Cost-Sharing Estimator (CostShEs) 1.0 is a prototype for
health-insurance knowledge representation, querying, and visualization. CostShEs 1.0 allows client-side consultation about the cost of a healthcare service for a client with/without insurance. The input to this prototype system is a form-based inquiry by a client about the cost of a service. Based on the client’s entered membership number, desired service, etc., CostShEs 1.0 generates a logical query. The output is the calculated cost of the service, optionally as a text or a visualization. For logical reasoning, CostShEs 1.0 requires (1) an RDFS taxonomy, (2) webized Rule Markup Language (RuleML) data, and (3) RuleML decision rules. The
taxonomy helps the system to find the client’s relatives that can take advantage of his/her insurance. The webized RuleML data are relational ground facts that link to the relevant information on web pages, such as a client’s web page. Based on (1) and (2), the decision rules answer the query. These three components form the knowledge base that CostShEs uses for reasoning with the Object Oriented Java Deductive Reasoning Engine for the Web (OO jDREW). The visualization of the system’s output
in Graph Inscribed Logic (Grailog) is obtained via GrailogKSViz 2.0 translation of RuleML/XML to Scalable Vector Graphics (SVG/XML).
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-Sharing Estimator
1. Representing, Querying, and Visualizing
Health-Insurance Knowledge in a
Cost-Sharing Estimator
Zainab Almugbel1 Harold Boley2
1Department of Computing
University of Dammam
2Faculty of Computer Science
University of New Brunswick
11th International Rule Challenge, RuleML+RR 2017
July 12-15, 2017
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 1 / 29
2. Outline
1 Introduction
2 CostShEs 1.0 Design
3 CostShEs 1.0 Implementation
4 Querying the CostShEs 1.0 Knowledge Base
5 Future Work
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 2 / 29
3. Introduction
Outline
1 Introduction
2 CostShEs 1.0 Design
3 CostShEs 1.0 Implementation
4 Querying the CostShEs 1.0 Knowledge Base
5 Future Work
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 3 / 29
4. Introduction
Objectives
Cost-Sharing Estimator prototype system CostShEs 1.0 attempts to:
Determine how much a client would pay for the costs of a health
service: full, partial, or none.
Apply linked-data concept in RuleML language
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 4 / 29
5. Introduction
Assumptions
Data is available for:
Describing the required aspects of insurers, hospitals, and insureds.
Specifying the classes and web references of the represented RuleML
data facts from the websites.
A taxonomy of relatives is used in deciding which client’s relatives are
covered by his/her insurance.
Knowledge about cost sharing is formalized as RuleML rules for the
system.
The Naf Hornlog RuleML engine OO jDREW is used for reasoning.
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 5 / 29
6. CostShEs 1.0 Design
Outline
1 Introduction
2 CostShEs 1.0 Design
3 CostShEs 1.0 Implementation
4 Querying the CostShEs 1.0 Knowledge Base
5 Future Work
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 6 / 29
7. CostShEs 1.0 Design
CostShEs 1.0 Design
The design consists of two parts:
System design (Windows-based application).
Input: a form-based inquiry by a client about the cost of a service.
Output: data facts about the calculated cost of the service, optionally
as RuleML answers or Grailog visualizations.
Decision model (rules as part of the knowledge base).
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 7 / 29
8. CostShEs 1.0 Design
System Design Preliminaries
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 8 / 29
9. CostShEs 1.0 Design
System Design
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 9 / 29
10. CostShEs 1.0 Design
Decision Model
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 10 / 29
11. CostShEs 1.0 Implementation
Outline
1 Introduction
2 CostShEs 1.0 Design
3 CostShEs 1.0 Implementation
4 Querying the CostShEs 1.0 Knowledge Base
5 Future Work
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 11 / 29
12. CostShEs 1.0 Implementation
CostShEs 1.0 Implementation
CostShEs 1.0 uses:
an RDFS taxonomy.
a webized RuleML subset for relational data.
Naf Datalog RuleML for decision rules (decision model).
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 12 / 29
13. CostShEs 1.0 Implementation
Taxonomy
The RDFS taxonomy represents the categories
of insurance coverage beneficiaries listed in
“Implementing Regulation of the Coopera-
tive Health Insurance Law”. It states that
the client’s insurance must cover his/her
dependent(s). Such eligible dependents are
partitioned into:
“spouse”.
“son up to the age of twenty-five”.
“unmarried daughter”.
The age and marital-status constraints on the
latter two are handled by the decision rules.
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 13 / 29
14. Web Page Data Facts
The data facts include medicalExpenses, insuranceOffer, client, and
partner. They are represented in RuleML/XML for linked-data access on
three kinds of web pages (for hospitals, insurance companies, and clients).
