. The definition of elective surgery from hospital information systems usually relates to patients admitted via the outpatient / waiting list system. This gives an elective mortality of 0.17% in this dataset.

The disadvantage of using only elective mortality as the benchmark figure is that it does not account for the ‘buck stops here’ role that neurosurgery provides for medicine as a whole. This is a role that is becoming more prominent in the modern era of trauma centres and multidisciplinary team working when increasing numbers of patient are admitted under neurosurgery without any reference to the admitting neurosurgeon and where case management is becoming increasingly protocol driven. It is important that this role is recognised and acknowledged by both the public and the medical profession. These figures suggest that a ‘death a month’ is a reasonable figure that accounts for this element of modern neurosurgical practice, but this is a figure that will need to be reviewed in the future as one consequence of using crude mortality to define individual competence will be an increase in referral of patients in extremis to neurosurgery.