Using SNOMED An update on the work of the Common User Interface programme

1
Using SNOMEDAn update on the work of the Common
User Interface programme

• Kit Lewis
• CFH Technology Office
• 06-Mar-2007

2
Common User Interface programme

• Partnership between NHS Microsoft
• Assists adoption of NHS IT in 4 ways
• Common Windows XP desktop
• Optimised productivity tools (MS Word, Excel
  etc.)
• Consistent user interface for clinical software
• Toolkit of components for software suppliers
• Long-term aims
• Programme started 2005, will run into 2009
• Adoption will take time

3
Design Guide Developers Toolkit

• AIM to provide a highly mobile workforce with
  a degree of familiarity between different systems
  sufficient for the safe and effective delivery of
  care without extensive re-training
• BY CREATING USER INTERFACE DESIGN STANDARDS

4
(No Transcript)
5
3 types of CUI product

• Design Guidance
• Interactive demonstrators
• Toolkit components
• Preview version for Microsoft-based software
  suppliers available Apr-2007

6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
Design Guide for clinical software

• Massive, difficult task
• Balance patient safety clinical usability
• User-centred, iterative design approach
• 4 major releases, iterations within each
  release
• Over 150 user research sessions so far
• Tackle high-priority areas first
• Patient safety NPfIT implementation urgency
• In conjunction with other CFH programmes

10
Supplier engagement

• SystemC
• EMIS
• The Learning Clinic
• CSW
• Kodak Dental
• InPS
• Axsys

• Map of Medicine
• Royal Pharmaceutical Society / BMJ Publishing
  Group
• NHS.net
• Healthspace
• NHS Direct

In detailed discussion with many others
11
Design Guide phase 1

• Work to date (end March 2007)
• Terminology (Using SNOMED) clinical noting
• Medications management
• Information display entry
• Date and time, patient identifiers incl. patient
  banner
• Decision knowledge support
• Consistent navigation
• Handover
• Clinical Spine Application (CSA)

12
Using SNOMED

• Core team
• Ben Luff (Microsoft) user interface design
  analysis
• Dr. Peter Johnson search capabilities, SNOMED
  guru clinical advice (primary care)
• Dr. Kate Verrier-Jones clinical advice (acute
  care)
• David Allan-Smith clinical advice (nursing)
• Ed Cheetham SNOMED guru
• Mike Carey realisation in Toolkit
• Kit Lewis design strategy direction

13
SNOMED within CFH
Developed by LSPs and their suppliers
Local USER INTERFACE Complying to CUI standards
Developed by LSPs and their suppliers
Developed by CFH SSeRP in collaboration with
local organisations (NHS LSP)
Local IMPLEMENTATION Search engine, usage
reporting etc.
National / local SNOMED ADD-ONS Sub-sets, etc
Improved by CFH Terminology Service
National SNOMED CONTENT Concepts, descriptions,
attributes, hierarchies relationships
Fixed
SNOMED model
14
Consistent implementation

• SNOMED (by design) allows a lot of richness
  flexibility
• User interface has significant influence
• Ease, speed accuracy of selecting the right
  concept
• and therefore the extent and quality of
  structured information within a clinical record
• Disparity of user interfaces leads to
  unpredictable data quality ( to reduced
  benefits of SNOMED)

15
Overall investigation

• SNOMED allows great richness of expression
• But can this actually be made use of in real
  life?
• SNOMED has lots of out-of-the-box features
• But is this enough for real life usage?
• Additions to SNOMED are expensive
• Must justify using real life assessment of value
• Is SNOMED usable in free-form clinical noting?

16
Different methods of data entry
17
Partially constrained data entry

• Code elaborate
• Method used in current GP systems
• Works well with READ codes (esp. v2)
• How will it cope with greater complexity of
  SNOMED?
• Free text parse
• Allows greater freedom of expression
• But must be used carefully

18
Code elaborate
1. User types in their search term
2. System presents possible options plus
modification controls and a field for the user to
type in additional text
3. When the user selects the option (and
modifiers), this information is encoded and
displayed in the relevant notes fields
19
Free text parse
1. User types in their notes. The system marks up
the text for encoding and chunks modifiers
2. User triggers the encoding process for any of
the highlighted text. The system presents
options and modifiers
3. When the user selects the code (and related
modifiers), they are displayed as encoded in the
right-hand pane
20
Early design iterations covered

