AUM SRI SAIRAM

Optimized social-technical intervention protocols for national

healthcare services enrichment

Abstract: Healthcare management is a challenge globally with first-world countries having

significant limitations on operating and maintaining healthcare systems. Contemporary Global
healthcare is found to have various challenges, constraints, and opportunities. India, as an
emerging economy, is unique with a large population basis, low service costs, and a very diverse
service offering. An ambitious national healthcare program, Ayushman Bharat Pradhan Mantri Jan
Arogya Yojana (AB-PMJAY) plan, for the economically challenged, was introduced to “leave no one
behind” moving to a needs-based healthcare model. The ambitious plan seeks to provide services,
at scale on a national health basis, funded by both national and state governments. This ambitious
plan, like many global health initiatives, has significant challenges, most especially operational
challenges. Every challenge is an opportunity for the application of optimization techniques and
ensuring awareness among Indian beneficiaries about the scheme, benefits and procedural
intricacies at various stages is critical to success. The opportunity to develop a socio-technical
protocol to optimize healthcare delivery is propositioned in this study. This work focuses on the
awareness levels of beneficiaries of the AB-PMJAY scheme in the Anantapur District of Andhra
Pradesh. We analyze data from 8,000 rural families using a novel expert model that generates
actionable knowledge by deploying three models in sequence to explore and identify gaps related
to awareness of enrolment procedures, treatment protocols, and claim reimbursement processes.
Ultimately, the actionable knowledge is presented as a set of actionable rules for stakeholders to
improve communication channels, enhance media engagement strategies, and optimize service
centre operations, thereby increasing awareness of the scheme among beneficiaries and
promoting proper utilization of its benefits.

1. Introduction
The concept of universal health coverage (UHC) is integral to achieving the
sustainable development goals (SDGs), particularly SDG3[1], which aims to ensure
healthy lives and promote well-being for all. Promoting UHC is crucial for ensuring access
to quality health services for all individuals without causing financial hardship. This
principle, rooted in the notion of ‘Leaving No One Behind’[2] underscores the importance
of ensuring access to health services and social health protection for people regardless of
their socio-economic status. Reforms in health financing structures including the
transition from direct payments to prepayment and pooling mechanisms are vital steps
toward achieving equitable access to healthcare services and financial security.
Publicly funded health insurance (PFHI) schemes, prevalent in many countries across
Latin America, Asia, and Africa, play a pivotal role in expanding social health
protection[3]. These schemes, often fully subsidized by governments, aim to enhance
access to care and financial protection by removing financial barriers at the point of care.
India, for example, has made significant strides towards UHC with initiatives such as the
Rashtriya Swasthya Bima Yojana (RSBY) and the Ayushman Bharat-Pradhan Mantri Jan

1
Arogya Yojana (AB-PMJAY). AB-PMJAY launched in 2018, stands as the world’s largest fully
government-funded health scheme, providing secondary and tertiary inpatient
healthcare services[4]. However, the success of such schemes hinges on the awareness
among beneficiaries. A lack of awareness can prevent individuals from accessing entitled
benefits, thereby compromising their fundamental right to health [5]. Despite the
availability of schemes like RSBY and PM-JAY, evidence indicates significant awareness
gaps, particularly among vulnerable populations and in states with higher poverty
levels[5]. While PM-JAY simplifies enrollment processes compared to RSBY, evidence
suggests low awareness levels, especially in rural areas. Understanding beneficiaries’
awareness and knowledge about scheme features is crucial for advancing progress
towards UHC.
The above leads to our problem statement, so add problem statement and we must
try to integrate the broader question of developing a similar expert model that can be
used in similar context, beyond India
Therefore, our objective would be: “Develop a systematic approach to enhance
awareness among beneficiary groups using government-initiated campaigns to optimize
the utilization of social security schemes.”

2. Literature Review

2.1. Global Health Scenario

Global health is at a crossroads, influenced by dynamic factors such as population

growth, aging demographics, and globalization[6]. Increasing global connectivity has
accelerated the spread of infectious diseases, while noncommunicable diseases continue
to place immense pressure on health systems. The demand for sustainable practices has
become critical, especially in addressing chronic hunger, water scarcity, and
environmental degradation. Access to clean drinking water remains a pressing issue,
directly impacting education, economic stability, and disease prevention. Furthermore,
high fertility rates in developing nations contribute to resource challenges and hinder
efforts to achieve health equity[7].

The global health landscape faces significant barriers, including inadequate

investment in healthcare infrastructure and workforce development. Aging populations
and limited resources strain healthcare systems, reducing their capacity to meet growing
demands. Water scarcity and chronic hunger exacerbate existing disparities, particularly
in underserved regions[7]. Antimicrobial resistance presents a rising threat to global
health efforts, while poor health literacy impedes effective communication and decision-
making. Additionally, educational and economic progress is often stymied by large
household sizes and insufficient access to sustainable practices, highlighting the need for
comprehensive and inclusive solutions[6].

2
2.2. Indian Healthcare

A robust healthcare system is integral to national development, fostering a healthy

population and driving socio-economic progress [8].India’s healthcare policies have
evolved significantly since the Bhore Committee in 1946, which stressed social
orientation and community participation in healthcare. Subsequent milestones included
the Mudaliar Committee, the National Tuberculosis Program, and the Alma Ata
Declaration, all of which shaped public health priorities. The first national health policy
(NHP) [9] in 1983 emphasised preventive, promotive, and rehabilitative care, succeeded
by NHP-2002 [10], which focused on primary healthcare and regulating the private
sector, and NHP-2017 [11], reflecting on changing healthcare needs.
Major policy initiatives included the Pradhan Mantri Swasthya Suraksha Yojana
(PMSSY), the National Rural Health Mission (NRHM), later integrated into the National
Health Mission (NHM) [12] in 2015, targeting health system strengthening in rural and
urban areas, Reproductive, Maternal, Neonatal, Child, and Adolescent Health
(RMNCH+A), and Communicable and Non-Communicable Diseases. Other initiatives
include the RSBY, providing social security to rural workers, and AB-PMJAY. These
initiatives aim to increase public health spending, reduce out-of-pocket expenses,
address rural-urban disparities, strengthen primary healthcare, and achieve universal
health coverage.

