Introduction

Few patients do not survive an operation, but many will incur significant morbidity around the time of surgery. These adverse outcomes are considered together as perioperative risk . Patient assessment is the process of gathering information to determine the particular risk for a given individual undergoing an operation, which carries distinct procedural risks . Care of the patient is also dependent on the system in which that care is being delivered. Managing these system risks is key, and one of the most important functions of multidisciplinary preoperative risk assessment is to inform perioperative care planning for the patient.

Why is assessing risk important?

Informed consent cannot be obtained without an estimation and communication of risk. This must include both the risk of the procedure itself and the competing risk of non-intervention. Conveying risk should also help the patient to form reasonable expectations for the perioperative and recovery process. Assessment of risk helps clinicians, in discussion with patients, to make better decisions when considering if a surgical procedure or intervention is appropriate for a given patient with a particular clinical need. Surgical technique has advanced significantly resulting in shorter and less invasive procedures, allowing patients with higher levels of comorbidity to be considered for surgery. However, it is important to remember that surgery is not always appropriate for all patients.

Despite improvements in surgical tools and technique, complications remain common. Planning the patient’s journey through the hospital to allow timely recognition and management of these complications will reduce resulting morbidity. The best performing high-volume centres often have similar complication rates to low-volume centres, but achieve better outcomes due to robust early warning systems when a patient deteriorates triggering prompt clinical action, i.e. a low rate of ‘failure to rescue’. Risk assessment informs pre-planned perioperative management strategies, e.g. provision of postoperative critical care resources. As part of this process, identification of specific risks may also allow for pre-optimisation (also known as pre-habilitation) covered elsewhere in this book.

Adjusting for the level of risk present within a case-mix also allows for a fairer comparison of local or individual surgical outcomes against national standards.

It is therefore essential that clinicians endeavour to accurately estimate risk on an individual patient basis for every patient, particularly for those patients who are either comorbid and/or undergoing a major surgical intervention and are therefore considered to be high risk.

What are we aiming to identify in risk assessment?

Our aim is to identify the higher-risk patient, generally accepted to be any patient who has an estimated perioperative 30-day mortality of greater than 5%. Perioperative risk is a broad term which can be thought of as a combination of:

  • Patient risks : not related to the surgery per se, but specific to that individual patient

  • Procedural risks : related to the surgery itself and the anaesthesia required, and which would apply to any patient undergoing that procedure

  • Systems risks : quality of preoperative investigation/optimisation and perioperative care

Procedure-related risks are generally easier to identify and communicate. The development of surgical quality improvement programmes has allowed a more accurate appreciation of procedural complications and the rate at which they are encountered on a national, local or even individual surgeon level. Examples include rates of haemorrhage, infection, nerve damage or anastomotic leak. These procedure-related risks are also dependent on other factors, such as urgency and duration of the procedure, volume of blood loss and type of surgery undertaken. The National Institute for Health and Care Excellence (NICE) has attempted to stratify surgical procedures into different grades of severity to provide guidance on the use of preoperative investigations and estimate perioperative risk ( Table 4.1 ).

Table 4.1
Examples of surgical procedures by severity grading (NICE)
Grade 1 Grade 2 Grade 3 Grade 4
Upper gastrointestinal endoscopy Haemorrhoidectomy Amputation Gastrectomy
Vasectomy Varicose vein surgery Mastectomy Colectomy
Tooth extraction Adenoidectomy Thyroidectomy Renal transplant
Excision skin lesion Reduction of dislocated joint Prostatectomy Hip replacement

Patient-related risks are more complex, and thus more challenging to identify and communicate, as they are usually due to individual disorders of physiology. This may lead to an increased risk of a complication less commonly observed in the healthy population. For example, patients with chronic obstructive lung disease (COPD) could reasonably be expected to encounter postoperative pulmonary complications (PPCs) more frequently. Alternatively, it may lead to a decreased ability to tolerate a known complication of a specific procedure, e.g. a patient with significant cardiac disease may fail to tolerate the expected degree of haemorrhage, even if relatively moderate. Patient-related risks may be broadly classified as:

  • Major adverse cardiac events (MACEs)

  • PPCs

  • Perioperative acute kidney injury (AKI)

  • Perioperative cerebrovascular events and cognitive dysfunction

  • Other morbidities such as urinary tract infection, surgical site infection (SSI) and venous thromboembolism

  • Other risks related to an individual’s other underlying disease processes, functional status or nutritional status

Systems-related risks include seniority/experience of clinicians, availability of diagnostic investigations such at CT imaging, rapid access to emergency operating theatre, and provision of perioperative critical care. These risks are modifiable and many healthcare systems now have robust standards and quality improvement projects to assess compliance with accepted standards of best practice. Examples include the UK National Confidential Enquiry into Patient Outcome and Death (NCEPOD), and more recently the National Emergency Laparotomy Audit (NELA). This subject is discussed in detail in Chapter 3 .

How do we assess risk in the elective setting?

