Modelling workflow processes for clinical information systems: impact on decision support and healthcare outcomes
1. Modelling workflow processes for clinical
information systems:
Impact on decision support and healthcare outcomes
2 November 2011
Phil Gooch
Centre for Health Informatics
School of Informatics
City University London
UK
2. Aims of this seminar
• Define and describe the relationships between clinical workflows,
guidelines, pathways and decision support
• Compare and contrast the evidence base for the effectiveness of
systems that implement these components
• Summarise the findings of a recent review of how these components
can be integrated as process-oriented systems and what challenges
need to be overcome
3. Workflow
“The automation of a business process ... during which documents,
information or tasks are passed from one participant to another for
action, according to a set of procedural rules” (WfMC 1999)
• A workflow (Wf) process definition (PD) identifies the various activities,
rules and associated control data.
• A Wf enactment engine interprets the PD and schedules Wf activities
• A workflow management system (WfMS) stores Wf PDs, creates,
executes and manages Wf instances and controls their interaction with
Wf participants and applications.
6. Clinical workflow
“The flow of care-related tasks [for] the management of a patient
trajectory: the allocation of multiple tasks of a provider or of co-working
providers in the processes of care and the way they collaborate”
(Niazkhani et al 2009)
• Patient care as a structured, collaborative process
• Co-ordination of work - scheduling, synchronisation, roles, resource
allocation, temporal constraints
• Information flow – integrating guidelines with info in the medical record
• Monitoring – dynamic task changes in the light of new information
9. Clinical decision support systems (CDSS)
Aim to provide diagnostic and treatment recommendations and advice at the point
of care, i.e. information tailored for the specific patient under consideration by
the clinician during a consultation
• Often implemented as part of computerized physician order entry system
(CPOE)
• Active - provide automated advice in the form of alerts, commentary and
recommendations in response to events occurring within the application while
the user works
• Passive - manually invoke or consult the system first before receiving decision
support (e.g. ‘Infobuttons’)
10. Clinical guidelines
“Systematically developed statements to assist practitioner and patient
decisions about appropriate health care for specific clinical
circumstances” (Field & Lohr 1990)
• Contain recommendations for best practice based on systematic
reviews of clinical evidence, consensus statements and expert opinion.
• Goal is to reduce variation in medical care by promoting the most
effective treatments, and to provide a means of quality control in clinical
practice via audit
11. Care plans
Goal-directed treatment plans, specific to a patient’s needs, which are
signed and time-stamped (Fox et al. 2006)
• Should include best practice treatment from clinical guidelines.
• But must be specific to the given patient, and include a temporal
dimension (‘by [date], patient X will receive [treatment], will be able to
perform [goal]’)
12. Clinical pathways
“Structured multidisciplinary care plans ... for a specific clinical
problem ...implementing local protocols based on clinical guidelines”
(Campbell et al 1998)
• Describe tasks to be carried out together the timing, task sequence and
role that completes each task.
• Should form a single, multidisciplinary record that becomes part of the
patient’s overall clinical record.
• Task timing, sequencing, and role-based task enactment are all
features of workflow systems.
13. Modelling clinical workflow and pathways: activity on node
• Clinical pathways arose from the application of project management techniques
to the management of clinical processes (Zander 1988)
• activity-on-node network consisting of nodes, representing project activities, and
arcs, representing the precedence relationship between activities.
• Duration (and usually role) assigned to each node.
- Earliest start time (EST), Earliest finish time (EFT)
- Latest start time (LST), latest finish time (LFT)
(Hillier & Lieberman 2010)
17. Decision support systems: systematic reviews of evidence for
effectiveness
• Clinical practice (reduction in errors, quality of documentation, patient
outcomes) is improved by active rather than passive CDSS
- CDSS success factors include: recommendations actionable via a computer,
availability at the point of care, and integrated with clinical workflow
(Kawamoto et al. 2005)
• Guideline adherence is improved by CDSS that use a knowledge base derived
from clinical guidelines (Garg et al 2005)
• Little evidence of effect on patient outcomes, but positive impact on practitioner
performance (Jaspers 2011)
19. Computerized clinical guidelines: systematic reviews of evidence
for effectiveness
• Two systematic reviews: Shiffman (1999) and Damiani (2010)
• Improved documentation and guideline adherence (Shiffman) but no meta-
analysis as systems were too heterogeneous in terms of outcome measures
and study types
• Improved the ‘process of care’ (Damiani) - somewhat vague, post-hoc binary
intervention variable based on the conclusion of each study
• Neither review distinguished between systems that simply present guideline-
based recommendations on a computer - i.e. for individual clinical decisions -
from systems that model and support longitudinal, longer-term clinical
processes
20. Clinical pathways: systematic reviews of evidence for effectiveness
• Clinical pathways are associated with reduced in-hospital
complications, improved documentation but do not increase length of
stay or hospital costs.