2.3. AB-PMJAY
A study [4] on the PM-JAY aimed to advance UHC and the SDGs by providing health
coverage of Rs. 5 lakh (approximately 5,904.88 USD) per family annually for secondary
and tertiary care hospitalisation. Targeting 10.74 crore poor and deprived families
(around 50 crore beneficiaries), the PM-JAY ensures cashless, paperless access to
treatment in public and empaneled private hospitals, with a special focus on the girl child
and senior citizens. The scheme places no restrictions on family size or age and identifies
beneficiaries using Socio-Economic Caste Census(SECC) 2011 data or RSBY enrolment. It
integrates state health programs and ensures national portability for beneficiaries. With
coverage for 1,929 medical packages, it encompasses surgeries, medications,
diagnostics, and pre-existing conditions. The PM-JAY is jointly implemented by the central
and state governments on a 60:40 funding basis, it operates through insurance, trust, or
mixed modes under standardized guidelines. This integrated approach promises to build
a robust healthcare ecosystem across India.

2.4. Heath Information challenges in the Indian context

India's healthcare system faces critical challenges related to awareness, access,
affordability, and accountability. The dominance of private healthcare expenditure
exacerbates affordability issues, while public healthcare remains underfunded and often
unreliable. Weak integration between public and private sectors, as noted by

3
Venkateswaran [13], fragments healthcare financing and delivery, hindering efficiency
and accessibility. Political economy and governance issues further complicate reforms,
with limited political incentives and insufficient public investment. The uneven
distribution of healthcare workers, particularly in poorer regions, coupled with inadequate
infrastructure in rural areas impedes service delivery. Barriers such as high out-of-pocket
expenses, geographic disparities, and social inequities exacerbate these problems [14].
Traditional systems like Ayurveda offer complementary solutions, while emerging
technologies from Industry 4.0 could revolutionize healthcare delivery, albeit with
challenges related to data privacy and ethics [15].
Publicly funded health insurance schemes play a crucial role in India, addressing
awareness gaps and overcoming enrolment barriers are imperative for their success. For
example, Karnataka has been actively implementing the health scheme known as
Ayushman Bharat-Arogya Karnataka (ABArK) [16] through the Suvarna Arogya Suraksha
Trust (SAST). In a study [17], in Chamarajanagar Taluk, 1027 households were included,
of which 452 (44 percent) were cardholders under ABArK, most cardholders (96 percent)
were below the poverty line and resided in rural areas (60 percent). While 65 percent of
the participants were aware of the scheme, only 3 percent of cardholders had used it.
Lack of awareness and poor communication with health workers were identified as major
barriers to the scheme utilisation. The same study employed simple random sampling
and probability proportional to population size (PPPS) to reach its sample size. Data
collection was done using a validated questionnaire, and descriptive statistics were used
for analysis.
A cross-sectional study [18] using descriptive statistics, aiming to evaluate the
awareness of government social security schemes among rural populations, was
conducted at a rural healthcare center in Kanpur. Out of 250 individuals surveyed, 212
were aware of at least one of the social security schemes. Awareness was higher among
those aged over 30, males, literate, and those in lower economic classes. The study
highlights the importance of age, gender, literacy, and socioeconomic status in
influencing awareness levels [18].
Individual states have also taken steps to implement similar health insurance
schemes, later integrating them with AB-PMJAY to work towards achieving UHC. For
example, Andhra Pradesh successfully integrated schemes like Arogya Shree and
Ayushman Bharat, to reduce out-of-pocket expenditure on healthcare [19]. Studies have
been conducted on the awareness of ABArK in Karnataka [17] and Mahatma Jyotiba Phule
Jan Arogya Yojana (MJPJAY) in Maharashtra [20], both of which aim to provide
comprehensive coverage for medical procedures and hospitalization expenses.
Studies have shown positive associations between awareness of health schemes and
enrollment, but significant gaps persist, particularly regarding AB-PMJAY [14], [15].
Addressing barriers such as illiteracy, ignorance, and lack of proper knowledge about
schemes is essential for successful implementation. Information, Education, and

4
Communication (IEC) initiatives are vital in motivating people to utilise health schemes
effectively.
A study by [5] examining awareness of the PM-JAY in six different states namely,
Bihar, Chhattisgarh, Gujarat, Meghalaya, Tamil Nadu, and Uttar Pradesh was carried out.
A cross-sectional household survey was done for the households between 2019 and
2020. Representative data from 11,618 households eligible for PM-JAY were analyzed,
with one adult representative from each surveyed household. Descriptive statistics and
multivariable logistic regression models were employed to explore the association
between awareness of PM-JAY and one’s eligibility, and socio-economic and demographic
characteristics. State-specific differences in awareness were observed, indicating the
need for tailored information dissemination strategies. The study revealed that 62
percent of respondents were aware of PM-JAY and among those aware, 78 percent were
knowledgeable about their eligibility for the scheme. Regression analysis confirmed that
older respondents with higher educational levels and salaried jobs were more likely to be
aware of PM-JAY. However, awareness was lower among respondents from Meghalaya
and Tamil Nadu. Additionally, respondents from Other Backward Classes (OBC), wealthier
socio-economic backgrounds, or from Meghalaya or Gujarat were more likely to be aware
of their eligibility status, while respondents from Chhattisgarh were less likely to know
about their eligibility.
Furthermore, a study [5] in Gujarat assessed awareness, enrolment, and utilization of
AB-PMJAY. Among 1152 households surveyed, 82.9 percent possessed AB-PMJAY cards,
but only 43.3 percent had utilized the benefits. Factors affecting awareness included
location, religion, and caste. Despite impressive awareness and enrolment rates,
challenges such as out-of-pocket expenses and underutilization persisted, highlighting
the need for further intervention across all states.
The study [21] highlights critical issues surrounding three major processes—
awareness, enrolment, and utilisation—of the AB-PMJAY scheme. Despite moderate
awareness levels, utilisation remains significantly low due to insufficient knowledge of
the scheme's features and benefits. Poor understanding of empaneled hospitals, their
inaccessibility due to distance, non-coverage of specific treatment facilities, and the
unwillingness of private hospitals to participate in the scheme are among the primary
reasons cited by respondents for non-utilisation. Additionally, the lack of impactful IEC
dissemination and inadequate interaction between health workers and beneficiaries
exacerbate the issue. OOP expenses remain a critical challenge, undermining the
scheme's aim of reducing financial burdens on impoverished populations. While urban
residents exhibit relatively better awareness of social security schemes than rural
counterparts, systemic barriers such as inadequate promotion of cashless treatments
and weak community engagement hinder their participation. Addressing these
challenges requires empaneling more specialty hospitals, improving the accessibility of
healthcare facilities, and equipping key health workers like ASHA workers, Ayushman

5
Mitras, and Sarpanchs with adequate training and incentives.

Maybe end this by including the objective of the current study which aims at
addressing awareness issues using its 3 models?
We therefore propose a generic approach to build a suitable expert model in three
phases to achieve our objective:
Phase1: Segment and identify beneficiary groups eligible for government schemes by
analyzing socio-economic parameters.
Phase 2: To perform gap analysis, assessing the effectiveness of various campaign
sources in enhancing the awareness levels of the identified beneficiary groups.
Phase 3: Generate actionable recommendations to guide stakeholders in enhancing
strategies and optimizing the deployment of campaign resources, ensuring maximum
outreach and impact.