The timescale of non-urgent operating allows for robust risk calculation. For many patients, the process can be completed at the initial clinic consultation, based on an accurate history and examination augmented with basic point of care, laboratory and radiological investigations. The majority of low-risk patients will not require more than this and can be quoted the usual procedural risks. Although there is evidence that the gut instinct of an experienced surgeon may be more effective than risk modelling tools, , modelling has become more sophisticated since these studies were undertaken and many tools are now used routinely to help predict risk for patients with chronic conditions undergoing major complex procedures.

It is important to remember that most available tools predict risk as a 28-day or 30-day mortality and/or morbidity figure. This is an accepted standard within surgical practice and is useful for quality improvement work, but most patients will expect to survive for more than 28 days after their operation and will expect to return to at least their preoperative quality of life. The risk of longer term or permanent impairment to quality of life is a key consideration in undertaking the process of informed consent, particularly after the landmark Montgomery vs. Lanarkshire Health Board UK Supreme Court decision. This ruling has changed the requirements of the process of informed consent, such that it must now include communication of all risks, no matter how small, if they could be deemed to be of significance to the patient.

Risk prediction models and scoring systems

Risk assessment is often subjective, leading to significant variation in the prediction and interpretation of risk. Risk prediction models have been created to reduce this subjectivity. These range from the most basic which define population-level risk for a given severity of concurrent disease, to vastly more complex risk models which account for individual variation in risk. The reality is that quoting a patient a numerical risk is often unhelpful as few patients ever truly believe they may be one of those who contribute to mortality statistics. These models are therefore more useful to the operating surgeon when deciding if offering surgery is appropriate, and clinical experience must be used in conjunction with the numbers produced by these tools. Nevertheless, their use is widespread and has value, so some of the more commonly used tools are described in the following text. This review is by no means exhaustive and many other models are in use. When evaluating accuracy of risk prediction models, we present statistical data where available. The median value for concordance statistic (c-statistic) or area under the receiver operating characteristic curve (AUROC) value is used to compare discrimination of outcomes, with 95% confidence intervals (95% CIs) where available. For these tests, a value of 1.0 would represent perfect performance and a value of 0.5 reflects the score performing no better than random chance. When describing calibration, an observed/expected ratio (O:E) is provided, with a value of less than 1.0 representing overestimation of the predicted outcome and greater than 1.0 underestimation.

General scoring tools

American Society of Anesthesiologists classification

This five-point classification system for assessing the fitness of a patient prior to elective surgery was developed by the American Society of Anesthesiologists (ASA) in 1963 and is used worldwide. The five grades each represent an increasing level of comorbidity and physiological derangement ( Table 4.2 ). A sixth category was later added to encompass the patient who has had death diagnosed by neurological criteria and is brought to theatre for heart beating organ donation. The tool correlates with in-hospital mortality but predicts an average risk defined from population level data rather than individual risk, and does not account for the severity of the operative procedure planned. It has been validated in several studies across multiple surgical specialties, which report a c-statistic for accuracy of 0.77 (95% CI 0.59–0.93) and a calibration O:E ratio of 1.08. However, ASA scoring is a markedly subjective assessment; one study reported such marked variation in inter-individual assessment that the authors concluded that ASA score alone should not be used to predict risk. Although of limited use as a risk prediction tool, ASA score does have value as a simple, easy-to-use and easy-to-understand measure to convey a patient’s degree of preoperative ill health to the wider theatre team at the beginning of an operation.

The ASA classification remains a quick, simple, widely used and reasonably accurate assessment of surgical risk in both the elective and emergency settings.

Table 4.2
American Society of Anesthesiologists grades
I Healthy
II Mild to moderate systemic disease caused by the surgical condition or other comorbidity, but medically well controlled and not affecting daily life
III Severe disease process which limits activity but is not incapacitating
IV Severe incapacitating disease process which is a constant threat to life
V Moribund patient not expected to survive with or without the operation
VI Patient diagnosed dead by neurological criteria and whose organs are being removed for donation

Surgical Outcome Risk Tool – SORT (v2)

The original Surgical Outcome Risk Tool (SORT) was a collaboration between researchers at NCEPOD and the University College London Surgical Outcomes Research Centre in the UK. It was developed from analysis of 19 000 patient outcomes, estimating risk of 30-day mortality from six variables: ASA, complexity of procedure, urgency, patient age, surgical risk specialty (gastrointestinal, thoracic or vascular surgery) and the anticipated presence of malignancy. The original version demonstrated a c-statistic of 0.8 19 for accuracy when discriminating for prediction of mortality. The latest update of the score in 2020 to SORT version 2 (SORTv2) has been validated in patients from the UK, Australia and New Zealand, updated to take account of the experienced physician’s estimation of risk and to include surgery on all parts of the body. The newer model outperforms the older version with an AUROC value of 0.91, and is accurate in predicting outcomes in neurological and cardiothoracic surgery.

