As diagnostic pathologists, our primary activity is the analysis of human fluids, cells and tissues. Our primary function, however, is to gather, process, synthesise and communicate information which is required for, or useful in the management of patient health. In diagnostic pathology, information is everything. It is our raison d’être, and our reports, by encapsulating this information, represent our contribution to patient management and public health.
In this issue and the previous issue of Pathology, two papers present evidence and a detailed analysis of what constitutes clinically useful information in the reporting of soft tissue tumours and radical prostatectomy specimens. 1,2 The multidisciplinary authorship includes lead clinicians and pathologists from Australasia and was convened as part of the National Structured Pathology Reporting for Cancer Project (NSPRCP) under the Royal College of Pathologists of Australasia (RCPA), Cancer Institute NSW, and Cancer Australia, funded by the Commonwealth Department of Health and Ageing. 3
The NSPRCP initially published protocols for the structured reporting of the six most common tumours in our community (breast, colorectal, prostate, melanoma, lung and lymphoma) but a further six followed and more are currently in development. Experience of related efforts in the USA, 4 UK, 5 Europe and elsewhere over the last two decades indicates that over 70 different tumour protocols will need to be developed and their currency maintained. This will be a significant burden on the pathology community but it reflects advances in medical knowledge, rapidly evolving targeted therapies and coupled bioassays and a corresponding clinical need for much more detailed and complex information in our pathology reports.
As presented in the two papers, the evidence-based assessment of prognostic factors for inclusion in pathology reports is a fundamental need and a pre-requisite, but it is only one part of the jigsaw puzzle which follows from computerisation. This includes the rapidly evolving eHealth environment, the need to develop and improve the functionality of Laboratory Information Systems (LIS) and the need to optimise communication of our reports, by improving electronic standards for messaging and developing standards for the display or rendering of reports more clearly and consistently in paper and electronic media.
Pathologists worldwide are under pressure to provide more for less, and the clinical need for more detailed reporting will require a re-examination of how we work and a significant and enlightened investment in improved information technology (IT) infrastructure.
THE SURGICAL PATHOLOGY REPORT
Our current model for reporting in surgical pathology has evolved from hand or typewritten ‘narrative’ reports of the last century. Apart from patient demographics, the standard headings included ‘Clinical History’, ‘Macroscopic Description’, ‘Microscopic Description’ and a final section which may be given as ‘Conclusion’, ‘Opinion’, ‘Diagnosis’ or ‘Summary’. This structure has withstood the test of time, and with the adoption of computers, pathologists worldwide opted to persist with the narrative paradigm by using word processors rather than structured data. Thus, computers facilitated report preparation and distribution but had little or no impact upon the structure or usefulness of our reports and, unlike clinical pathology testing, surgical pathology could not reap the benefits of analysis and decision support intrinsic to medical informatics.
Moreover, as we increasingly used clinical information and ancillary studies such as immunohistochemistry, cytogenetics, molecular analyses and flow studies to derive our diagnoses, two issues obfuscated our thinking. Firstly, the method for incorporation of ancillary studies is not clearly defined within this structure, and some pathologists have been reluctant to incorporate results from other laboratories, even when that information provides a critical part of their final conclusion. Secondly, once pathology data are derived from multiple modalities other than simply morphology, the traditional report structure does not make logical provision for a ‘Synthesis’ in which information from multiple modalities, often of varying predictive value, can be combined and weighed inferentially to derive higher elements, often part of the conclusion. Some have used ‘Comment’ or ‘Commentary’ as an additional heading to define this component. Others have developed their synthesis within the ‘Conclusion’ section. This represents a lack of consistency in the use of this section since most pathologists use it as a summary, i.e., a shortened repetition of what is already in the rest of the report, and increasingly it is used as a banner in the first part of a report, rather as a newspaper headline.
There are some advantages to the word processing paradigm. It is universally understood, intuitive and laissez-faire, allowing for ultimate reporting flexibility. In diagnostic pathology this flexibility is a double-edged sword, and has now become a primary impetus for change.
DRIVERS FOR CHANGE
Clinical patient care
Although many pathologists have enthusiastically adopted structured reporting in Australasia, others have eschewed it, and clinicians have been left with a mixture of reporting styles. Thus, it is our clinicians who have been the primary driving force behind the move to the uniform adoption of structured reporting. In the USA also, it was the College of Surgeons Commission on Cancer that initially mandated synoptic reporting for cancer program accreditation.