3. Theoretical grounding

3.1 Social theory for health information sharing

There are prominent theories implemented by several countries for sharing health
information among individuals in society based on their intentions, behaviors, and
relationships. These theories are illustrated in Figure 1. The Social Support Theory (SST),
developed by Sidney Cobb (1976), suggests that peer emotional and informational
support encourages health-related behaviors. It emphasizes how support from close
relationships fosters individuals’ well-being. However, its effectiveness may vary in
complex networks, where the support provided can be diluted or less impactful due to
weaker connections [22]. In [23] the extensive use of online forums was observed as a
means of supporting mental health patients.

6
 Figure 1: Health information sharing theories.

The Social Capital Theory (SCT), proposed by Pierre Bourdieu (1985), emphasizes the
role of social networks in facilitating information sharing. It suggests that trust, social
interactions, and shared goals within networks enhance cooperation and the flow of
information. In the context of health information sharing, social capital motivates
individuals to share knowledge based on these relational bonds [24]. In [25], it was
suggested that community-based organizations (CBOs) play a key role in sharing health
information due to their strong relationships with community members and stakeholders.
The Theory of Reasoned Action (TRA), developed by Fishbein and Ajzen (1975), posits
that individual behavior is influenced by intentions, which are shaped by attitudes and
subjective norms. It suggests that individuals act based on their evaluation of outcomes

7
and social expectations. The theory has been widely applied to health-related behaviors
[26], such as using an online video intervention to educate surgical patients on proper
disposal methods for unused opioids.
The study in [24] investigated health information sharing on WeChat, developing an
integrated model based on the above three theories to analyze factors that influenced
sharing intentions. The findings revealed that subjective norms and attitudes significantly
impacted sharing intentions, while social capital dimensions positively influenced these
norms and attitudes. However, trust did not have a significant effect on sharing
intentions, and the role of emotional support was nuanced and context dependent. The
study’s cross-sectional design limited its ability to establish causal relationships, and
regional differences in health information sharing were not examined, highlighting the
need for longitudinal studies to better understand sharing intentions over time.

3.2 Technology adoption

The study [27] examines the impact of digitalization on public health using data from
81 developing countries between 2002 and 2019, showing significant improvements in
health outcomes, especially in Africa and middle-income nations. It identifies income
inequality as a mediating factor, offering insights to guide policies that leverage
digitalization for better public health. The study [27] employs empirical analysis with
panel data, using classical static panel regression, stepwise regression for mediating
effects, and robustness tests with varied estimation methods. Data is sourced primarily
from the World Bank, with additional inputs from the World Income Inequality and
UNCTAD databases, and missing data addressed using the moving average method. It
highlights gaps in understanding, including limited focus on macro-level public health
impacts, insufficient explanation of digitalization’s mechanisms, and a lack of regional
variation analysis. The study emphasises the need for further exploration of
digitalisation’s macro-level effects, its mechanisms, and regional differences.
The paper [28] examines the meaningfulness of work in digitalized health services,
focusing on co-creation among long-term care professionals through interviews with
eighteen participants, and offers recommendations to improve digital platform success
and collaboration within stakeholder ecosystems. Utilizing confidential data from the
H2020 CXP project, the study employed a qualitative case study methodology, with
thematic content analysis of the interviews. Audio recordings were transcribed and
analysed through an abductive coding process, involving three phases of thematic
grouping. The CXP platform did not meet functionality and agility expectations, and its
overall advantage remained unclear due to a lack of understanding of stakeholder
ecosystem needs. Previous studies focused on technical success, neglecting the
importance of fostering social learning and addressing co-creation platform failures.
Community leaders are crucial in building trust and combating misinformation,
ensuring accurate health information reaches underserved populations. Targeted training

8
enhances their ability to address public health challenges and strengthen community
trust. The study [29] used semi-structured focus group discussions, seven conducted
online, with demographic questionnaires, audio recordings, transcriptions, and WHO-
based thematic analysis. Community leaders struggled to interpret COVID-19 health
information due to limited training, information overload, and difficulty identifying
credible sources, compounded by historical mistrust. Further research is needed on
community leaders' information discernment, communication strategies, and support
mechanisms, along with training to counter misinformation.
The above shows no connection to the study

3.3 Data mining and expert modelling

Organizations use data mining to uncover hidden patterns in transaction data, but
existing methods often produce irrelevant results and fail to incorporate prior domain
knowledge [30]. This research addresses this gap by integrating managers’ beliefs to
generate unexpected patterns that challenge intuition, enhancing the relevance of
findings. The approach focuses on developing decision-support systems that provide
meaningful insights to improve managerial decision-making. The primary goal of data
mining is to identify significant patterns within data. Conventional knowledge discovery
methods typically rely on objective measures of interestingness, such as confidence and
support, to assess the statistical significance of patterns. These methods analyze
associations between data items at a binary level. However, subjective measures of
interestingness have gained recognition as equally important [31], [32]. Subjective
measures evaluate unexpected association rules that diverge from stakeholders’
expectations. Such unexpected patterns are often actionable, providing stakeholders
with valuable insights for decision-making[30].

3.3.1 Association Rule Mining (ARM)

Association Rule Mining (ARM) is a data mining technique used to uncover
relationships or patterns between items in large datasets[30], [31]. It is widely applied in
market basket analysis, recommendation systems, healthcare, and other domains where
understanding co-occurrence relationships is valuable [31]. The primary goal of ARM is to
identify association rules, which are statements of the form A → B . This means that when
item A occurs, item B is likely to occur based on observed patterns in the data. A
represents the antecedent (the ‘if’ part of the rule), while B represents the consequent
(the ‘then’ part of the rule). An example of ARM in healthcare:

Dataset: {Patient symptoms and diagnose}

Rule: A → B
A={ fever, cough}: If a patient has fever and cough
B={ flu}: Then they are likely to have the flu

9
[Link] Unexpected/Actionable Rule
An association rule A → B is deemed unexpected relative to the belief b recall of X → Y

on dataset D , if the following conditions hold [30]:

(i) Logical contradiction:

B∧ Y ⊨ FALSE , thus B and Y logically contradict each other, challenging the
original belief.
(ii) Significant co-occurrence:
A ∧ X holds with respect to the support of A → X , on a statistically large
subset of tuples in D .
(iii) Refined Rule Validity:
The refined rule A , X → B holds true in D .
These conditions ensure that the unexpected rule highlights a novel and meaningful

deviation from prior beliefs [30]. As suggested in [30] this belief X → Y can be either
initially elicited from the decision maker or derived from data using machine learning
methods.
[33]

