<scp>ASA</scp> physical status classification: a relic of a bygone time?

The American Society of Anaesthesiologists (ASA) physical status classification system is a recognised tool to classify the physical fitness of patients before surgery [1]. Although the ASA physical status was not designed to predict outcomes, higher scores are consistently associated with higher mortality, morbidity and incidence of complications. With surgical cohorts increasing in age, comorbidity and complexity, there is an ever-increasing need to individualise the assessment of risk of adverse outcomes. The ASA physical status features alongside other patient, anaesthetic and surgical factors in several well-validated clinical risk prediction models, including the Surgical Outcome Risk Tool (SORT) and the National Emergency Laparotomy Audit (NELA) risk calculator [2, 3]. Whilst models are becoming increasingly sophisticated, they are blind to the individual patient and healthcare setting. In the UK, population-level data about surgical patients and anaesthetic practice remain scarce. The Royal College of Anaesthetists National Audit Project (NAP) activity surveys have perhaps the broadest denominator, albeit for a brief data collection period every 3–4 years. Between 2013 (NAP5) and 2021 (NAP7), the median age of patients having surgery increased from 50.5 y to 52.8 y, and there was a notable increase in the proportion of older (aged > 75 y) patients [4]. There was also an increase in both the absolute number and proportion of patients who were ASA physical status 3 or 4 undergoing surgery in almost every cohort. National outcome data are not available, so it is unknown whether these increases in age and comorbidity have led to increased risk. However, this period certainly coincided with major developments in peri-operative and enhanced recovery pathways, especially for patients who are older or at higher risk. Nevertheless, pathology- or procedure-specific data are collected and audited routinely in the UK. For example, a study of the National Joint Registry of elective hip and knee replacements in England and Wales linked individual patient data to other databases (including Hospital Episode Statistics and civil mortality data) [5]. This study showed that comorbidity data added little improvement to mortality prediction based on age, sex and ASA physical status. However, despite case ascertainment rates of > 95% for the registry, 23.4% of cases were not included due to failure to link to a Hospital Episode Statistics record and a further 5.7% of cases were omitted due to lack of consent for involvement in research. For comparison, NELA does not require patient consent and the most recent case ascertainment rate (using coded data as a denominator) was 72.1% [3]. It is plausible that these cohorts represent the population incompletely and that inclusion bias may ultimately affect subjective and objective assessment of peri-operative risk. Kilhamn et al. present a methodical and comprehensive analysis of a truly national peri-operative cohort derived from the Swedish Peri-operative Register [6]. The main conclusion of the study was that ASA physical status remains a useful and reliable predictor of mortality after major surgery, even after adjusting for baseline characteristics including age, comorbidities and socio-economic factors. For many clinicians, we expect that the main take-home message will be the clear discrimination of mortality risk between patients of all ages when stratified by ASA physical status 2, 3 or 4. The major strength of the analysis by Kilhamn et al. is the near-universal coverage of the nationwide surgical population and linkage with national health registers, with low rates of missing data and loss to follow-up. This reflects the completeness of Swedish national databases with an ‘opt-out’ approach to consent [7]. The inclusion of virtually all major, adult, non-cardiac, non-obstetric surgery from early 2019 to 2023 (discounting the period of COVID-19 restrictions) provides an updated understanding of mortality risk in the context of modern demographics. The use of days at home up to 30 days after surgery (DAH30) as a secondary outcome measure is progressive. This metric derives from chronic disease management research and was first evaluated in peri-operative care in 2017 [8]. A large Swedish multicentre study validated DAH30 and showed a strong correlation with measures such as 1-year postoperative mortality and incidence of other complications [9]. Compared with 30-day and 365-day mortality data alone, DAH30 provides a useful composite indication of the quality of patients' survival. Higher DAH30 values reflect successful peri-operative care, in that significant early complications including readmissions, discharges to rehabilitation or nursing facilities and deaths, have been avoided. These