Disease coding for anaesthetic and peri‐operative practice: an opportunity not to be missed

Diagnoses and outcomes in UK primary and secondary care systems are currently coded using the International Classification of Diseases (1CD-10) 1, These and other data are submitted to the Secondary Users Service 2, which provides information to clinical commissioners and stores data in the Hospital Episode Statistics database 3. Hospital Episode Statistics data are used to calculate hospital payment for services, such as payment by results 4, and to assess effectiveness of care, monitor activity trends, measure clinical quality and even to offer patients a choice of care provider 5. One might expect that anaesthetic interventions and complications are explicitly coded, but despite anaesthesia generally being an early adopter of technology, this is not the case. Here, we explore the history of coding, why and how anaesthesia has missed out, and the possible solutions. Disease and outcome coding was probably first attempted in the 17th century by John Graunt, whose London Bills of Mortality recorded causes of death and childhood survival rates 6. Florence Nightingale subsequently used outcome statistics (and innovative charting) during the Crimean war (1853–1856) when lobbying for better equipment which thereafter decreased infection rates 7. In 1853, the International Statistical Congress (later Institute) recommended an international classification of causes of death. Early efforts by Marc d'Espine of Geneva linked cause to process (gouty, herpetic, degenerative, etc.), while William Farr of Shropshire used anatomical causation. Farr's system led in 1900 to the adoption of Jacques Bertillon's International List of Causes of Death, the system now known as the International Classification of Diseases (ICD) 8. This was formalised and responsibility devolved to the World Health Organization (WHO) in 1948 9. The current iteration used in the UK, ICD-10, contains 17,935 disease and procedure codes, whereas the modified system used for billing by Centers for Medicare and Medicaid Services (CMS) and the National Center for Health Statistics (NCHS) in the USA, ICD10-CM, has about 150,000; an important difference to which we will return later. Other systems arose in addition to ICD-10. In the 1980s, Dr James Read, a Loughborough GP, created the eponymous ‘Read Codes’. The Royal College of General Practitioners and the British Medical Association recommended national adoption by GPs and this was implemented by the Department of Health in 1990 10, 11. The initial system was ‘4-byte’ (four levels of alphanumeric hierarchy) superseded by the Read v2 ‘5-byte’ version and then Clinical Terms Version 3 (CTV3). The alphanumeric construction and persistence of obsolete terms are limiting factors, and CTV3 is due for retirement by 2018 12. SNOMED-Clinical Terms is an encyclopaedic system of clinical and pathological terms published bi-annually and its use as the single terminology for healthcare in the UK was endorsed in ‘Personalised Health and Care 2020’ 12, 13. SNOMED was developed from the 1965 Systematized Nomenclature of Pathology (SNOP) and after merging with CTV3 formed SNOMED-CT in 2002 14. Currently, UK hospitals use the coding systems ICD-10 and OPCS -4 (Office of Population Census and Surveys Classification of Interventions and Procedures (formerly London, UK)). Both classifications suffer from limited lexicons and long delays for new content insertion. Neither is suited for recording anaesthetic outcomes, diagnoses or complications. OPCS-4 includes anaesthesia only as part of the surgical procedure codes; the one anaesthesia-specific outcome is ‘X59.1 Pre-operative anaesthetic death’ i.e. why surgery didn't happen! ICD-10 categorises diseases and complications by letter and number and is somewhat Victorian in terminology 1. Chapter I concerns ‘certain infectious and parasitic diseases’ coded A00 to B99. Potentially, in-theatre complications fall into Chapter XIX ‘Injury, poisoning and certain other consequences of external causes’ coded S00–T98, but anaesthesia is relegated to ‘Complications of surgical and medical care, not elsewhere classified’ (T80–T88; Box 1). An external cause code from Chapter XX ‘External causes of morbidity and mortality’ (V01–Y98) identifies the circumstance. During the 5th National Audit Project (NAP5), the acronym AAGA (accidental awareness under general anaesthesia) was coined by one of us (TC) and the NAP5 report recommended the creation of a database of future UK AAGA reports 15. Given that AAGA is a complication that may reflect human error, equipment failure or pharmacogenomic variability in patient drug response, we looked for an ICD-10 code, but without success. We therefore asked the WHO, via the Clinical Classifications Service (part of NHS Digital and the World Health Organization-Family of Classifications (WHO-FIC) Collaborating Centre for the UK) for AAGA to be included in ICD-10. Although the WHO agreed that such an outcome could be classified, they did not accept that it required separate coding, classifying it instead as a subset of T88.5 (Box 2). Note that in this coding system, if AAGA occurs during obstetric anaesthesia, the obstetric nature ‘trumps’ the anaesthetic aspect, even if the cause is secondary to equipment failure. These developments were both heartening and disappointing. Yes, AAGA reached the bar for ICD-10 recognition, but it didn't warrant a named separate entity, despite the fact that the much larger US version (ICD-10- CM) already contains the codes: T88.53 Unintended awareness under general anesthesia during procedure. T88.53XA …… initial encounter, T88.53XD …… subsequent encounter, T88.53XS …… sequela. Because of the separate nature of ICD-10 and ICD10-CM, these codes cannot be used in the UK, so trawling ICD-10 data for this specific complication is impossible; the same UK codes could equally apply