Future of clinical coding
BMJ 2016; 353 doi: https://doi.org/10.1136/bmj.i2875 (Published 26 May 2016) Cite this as: BMJ 2016;353:i2875
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
I would like to thank Roger Weeks and John Mason for their reviews. Both make mention of the role of clinical coders, with one correspondent pointing out the importance of their work and the other questioning whether the system they have supported has served the NHS well. In this context it is important to appreciate that ICD10 is reported to the World Health Organisation. Commissioning relies on the grouping of this data into Health Resource Groups. Hospital Episode Statistics (HES) are extracted from the data annually and are linked to other important data sources such as the Office of National Statistics for mortality and the cancer register. Dr Foster intelligence including Hospital Standardised Mortality Ratios, NHS Choices and Patient Reported Outcome Measures are all users of this data. Universities and academic institutions are able to request HES data leading to numerous academic publications each year.
The second important theme is one of clinical engagement and how this might compare between the two systems. The current system does not require clinical engagement, a survey of consultants showed that many are oblivious to the data collection or they do not think it is any part of their responsibility1. The juniors are often involved by completing a Korner form at the time of discharge, which is now usually synonymous with the interim discharge summary. They are often unaware of the data extraction taking place. Furthermore, clinicians perceive the data to be inaccurate and unsuitable for audit and quality improvement2. In some instances, this is due to lack of clarity in the written notes but more importantly it is due to a lack of understanding of disease classification. Diseases are grouped together, but the outcomes for the diseases in the group might be different. Inevitably there are pockets of good practice where clinicians get actively involved in overseeing the clinical coding, this is often done in high cost specialties to ensure that income to the department is maximised.
The proposed use of SNOMED CT will require clinicians to be actively engaged in the accurate recording of clinical information at the bedside. As has been pointed out this may include symptoms and clinical signs as well as firm diagnosis, treatment and medication. The work that has already commenced with expert clinical coders and Royal College endorsed speciality groups on SNOMED CT content will ensure that the appropriate terms are available, easy to access and free of hierarchical errors. My assertion that clinicians will be able to use SNOMED CT and not ICD10 is based on the knowledge that SNOMED CT will contain the terminology used in everyday practice and as such is suitable for the medical records and all clinical communication. ICD10 uses terminology that is often obscure and because it is a classification system the exact meaning is lost as the condition is grouped with others.
As has been pointed out, training is also very important and to this end the RCPCH are including clinical coding and SNOMED CT in the new college training curriculum which is currently under development. It is hoped that other colleges will adopt this approach as curricula come up for review. Trusts will also need to provide training as part of the role out of electronic health records.
I am grateful to Weeks for highlighting the fact that the use of SNOMED CT has the potential to hugely improve patient care. Not only through the use of decision support in well advanced EHR’s, but also by providing data which is sufficiently granular to support monitoring of outcomes and quality improvement. Interest from the Royal Colleges and other professional groups in pursuing this agenda will do more to engage clinicians than any amount of propaganda.
(1) Spencer SA, Davies MP. Hospital episode statistics: improving the quality and value of hospital data: a national internet e-survey of hospital consultants. BMJ Open 2012; 2(6 bmjopen-2012-001651 [pii];10.1136/bmjopen-2012-001651 [doi]).
(2) Spencer, S A. Hospital Episode Statisitics (HES): Improving the quality and value of hospital data: A discussion document. 2011; http://www.aomrc.org.uk/doc_download/9379-hospital-episode-statistics-im...
Competing interests: No competing interests
Stephen Spencer's timely editorial on future clinical coding contains some errors which diminish the importance of his message that Clinicians (in primary and secondary care) and Social workers must all use the Systematized Nomenclature of Human and Veterinary Medicine Clinical Terms (SNOMED CT) concepts to populate a single transferable patient centred Electronic Health Record (EHR) (which he omits to mention) used and shared by everyone concerned including the patient or client for the primary purpose of high quality care (not mentioned either).
They are:
1) Describing the current activity of coders (who never see the patients) in hospitals as 'clinical coding' (clinical literally means 'bedside') whereas their activity is historical, i.e. not contemporaneous, and the coders never see the patients. This coding is definitely not clinical and its usefulness is highly questionable as it is not verified by clinicians. Also the need for hospitals to maximise their income sometimes leads to 'upcoding'. (see https://chpi.org.uk/nhs-greater-risk-fraud-new-private-providers-system-... ) which produces even more unreliable data.