• Useful add-ons
• Richness of the SNOMED model
• Including archetypes
• Additional elaboration with free-text
• Learning from real usage

21
Useful add-ons

• Sub-sets to limit search results
• Overall not returning non-relevant branches
  terms
• Context-specific at varying levels of
  granularity
• Alternative top-level categories navigation
  structures
• End-users dont relate to some of the
  out-of-the-box SNOMED top level categories
• Current structures can be confusing to browse
• Would enable auto-filing of items
• READ v2 very good at this for Primary Care

22
Search result lists should be constrained not to
show certain branches, terms or synonyms
Clinical finding in particular is not a very
intuitive top-level category
There must always be some filtering
23
Useful add-ons (contd.)

• Ordering of attribute value lists
• SNOMED does not provide a default order becomes
  an implementation question
• eg. Severity attribute values ordered
  alphabetically
• fatal, mild, mild to moderate, moderate, moderate
  to severe, severe
• Clinical archetypes eg. Blood pressure
• Consider with reference to OpenEHR, ENV 13606,
  HL7 v3
• Integrated spell-checking

24
Having fatal at the top of this list is clearly
a bad idea
25
(No Transcript)
26
Richness of the SNOMED model
27
Cognitive overhead to process these choices
28
Just presenting these options causes people to
stop and think
29
Counter-intuitive to search for a (positive)
concept and then negate
Its all getting too much!
30
Most recent design guidance

• Keep the user interface simple
• Help the user structure as much as possible
• Both code elaborate and free text parse
  approaches possible
• But provide structure only when requested
• Requires certain search matching capabilities

31
Basic noting scenario
32
(No Transcript)
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
(No Transcript)
41
(No Transcript)
42
Browsing for related concepts
43
(No Transcript)
44
(No Transcript)
45
(No Transcript)
46
(No Transcript)
47
There must be no disconnect between the codes and
the display text
48
(No Transcript)
49
Working with headings subsets
50
(No Transcript)
51
(No Transcript)
52
Post-coordination
53
(No Transcript)
54
(No Transcript)
55
(No Transcript)
56
(No Transcript)
57
This could be more readable!
58
(No Transcript)
59
Significant modification
60
(No Transcript)
61
(No Transcript)
62
Recognition of commonly-used shorthand
63
What does No rash or vomiting actually mean?
64
(No Transcript)
65
(No Transcript)
66
Chunking text
67
(No Transcript)
68
(No Transcript)
69
(No Transcript)
70
(No Transcript)
71
Using archetypes
72
(No Transcript)
73
(No Transcript)
74
(No Transcript)
75
(No Transcript)
76
Useful add-ons
77
Hard to know if this is a good thing without
real-life usage studies
78
Word equivalence detection is a very useful add-on
79
Disambiguation of commonly-used shorthand
80
Learn from real-life usage
81
Patient safety considerations
82
(No Transcript)
83
(No Transcript)
84
(No Transcript)
85
(No Transcript)
86
(No Transcript)
87
(No Transcript)
88
Search matching capabilites

• Include key hierarchies
• Clinical findings, procedures, observable
  entities pharmaceutical / biological product
• Exclude concepts from other hierarchies unless
  chosen concept post-coordinates with them
• Use statistical word matching
• Matches exceed a similarity score threshold
• Exclude common words (eg. from, to)
• Recognise negation and qualifiers

89
Search matching capabilites (contd.)

• Use SNOMEDs word equivalence matching tables
• Use prefix matching
• Use variable word order matching
• Parse for authorised structure
• Qualification
• Modification
• Post-coordination

90
Conclusions

• It is possible to make SNOMED usable at the point
  of care
• Useful
• Effective
• Satisfying
• CTP (Community Technology Preview) release
  version of SNOMED UI component in April
• Contains much of the functionality shown here
• For Microsoft-based software suppliers

91
Feedback

• Latest design guidance available early April
• Register at www.cui.nhs.uk
• Available to NHS organisations suppliers
• Manual authentication process (please be
  patient!)
• All feedback welcome
• Contact us at cuiprogramme.mailbox_at_nhs.net

92
Participation

• We need people to participate in help us
  organise user research sessions, later this year

cuiprogramme.mailbox_at_nhs.net