1 Institute for Research on Population and Social Policies 2 - - PDF document

▶
1 institute for research on population and social
SMART_READER_LITE
LIVE PREVIEW

1 Institute for Research on Population and Social Policies 2 - - PDF document

Oct 07, 2022 257 likes •636 views

A conceptual framework to design a dimensional model based on the HL7 Clinical Document Architecture Fabrizio Pecoraro1, Daniela Luzi1, Fabrizio L. Ricci2 1 Institute for Research on Population and Social Policies 2 Institute of Molecular

slide-1
SLIDE 1

A conceptual framework to design a dimensional model based on the HL7 Clinical Document Architecture

Fabrizio Pecoraro1, Daniela Luzi1, Fabrizio L. Ricci2

1 Institute for Research on Population and Social Policies 2 Institute of Molecular Bioimaging and Physiology National Research Council, Rome – Italy

slide-2
SLIDE 2

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Background

There is a growing attention in the use of healthcare data for secondary purposes (i.e. “non-direct care use of personal health information”) to:

  • improve patient care, decision making and communication processes
  • predict public health trends and reduce healthcare costs

Integration of data provided by heterogeneous information systems developed for different specialties and purposes and by different organizations in a common central Data Warehouse Implementation of transformation procedures to convert data from source systems to a common data model challeng needs

slide-3
SLIDE 3

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Background

Integration of information system has been addressed in healthcare to improve semantic interoperability Data modelling In our vision, HL7 standards and in particular the Clinical Document Architecture (CDA) can be the basis to define a common schema to represent clinical information in a data warehouse HL7 represents one of the main candidates for the exchange of information among different stakeholders mainly used for patient’s care delivery purposes (i.e. primary uses)

slide-4
SLIDE 4

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Objective

Propose a conceptual framework to map the dimensional model concepts with the HL7 CDA components to facilitate the definition and implementation of ETL tools developed in a data warehouse architecture

slide-5
SLIDE 5

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Dimensional model

Fact: measurements of the performance of a business process in a qualitative and/or quantitative way (e.g. episodes of care, clinical outcome) Dimension: descriptive information about the fact (e.g. time, patient, location) Formalism to support the conceptual modelling phase in a Data Warehouse project It is represented by a Fact surrounded by independent Dimensions

slide-6
SLIDE 6

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

HL7 Clinical Document Architecture (CDA)

Records clinical observations and services in a mark-up standard document has a textual part (human readable) and structured part (software processable) The structured part relies on coding systems to represent concepts Based on the Reference Information Model (RIM)

slide-7
SLIDE 7

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

CDA main components based on RIM classes

CDA Backbone HL7 Hierarchy The set of specialized Acts and their relationships are the basis to define a CDA backbone Entities, Role and Participation classes define a HL7 hierarchy that describes subjects/objects involved in the process and their role within the action Describes health business processes decomposing it into elementary descriptions

slide-8
SLIDE 8

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Methods

HL7 Hierarchy Fact Dimensions

Map the HL7 CDA components (HL7 Hierarchy, CDA Backbone) with the dimensional model concepts (Fact, Dimension)

CDA Backbone

slide-9
SLIDE 9

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Mapping procedure

Dimensional model lifecycle: 1. Choose the business process 2. Declare the grain 3. Identify the Fact & Measures 4. Identify the Dimensions 5. Refine the dimensional model Based on a case study: Continuity of Care Document (CCD) to define clinical

  • utcome indicators for quality assessment.
slide-10
SLIDE 10

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Mapping procedure

Dimensional model lifecycle: 1. Choose the business process 2. Declare the grain 3. Identify the Fact & Measures 4. Identify the Dimensions 5. Refine the dimensional model

slide-11
SLIDE 11

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Choose the business process

Candidate Dimensions Business Process Description D a t e P at ie nt P er fo r m er L a b te st Di a g n

  • si

s Pr

  • c

e d ur e Pr

  • d

u ct R e a cti

  • n

Alerts allergies, adverse reactions and alerts X X X X X Medications patient’s current medications and pertinent medication history X X X Results results of observations generated by laboratories, imaging procedures, and other procedures X X X X Procedures interventional, surgical, diagnostic, therapeutic procedures, treatments pertinent to the patient historically X X X X

Determining and prioritizing processes to be analysed

slide-12
SLIDE 12

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Continuity of Care Document – Result Section

CDA Backbone HL7 Hierarchy

CDA Components

slide-13
SLIDE 13

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Mapping procedure

Dimensional model lifecycle: 1. Choose the business process 2. Declare the grain 3. Identify the Fact & Measures 4. Identify the Dimensions 5. Refine the dimensional model

slide-14
SLIDE 14

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Declare the grain

Specifying the level of detail of the dimensional model (i.e. what an individual fact table row represents) High level atomic information captured by a business process (e.g. value of a vital sign analysis observed during a test) Low level transactions are aggregated

  • ver dimensions

(e.g. average value of a vital sign over a specific time interval) The grain declaration is subjected to the granularity of data contained in the clinical document

slide-15
SLIDE 15

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Mapping procedure

Dimensional model lifecycle: 1. Choose the business process 2. Declare the grain 3. Identify the Fact & Measures 4. Identify the Dimensions 5. Refine the dimensional model

slide-16
SLIDE 16

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Identify the Fact

Identify a CDA concept that represents a “measurement of the healthcare business process”. Acts that define the CDA Backbone are suitable candidates to identify Fact An action performed to determine an answer or a result value

slide-17
SLIDE 17

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Identify the Measures

Attributes of the chosen Act are used to identify measures of the Fact table

value and interpretationCode attributes that represent a quantitative and qualitative description of the event observed.

