Keywords
artificial intelligence - diabetes mellitus - Ramadan fasting - physician attitudes
- continuous glucose monitoring - clinical practice
Introduction
Ramadan fasting (RF) is an important religious obligation for Muslims. However, it
poses considerable health challenges for individuals with diabetes due to prolonged
periods of fasting, altered meal timings, and changes in sleep patterns.[1]
[2] These changes may increase the risk of hypoglycemia, hyperglycemia, dehydration,
and ketoacidosis if not managed carefully. While evidence-based guidelines such as
the International Diabetes Federation-Diabetes and Ramadan Alliance (IDF-DAR) practical
guidelines provide stratified risk assessments and structured care models for Ramadan-focused
diabetes management,[3] the integration of novel technologies, including artificial intelligence (AI), into
these models remains underexplored.[4]
[5]
[6]
AI has emerged as a transformative tool in diabetes care, offering applications such
as continuous glucose monitoring (CGM) with predictive alerts, automated insulin delivery
(AID) systems, and machine learning-based risk calculators for glycemic variability.
These tools have demonstrated efficacy in improving glycemic control and reducing
complications during RF in both type 2 diabetes (T2D) and type 1 diabetes (T1D) populations.
However, the successful implementation of AI in clinical care requires not only technological
readiness but also clinician awareness, training, and acceptance.[4]
[5]
[6]
Several studies have examined the use of AI-supported applications for managing diabetes.[7]
[8]
[9]
[10]
[11] AI-based machine learning models were used to predict glucose variability and hypoglycemia
risk in patients with T2D on a multiple drug regimen who fast during Ramadan in the
PROFAST-IT Ramadan study.[7] Also, recent studies evaluating AID systems have demonstrated improved safety and
glycemic outcomes during Ramadan across adolescents, adults, and pediatric populations.[8]
[9]
[10]
[11] These findings highlight the promise of AI but underscore the need to understand
clinicians' readiness to adopt such tools.
Several surveys have examined health care providers' general attitudes toward diabetes
management during the Ramadan period.[12]
[13] Only limited data exist related to AI in this specific context.[14] Understanding these aspects is crucial for identifying gaps in implementation, training
needs, and potential barriers, such as cost, regulatory concerns, and cultural acceptability.
This is the first survey focusing specifically on physician perceptions of AI in care
during Ramadan.
Materials and Methods
Study Design and Setting
An exploratory, descriptive, cross-sectional survey was conducted to assess doctors'
knowledge, attitudes, and practices (KAP) regarding the use of AI in managing diabetes
during RF. The survey was administered online from July 18 to August 31, 2025.
Target Population and Recruitment
A convenience sample of doctors with an interest in diabetes care, specifically those
with a focus on diabetes management during Ramadan were targeted. A contact list was
compiled from professional mailing groups and networks were used to ensure a basic
understanding of the subject. Invitations were distributed via email and professional
messaging groups. Because the survey relied on convenience sampling within these networks,
the findings may overrepresent doctors who are already engaged or interested in this
field.
Survey Instrument
The descriptive questionnaire was developed following a literature review on AI applications
in diabetes care during Ramadan, but no formal psychometric validation was undertaken.
The instrument comprised multiple-choice and Likert-scale questions covering four
domains: (1) demographics and professional background, (2) knowledge and awareness
of AI tools, (3) clinical attitudes and practices, and (4) perceived barriers and
training needs. Additional open-ended questions elicited narrative responses regarding
perceived benefits, concerns, and suggestions for implementing AI. The full questionnaire
text and layout is provided in [Supplementary Table S1].
Survey Administration
The questionnaire was hosted on Google Forms, a secure, Web-based platform that ensures
data privacy and confidentiality. An introductory page explained the purpose of the
study and requested informed consent. The survey link was distributed with two reminder
messages during the study period. Responses were collected anonymously, and no personal
identifiers were stored. All completed responses were included. The Google Forms tool
only captures full responses and does not collect data on incomplete responses or
nonresponders.
Data Management and Analysis
Responses were automatically collated in electronic format for analysis. Descriptive
statistics (frequencies and percentages) were generated for categorical variables,
using item-level denominators to account for missing responses. Open-text responses
were reviewed and summarized thematically; however, no formal qualitative coding framework
was applied. Given the exploratory nature of the study, no hypothesis testing or subgroup
analyses were undertaken.
Ethical Considerations
The study was deemed to carry no risk as it involved an anonymous survey of doctors
on a nonsensitive topic. A General Data Protection Regulation statement was included
in the survey, and specific ethics approval was not sought. All participants provided
electronic informed consent prior to participation.