For instance, client (insured) pages contain the client predicate that
stores information about the membership program he/she may have with a
specific insurance company.
<Atom>
<Rel>client</Rel>
<Ind iri="pm:membershipNumber">11122</Ind>
<Ind iri="p:name">ali</Ind>
<Ind iri="pm:hostingOrganization">tawuniya</Ind>
<Ind iri="pm:url">http://localhost:90/tawuniya.xml</Ind>
<Ind iri="pm:programName">silver</Ind>
<Ind iri="pm:description">health Insurance membership</Ind>
</Atom>
16. Decision Rules
The rules are defined for decision-making (based on the decision model).
They are used to estimate the cost of a health service. The main predicate
is named costSharingEstimator. It uses seven named variables
(indicated by a “?” prefix) for estimating the cost-sharing ?Payment for a
client ?Name with a ?clientMembershipID, who may have medical
insurance at ?InsCompany and needs a ?RequiredTreatment for a
?BodyPart in a ?Hospital.
The costSharingEstimator rules have this head:
<Atom>
<Rel>costSharingEstimator</Rel>
<Var>clientMembershipID</Var>
<slot><Ind>insured</Ind><Var type="Self"/></slot>
<Var>Name</Var>
<Var>InsCompany</Var>
<Var>Hospital</Var>
<Var>BodyPart</Var>
<Var>RequiredTreatment</Var>
<Var type="Integer">Payment</Var>
</Atom>
17. CostShEs 1.0 Implementation
Decision Rules (Cont’d)
The costSharingEstimator predicate is defined by three rules based on the
type of the insurance-coverage beneficiaries (Taxonomy):
1. type="Self" means the client is insured and the treatment is for
him/herself (Rule 1).
2. type="Dependent" means the client is insured and the treatment is
for one of his/her dependents (Spouse, Son, Daughter) (Rule 2).
3. type="NonInsured" means the client is not insured (Rule 3).
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 17 / 29
18. Querying the CostShEs 1.0 Knowledge Base
Outline
1 Introduction
2 CostShEs 1.0 Design
3 CostShEs 1.0 Implementation
4 Querying the CostShEs 1.0 Knowledge Base
5 Future Work
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 18 / 29
19. Querying the CostShEs 1.0 Knowledge Base
Sample Query (1)
In this query, (Rule 1) of the costSharingEstimator predicate is called.
An insured client with membership id= 11122 checks all possible hospitals
and payments for treating him/herself.
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 19 / 29
20. Querying the CostShEs 1.0 Knowledge Base
Sample Query (1): RuleML Answers
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 20 / 29
21. Querying the CostShEs 1.0 Knowledge Base
Sample Query (1): Grailog Visualization of Answers
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 21 / 29
22. Sample Query (2)
In this query, (Rule 2) of the costSharingEstimator predicate is called.
An insured client with membership id= 11122 checks which of his/her
daughters (Sara and Nourh) are covered by his/her insurance. After
submitting the query, CostShEs first retrieves the client’s daughters who are
registered in his/her web page as female children. Then, it uses the <Naf>
element and a built-in to only retain the unmarried daughters as follows:
<Naf>
<Atom>
<Rel>stringEqualIgnoreCase</Rel>
<Ind>married</Ind>
<Var>status</Var>
</Atom>
</Naf>
Here, the answers show that the only unmarried daughter (Sara) has the
client’s insurance coverage.
23. Querying the CostShEs 1.0 Knowledge Base
Sample Query (2): RuleML Answers
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 23 / 29
24. Querying the CostShEs 1.0 Knowledge Base
Sample Query (2): Grailog Visualization of Answers
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 24 / 29
25. Querying the CostShEs 1.0 Knowledge Base
Sample Query (3)
In this query, (Rule 3) of the costSharingEstimator predicate is called.
A non-insured client with membership id= 89001 checks all possible
hospitals and payments for treating him/her.
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 25 / 29
26. Querying the CostShEs 1.0 Knowledge Base
Sample Query (3): RuleML Answers
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 26 / 29
27. Querying the CostShEs 1.0 Knowledge Base
Sample Query (3): Grailog Visualization of Answers
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 27 / 29
28. Future Work
Outline
1 Introduction
2 CostShEs 1.0 Design
3 CostShEs 1.0 Implementation
4 Querying the CostShEs 1.0 Knowledge Base
5 Future Work
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 28 / 29
29. Future Work
Future Work
CostShEs 1.0 can be improved in various ways, e.g.:
The knowledge base can be extended to contain more facts about
different medical insurance companies and rules about further cases of
decision-making.
GrailogKSViz 2.0 could be extended to visualize not only positional
but also slotted answer atoms.
Zainab Almugbel, Harold Boley Representing, Querying, and Visualizing Health-Insurance Knowledge in a Cost-ShRuleML+RR 2017 29 / 29