The clinical needs relate to both the clarity and the completeness of pathology reports.
Narrative documents are inherently impenetrable unless structured with headings and formatting devices to improve assimilation of key facts. 6 In 1991, Markel and Hirsch 7 reported that their surgeons had complained that key information was missing from their reports when it was in fact present. They described the now familiar ‘synoptic’ reporting style in which the key information of a report was included separately as a columnar list, in the form of a question and an answer or value. This practice has now been widely adopted, with the general approval of clinicians.
Most importantly, many studies have now conclusively shown that the use of synoptic reporting templates significantly improves the completeness of cancer reports across a broad range of tumour type. 8–12 It is likely that this improvement is two-fold. Whilst reporting protocols or checklists have a ‘reminder’ function akin to the checklists used by airline pilots prior to take off, there is also an educational component. The rapid growth of knowledge and ancillary testing required for pathological assessment of cancers has made the task of keeping abreast of advances across the broad range of specific cancers extremely difficult for pathologists. Reporting protocols ensure that all pathologists are kept abreast of the latest minimum reporting standards for all tumours.
It is not only the management of major internal malignancies that has driven this process. As all clinicians develop their practice management information systems and embrace eHealth, there is an increasing thirst for structured reports for all specimens, particularly those in which there is a degree of predictability and repetition, or restricted range of responses. This particularly includes skin excisions for tumours but it is not intrinsically limited to any site or neoplasia.
In addition to primary patient care, cancer registries across all Australian States and in New Zealand spend significant resources in manual collation and interpretation of cancer reports in an attempt to extract meaningful staging and prognostic data from narrative reports. Cancer data from pathology, particularly TNM staging, is a fundamental part of the Cancer Data Strategy of Cancer Australia 13 and it underpins the aims of the Australian Institute of Health and Welfare (AIHW) and the National Cancer Monitoring Advisory Group to monitor cancer nationwide, 14 yet TNM staging has not been universally adopted by pathology laboratories.
Atomic versus narrative reporting
Narrative reports are currently located in our computers as an indivisible string or block of text, unintelligible to the computer and incapable of manipulation for extraction of data except by the use of complex computer programs using advanced linguistic analysis. 15,16 By recording and storing data elements as discrete items (often referred to as ‘atomic’ or ‘discrete’ data) within the LIS at the outset, however, surgical pathologists can unlock the power of medical informatics and the myriad benefits that follow.
For end users, information from aggregated reports can be automatically searched, extracted, compared, sorted, and viewed as cumulative summaries. For example, we can then automatically provide plastic surgeons with monthly summaries of tumour types and clearances including average margins of excision as well as comparison with peer group norms, similarly extracted across data from multiple providers. These can be further sub-analysed by tumour type, or any other factors provided by the pathology report. Srigley et al. 11 have recently reported the Cancer Care Ontario experience, in which atomic data on prostate cancer resections was fed to central registries. By feeding this information back to individual institutions and closing the loop through providing knowledge transfer sessions, they were subsequently able to show a significant improvement in tumour clearance rates. This powerful and fundamental change, in which patient management and public health are continuously improved, will not escape the attention of healthcare funders.
For surgical pathologists, atomic data structures have many benefits. Firstly, we can simplify and verify the recording of information. Much can be captured by drop-down menus, tick boxes or other devices, thereby enabling conformity and reducing the need for narrative description. As a corollary, when narrative is required, it becomes intrinsically more powerful by nature of its rarity within a report and juxtaposition to a specific heading.
Secondly, use of atomic data enables dissociation of data entry from report formatting. Recording of data should follow pathologist workflow and convenience. Report formatting, however, should follow principles suited to efficient assimilation of key data by the clinician, and the two are often at odds. Moreover, this dissociation frees the pathologist from a need to consider formatting, since this can be pre-defined and performed automatically by the LIS. Most significantly, the requirement for the pathologist to repeat the dictation of Summary or Conclusion data in every report, of every working day, is finally obviated, since this can be generated automatically by the LIS using pre-defined syntax.
Thirdly, even more than for end users, the ability for the pathologist to extract, search and manipulate aggregated data from our cumulative life work will be invaluable in very many ways.