3.3.2 Subjective Logic and the Binomial Opinion Tuple

Data-driven models used to classify data into different categories make use of one
essential tool called a confusion matrix . This matrix gives the number of instances the
model produced on test data:
 True Positives (TP): Number of correctly predicted instances as positive
outcomes.
 True Negatives (TN): Number of correctly predicted instances as negative
outcomes.
 False Positives (FP): Number of incorrectly predicted instances as positive
outcomes.
 False Negatives (FN): Number of incorrectly predicted instances as negative
outcomes.
The recall measure[33] (also known as the true positive rate) is a widely recognized
metric derived from the confusion matrix to evaluate the prediction performance of a
classification model. It is calculated as: 𝑇𝑃/ (𝑇𝑃 + 𝐹𝑁).
To address uncertainty in association rules, subjective logic provides a probabilistic
framework [32]. It represents opinions about relationships in the form of a binomial
opinion tuple (b recall, d recall , urecall ) where:

 b recall: Belief, the proportion of evidence supporting the rule.

 d recall : Disbelief, the proportion of evidence against the rule.

10
  urecall : Uncertainty, the uncommitted belief mass.
The components satisfy the following conditions:

b recall, d recall , urecall ∈[0 , 1] (2)

b recall +d recall +¿ urecall =[0 , 1] (3)

In ref [32], recall measurement was formalized as a binomial opinion, with subjective
measures b recall, d recall , and urecall calculated as follows:

TP
b recall= (4)
TP+ FN +W
FN
d recall = (5)
TP+ FN +W
urecall =1−b recall−d recall (6)

In the binary domain, W is assigned a value of 2 to maintain a balance among belief,

disbelief, and uncertainty, ensuring the model is not overly influenced by limited
evidence.

3.3.3 Fused Association Rule Mining

Fused Association Rule Mining (FARM) is an extension of the ARM methodology. FARM
extends ARM by incorporating fused attributes and fused sources to account for domain-
specific knowledge and managerial beliefs. This enhancement allows for the discovery of
unexpected and actionable rules that provide greater relevance and insight.

To illustrate the concept of FARM, consider a company assessing the awareness

levels of two categories of customers about its products as shown in Table 1.
Table 1: Toy example data
Category of Customer Sex Age Q1Source Q2Source Awareness Level
C1 Female 19 S1 S2 L 12
C1 Female 19 S1 Null L1
C2 Male 19 S2 Null L1
C1 Male 19 S2 S1 L 12
C1 Female 19 S2 S2 L 12
C2 Female 19 S1 S2 L 12
C2 Male 30 S1 S2 L 12
C2 Male 19 S2 S1 L 12
C1 Female 19 S1 S2 L 12

A stepwise procedure to form the transaction dataset is as follows:

Step 1: Customers are classified as C1 and C2, based on economic conditions.
Step2: Binarization of factors (Sex and Age) associated with the economic condition

11
  Le the us denote the two factors — Sex as F1 and Age as F2—which are shared
across customer categories, with
 F 1=1: Male; F 1=0 : Female.
 F 2=1: Age 18 – 35 ; F 2=0: Age ¿ 35.
Step 3: Formation of binary pattern F1F2 that represents the four possible groups of
customers.
 F1F2 can take the following patterns: 00,01,10 and 11.
 For example, the bit pattern ‘01’ represents “all the Female customers in the age
group 18-35”.
Step 4: Obtain the decimal equivalent of the bit patterns ( F 1 F 2 ) as (F 1 F 2)d
 Example: The pattern ‘11’ is represented as 3.
It can be generalized here that for ‘n’ factors affecting the customers 2n patterns and
hence ¿ ¿) decimal equivalents can be generated.
Step 5: The company invested in two primary campaign sources S 1 and S 2, and
customers were asked two logically sequenced questions Q 1 and Q 2 to identify the
sources from which they learned about the product. Awareness is categorized as
 L 12 (correct responses to Q 1and Q 2 ).
 L 1 (correct response to Q 1 only),
 it should be noted that, the correct response to Q 1 must precede the
response to Q 2.
Table 2 provides an example detailing how customer demographics, campaign
responses, and awareness levels are recorded. For instance, the first row describes a
female customer from the category C 1, aged 18 – 35 , who was exposed to two different
campaign sources ( S 1 identified in response to Q 1 and S 2 identified in response to Q 2)
and correctly recognized both, achieving an awareness level of L 12, which reflects full
campaign awareness.

Table 2: Toy example for fused attribute formation.

Category of Customer F1 F2 (F 1 F 2) (F 1 F 2)d Q1Source Q2Source Awareness
Level
C1 0 1 01 1 S1 S2 L 12
C1 0 1 01 1 S1 Null L1
C2 1 1 11 3 S2 Null L1
C1 1 1 11 3 S2 S1 L 12
C1 0 1 01 1 S2 S2 L 12
C2 0 1 01 1 S1 S2 L 12
C2 1 0 10 2 S1 S2 L 12

12
 C2 1 1 11 3 S2 S1 L 12
C1 0 1 01 1 S1 S2 L 12

[Link] Fused Attribute

Table 2 also serves as the foundation for creating fused attributes as shown in Table
3. The first column in Table 3 represents the fused attribute, created by concatenating
the customer category column, the decimal equivalent of the binary pattern of affecting
factors, and the awareness level column in sequence.

Table 3: Transformed dataset with fused attributes.

Fused Attribute Q1Source Q2Source

C 1−1−L12 S1 S2
C 1−1−L1 S1 Null
C 2−3−L 1 S2 Null
C 1−3−L 12 S2 S1
C 1−1−L12 S2 S2
C 2−1−L 12 S1 S2
C 2−2−L 12 S1 S2
C 2−3−L 12 S2 S1
C 1−1−L12 S1 S2

[Link] Fused Source

It can be observed that Table 3 consists of categorical variables. Fused sources are
generated by applying one-hot encoding to these categorical variables. This process
creates a combined representation where each entry links a specific question number ( Q
) to its corresponding campaign source ( S), as shown in Table 4.

For example: Q1_S1 represents the response for Q 1 as S 1 from the customer. In our

paper, we represent this as AQ 1 , S 1 which can be interpreted as “Awareness about Q1 was

acquired from source S1.”

Table 4: A snapshot of transaction data After one-hot encoding.