• However, the clinical pathway development and implementation
process is poorly reported, so the key factors critical to success cannot
be determined (Rotter et al 2010).
• Stroke patients on a clinical pathway may have lower QoL and
satisfaction scores (Kwan 2003)
21. Process-oriented health information systems
• ‘Computer-aided healthcare workflows’: integration of guidelines and
protocols with a health information system (HIS) (Song et al 2006)
• ‘Process-oriented health information systems’: formally models guidelines,
workflows, or clinical pathways and provides support for clinical
decisions that extend over time
23. Findings
• Modelling clinical workflow does not guarantee clinical workflow
integration or point-of-care use
• ‘Idealised’ workflow needs to adapt to actual workflow for a given
patient
• Workflow integration to provide point-of-care support tends to
involve:
• use of an integrated device for data collection, display and
decision support (e.g. mobile)
• use of electronic encounter forms that mirror paper-based forms
• augmented use of paper for data input and/or output
24. Findings
• Web technologies are being used to integrate guidelines, workflows,
pathways and clinical decision support
• Use of formal models, shared knowledge resources and ontologies
• Decomposition of clinical processes into discrete workflow steps is often at
odds with the collaborative nature of clinical work.
• Challenge is to provide adaptive workflow that allows dynamic modification of
tasks, roles, and activity sequencing in response to changing conditions
• Evidence-base for process-oriented systems is in its infancy - perhaps
because it is a potential enabler of intervention, rather than an intervention
itself
25. Going further …
• JAMIA paper: http://jamia.bmj.com/content/18/6/738.full
• Recent conference workshops:
• http://www.uni-ulm.de/in/prohealth-11.html
• http://aimedicine.info/aime11/AIME_11_Keynote_Manfred.pdf
• Process mining: http://processmining.org/
My aims today are to define what we mean by process-oriented healthcare systems. I’ll also discuss the evidence for the effectiveness of these systems in terms of improving clinical practice, and present the results of a systematic review of the literature. These terms are inter-related so it’s difficult to start with a definition of one without discussing the others, nevertheless, let’s start with workflow
Workflow involves the automation (in whole or in part) of a business process, where tasks are sequenced according to a set of procedural rules. The idea is that the task definitions and data flows are separate from the application logic and the actual execution of the workflow, so that process changes can be made without changing the application code. A WfMS stores workflow process definitions, and allows a process definition to be instantiated as a running workflow. The WfMS also orchestrates the interaction of applications invoked by the Wf. Principle of separation of concerns: separate the flow of activities from the activities themselves. And the activity definitions from tasks associated with each activity. And decision support that may be available for each task
Single process definition, multiple running instances in different states
BPM - swimlanes to represent roles. Informal notation that needs to be converted to a form that can be executed - I.e. run - by a computer.
Niazkhani defined clinical workflow as the flow of collaborative, care related tasks for the management of the patient. This involves co-ordination of work items, and the temporal constraints that bind them. It involves the application of guidelines in the light of knowledge about the patient (from data in the EHR), and the ability to monitor and to make dynamic changes to planned or in-progress tasks as a result. Niazkhani found that the formal, stepwise and role-based nature of workflow imposed by CPOE systems was incompatible with both their conceptual model and with the real-life clinical workflow as reported in the studies reviewed. They recommended that system developers pay more head to the collaborative nature of clinical processes. Others have also questioned the application of computerised workflow systems to the complex, contextual nature of clinical workflow. It may not always possible to decompose care-related tasks into a sequence of discrete workflow steps. Some tasks may be partially, or provisionally, completed while other tasks are carried out in parallel. New knowledge gained from downstream or parallel clinical processes may allow the remainder of the provisionally undertaken task to be completed.
Stroke protocol. Notice activity node expansion into a sub-net workflow
Temporal abstraction for modelling clinical processes (Asbru) Typically these constraints are not present in the guidelines from which the workflow is modelled and need to be added from local protocols or practice.
What do we mean by clinical decision support? We will focus on clinician-mediated decision support within the context of a health information system Two types: active and passive.
In the healthcare domain, clinical guidelines aim to provide advice on the most effective treatment for a given patient group with a given condition, based on the best evidence available. The goal is to reduce variation in healthcare outcomes - for example, anyone admitted to hospital with a suspected stroke, no matter where they are in the country, should be able to expect the same baseline level of treatment, and if this treatment process has been shown to be effective, their opportunities for recovery should be improved.
