The Safe and Sound project is exploring two ways of delivering clinical decision support and other knowledge based services to the point of care
We believe there are three key problems with centralised orchestration of clinical services that the novel framework of Safe and Sound aims to address:
Centralised systems are inherently impractical when extended to very large scales.
One-size-fits all
We see three issues under this heading:
1. Traditional clinical information systems are often under-used or rejected because of their poor fit to users' needs and expectations.
Baseline: In the traditional centralised approach a single standard guideline is developed to cover all institutions (e.g. a national guideline covering all health authorities, or a trust-level guideline for all hospitals in a trust). The users of the guideline in the various institutions must adapt to its assumptions, because the guideline cannot adapt to the idiosyncratic workflow of the users while remaining "standard". The resulting burdens and resentments are well known.
Experimental: In the choreography approach standard guidelines cover only the general, invariant aspects of care common to every institution. Lower-level details of how these invariant tasks are to be carried out are not specified as part of the high-level guideline but are determined locally. An institution may comply with the standard guideline while staff follow a locally determined workflow, possibly different to that at other institutions where the guideline is deployed.
2. Centralised guidelines do not adapt well to the fact that the structure of the care team carrying out the guideline may change radically from institution to institution.
Baseline: A centralised guideline either specifies all clinical tasks without regard for who carries them out, or specifies a particular "standard" division of jobs among team members which may well not fit the way things are done in a particular institution. Many medical errors are associated with poor communication or distribution of tasks among members of the care team, so this mismatch increases the potential for error.
Experimental: In the Safe & Sound approach high-level guidelines are specified in terms of clinical roles, without defining which individuals take those roles on a particular occasion. In one institution a single individual may fill several different roles; in another the roles may correspond to different individuals. Clinical tasks are distributed to whoever happens to fill the corresponding role while maintaining compliance with the overall guideline.
3. Centralised systems do not readily adapt to real-world healthcare processes, which are highly dynamic and include a great deal of uncertainty. Clinical staff can change frequently and unpredictably, and the clinical situation may evolve rapidly and unexpectedly.
Baseline: Standard pathways may cover the "nominal" clinical process, but it may be rare for a patient's journey through the healthcare system to be entirely nominal. The clinic team may change unexpectedly, the patient may not respond as expected, and the pathway or clinical guidelines may not be followed to the letter. Attempting to cover for all contingencies contributes to complexity, lack of portability, difficulty of review and audit, and increased probability of error.
Experimental: In the Safe & Sound approach it is possible for high-level guidelines or pathways to define the standard clinical process with subsidiary guidelines to resolve exceptional situations. These subsidiary guidelines can themselves be made sufficiently general that they can be easily checked, standardised and shared across institutions and across medical conditions.
Centralised systems do not easily handle more complex situations like co-morbidities and third-party participants.
1. Co-morbidities.
Baseline: When a patient is under treatment for two or more conditions it is entirely possible that the specific combination of treatment pathways will not have been developed together or tested in combination. Mismatches in terminology & responsibility and clashes between incompatible treatments are likely, leading to medical error.
Experimental: Multiple guidelines may be active in parallel. The lower level details of clinical tasks will be instantiated with respect to the current clinical context, allowing interactions to be checked and responsibility for each aspect of the patient's care to be determined in the context of the particular patient, the clinical situation and the clinical team.
2. Third-party participants.
Baseline: Suppose a foreign national is taken ill while on holiday or business, or a new third-party service is instroduced standard centralised guidelines are unlikely to have been designed with these new circumstances in mind.
Experimental: Introduction of an unanticipated participant or non-standard service requirement is no problem provided the participant is fulfilling a recognised service role in a pathway. Provided the participant is properly accredited, the details of how they fulfil their role may be left to them.
Centralised service management does not scale
1. Centralised systems require overall standardisation of all technical aspects. This is highly impractical at large scales.
Baseline: Because control is centralised, everything in the healthcare system must interoperate with a central hub. Thus technical standards (image formats, database structures, medical vocabularies) must be imposed universally. If national or international guidelines are to be deployed this is a major technical undertaking because every healthcare organisation, region, trust, institution and department must change all its computer systems to use the common standards.
Experimental: Under the Safe & Sound framework it is only necessary to standardise technical details between the small group of participants interacting in a particular case. Clinical process can be standardised more widely. Thus for example a national treatment guideline may be followed in Gloucestershire using image format F1 and medical vocabulary V1, while Nottinghamshire follows the same guideline using image format F2 and vocabulary V2.
2. Centralised systems lead to a drive to control access and boundaries tightly. This is increasingly difficult as systems get larger, and has a stifling effect on innovation and research.
Baseline: Monolithic repositories of medical knowledge, guidelines and data with tight control over system perimeters and access are seen as the only way to maintain safety, confidentiality and quality of information.
Experimental: Individuals in the system (e.g. a doctor making a decision) decide who they want to interact with (e.g. an online decision service, a colleague in another institution), and on what basis. Centralised control can be imposed, if desired, over the contractual requirements of e.g. the delegation of a task, but the choice of who to work with is otherwise left open.
3. Centralised systems are highly non-portable.
Baseline: Taking a computerised guideline developed for one institution and deploying it in the same setting in another institution is well known to require substantial effort (there are estimates of up to 90% of the effort required to develop the original version), because in traditional centralised approaches almost every aspect of the guideline is customised to suit the host institution's systems, working practices etc.
Experimental: A high-level guideline may be defined covering the essential tasks in outline at a level that can be implemented at any institution (the "invariant" aspects of the guideline). This can be used unmodified in any institution. The individual tasks in the guideline may then be implemented in different ways at different institutions.