to postoperative nausea and vomiting, prolonged drug effect or inadequate analgesia, etc. Moreover, the same barrier applies to other anaesthetic codes. In ICD-10, T88.4 tells us tracheal intubation was difficult but not why, although association with, for example, an obstetric code might suggest the cause. T88.3 records malignant hyperthermia, but not the triggering agent. Even where codes exist, outcome recording can fail or be inaccurate 16, 17. The clinician may record the outcome inaccurately, not at all, or in a way that cannot be identified by a clinical coder, so unless a subsequent hospital letter prompts a GP to code the information, it disappears and cannot be recalled or interrogated. Interestingly, despite the unfamiliarity of the UK anaesthetic community with coding, T88.4 (failed or difficult intubation) was coded 940 times in 2014–2015, suggesting that this complication is either recognised by coders or recorded by clinicians 18. There are very few national anaesthesia databases other than the malignant hyperthermia database in Leeds 19 (which uses ICD-10 to code for billing purposes) and the National Reporting and Learning System (NRLS), which analyses patient safety reports 20. Herein lies another clue as to why anaesthesia has missed out in coding terms; coding is arguably driven by billing (hence the huge number of codes in the US version of ICD-10) and the lack of anaesthesia's direct involvement in billing may explain its absence from coding lexicons. Whether the nature of paper-only anaesthetic records has contributed to this lack of coding/billing is a matter for debate, but the advent of electronic patient records (EPR) changes the potential for coding in anaesthesia. Specific codes for adverse events such as malignant hyperthermia, difficult and failed tracheal intubation, or complications of peripheral and neuraxial blockade, can be preselected within an EPR to make coding easy for the clinician. As well as enabling hospitals to bill for the increased clinical activity required in these situations, it also permits continuous data collection and analysis, highlighting regions, hospitals, or even individuals with different reporting rates. The main problem with ICD-10 is that it does not have the detail needed to capture, identify and return rates of anaesthetic complications in a way that allows meaningful retrospective analysis. A complication may be correctly recorded and identified but the generic nature of the codes (except possibly malignant hyperthermia) makes subsequent analysis meaningless. The solution to this problem may lie with SNOMED-CT, which already contains multiple terms relating to anaesthetic complications and outcomes as concepts (Box 3). These terms are the ‘fully specified names’ for each concept and each includes a ‘semantic tag’ (in brackets) which indicates the type of concept it is (i.e. in which hierarchy it sits). Concepts also have ‘preferred terms’ – names they are commonly known by, without the semantic tag, and may also have one or more synonyms. The concepts and terms are linked together with attributes. Thus, the finding of ‘awareness during anaesthesia’ can stand alone or be part of the ‘situation’ of a patient history of awareness (Fig. 1). The system allows a report of awareness (699030008) during apparently adequate anaesthesia to be linked in time to terms 250806001, 364676005, 250840001 relating to end-tidal anaesthetic agent concentration, observation of anaesthesia and depth of anaesthesia, respectively. The authors are liaising with SNOMED-CT to explore how the specialty can use the existing SNOMED-CT content for these purposes. There is a strong argument that it is time for UK anaesthesia to adopt SNOMED-CT to enable reporting, recording and analysis of its own outcome data and thereby address the weakness of specialty-relevant data collection. Outcomes (and associated costs) should drive service improvement, and coding allows both local and national identification, analysis and comparison of clinical activity and outcome. Accidental awareness during general anaesthesia is an adverse outcome, the documentation of which is now recommended by a national report (NAP5). Current UK coding systems (ICD-10, CTV3) do not include this outcome and although AAGA will be categorised within ICD-10 in 2017, it will be as a generic complication in relation to drug administration and not identifiable as the cause. SNOMED-CT, the world's largest clinical terminology system, will replace CTV3 and ICD-10 by 2020 and already includes terms that relate to many anaesthetic complications. If the specialty is to achieve its stated aims, national bodies that lead UK anaesthesia will need to formally engage with coding systems to ensure that anaesthetic outcomes are represented. Coding for peri-operative medicine and intensive care should also be considered 21. An historical barrier has been our collective ignorance of the subject (which this paper hopes to rectify), but one future driver is that where a complication is coded, it can be billed for. An opportunity presents itself with the introduction of EPR and electronic records to anaesthesia. The fundamental principle would be to adopt SNOMED-CT terms in official documents such as the Royal College's Guidance for Professional Anaesthetic Services (GPAS) 22 and the many and various AAGBI's guidance documents about anaesthetic practice. Once recognised as a driver for change, electronic records should be embedded in GPAS and used as a level-one standard for Anaesthesia Clinical Services Accreditation by the Royal College of Anaesthetists 23. The authors gratefully acknowledge the assistance and advice given to them by Ms Sheree Hemingway, Advanced Terminology Specialist and Andrew Brooke, Senior Classifications Manager at NHS Digital. No funding was received for this work. JJP is an Editor of Anaesthesia; the other authors declare no competing interests.