outcomes are considered consistently important by clinicians, patients and carers [10]. It has also been argued that DAH30 provides more statistical power than multiple low-incidence binary outcomes [9]. Conversely, when compared with other patient-centred or patient-reported outcome measures, DAH30 could prove more efficient and less prone to bias. It is difficult to imagine, for instance, how a national cohort study on this scale could collect and analyse peri-operative quality of life outcomes practically. The similar days alive and out of hospital at day 30 after surgery (DAOH30) measure was one of three patient-centred outcome measures recommended by the Standardised Endpoints in Perioperative Medicine initiative [11]. This has been validated in England through retrospective linkage of NELA patients with Hospital Episode Statistics admissions data [12]. In contrast, Kilhamn et al. omitted days spent in rehabilitation and nursing facilities from DAH30 values, which aligns with earlier studies. This is an important distinction, because discharge to such facilities has consequences for healthcare services and is considered an undesirable outcome by patients [13]. In addition to differences in reporting, there remain concerns that social and systemic factors could confound DAH30 and similar outcomes. In individualised risk prediction, most existing tools are designed to assess mortality risk, but some report multiple outcomes. The American College of Surgeons Risk Calculator presents the percentage risk of 13 outcomes, including specific complications (e.g. surgical site infection, pneumonia); return to the operating theatre; readmission; and discharge to nursing or rehabilitation facility [14]. Whilst this presentation might benefit some patients, DAH30 could be easier to communicate and interpret as a single composite measure of functional recovery. Kilhamn et al. showed clinically meaningful differences in DAH30 between groups of patients. For example, median DAH30 after acute surgery was 22 days for patients who were ASA physical status 3, but 11 days for those who were ASA physical status ≥ 4. Conversely, in surgery types with a high 30-day mortality (e.g. acute thoracic and neurosurgery), the DAH30 becomes less specific. The steady reduction in DAH30 with increasing age vs. the marked reduction with increasing ASA physical status adds further weight to the utility of ASA physical status as a useful classification tool in older patients. We can envisage a scenario in which more individualised prediction of DAH30 values informs the shared decision-making process. Kilhamn et al. defined major surgical procedures as “requiring anaesthetic personnel and lasting > 60 min and/or a total time in the operating theatre > 120 min”. Selecting cases by these criteria is made possible by the granularity of Swedish Perioperative Register data but potentially limits comparability with other studies. Beyond these inclusion criteria, the only surgical factor analysed as an exposure is surgical type, which shows a clear association with mortality risk. There is no ideal way of comparing procedures of different types, and any population-level analysis necessarily sacrifices granularity for scope. However, we can draw comparison with two important pieces of research. First, SORT uses a lookup table from the Specialist Procedure Codes database to code all procedures by surgical severity from ‘minor’ to ‘complex-major’ [15]. The mortality risk model based on this classification, when combined with subjective assessment by a clinician, probably remains the most precise model in widespread use [2]. More recently, a large study from the USA developed a novel Operative Stress Score [16]. The authors used a modified Delphi consensus to assign an ordinal score from 1 (‘very low’) to 5 (‘very high’) to quantify physiological stress for common surgical procedures in the context of pre-operative frailty. The observed interaction between Operative Stress Score and frailty (here assessed by the Risk Analysis Index) supports current concepts of peri-operative risk as an interaction between the physiological disturbance of an operation and the susceptibility of an individual to this. Given the granularity of procedure and timing data available to researchers, we are excited about the possibility that further work in this area will explore the relationship between surgical factors and outcomes on a similar scale. Kilhamn et al. acknowledge that including data on frailty would have added to the study, but perhaps this statement is more complex than it seems. Frailty is defined as “decrease in physiological reserve across multiple organ systems leading to increased vulnerability to external stressors” [17]. In peri-operative care, this manifests as a reduction in resilience that contributes to adverse outcomes, including