Fused Attribute _C1-1-L1 Fused Attribute_C1-1-L12 Q1_S1 Q1_S2 Q2_S1 Q2_S1

FALSE TRUE TRUE FALSE FALSE FALSE

TRUE FALSE TRUE FALSE TRUE FALSE
FALSE FALSE FALSE TRUE TRUE FALSE
FALSE FALSE FALSE TRUE FALSE TRUE
FALSE TRUE TRUE FALSE FALSE FALSE
FALSE FALSE FALSE FALSE FALSE TRUE
FALSE FALSE TRUE FALSE FALSE FALSE

[Link] Fused Association Rule

A fused association rule is a rule where the antecedent is a fused attribute, and the
consequent is a fused source, thus “ Fused Attribute → Fused Source . It represents the
association between customer groups based on similar characteristics and the campaign

13
source that influenced their awareness. For an example:

C 1−1−L12 → Q1 S 1 (1)

This can be interpreted as: ‘Female customers aged 18 – 35 , belonging to Category 1,

have awareness about Q 1 from S 1.”

[Link]. Fused Source Association Rule

The rule between two fused sources is referred to as the Fused source association rule.

For an Example:

Q 1S 1 →Q 2S 2

It can be interpreted as “The source of awareness for Q1 is S1 could result in source S2

for Q2”. The inter-source influence of S1 and S2 on customers contributes eventually to
the customer’s awareness level, either increasing or decreasing it. “

4. Methodology
4.1 Data Collection
The study explored the challenges and barriers affecting the optimal utilization of the
AB-PMJAY scheme. Key challenges include
 Awareness Gaps: Limited understanding of beneficiary entitlements, such as
portability, free post-treatment medicines, and coverage for pre-and post-
hospitalization expenses [34].
 Barriers: Socio-economic conditions, insufficient communication between
stakeholders, long-distance travel, and inadequate digital promotion[34], [35].
These challenges necessitate targeted public awareness campaigns tailored to both rural
and urban beneficiaries. To evaluate awareness levels, enrolment processes, and
utilization challenges, a structured questionnaire approach was employed. The survey
was designed based on inputs from AB-PMJAY designated officials, including Ayushman
Mitras, sarpanch, and doctors, keeping the intentions, behavior, and relationships of the
beneficiaries. The questions were organized in a logical sequence. The survey aimed to
assess the awareness levels of beneficiaries regarding three main processes outlined in
the AB-PMJAY guideline documents namely: enrolment, treatment, and claim processes
under Set-1, Set-2, and Set-3 respectively as given in Table 5. The dataset was compiled
from a survey conducted by interacting with eligible representatives from 8000 families
in Anantapur District, Andhra Pradesh who have already enrolled themselves into the AB-
PMJAY scheme.

Table 5: Questionnaires to assess the awareness levels.

Set No Questions

14
 1 QE1. Do you know about PMJAY or not?
QE2. Are you eligible or not?
QE3. Do you know about the Enrolment Location?
QE4. Do you know there is no cap on the family size who can enroll under the AB-PMJAY scheme?
QE5. Are you aware of the minimum time required to Register and raise queries?
2 QT1. Are you aware that the Treatment under AB-PMJAY is for IPD only?
QT2. Are you aware that treatment can be availed in Empaneled Hospitals only?
QT3. Are you aware that the treatment is portable among Empaneled Hospitals?
QT4. Are you aware that the treatment includes daycare and follow-up care?
QT5. Are you aware that post-treatment medicines are provided for 15 days free of cost?
3 QU1. Does the treatment reduce out-of-pocket expenses?
QU2. Are you aware that the scheme covers pre and post treatment costs?
QU3. Do you know that the packages and procedure costs of the treatment are available at the points of service?
QU4. Are you aware that empaneled hospital will claim the treatment cost?
QU5. Are you aware that all these facts can be shared as feedback to the stakeholders at the points of
service/online?
4 Q1. Which Place do you belong to? (Rural/Urban)
Q2. What is your age?
Q3. Number of family members, their livelihood?
Q4. What is your Caste/Category?
Q5. What is the highest education qualification among the family members?

Based on the health information sharing cycle, the survey identified three major
sources influencing beneficiaries
 Social media: (Based on TRA theory).
 Service centers: (Based on SST theory).
 Individuals and Community Networks: (Based on SCT theory), including
interactions during health camps, Ayushman Bharat Divas, and health melas.
Responses to Set-1, Set-2, and Set-3 questions were derived from three possible sources,
as shown in Table 5. The responses collected through the survey were categorized into
variables, as depicted in Table 5, to facilitate analysis of the awareness levels, socio-
demographic factors, and associated challenges.

Table 6: Sources and points of contact. Table-9 Fused association rules

Source Variable Points of Contact
People Asha Workers*, Anganbadi Workers*,
Ayushman Mitra*, Neighbors and Friends
Social-Media TV, Mobile, PM’s Letter, Toll-Free Number,
Radio
Service-Center Empaneled Hospital, Common Service Centre
(CSC), Panchayat Office

Table 6: Binarization demographic and economic factors affecting the beneficiaries.

Variable Measure
Age Age between 18-50 as 0
Age greater than 60 as 1
Sex Male as 0
Female as 1
Household Size Less than 5 members as 0
Greater than 5 members as 1
Deprivation Category D1 (Households with only one room with kucha walls and

15
 roof) [36]
D5 (Scheduled Caste (SC) or Scheduled Tribe (ST)
households)[36]

The AB-PMJAY scheme categorizes deprived beneficiaries into five groups, D1-D5 [36],
based on criteria such as housing conditions, demographics, disability, and caste status.
Based on the demographic and caste information gathered through Set-4 questions, it
was determined that beneficiary families predominantly fall into categories D1 and D5,
as detailed in Table 6. While approximately 97% of the population is aware of scheme
enrollment-related questions, a significant portion is either partially or completely
unaware of the questions in Set-2 and Set-3. This indicates the need for a comprehensive
investigation and analysis to enhance scheme utilization[37].

4.2 Data Pre-processing

During the pre-processing phase, data collected from beneficiary responses
undergoes key transformations which are listed below
 Transactions deemed irrelevant or too infrequent are filtered out to streamline the
dataset, ensuring that the subsequent steps operate on meaningful patterns.
 Missing values are managed by employing the mode of the feature and
interpolation techniques, ensuring data completeness.
 The categorical variables, people, social-media and service-center mapped to the
responses to questions in Set-1, Set-2, and Set-3 are encoded into binary format
using one-hot encoding.

The summary of the pre-processing phase is shown using the Venn diagram in Figure 1 in
which three levels of awareness namely L 1, L 12, and L 123 used in this paper.
Beneficiaries who gave positive responses to all questions in Set-1 are categorized as L 1;
those who responded positively to all questions in Set-1 and Set-2 are categorized as

L 12; and those who provided positive responses to all questions in Set-1, Set-2, and Set-
3 are categorized as L 123.