Finally, most importantly, and as demonstrated within the two index papers published in Pathology, 1,2 there is an increasing range of data required in our reports with a commensurate increase in complexity, number of decision points, and knowledge required by the reporting pathologist. Only by the use of atomic data can we harness the potential for decision support and verification by the LIS or expert system at our workstations. This can take many forms and its potential has barely begun to be explored (see Laboratory Information System Performance below).
THE AUSTRALASIAN NSPRC PROJECT
At the outset, the Framework Committee and Project Group were concerned that if structured reporting were to be adopted widely by pathologists, it would have to improve their reporting experience. Underlying this was the belief that enlightened IT implementation of well thought out protocols could actually speed up reporting. In particular, we did not believe that the current practice of creating a traditional narrative report, repeating it in ‘synoptic’ format, and then repeating the key elements again in the summary section of the report, made any sense – but it did offer considerable scope for improvement.
Several key principles follow directly from this approach, and account for the unique nature of the Australian Project. 17
1. The structured reports produced would supplant the traditional report.
2. Therefore, both Mandatory (Core) and non-Mandatory elements would need to be available for inclusion.
3. The sum of the Mandatory elements would be equal to a Minimum Dataset.
4. The protocols must be flexible, defining a universal report structure which may expand or contract on a case by case basis at the pathologist's discretion from a Minimum Dataset of mandatory elements to a full comprehensive report or Maximum Dataset.
5. There should be no repetition of data entry – repetition is reserved for the report output, as a formatting device and generated automatically by the LIS.
6. Discrete data responses should be used wherever possible but narrative should be available where necessary to record information that is beyond synoptic capture.
7. The data elements recorded should be stored and processed atomically as individual discrete elements.
8. The whole process must be electronically enabled and enhanced.
9. Inclusion of elements should be evidence-based wherever possible. Evidence in support of prognostic significance is assessed using the recently published National Health and Medical Research Council (NHMRC) Revised criteria. 18
10. Reporting would embrace multidisciplinary and clinico-pathological integration, with recognition that this would require an overarching component to the report or ‘Synthesis’.
11. There should be emphasis on report format and legibility.
Thus, whilst the RCPA/NSPRCP has followed the College of American Pathologists (CAP) and Royal College of Pathologists (RCPath) in developing cancer datasets that could be used in ‘synoptic’ reports, the RCPA protocols further attempt to define a universal information structure which allows the pathologist the flexibility to report a case using a minimum dataset or a maximum dataset with unlimited, but structured use of narrative – or anything in between – whilst preserving data conformity and universality. This is the essence of ‘structured’ rather than ‘synoptic’ reporting.
Structured versus synoptic reporting
Whereas ‘synoptic reporting’, as described by Markel and Hirsch 7 and subsequently widely adopted worldwide, could be viewed as a formatting device to make key report information more accessible to clinicians, ‘structured reporting’ is a wholistic approach that adopts all of the principles listed above and defines the underlying data structure rather than just the report content and format. A structured report is neither a synopsis nor necessarily a minimum dataset. It is whatever the pathologist chooses to report, provided that it includes at least the minimum dataset, and provided it is done by populating a predefined data structure with observations and inferences, both atomic and narrative as required by the protocols. There is no inherent limitation on the amount of narrative prose a pathologist may choose to use, but its use is inherently unnecessary unless required for subtle nuance, or conveying uncertainty, unexpected findings or any other information that is beyond synoptic capture.
Synthesis may be defined as ‘the combination of components or elements to form a connected whole’. 19
Although the term is not commonly used in pathology, the concept has always been an implied component of our reporting. Increasingly, as we use clinical information and multiple investigational modalities to make diagnostic decisions or inferences, these synthetic elements do not sit comfortably within any one section of the traditional report structure.
Synthesis is the term chosen to define the integration of all these composite findings into a coherent, balanced whole. As an example, in soft tissue tumours and in lymphoma, the tumour type is itself a synthesis since it depends upon integration of many findings, clinical, morphological and ancillary, and it is logically inconsistent to include it as a ‘Microscopic’ finding.
THE eHEALTH ENVIRONMENT
Structured pathology reporting cannot operate independently of the eHealth environment. Discrete or atomic data elements are of limited use if there are no standards for messaging the discrete data within health, or if hospital and medical practice software cannot make sense of the message.
In the long term, the goal of eHealth both in Australia and internationally, is that pathology reports and clinical messaging in general will be both machine readable and human readable.