Results
Respondents' Demographic and Professional Profiles
A total of 134 doctors completed the survey. The largest proportion of respondents
was from the Arabian Gulf region (33.6%), followed by the Middle East (23.1%), North
Africa (14.2%), Europe and North America (14.9%), South East Asia (11.9%), and sub-Saharan
Africa (2.2%). The top reporting countries were Türkiye (n = 28), the United Arab Emirates (n = 22), Malaysia (n = 14), Saudi Arabia (n = 12), Tunisia (n = 11), Jordan (n = 8), Iraq (n = 7), and the United Kingdom (n = 7). Nearly three-quarters (74.6%) were senior consultants, 14.9% were mid-grade
doctors, and 10.4% were trainees. Most specialized in adult endocrinology (60.4%),
with internal medicine (22.4%) and family medicine (13.4%) also represented. The majority
worked in tertiary academic centers (67.2%) and the public sector (76.9%). Two-thirds
had more than 10 years of specialty experience. While 62.7% had received Ramadan-specific
diabetes training, only 23.9% had training related to AI.
Knowledge
Familiarity with AI tools was mixed. CGM with predictive alerts was the most recognized
(55.2%), followed by AID systems (35.1%), machine learning-based risk calculators
(20.1%), and AI-powered decision support systems (14.9%). Over one-third (38.8%) reported
no familiarity with any AI applications ([Fig. 1]). Awareness of AI technologies showed a similar trend, with 55.0% reporting awareness
of CGM, 48.9% of AI systems, and 45.8% of risk calculators. However, only 19.1% had
heard of machine learning predictive tools, while 26.0% were unaware of any such technologies
([Fig. 1]).
Fig. 1 Respondents' awareness of existing artificial intelligence (AI)-based technologies
used in diabetes care in general (A) and their familiarity with AI applications with potential for diabetes care during
Ramadan (B). Responses are not mutually exclusive. The graphic displays absolute numbers and
labels indicating the percentage for each response.
Regarding accuracy, 47.0% believed that AI predicted both hypo- and hyperglycemia
equally, 20.9% favored the prediction of hypoglycemia, and 26.9% were unsure ([Fig. 2]). When asked which patient group benefits most, 41.0% cited equal benefits for T1D
and T2D while 35.8% favored T1D. Only 45.5% were aware of guidelines recommending
the use of AI during Ramadan.
Fig. 2 Respondents' perception of the potential applications of artificial intelligence
(AI) during Ramadan. Responses are not mutually exclusive. The graphic displays absolute
numbers, accompanied by labels indicating the percentage for each response.
Attitudes
Overall, 73.9% agreed that AI enhances safety during fasting; however, less than half
(48.5%) felt confident in using AI tools for decision-making. Perceptions of patient
acceptance were mixed: 40.3% believed patients were receptive, while 39.6% were neutral.
On cultural considerations, 20.1% felt that religious beliefs might conflict with
AI, though most were neutral (39.6%). Just under half (45.5%) agreed that AI respects
patient autonomy, while 40.3% were undecided ([Fig. 3]).
Fig. 3 Respondents' attitudes to adopting artificial intelligence (AI) in diabetes care
during Ramadan. Responses are not mutually exclusive. The graphic displays absolute
numbers, accompanied by labels indicating the percentage for each response.
Clinical Practices and Barriers
In clinical practice, 26.9% reported often recommending AI tools, 32.1% sometimes,
17.9% rarely, and 9.9% never, while others considered AI irrelevant to their setting.
Only 28.4% reported direct clinical benefits, while 41.0% did not observe any. Barriers
were prominent, with affordability (59.7%), limited access (56.0%), and insufficient
training (54.5%) most frequently cited. Additional concerns included patient resistance,
regulatory hurdles, and cultural acceptability ([Fig. 4]).
Fig. 4 Respondents' perceived barriers to adopting artificial intelligence (AI)-assisted
applications with potential for diabetes care during Ramadan. Responses are not mutually
exclusive. The graphic displays absolute numbers, accompanied by labels indicating
the percentage for each response.
Thematic analysis of open-ended responses reinforced these findings. Doctors most
often highlighted the prevention of hypoglycemia and improved safety, followed by
better insulin titration, improved glycemic control, and postprandial management.
Others emphasized the importance of patient empowerment and confidence, with AI tools
facilitating greater self-awareness and independence during fasting ([Tables 1] and [2]). A smaller number cited benefits for clinical decision-making, particularly through
the use of risk calculators. One illustrative example described a 32-year-old woman
who safely completed Ramadan using AI-based glucose monitoring ([Tables 1] and [2]).