Figure 2: Survey Analysis

16
 The Venn diagram provides the following breakdown:
 L 1: A total of 7,820 households responded positively to all questions in Set-1, but
not to those in Set-2 or Set-3, and are therefore categorized as having awareness
level L 1.
 L 12: Among these, 3,210 households were fully aware of the questions in both
Set-1 and Set-2 but not Set-3, earning them awareness level L 12.
 L 123: Out of the 3,210 households, 1,420 demonstrated full awareness of all
questions in Set-1, Set-2, and Set-3, and are designated with awareness level

L 123.

4.3 Expert Model

In this study, an expert model comprising three interconnected phases is presented,
as illustrated in Figure. The first phase employs the FARM model to identify patterns and
insights within the data. Building on this, the second phase introduces a belief model
grounded in Subjective Logic, as suggested in [32], to systematically evaluate patterns
through belief, disbelief, and uncertainty. Finally, the third phase develops an
unexpected pattern model, as provided in [30], leveraging outputs from the previous

phases to uncover actionable patterns that challenge managerial intuition.

Figure 3: Expert model phases for data analysis.

5. Results and Discussion

5.1 FARM Model
Refer to the section 3.3.3 Toy example. A snapshot of the Awareness data set
transaction data set generated for AB-PMJAY beneficiary with fused attributes and fused
sources s Set-1 to Set-4 is shown in Table 8. Three sample fused association rules with
respect to enrollment. treatment and utilization is shown in Table 9 for our

17
understanding. Similarly Table 10 , gives three sample fused source association rules or
belief rules.

Table 8: A snapshot of the Awareness transaction dataset D

Fused Attribute Fused Attribute QE 1 People QT 1Media QU 1 Service center
D1-1-L123 D5-1-L123
FALSE TRUE TRUE FALSE FALSE
TRUE FALSE TRUE FALSE FALSE
FALSE FALSE FALSE TRUE FALSE
FALSE FALSE FALSE TRUE FALSE
FALSE TRUE TRUE FALSE FALSE
FALSE FALSE FALSE FALSE TRUE
FALSE FALSE TRUE FALSE FALSE

Table 9: Fused association rules

Fused Attribute Fused Source Fused Association Rules Description
“Male beneficiaries in the age group
18-35 belonging to Deprivation
D 1−1−L 123 QE 1 People D 1−1−L 123 →QE 1People category D1 have the awareness
about AB-PMJAY scheme from
People” (Refer to Table 6)

"Male beneficiaries aged 18-35 from

Deprivation category D1 were aware,
D 1−1−L 123 QT 1¿ Media D 1−1−L 123 →QT 1¿ Media through media (Refer to Table-6), that
scheme benefits are exclusively
available for IPD patients."

"Male beneficiaries aged 18-35 from

Deprivation category D1 were aware,
D 1−1−L 123 QU 1 People D 1−1−L 123 →QU 1People through Service-Centre (Refer to
Table-6), that scheme benefits are
meant for reducing Out of Pocket
expenses.

Table 10: Fused source association

Fused source-1 Fused Source-2 Fused source Association Rules Description
“It gives rise to all the evidence
QE 1 M edia QT 1People QE 1 Media → QT 1People about enhanced beneficiary
awareness regarding AB-PMJAY
scheme and admission type”
“It gives rise to all the evidence
QT 1Media QU 1¿ M edia QT 1Media →QU 1Media about enhanced beneficiary
awareness regarding admission type
and any kind of expenses incurred
during the process”

5.2 Belief Model applied to the FARM Model

In this study, fused source association is modeled as a binary domain A={e , e ' },

where e represents “enhanced awareness ”of the beneficiary and e ' represents “non-
enhanced awareness” of the beneficiary . For the transaction dataset, consider the fused
association rule

18
 QT 4 ¿ M edia→ QU 4 ¿ M edia

where,
 QT4 : “Are you aware that the treatment includes day-care and follow-up care?”
 QU4: “Are you aware that the empanelled hospital will claim the treatment cost?”

The required action is that if a beneficiary is admitted under the scheme for at least one
day, they should not pay anything toward the treatment.
Due to inter-source influence on beneficiaries, each association provides evidence
(either positive or negative) that contributes to increasing or decreasing the beneficiary’s
awareness level. The confusion matrix generated for the fused source association rule is
shown in Figure 4, with the data provided in Table 11.

Figure 4: Confusion Matrix.

Table 11: Confusion matrix data.
Predicted/Actual Enhanced Awareness (Positive, Not Enhanced Awareness
QU4) (Negative, QU4)
Enhanced Awareness (Positive, QT4) True Positive (TP = 1490) QU4 = False Positive (FP = 704) QU4 = No,
Yes, QT4 = Yes QT4 = Yes
Not Enhanced Awareness (Negative, False Negative (FN = 1163) QU4 = True Negative (TN = 4642): QU4 =
QT4) Yes, QT4 = No No, QT4 = No

The modeled binary domain is further characterized using a binomial opinion tuple Ba

(b recall, d recall , urecall ) based on the scenario. The binomial opinion tuple ( b recall, d recall , urecall ) is

calculated as (0.56 , 0.43 , 0.01). As a first step, the fused source association rules are

filtered using the b recall ¿ 0.5 . The Top five belief rules identified are shown in Table 12.

Table 12: Top five belief rules.

Belief Rules b recall

19
 QT 4 M edia →QU 4 M edia 0.56
QT 2Media →QU 2People 0.55
QT 3 people → QU 3 People 0.55
QT1_people → QU1_People 0.54
QT 5¿ → QU 5 People 0.52

5.3 Actionable pattern model based on the Belief model

In this study, unexpected association rules or discovered patterns are referred to as
actionable rules. As suggested in [30], the belief measure was calculated in the previous
phase using machine learning techniques, and the corresponding actionable rules were
derived for the dataset, as shown in Table 13.

Table 13: The actionable rules generated from the fused source association rules.
Rule as per the The rule defined for our data Description
definition [30][31]
Belief Rule QT 4 ¿ M edia→ QU 4 ¿ M edia Belief (b recall ¿ with a minimum threshold value
> 0.5, which can be stated as” The beneficiaries
X →Y who got the awareness about admission and
payment through media sources”
Refined Rule Refinement of the Belief using ZOOMINAUR
algorithm[30] which states “All male
beneficiaries aged 18-35 belonging to
A, X →B D1-1-L123,QT 4 ¿ M edia→ QT 4 ¿ People deprivation category D1 who got the awareness
about admission procedures from media sources
tend to rely on payment procedures from friends
or neighbors.
Actionable Rule The discovered, Unexpected and Actionable
Rule using the ZOOMOUTAUR Algorithm[30]
states ” “All male beneficiaries aged 18-35
A→B D1-1-L123→ QT 4 ¿ People belonging to deprivation category D1 should be
informed through direct communication
channels, such as community representatives or
individuals, to ensure they are aware that no
payment (QU4) is required for treatment under
the AB-PMJAY scheme."