Surgical risk scale

The Surgical Risk Scale (SRS) identifies three factors as the main determinants of surgical outcome prediction: complexity of procedure (minor, intermediate, major and major-complex), urgency of procedure (elective, urgent, emergency) and ASA grade. It is simple to use, requiring only these three variables, but predicts average population outcome rather than individual risk. For mortality, it has a c-statistic of 0.85 (95% CI 0.66–0.95) and O:E of 0.81. Most of the more complex models are built on the three variables used in the SRS.

Physiological and Operative Severity Score for the enumeration of Mortality models

The original Physiological and Operative Severity Score for enUmeration of Mortality (POSSUM) score was developed in 1991 through multivariate analysis of 62 parameters from a heterogenous general surgical population, from which the most statistically powerful outcome predictors were selected. It uses 6 operative variables to estimate the risk of both mortality and morbidity following a specific surgical procedure, with adjustment for the individual patient by the inclusion of 12 physiological parameters ( Table 4.3 ). Values for each variable are bracketed and given a weighting of 1, 2, 4 or 8 to calculate an Operative Severity Score and a Physiological Score. The use of additional operative variables such as blood loss and the presence of peritoneal soiling mean that these must be predicted by the surgeon if the score is to be used preoperatively, but despite this limitation it has been one of the most commonly used risk scoring tools and has been extensively validated. It demonstrates a c-statistic of 0.82 (95% CI 0.47–0.95) and calibration O:E of 0.86 (95% CI 0–1.73) for mortality. For morbidity, the c-statistic is 0.75 (95% CI 0.56–0.84) and calibration O:E 1.0 (95% CI 0.8–1.44).

Table 4.3
Variables used for the calculation of the Physiological and Operative Severity Score for enUmeration of Mortality (POSSUM) score and its derivatives
Operative variables Physiological variables
Operation severity class Age
Number of procedures Cardiac disease
Blood loss Respiratory disease
Peritoneal contamination Electrocardiogram
Malignancy status Systolic blood pressure
Urgency Pulse rate
Haemoglobin concentration
White cell count
Serum urea concentration
Serum sodium concentration
Serum potassium concentration
Glasgow coma scale score

Portsmouth POSSUM

The original POSSUM model was found to overpredict mortality significantly, particularly in the lowest risk groups, and was revised by Whiteley et al. from Portsmouth, UK, to produce an amended predictor calculation termed Portsmouth Physiological and Operative Severity Score for the enUmeration of Mortality (P-POSSUM) with a closer fit to observed in-hospital mortality.

The P-POSSUM model has been extensively validated in many large cohorts , and demonstrates a c-statistic for mortality of 0.81 (95% CI 0.56–0.94), calibration O:E for mortality of 1.03 (95% CI 0.56–15.87), and a c-statistic for morbidity of 0.61. Despite this improved fit, the score still overestimates mortality in the elderly, and in lower-risk groups, e.g. elective surgery. It also overestimates mortality in certain surgical subspecialties, which has prompted development of specialty-specific POSSUM-based scores for major elective colorectal (CR-POSSUM), oesophagogastric (O-POSSUM) and vascular (V-POSSUM) surgery.

Colorectal POSSUM

Minimally invasive laparoscopic techniques and Enhanced Recovery After Surgery (ERAS) protocols have changed the risk profile of major elective colorectal surgery, widening consideration of surgery in an older population, and are increasingly applied to emergency work.

POSSUM and P-POSSUM models underpredict mortality in emergency CR surgery and lack calibration at the extremes of age in elective work. CR-POSSUM was developed to include elective and emergency procedures. CR-POSSUM was superior to P-POSSUM when predicting perioperative mortality in index and validation cohorts of almost 23 000 UK patients , across a variety of emergency and elective colorectal procedures, but a subsequent systematic review of 18 studies comparing POSSUM, P-POSSUM and CR-POSSUM in surgery for colorectal malignancy found P-POSSUM to have the greatest predictive accuracy for mortality. A study of patients undergoing elective sigmoid resection for either carcinoma or diverticular disease found that CR-POSSUM overpredicted mortality for malignant disease and underpredicted in benign disease, despite good accuracy for the group as a whole.

Risk prediction for emergency laparotomy is a unique challenge and is covered in greater detail later in the chapter.

Oesophagogastric POSSUM

There has been significant interest in developing a risk prediction model specific to oesophagogastric surgery, particularly in oesophagectomy where the rates of both perioperative mortality (5%) and major morbidity (∼60%) remain high.

The original POSSUM model discriminates poorly for morbidity or mortality in patients undergoing oesophagectomy and an adjusted O-POSSUM model (which includes age, physiological status, mode of surgery, type of surgery and histological stage) was developed, demonstrating improved accuracy for risk-adjusted prediction of death in the index cohort. However, a systematic review of further studies found that O-POSSUM overpredicted in-hospital mortality and did not identify patients at higher risk.

Several other models have been proposed, but a systematic review of the evidence for all these scores, including O-POSSUM, has concluded that no currently available model can be applied to clinical practice with any confidence.

You're Reading a Preview

Become a Clinical Tree membership for Full access and enjoy Unlimited articles

Become membership

If you are a member. Log in here