slide-18
SLIDE 18

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Mapping procedure

Dimensional model lifecycle: 1. Choose the business process 2. Declare the grain 3. Identify the Fact & Measures 4. Identify the Dimensions 5. Refine the dimensional model

slide-19
SLIDE 19

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Identify Dimensions

Dimensions are determined by answering the following questions:

  • Who participates (person): patient, provider, responsible parties
  • What is studied (fact): encounter, hospitalization, adverse reaction
  • When has been performed (time): date, year, month, week-day
  • Where has been placed (location): facility, hospital, patient’s home
  • Why has been performed (reason): pathology, adverse reaction
  • How has been measured (manner): visit, hospitalization

(Zachman framework, Inmon et al., 1997)

Identify CDA concepts able to answer these questions (CDA Backbone, HL7 Hierarchies, Classes and Attributes) Our approach,

slide-20
SLIDE 20

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Identify Dimensions: CDA Backbone

Healthcare service provided over a period of time (what) Range of values

  • f the observation

(what) ClinicalDocument and Section can be included in the model to provide low-level of details additional information

slide-21
SLIDE 21

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Identify Dimensions: CDA Backbone

Other Dimensions (why): Has reason (RSON): reason or rational for a service (e.g. “treadmill test” has reason “chest pain”) Is etiology for (CAUS): what caused the observation (e.g. “diabetes mellitus” is the cause of “kidney disease”)

slide-22
SLIDE 22

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Identify Dimensions: HL7 Hierarchies

Performer (who): the physician(s) that carried out the clinical event RecordTarget (who): represents the medical record that this document belongs to Both are related with the “scoper” Entity Organization through a Role. This class can be used to identify where the observation has been placed

slide-23
SLIDE 23

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Identify Dimensions: HL7 Hierarchies

Participant (who):

  • Authenticator: A verifier who legally authenticates the accuracy of an act (who)
  • Consultant: An advisor participating in the service by performing evaluations

and making recommendations

  • Responsible party: the provider (person or organization) who has primary

responsibility for the act

slide-24
SLIDE 24

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Identify Dimensions: HL7 Hierarchies

Specimen (what):

  • part of some entity, typically the subject, that is the target of focused laboratory,

radiology or other observations

  • used when observations are made against some substances or objects that are

taken or derived from the subject

slide-25
SLIDE 25

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Other Dimensions: HL7 Hierarchies

  • location (where): healthcare facility where the event occurred
  • consumable (what): substance taken up or consumed as part of

the administration

  • product (what): a material target that is brought forth (e.g.

dispensed) in the service

  • author (who): the humans and/or machines that authored the

document

  • informant (who): is a person that provides relevant information,

such as the parent of a comatose patient who describes the patient’s behavior prior to the onset of coma

slide-26
SLIDE 26

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Identify Dimensions: Act attributes

  • code (what): particular kind of event/act

(e.g. hemoglobin, white blood cell, platelets, electrolytes)

  • effectiveTime (when): time when the event has been performed
  • statusCode (how): state of the clinical statement

(e.g. completed, cancelled, active, nullified)

  • methodCode (how): technique used to make an observation

(e.g. ultrasound, x-ray, computed axial tomography)

Degenerate Dimensions

dimension keys in the fact class that is not related to a dimension table

slide-27
SLIDE 27

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Dimensional Model

slide-28
SLIDE 28

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Mapping procedure

Dimensional model lifecycle: 1. Choose the business process 2. Declare the grain 3. Identify the Fact & Measures 4. Identify the Dimensions 5. Refine the dimensional model

slide-29
SLIDE 29

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

The design of a dimensional model based on the CDA elements results in a high-level normalized data model This representation is typically adopted in transactional database where an high volume of transactions (insert, update, delete) is performed Conversely, Analytical processing is characterized by a low volume of transactions (insert) with complex queries to be executed

Denormalization of HL7 hierarchies

Refine the Dimensional Model

slide-30
SLIDE 30

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

attributes of all classes are collapsed in one class However, requirements that have to be represented with a many-to-many relationship require a bridge table

Refine the Dimensional Model

Denormalization of HL7 Hierarchies

slide-31
SLIDE 31

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Refine the Dimensional Model

Resolving attributes’ Data Types

value: real unit: string

QTY

The two elements are included in the Observation class as independent attributes

Solution

slide-32
SLIDE 32

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Refine the Dimensional Model

Resolving attributes’ Data Types

code: string codeSystem: uid codeSystemName: string codeSystemVersion: string displayName: string

SET<CE> Solution #2

The five elements are included in the Observation class as independent attributes Only the first entry of the interpretationCode is collected

Solution #1

A separate table is created with the five elements as independent attributes The table is linked with a one-to-many relationship with the Observation to store multiple valued attribute

slide-33
SLIDE 33

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Dimensional Model – Logical version

measures

slide-34
SLIDE 34

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Conclusions

The use of HL7 CDA elements to define a dimensional model has many advantages:

  • Makes the integration of heterogeneous systems more robust

limiting the ambiguity in the semantics of messages and avoiding the proliferation of dialects

  • Makes data transformation a simplified straightforward process

thus reducing the resources to be invested to implement ETL tools

  • Enhances the extensibility of the data warehouse architecture

facilitating the integration of other systems in the future The framework proposed of this study will be implemented in the Smart Health 2.0 project, a regional EHR infrastructure

slide-35
SLIDE 35

MIE 2014, 30 August – 3 September 2014, Istanbul, Turkey

Thank you for your attention

Fabrizio Pecoraro f.pecoraro@irpps.cnr.it