It can be observed that the belief rules are fused source association rules, the
antecedent of the refined rules and the final actionable rules are fused association rules
desirable for each beneficiary group required for increasing the awareness level. The top
five support measures of the antecedent of refined rules are given in Table 14.

Table 14: The support measure of the antecedent of five refined rules
Antecedent of Refined Rule Support
D1-1-L123, QT 4 ¿ Source−media 0.53
D1-1-L123, QT 3¿ People 0.52
D1-1-L123, QT 1¿ People 0.52
D1-1-L123, QT 2¿ Source−media 0.5
D1-1-L123, QT 5¿ People 0.45

20
The possible actionable rules about treatment and utilization for the beneficiary group
D1-1-L123 are given in Table 15. Similarly, other actionable rules for the other
beneficiary groups can also be generated using the suggested model and actionable
knowledge can be provided to stakeholders for decision making.
The results of the study provide actionable recommendations to enhance stakeholder
communication, optimize media outreach strategies, and improve service center
workflows, ultimately increasing beneficiary awareness and enabling more effective
utilization of the scheme's benefits.

Table 15: Actionable rules generated from the refined rules.

Refined Rules Actionable Rule

D1-1-L123, QT4_Media→ QU4_Source_People D11123 → QU4_ People

D1-1-L123, QT4_Media→ QU4_Service Center D11123 → QU4_ Service Center

D1-1-L123, QT3_ people→ QU3_ Media D11123 → QU3_ Media

D1-1-L123, QT3_ people→ QU3_Service Center D11123 → QU3_Service Center

D1-1-L123, QT1_ people→ QU1_ Media D11123 → QU1_ Media

D1-1-L123, QT2_ Media→QU2_People D11123 →QU2_Service Center

[Link] and Consistency in Questionaries

To avoid randomness and minimize biased responses while ensuring the reliability
and consistency of the questionnaire, the measures were rigorously evaluated using
Cronbach’s alpha test. This statistical method was employed to assess the internal
consistency of the questionnaire, determining whether the items within the survey
reliably measured the same underlying construct. By applying Cronbach’s alpha, the data
collected was ensured to be consistent and trustworthy, as shown in Table 16, thereby
reducing the potential for errors and increasing the validity of the findings

Table 16: Reliability measure for the questionaries.

Variable Cronbach Alpha Value
Enrolment-Wise 0.89
Treatment and Utilization-Wise 0.88
For Beneficiary group D1-1-L123 and Treatment and Utilization-wise 0.85

For Beneficiary group D5-1-L123 and Treatment and Utilization-wise 0.79

Although the AB-PMJAY scheme deployed diverse strategies, as outlined in Table 6, to

boost awareness and optimize utilization among the populace, the survey revealed
suboptimal utilization, aligning with national-level data [36]. Proportionate sampling was
employed to determine the sample size. This approach adjusts the required sample size
based on the population’s characteristics and size, ensuring proportional representation

21
across subgroups. The sampling process was defined by the following statistics: a
confidence level of 99 % (corresponding to a Z-score of 2.57 ), a margin of error of

± 1.5 % , and an estimated population proportion of 50 % (commonly used to account for

maximum variability). The population size of 619,000 was used to calculate the sample
size. After applying population correction, this method yielded an efficient sample size of
7,285.

6. Conclusion
In conclusion, the adoption of this model that generates actionable knowledge from
subjective measures defined in a binary domain offers a robust approach to enhance the
implementation and awareness of social security schemes like AB-PMJAY. By leveraging
the fused association rule mining model, belief model, and actionable pattern model this
study fills the gaps in awareness linked to specific demographics or communication
channels and highlights critical areas where targeted interventions are needed.
Ultimately, this approach not only enhances the utilization of AB-PMJAY benefits but
also provides a replicable model for addressing challenges in diverse populations,
ensuring equitable healthcare access and contributing to universal health coverage and
sustainable development goals.
The study explores the challenges and barriers affecting the optimal utilization of the
AB-PMJAY scheme. Key challenges include strategic deployment of sources of
instruments to increase the awareness of the scheme details.

Acknowledgment
We express our deep sense of gratitude to the founder chancellor of our university,
Bhagawan Sri Sathya Sai Baba for His inspiration and message for life which becomes the
ultimate purpose of this proposed work.

References

[1] M. P. Kieny et al., “Strengthening health systems for universal health coverage and sustainable
development,” Bull. World Health Organ., vol. 95, no. 7, pp. 537–539, 2017, doi:
10.2471/BLT.16.187476.

[2] WHO, “ Leave no one behind strengthening health systems for UHC and the SDGs in Africa.
Brazzaville,” WHO,Regional Off. Africa, 2017.

[3] D. Parisi et al., “Awareness of India’s national health insurance scheme (PM-JAY): A cross-sectional
study across six states,” Health Policy Plan., vol. 38, Dec. 2022, doi: 10.1093/heapol/czac106.

22
[4] H. Grewal, P. Sharma, G. Dhillon, R. S. Munjal, and R. K. Verma, “Universal Health Care System in
India : An In- Depth Examination of the Ayushman Bharat Initiative,” Cureus, vol. 15, no. 6, pp. 1–5,
2023, doi: 10.7759/cureus.40733.

[5] D. Parisi et al., “Awareness of India’s national health insurance scheme (PM-JAY): A cross-sectional
study across six states,” Health Policy Plan., vol. 38, no. 3, pp. 289–300, Apr. 2023, doi:
10.1093/heapol/czac106.

[6] S. Kuriakose, “Global Health: Issues and Challenges Corresponding author,” vol. 27, no. 1, pp. 1–4,
2020, [Online]. Available: [Link]

[7] S. Aremu, A. Ibrahim et al.., and undefined 2021, “Global Health: Challenges And Responses,”
[Link], vol. 8, no. 4, pp. 11351–11361, 2021, [Online]. Available:
[Link]

[8] G. A. Raj Janak,Dalal Shauryavir, “Evolution of the Healthcare Policy Framework in India,” 2024.

[9] GOI, “NHP-1983.” [Online]. Available: [Link]

[10] GOI, “National Health Policy-2002”, [Online]. Available:

[Link]

[11] GOI, “National Health policy-2017.” [Online]. Available:

[Link]

[12] GOI, “National Health Mission-2015”, [Online]. Available:

[Link]

[13] Venkateswaran S., “Health Status in India: Challenges and Opportunities (CSEP Working Paper 25).,”
Cent. Soc. Econ. Prog., no. January, 2022.