2) Says that the coder system has served us well. Surely using a model based on US billing systems is questionable in a solidarity based healthcare system even when we have so-called payment by results (presumably a result is a treatment not an outcome). No actual bills are sent to payers so the elaborate and expensive solely administrative system would seem to be an expensive but entirely academic exercise.
3) Says that clinicians cannot be expected to record data using the International Classification of diseases 10 (ICD-10) for symptoms and diagnoses1 and the Office of Population Censuses and Surveys (OPCS) Classification of Interventions and Procedures version 4 (OPCS-4) for procedures. On the contrary clinicians are fully cognisant with the mainly diagnostic terms used in ICD 10 (particularly in their speciality domain) and surgeons must be the only clinicians really able to identify the actual operation that they performed in OPCS 4 . It should also be noted that all the diseases in ICD 10 and procedures in OPCS 4 are actually contained in SNOMED CT. The reason for moving to SNOMED CT is based on what it has in addition to diseases and operations.
4) Mentions the symptoms in SNOMED CT which have been developed from the rich set of such concepts in Read codes versions 2 and Clinical Terms Version 3 (CTV3) and the original SNOMED but omits to mention physical signs (found at examination of the patient), family history, social history, patient environment, adverse reactions and allergies, tests and their results, reports and interpretation drug therapies, physical therapies, and lots more.
5) Says SNOMED CT has been developed from an amalgamation of an American pathology database and Read codes. Actually SNOMED CT was created by combining the two versions of the Read codes with the Systematized Nomenclature of Human and Veterinary Medicine (SNOMED) and SNOMED RT (Reference Terminology) not the Systematized Nomenclature of Pathology (SNOP). This is important to know as SNOMED in any of its iterations was little (if ever) used in EHRs or medical computer systems. This is not to say that SNOMED is not an excellent piece of work, simply that it has not been tried and tested in the same way as Read codes which have been used in GP systems (covering virtually all UK patients now) since 1987.
6) Repeats statements from International Health Terminology Standards Development Organisation - responsible for SNOMED CT (ITHSDO) and Health and Social Care Information Centre (HSCIC) that SNOMED CT is not a disease classification but a terminology. However the presence of (potentially highly useful) multi-hierarchies (concepts can have more than one parent) means that that concepts are actually classified. Without these hierarchical classifications navigation through the 450,000 concepts would be even more difficult. Work done by myself has shown that there are many hierarchical errors which need to be corrected before we can rely on SNOMED CT for clinical coding.
7) Says that Pneumococcal pneumonis [pneumonitis Syn: pulmonitis - actually means 'Inflammation of the lungs' so I presume (Pneumococcal) pneumonia is what is meant ] is a child of Bacterial Pneumonia whereas it is actually a grandchild as shown in the simplified hierarchy/classification below:
Bacterial pneumonia (disorder)
Pneumonia caused by Streptococcus (disorder)
Pneumonia caused by Streptococcus (disorder)
Bronchopneumonia caused by Streptococcus (disorder)
Congenital group A hemolytic streptococcal pneumonia (disorder)
Group B streptococcal pneumonia (disorder)
Pneumococcal pneumonia (disorder)
Pneumonia caused by infection caused by Streptococcus pyogenes (disorder)
( see http://www.diseasesdatabase.com/snomed/53084003 )
This example demonstrates the value and existence of classification by hierarchy.
8) Whilst recommending SNOMED CT unequivocally we are not told how:
a) Clinicians and others are going to be trained to use SNOMED CT and
b) Errors are going to be recognised. As in any essentially untried, untested and unpiloted new technology errors in e.g. hierarchies (thus classification) and mapping from legacy systems are bound to (and do) occur. Read CTV3 maps are abundant in errors of commission (huge numbers) and omission (significant numbers) making transition from old systems to new fraught with danger to data accuracy.
The main potential value in the move to SNOMED CT coding in EHRs is support it can give for vastly improved patient care. Accurate coding of all concepts from patients' symptoms and signs to diagnoses, and treatments so that we can use intelligent systems to warn us of potential adverse events, interactions (drug - drug and drug - disorder) and evaluate the real outcomes (not procedures) of our therapeutic interventions. All other use-case needs can piggy-back off the data held in the accurately coded EHR.