The NSPRC Project is timely as the National eHealth Transition Authority (NeHTA) in Australia and many others internationally are actively engaged in defining the eHealth environment, and many pieces of the electronic jigsaw are in development simultaneously:
1. The RCPA in conjunction with NeHTA, Standards Australia and the Quality Use of Pathology Program (QUPP) and a broad range of stakeholders has begun the Australian Pathology Terminology and Information Standardisation Plan (APTISP). This large-scale project aims to develop information and terminology structures for all pathology reporting and requesting. This will include the binding of unique identifiers to every possible individual data element in our reports, using universal terminologies such as SNOMED CT, LOINC and others. As a preliminary, NeHTA has already commissioned the development of tumour-specific archetypes, which define the information structures and interrelationships of all the elements within the NSPRCP cancer protocols.
2. Standards Australia, in conjunction with a wide range of stakeholders in the IT industry is reviewing and re-defining the electronic diagnostic messaging standard AS4700.2 and the accompanying handbook HB262. These are respectively, an implementation and guide for the use of HL7v2.4 for diagnostics messaging. They will ensure that the structured pathology data, through conforming with the above Australian Terminology and Information Standards, can be distributed as discrete data elements and displayed to end users in a consistent and standardised fashion, whether it be electronic or on paper.
Internationally, Integrating the Health Enterprise (IHE), an organisation aimed at describing how to use standards together to achieve interoperability, has recently modelled the structure of Anatomical Pathology reporting 20 with a view to providing for their incorporation into the global health network. Thus far, the model is based upon the traditional report paradigm with additional synoptic elements. 21
As discussed in the preceding sections, this report structure, being derived from morphological approaches of the last century, is ill equipped to deal with our present multidisciplinary approach to investigation or the challenges of the future and it will be critical to review and develop this model further as surgical pathology reporting evolves.
INTERNATIONAL COLLABORATION ON CANCER REPORTING (ICCR)
Following the signing of Memoranda of Understanding between CAP and RCPA and between CAP and the Association of Canadian Pathologists (ACP) in 2009, there has been reciprocal representation on the expert panels from all three countries. In 2011, recognising that normalisation of cancer datasets is a prerequisite for international benchmarking, the CAP, RCPA, ACP and RCPath (UK) signed an agreement to develop future cancer protocols collaboratively. The goal of this collaboration is to produce common, internationally agreed datasets for cancer using the best approaches of each of the countries involved, together representing a population of over 450 million people. As a trial project the ICCR is assessing international core data sets, naming conventions and value lists for lung, melanoma, prostate and endometrial cancers, with the aim of reviewing these results at the Congress of the European Society of Pathology in Helsinki in August 2011.
Although important for international cancer statistics and benchmarking, there are other reasons why these efforts must succeed.
1. The burden of developing and maintaining datasets for over 70 different cancers will be better shared across the global expert community.
2. There are many countries in the early stages of adoption of structured reporting. If the ICCR datasets are developed sympathetically with the developing world, this could accelerate adoption of structured reporting worldwide.
3. Electronic implementation of datasets is expensive, and universal international interoperability of data structure and terminology will greatly simplify implementation. This applies not only to messaging and reporting but equally to laboratory information system functionality (see below).