Table 1
Narrative summary of the perceived greatest potential benefit of AI in managing diabetes
during Ramadan, their concerns and reservations about using AI in this context, and
their suggestions for improving AI implementation in diabetes care during Ramadan
|
Theme
|
N
|
Illustrative Quotes
|
|
Hypoglycemia prevention/reduction
|
18
|
“Reduced risk of hypoglycemia”; “CGM helped markedly in predicting hypoglycemia and
titrating insulin doses”; “Hybrid closed loop systems prevented hypoglycemia during
fasting.”
|
|
Improved glycemic control and treatment adjustment
|
10
|
“Better adjustment of the insulin dosing during using Freestyle Libre sensors”; “Excellent
glycemic control during Ramadan”; “To prevent hypoglycemia, to adjust doses of insulin
and other hypoglycemic medications.”
|
|
Patient empowerment and confidence
|
7
|
“Healthy life and feel more awareness”; “CGM makes the patient aware about their glucose
pattern, hence lowers the risk of hypoglycemia”; “AID and CGM gave confidence to many
of my patients with T1D to fast safely.”
|
|
Decision-support tools (risk calculators, counseling)
|
5
|
“Risk calculator can guide education and treatment adjustment”; “Risk calculator improves
RF counseling”; “Risk calculator and CGM help to minimize emergencies.”
|
|
Case-specific anecdote
|
1
|
“A 32-year-old woman with diabetes… used an AI-based glucose monitoring tool that
helped us adjust her medication safely. She completed Ramadan without complications.”
|
Abbreviations: AI, artificial intelligence; AID, automated insulin delivery; CGM,
continuous glucose monitoring; T1D, type 1 diabetes.
Table 2
Reported clinical benefits of AI in diabetes care during Ramadan by those who identified
with observed benefits from AI use in diabetes care during Ramadan
|
• I. Prompt: What are the perceived greatest potential benefits of AI in managing
diabetes during Ramadan?
• Responses:
•Ensuring safety during fasting by reducing the risk of hypoglycemia and hyperglycemia.
•Personalized care through optimized medication, insulin dosing, and dietary guidance.
•Enhanced patient self-management and independence in decision-making.
•Real-time monitoring and alerts to prevent complications like hypoglycemia.
•Predictive tools to foresee glucose fluctuations and prevent adverse events.
•Empowerment of patients to fast safely without constant medical supervision.
|
|
II. Prompt: What are the concerns and reservations about using AI in this context,
along with their suggestions for improving AI implementation for diabetes care in
Ramadan?
• Responses:
•Cost and accessibility of AI tools, including affordability and technological availability
for patients.
•Cultural and religious concerns, including the potential misalignment of AI recommendations
with fasting traditions.
•Lack of clinical training for health care providers and concerns about patient misunderstanding
or misinterpretation of AI guidance.
•Fears of overreliance on technology leading to a loss of essential human interaction
and clinical judgment.
• Issues with data privacy and confidentiality breaches.
• •The potential for misuse of AI applications and errors in algorithmic recommendations.
|
|
III. Prompt: What are the suggestions for enhancing AI implementation for diabetes
care in Ramadan?
• Responses:
•Emphasis on comprehensive training for both health care providers and patients to
enhance understanding and usage.
•Improving accessibility to AI tools and making them user-friendly.
•Development of culturally sensitive AI tools that respect religious practices during
Ramadan.
•Integration of AI into health care systems, ensuring alignment with clinician workflows.
•Public awareness campaigns and workshops for both patients and healthcare providers
to promote AI tools.
•Simplification of AI systems to make them more accessible and reduce fear or resistance
to technology.
|
Abbreviation: AI, artificial intelligence.
Support and AI-Training Needs
Most respondents (61.1%) felt training was needed across all professional groups,
including doctors, nurses, and pharmacists. Preferred training formats included workshops
(71.6%), simulation-based approaches (67.2%), and online modules (62.2%). Peer mentorship
(43.3%) and culturally adapted materials (30.6%) were also valued. Nearly 70% expressed
a strong interest in participating in structured training, with an additional 25.4%
open to considering it.
Discussion
This study provides unique insights into doctors' KAP regarding the use of AI in managing
diabetes during RF. While the majority of respondents recognized the potential of
AI to enhance glycemic safety and improve clinical decision-making, actual clinical
utilization of AI tools during Ramadan remains relatively low. The gap between awareness
and adoption may be driven by several modifiable barriers, most notably insufficient
training, limited access to technology, and affordability issues.
The acknowledgment of AI's potential role in predicting hypoglycemia and supporting
real-time glucose monitoring is consistent with published evidence demonstrating the
utility of CGM systems and hybrid closed-loop (HCL) technologies in improving glycemic
control during fasting periods. In particular, the MiniMed 780G and similar AID systems
have been shown to support safer fasting in both adolescents and adults with T1D.