In avoiding use of critical faculties in regard to the quality, implementation and training in the use of SNOMED CT Spencer fails to prepare us for the enormous task ahead. The danger is that clinical engagement in this laudable and potentially game changing process will be again fatally damaged by not forewarning us of the likely pitfalls.
Competing interests: I was the other author of the Read codes version 2 with the late James Read. My company Medical Intelligency Ltd. is devoted to working on implementation of SNOMED CT in our US based EHR 'Medbase'. Our current team of four are all volunteers but we hope to 'monetize' the fruits of our work in future.
It is timely to at last draw attention to the need for accurate coding. There will always have to be a structure to the data and hopefully clinical systems to aid this. Whether this data collection can be achieved without training and an understanding of, for instance, a primary diagnosis and conditions 'due to' will be another matter. Hospitals are required to return structured diagnostic and operative data using ICD10 and the UK OPCS operation classification. Professional coders are vital to support this and it is unfortunate that they are not mentioned.
Primary care has had the Read codes for years. Read3 became multiaxial and a more useful tool but was never implemented in primary care. The information from poor use of these tools can be disappointing. The easy option for Read users was to use the top of the hierarchy e.g. cardiac disease (and many prided themselves on 'knowing the code'), rather than coronary artery disease or mitral stenosis which would have been more accurate.
We need to improve the status of clinical coders, for instance with a degree course, and cross this work with medical education. Hopefully then the role of clinical coders could be of a QA officer whilst more mundane coding would pass through systems in both primary and secondary care abstracted from clinical notes made by informed clinicians. Des Spence your previous columnist could indeed 'draw up his list of codes on the back of an envelope' but it would be a picking list for speed of entry of material which experience showed to be common with the ability to search out the uncommon from the full dataset.
There is a golden opportunity to radically improve the transfer of summary information between primary, community and secondary care with information easily transferred between clinical systems. Structure would still be required e.g. family history means different things to a clinician and a social worker.
No mention was made of the dm+d Snomed drug and medical appliance codes which would, if properly implemented, allow accurate and vitally important drug lists to be exchanged without the endless re-entering of data in hospitals and primary care. Even better would be to connect the use of these codes with the supply of drugs to the NHS, requiring suppliers to use the codes as bar codes easily read e.g. on patient admission, again avoiding endless transcription.
Competing interests: Previously chair of the Academy of Medical Royal Colleges Information Group and member of Snomed Editorial Board
We GPs have been using the clinical coding systems based on Read codes for I suppose about 30 years. GP notes depend on them for searching , linking and auditing as well as remuneration through QOF etc.
It would help consistency if hospital records letters and discharge summaries incorporated these codes as they could be read and transcribed on to the primary care Systm one etc without the need to us to then transcribe them.
It is not new technology it has been around for 30 years but hospital colleagues are so utterly slow to cotton on to the power of search and audit using these even though the whole structure of general practice has depended on it for 30 years.So hardly the future just please catch up !
Yours sincerely
Angela Salter
Competing interests: No competing interests
Re: Future of clinical coding
This article articulates the challenges around moving towards the capture of coded data by clinicians at the point of care. The Professional Records Standards Body (PRSB) has worked hard over several years getting the views of its member Royal Colleges, developing and assuring clinical record standards. We are now talking to the Royal Colleges about how we can work together to support clinicians to implement the record standards.
For successful delivery of the vision described, several areas need addressing
1) Clinicians need to understand the need for accurate recording of data which will involve a degree of training and education in new ways of working
2) Suppliers need to work with clinicians to make the interface as simple as possible to use, thus reducing any resistance to adoption that a complex user interface may create
3) The clinical content for the record standards needs to be developed and assured so that information captured at the point of care accurately represents the clinical narrative, but at the same time is capable of being used for routine administrative purposes.
4) The reporting mechanisms need to be developed to recognise these changes and reduce the administrate burden of the routine data collections.
5) Widespread adoption of SNOMED-CT across health and developing informatics into social care will continue to improve the use of point of care information for secondary uses, importantly including integrated and long term health and social care
6) There is a need to encourage adoption through both the Royal Colleges and via CCIOs within local organisations to articulate the need for change, address any concerns and create a pull from front line clinicians who recognise the importance of capturing accurate information at the point of care to support the care of the individual patient, audits and clinical research
The PRSB is ideally placed to help drive forwards the challenges outlined about working with clinical and technical colleagues across the system. For more information on the work of the PRSB, go to the http://theprsb.org
Competing interests: Clinical Advisor to PRSB