As part of planning treatment, care plans are used both by individual professions and multidisciplinary teams, in order to encompass the use of guidelines in a way that is relevant to the specific needs of the patient. Unlike generic guidelines, care plans include specific goals and a timeframe during which these goals should be achieved. Having said that, increasingly, guidelines are becoming more goal-oriented (examples include Clinical Knowledge Summaries, NICE guidelines) Clinical (care) pathways aim to integrate care plans with guidelines and protocols for best practice
Clinical pathways aim to provide a longitudinal view of care planning that focuses on the right tasks, being done at the right time, by the right person. A care pathway document may form the entire record for one or more episodes of treatment. What distinguishes clinical pathways from care plans is the record of variance: that is, when treatment varies - for better or for worse - from that defined in the ideal pathway, this should be recorded, with the reasons for it. The interesting thing about care pathways is that they contain many of the concepts used by workflow systems as used in other industries: such as task scheduling, sequencing and role-based enactment. In my view, CPs can bring together Wf, CDSS and CIGs A clinical guideline provides recommendations for best practice for the clinical domain addressed by the guideline, but does not provide implementation details. A clinical protocol provides a local, consensus view of a guideline with explicit steps for implementation A clinical pathway is a versioned document of a process, and includes actions recommended by one or more protocols and guidelines, activity role constraints, and sequencing constraints; it has goals and it provides a record of care and information about the patient state and a ‘variance record,’ that is, a method for documenting and recording where deviations from the planned pathway have occurred. Limitations of paper-based clinical pathway documents. It is difficult to tailor clinical pathway forms to the needs of the individual patient, and interdependencies between different pathways are not made explicit: multiple paths tend to be merged into a simple list of tasks, leading to the claim that care pathways simply provide time-based ‘cookbook’ care.
Now, given that care pathways originated from the application of project management techniques to clinical processes, it is perhaps not surprising that initial modelling efforts used formalisms from the project management domain.
This hierarchy of modelling approaches is usefully illustrated by Dang. The top two levels show the core stages in a patient journey, which will involve consideration of delays, queues and rates of flow. General pathway of Admit Assess/Detect Treat Discharge
Precedence diagram showing temporal constraints. Separation of concerns between the flow of activities and the tasks associated with each node, each of which may have an opportunity for CDSS
Precedence diagram showing activity flow. Each activity may have an opportunity for decision support, e.g. may hook into a CDSS for assessment, diagnosis, CPOE, labs testing, e-prescribing … Again, general pathway of Admit Assess Treat Discharge
Do clinical decision support systems make a difference? Jasper - particular benefits for practitioner performance when drug ordering and preventive care reminder systems
Improved ‘process of care’ (Damiani) somewhat vague - as determined by each study author in the conclusions! e.g. less time searching or retrieving information, better information sharing. Similarly for Jaspers review - patient outcomes not defined. Garg breaks the outcomes down better as reduction in unnecessary testing, medication ordering, admissions, referrals, reduction in time from referral to consultation, time to treat etc, staying within therapeutic range of dose etc, better disease management, e.g. diabetes as measured by BP, BMI, HbA1c
Why do clinicians tend not to use guidelines at the point of care? Much clinical decision making involves implicit knowledge, intuition - hard to verbalise Adherence to clinical guidelines by clinicians is often poor, because: Information overload: quantity and complexity (Shahar 2004) Often available as free text only – hard to locate patient-specific information (Shahar) Often ambiguous; recommendations often either too general or too specific (Quaglini 2000) Do not fit local practice (Peleg 2006) Internal (clinician attitude), external (resources, time, patient consent) barriers (Goud 2010) Shiffman - wide variety of systems and domains: pressure ulcer prevention, vaccination reminders, diabetes risk alerts, some provided paper output, others on-screen
Do clinical pathways make a difference? E.g. for stroke, reduction in UTI. No differences in terms of death, dependency or discharge destination Kwan review noted lack of blinding and randomization in selected studies Rotter review did not distinguish between computerized and paper pathways.
We wanted to bring together the literature on workflows, pathways, guidelines and CDSS to identify common themes and to come up with an implementation model for process-oriented health information systems. This hadn't really been done before - plenty of reviews of each of these areas separately. Big problem, as reported by Lau et al 2011, is that the majority of successful system implementations are reported by those that developed the system, who were also involved in the evaluation. In our review, typical reporting biases included The use of interviews carried out by the development team; subjective evaluation and selective, vague reporting without quantification (e.g. ‘the system was tested on a number of clinical scenarios of different complexity levels and in each case the output was as per the original’, 'in all cases the system matched with the expected results’, ‘practitioners approached us ... to tell us the system had informed them of very important clinical data’); evaluation carried out by the development team who were also the study authors and study participants; and evaluation results reported as an addendum without details of the setting or data audit trail.
'CPOE systems tend to fragment and serialise tasks normally performed collaboratively and in parallel' (Niazkhani et al 2009) - Healthcare as a knowledge centric but weakly structured process. It may be that attempts to computerise clinical workflow need to acknowledge the existence of informal working and the use of provisional clinical judgements. Clinical decision-making processes may be provisional, collaborative, incremental and even mutable, whereas IT systems may ‘freeze’ data in unhelpful ways. Can use of these systems be considered as an intervention in itself? Or is it like telemedicine - an enabler of targeted and evidence-based intervention - see Jeremy Wyatt’s presentation on the evidence base for telemedicine and telecare at ITCH 2011