[14] M. Dhakad and N. Saikia, “Gender Disparities in Health-care Expenditure (HCE) and Financing
Strategies (HCFS) for In-patient Care in India,” SSM - Popul. Heal., vol. 9, p. 100372, Dec. 2019, doi:
10.1016/[Link].2019.100372.

[15] N. Kranthi, Y. Bahurupi, S. Kishore, M. Singh, P. Aggarwal, and B. Jain, “Awareness and readiness of
health care workers in implementing Pradhan Mantri Jan Arogya Yojana in a tertiary care hospital at
Rishikesh,” Nepal J. Epidemiol., vol. 10, pp. 865–870, Dec. 2020, doi: 10.3126/nje.v10i2.27941.

[16] K. GOI, “Ayushman Bharat-Arogya Karnataka.” [Online]. Available: [Link]

[17] Girish B., Jithin Surendran, Vishma B. K., Jyothika V., and Shifa Tahreem, “Study on awareness and
utilization of Ayushman Bharat Arogya Karnataka scheme in Chamarajanagar taluk: a cross-sectional
study,” Int. J. Res. Med. Sci., vol. 11, no. 2, pp. 544–550, Jan. 2023, doi: 10.18203/2320-
6012.ijrms20230161.

[18] S. S. S. A. R. S. Sachdev R Garg K, “Awareness of Indian government initiated social security schemes
utilization among villagers of Kanpur rural region: An evaluative cross-sectional study,” J Fam. Med
Prim Care, vol. 11, pp. 2456–2460, Dec. 2022, doi: 10.4103/jfmpc.jfmpc_1727_21.

[19] S. Reddy and I. Mary, “Aarogyasri Scheme in Andhra Pradesh, India: Some Critical Reflections,” Soc.
Change, vol. 43, no. 2, pp. 245–261, 2013, doi: 10.1177/0049085713492275.

[20] A. Sapkal and S. Deshpande, “Awareness about Mahatma Jyotiba Phule Jan Arogya Yojna among the
residential population of an urban community: a cross sectional study,” Int. J. Community Med. Public

23
 Heal., vol. 6, p. 3848, Dec. 2019, doi: 10.18203/2394-6040.ijcmph20193982.

[21] B. Thomas, D. Raykundaliya, S. Bhatt, and K. Vadhel, “Study of awareness, enrolment, and utilization
of Ayushman Bharat Pradhan Mantri Jan Arogya Yojana in Gujarat, India,” Int. J. Community Med.
Public Heal., vol. 10, Dec. 2023, doi: 10.18203/2394-6040.ijcmph20232151.

[22] M. Liu, Y. Yang, and Y. Sun, “Exploring Health Information Sharing Behavior Among Chinese Older
Adults: A Social Support Perspective,” Health Commun., vol. 34, no. 14, pp. 1824–1832, Dec. 2019,
doi: 10.1080/10410236.2018.1536950.

[23] P. Marshall, M. Booth, M. Coole, L. Fothergill, and Z. Glossop, “Understanding the Impacts of Online
Mental Health Peer Support Forums : Realist Synthesis,” no. May, 2024, doi: 10.2196/55750.

[24] Y. Hong, M. Wan, and Z. Li, “Understanding the health information sharing behavior of social media
users: An empirical study on WeChat,” J. Organ. End User Comput., vol. 33, no. 5, pp. 1–24, Sep. 2021,
doi: 10.4018/JOEUC.20210901.OA9.

[25] S. Ramanadhan, “Social media use by community-based organizations conducting health promotion : a
content analysis,” no. December 2013, 2014, doi: 10.1186/1471-2458-13-1129.

[26] J. Lewis, S. Crawford, S. Sullivan-Boylai, and R. Poza, “A Clinical Trial of a Video Intervention
Targeting Opioid Disposal After General Surgery: A Feasibility Study,” J. Surg. Res., vol. 262, pp. 6–13,
Jun. 2021, doi: 10.1016/[Link].2020.12.010.

[27] J. Wang and Y. Xu, “Digitalization, income inequality, and public health: Evidence from developing
countries,” Technol. Soc., vol. 73, May 2023, doi: 10.1016/[Link].2023.102210.

[28] P. Kuoppakangas, J. Stenvall, T. Kinder, J. Lindfors, and A. Talonen, “Detecting and managing the
mechanism of perceived meaningfulness of work and digital transformation in public sector health and
social care services,” Technol. Forecast. Soc. Change, vol. 194, Sep. 2023, doi:
10.1016/[Link].2023.122663.

[29] P. Githinji et al., “Public health communication during the COVID-19 health crisis: sustainable
pathways to improve health information access and reach among underserved communities,” Humanit.
Soc. Sci. Commun., vol. 11, no. 1, p. 1218, Sep. 2024, doi: 10.1057/s41599-024-03718-7.

[30] B. Padmanabhan and A. Tuzhilin, “Unexpectedness as a measure of interestingness in knowledge

discovery,” Decis. Support Syst., vol. 27, no. 3, pp. 303–318, Dec. 1999, doi: 10.1016/S0167-
9236(99)00053-6.

[31] R. Sharma, M. Kaushik, A. P. Sijo, B. Y. Sadok, and D. Draheim, “Expected vs. unexpected: Selecting
right measures of interestingness,” in Lecture Notes in Computer Science (including subseries Lecture
Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Springer Science and Business
Media Deutschland GmbH, 2020, pp. 38–47. doi: 10.1007/978-3-030-59065-9_4.

[32] B. Herd and S. Burton, “Can you trust your ML metrics? Using Subjective Logic to determine the true
contribution of ML metrics for safety,” in Proceedings of the ACM Symposium on Applied Computing,
Association for Computing Machinery, Apr. 2024, pp. 1579–1586. doi: 10.1145/3605098.3635966.

[33] P.-N. Tan, M. Steinbach, A. Karpatne, and V. Kumar, “Introduction to data mining,” p. 859, 2020.

[34] B. Thomas, D. P. Raykundaliya, S. Bhatt, and K. Vadhel, “Study of awareness, enrolment, and
utilization of Ayushman Bharat Pradhan Mantri Jan Arogya Yojana in Gujarat, India,” Int. J. Community
Med. Public Heal., vol. 10, no. 8, pp. 2741–2747, 2023, doi: 10.18203/2394-6040.ijcmph20232151.

24
[35] D. Parisi et al., “Awareness of India’s national health insurance scheme (PM-JAY): A cross-sectional
study across six states,” Health Policy Plan., vol. 38, no. 3, pp. 289–300, Dec. 2023, doi:
10.1093/heapol/czac106.

[36] NHA, “Official Website PMJAY.” [Online]. Available: [Link]

[37] NHA, “PMJAY Dashboard.” [Online]. Available: [Link]

25