LABORATORY INFORMATION SYSTEM (LIS) PERFORMANCE
Whilst data content standards such as those described in this issue, and the electronic standards for information structure, terminology and messaging are prerequisites, the key to rapid implementation of structured reporting is at the pathologist's workspace. Many currently available LISs for Anatomical Pathology have facility for structured data entry and storage, specifically for implementation of the CAP Synoptic Reporting Checklists. 4 In Australasia, however, most LISs are not designed to allow entry of the discrete components used in structured reporting. This will change as systems are upgraded or replaced in the near future, and functional LIS requirements for structured reporting in surgical pathology have been developed by the NSPRC/RCPA Project. 22 However if structured reporting implementation is to be successful, the pathologist/LIS interface must be improved so that reporting is not only more complete but faster. It is not sufficient to demonstrate that an LIS can accommodate structured reporting; it must be efficient, intuitive, value adding, and one of the best measures of performance will be pathologist satisfaction. This component has received insufficient attention to date; there are few published standards, and the look and feel of an LIS is often regarded as a trademark of the LIS vendor rather than a universal requirement. To begin to address these issues, the NSPRC/RCPA Project has developed Universal Design Requirements that are freely available and open for comment. 23
Decision support is a further key area for development which may be built into the LIS or interfaced between the pathologist and LIS as a stand-alone system. In its simplest form, decision support may consist of a simple hyperlink to the relevant RCPA protocol commentary which governs or advises on the way a cancer data element is measured or assessed. The repetition and predictability of structured data, however, lends itself to rules-based linking of data, where, for any given case, the diagnostic questions to be addressed and appropriate value lists are offered dynamically, based upon information as it accumulates during the reporting process. For example, the specimen type alone, recorded at the time of reception and before the lid has been removed, holds considerable predictive power for the likely questions and answers to be addressed. Subsequently acquired information then directs and redirects the process dynamically. This information cascade has the power to accelerate and improve reporting but it should not inhibit the use of structured narrative where required. Rules-based systems can be made much more powerful when the LIS is configured to accumulate agnostic patterns of data from large numbers of aggregated reports from which to infer rules. This latter method, sometimes including so-called ‘ripple-down rules’ (http://en.wikipedia.org/wiki/Ripple-down_rules) 24 has the advantage that it builds by itself. Thus every report can be informed by the accumulated experience of all previous reports.
ISSUES TO BE ADDRESSED
As with all NSPRCP/RCPA datasets, the two articles published in Pathology have included clinical information considered important or essential to the reporting of each cancer. 1,2 This is often lacking from pathology requests, and even the physician responsible for managing the patient may not be identified. The listing of this information is a first step in a process of education for our clinical colleagues, some of whom may not understand the consultative nature of pathology reporting. There are grounds for optimism that electronic requesting may enable ‘requesting support’ for clinicians, and the ‘Clinical’ section of the RCPA protocols is intended to provide the template for improvement.
Potentially, within any one patient episode, there may be many separate specimens. Any one tumour may reside within one or more of these specimens and any one specimen may have multiple tumours which may or may not be related to each other. The same issues apply to specimens from multiple patient episodes. Reporting in pathology has evolved as an episode-centric process and our reports reflect this. The end users, however, might benefit more from a ‘tumour-centric’ approach, more akin to the synthesis achieved at multidisciplinary team (tumour board) meetings. Perhaps there may be a case for a new class of pathology report; one which presents accumulated information relating to any given tumour, or tumour group. This would not be simply a cumulative report or summary function, since putting together information from multiple disparate specimens or episodes would require synthesis. Such a report would represent a sentinel event in the patient health record.
Asynchronous data and interim, supplementary and amended reports
Increasingly, our reports contain information from multiple investigational modalities and laboratories. Asynchronous accumulation of diagnostic information creates a dilemma for the pathologist in knowing what to report, and when. Agreement is needed regarding implementation of interim, supplementary and amended reports as this relates to cancer protocols.
Images can be integrated into structured reports and universal descriptors can be developed for the images, their annotations and captions, as well as linkage to appropriate report subsection or data element. 22 High level image analysis such as vector quantisation 25 has clearly shown that the images themselves contain powerful information. As these processes develop, we can anticipate that such metadata may also require their own fields as part of the report structure. It is early days, however, and a DICOM (Digital Images and Communications in Medicine) standard has only recently been devised for surgical pathology. 26,27
STRUCTURED REPORTING – THE FUTURE
The principles discussed above have been developed in response to clinical needs for the reporting of cancer. There is no intrinsic reason why this approach should be limited to malignancy and the same principles can be applied equally to the reporting of any surgical pathology specimen. In general, the benefits are greatest if there is sufficient predictability surrounding a specimen to devise rules governing likely or possible diagnostic questions and their responses, and especially if there are many such questions. Commercial expert systems are available, and are being further developed, which allow the creation and sharing of rules-based templates for structured reporting of any specimen type.
Evidence-based datasets such as those presented in this issue, are an essential starting point for the reporting of complex conditions, of which cancer is an exemplar. There are now many compelling reasons to take this information beyond Word processor-based ‘synoptic reporting’ as it is currently practised in Australia, to the adoption of structured reporting of discrete or atomic data, thereby bringing to surgical pathology the power of data aggregation, medical informatics and clinical feedback loops which are intrinsic to better patient care.
The author would like to acknowledge considerable contributions and support from Meagan Judge, Project Manager NSPRCP, and Professor Michael Legg, University of Wollongong.