increases in complication rates; duration of hospital stay; post-discharge dependence; and mortality rates [17]. It is accepted that frailty is a distinct concept from comorbidity and disability. However, there are numerous frailty assessment instruments, many of which blur the distinction between these constructs [18]. Furthermore, there is still no consensus on which is most appropriate in peri-operative research. The Clinical Frailty Scale (CFS) is relevant to anaesthetists because it has been recommended as a routine screening tool and has been integrated into routinely collected data (e.g. NELA case records). A recent revision of the CFS during the COVID-19 pandemic was accompanied by an excellent author commentary on its application [19]. The CFS summarises the judgement of a clinician on a scale from 1 (‘very fit’) to 8 (‘living with very severe frailty’) or 9 (‘terminally ill’). At CFS 2 an individual has ‘no active disease symptoms’, at 3 ‘medical problems are well-controlled’, but at 4, ‘symptoms limit activities’. Levels 5, 6 and 7 are distinguished by the order of daily activities requiring help. So, arguably, the CFS measures similar constructs to the ASA physical status at levels 1–4, i.e. the burden of comorbidities, but describes rather different phenotypes at level 5 and above. Numerous studies have shown an association between higher CFS values and adverse outcomes, but it is unclear whether adding CFS to existing risk models enhances their predictive ability. A Scottish study of 2246 patients undergoing emergency laparotomy found that the NELA mortality risk prediction model was not enhanced by addition of CFS [20]. Conversely, a more recent retrospective single-centre study found that CFS did enhance the NELA model in older patients [21]. Importantly, the NELA model used in both studies also included intra-operative findings. In contrast with tools based on clinical judgement, many instruments calculate a frailty index by counting deficits in health, often based on coded electronic health records or epidemiological datasets [22]. These include the Electronic Frailty Index, which was designed to identify people with frailty from UK primary care records based on 36 deficits [23]. Another example is the Pre-operative Frailty Index, which captures 30 deficits in Canadian registry data and has shown predictive validity in cardiac surgery and cancer surgery cohorts [24, 25]. In common with several other frailty index scores, at least half of the deficits are comorbidities (e.g. heart failure, cerebrovascular disease), which are assigned the same weighting as markers classically associated with frailty (e.g. weight loss, falls). Many studies have instead captured discrete markers of frailty. A broad meta-analysis showed that sarcopenia, the age-related decline in muscle mass and strength, was associated strongly with multiple adverse outcomes after surgery [26]. Radiological measurement of sarcopenia and other body composition markers like myosteatosis (intramuscular fat deposition) is practical in groups of patients with similar clinical presentations, for instance on cross-sectional abdominal imaging before laparotomy or colorectal resection. One study of 1043 patients undergoing emergency laparotomy showed that sarcopenia outperformed CFS as a predictor of 30-day mortality [27]. Another study of 610 patients having emergency laparotomy showed that these radiological data enhanced the discrimination of the NELA 30-day mortality risk prediction model [28]. Further work is needed to validate these models externally and examine the effect in broader surgical populations. The ASA physical status remains effective and used widely, in part, because it is a useful qualitative indicator of comorbidity: the classification of patients' comorbidities into mild, severe and life-threatening disease is simple and easily applied. However, longstanding concerns about accuracy and inter-rater reliability of ASA physical status have been upheld by recent work on the NAP7 activity survey, with evidence of underscoring in severe cardiovascular and cerebrovascular disease, obesity and acute presentations [29]. Despite this, it is unlikely that another peri-operative stratification tool will enjoy such widespread uptake and longstanding performance. Part of the beauty of the ASA physical status is its simplicity, both in application and understanding. Predicting outcomes with greater precision depends on an understanding of the interaction between the surgical insult and physiology of the individual patient. This kind of highly granular, bespoke data will be of interest to some patients; but it must be remembered that many patients will benefit from the simplicity of measures such as DAH30 during their consent journey. No competing interests declared.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.