Nonetheless, in this study, only a minority reported frequently recommending AI-based
tools, such as CGM or predictive apps, during Ramadan, suggesting a disconnect between
the theoretical value and clinical integration.[14] Several key findings help explain this discrepancy. First, while 60% of participants
had received formal training in Ramadan-specific diabetes care, only one-quarter reported
training in the use of AI in clinical practice. This aligns with previous global surveys
that underscore the limited inclusion of digital health and AI training in traditional
medical curricula. Encouragingly, two-thirds of respondents in this survey expressed
a willingness to engage in structured AI-focused education, with workshops, simulation-based
training, and online modules identified as preferred modalities. These findings suggest
a clear opportunity to bridge the adoption gap through targeted educational strategies.
The knowledge of specific AI tools varied widely. While over half of the respondents
were familiar with CGM-based systems and AID systems, awareness of other applications,
such as machine learning-based risk calculators (e.g., IDF-DAR), was limited.[3] Interestingly, nearly half of the participants indicated unfamiliarity with any
current AI applications in diabetes care during Ramadan. This echoes findings from
the DAR Global Survey, which noted significant variability in provider awareness and
comfort with emerging technologies in Ramadan-focused care.
Regarding the perceived benefit from AI in diabetes care during Ramadan, clinicians
most frequently reported that AI tools supported hypoglycemia prevention during RF.
CGM, HCL systems, and AID were described as enabling early prediction of glucose drops,
reducing severe episodes, and informing safe decisions about when to break the fast.
These perceptions align with established evidence that such technologies enhance safety
in T1D, a critical consideration in the context of prolonged fasting. A second theme
was improved glycemic control and treatment optimization, with respondents noting
more accurate insulin titration and management of postprandial hyperglycemia. At the
same time, AI-based monitoring and risk calculators were seen to enhance patient empowerment
and confidence, as individuals gained greater awareness of their glucose patterns
and felt more secure in undertaking RF. These benefits suggest that AI is viewed not
only as a clinical tool but also as a means of strengthening patient engagement in
self-management. Together, the responses suggest that clinicians perceive AI to enhance
Ramadan care through three interrelated domains: safety, treatment adjustment, and
patient confidence.
The qualitative data ([Tables 1] and [2]) in this study provided additional depth and insight. Many doctors noted the perceived
benefits of AI in reducing glycemic variability, preventing hypoglycemia, and improving
patient confidence in fasting. Others highlighted AI's potential to enhance personalized
treatment adjustments and support patient autonomy, aligning with the principles of
empowerment-based care models. However, some respondents expressed concerns regarding
overreliance on technology, potential erosion of the physician–patient relationship,
and the lack of culturally contextualized AI applications. These sentiments reinforce
calls in the literature for codesigning AI systems with input from patients, clinicians,
and religious scholars to ensure both clinical efficacy and cultural appropriateness.
These data may serve as the basis for a needs assessment exercise.
Cultural and regulatory barriers were perceived by some respondents, albeit at lower
frequencies than technical and cost-related constraints. This highlights an important
consideration: while technological solutions may be effective, their acceptance and
implementation must be tailored to diverse religious, cultural, and socioeconomic
contexts. Technical issues are open to worldwide contributions, whereas cultural issues
are unique to those observing the fast and must be led by experts with both Ramadan
and AI expertise, especially in Muslim-majority countries. For example, alerts or
recommendations from AI systems should be framed respectfully, perhaps using permissive
rather than directive language (e.g., “You may wish to consider breaking the fast”).[15]
This study also aligns with prior research that highlights the crucial role of physician
engagement in promoting the adoption of technology-enhanced diabetes care. Beyond
individual training, broader health system initiatives—such as endorsement by professional
societies, integration into guidelines (e.g., IDF-DAR, American Diabetes Association),
and provision of subsidized AI tools—may be needed to enable equitable access and
scalable implementation.[16]
To the best of our knowledge, this is the first focused survey on the role of AI in
managing diabetes during RF. A key strength of this study is its diverse sample, which
encompasses doctors from multiple regions, most of whom are senior clinicians. This
enabled a rich understanding of current perspectives and barriers. However, the study
has several limitations that should be mitigated in future studies. It relies on self-reported
data and is therefore subject to selection and reporting bias. It does not include
patient perspectives or objective outcomes related to the use of AI. As an exploratory
survey, no comparative or inferential statistics were applied, limiting the ability
to draw causal inferences. A Delphi method with a highly select group may advance
the understanding of the subject from an ethnic and socioeconomic angle.
Furthermore, the survey used a convenience sampling strategy, drawing participants
from professional networks known to be interested in Ramadan and diabetes care. This
may have introduced selection bias and led to the overrepresentation of doctors who
were already favorably disposed toward AI. Also, the cross-sectional design captures
perceptions at a single point in time and cannot establish causality or assess changes
over time. While a literature review informed the questionnaire, it was not psychometrically
validated, and the free-text responses were not analyzed using a formal qualitative
coding framework, which limits the depth and reproducibility of thematic interpretation.
Finally, the study excluded patient perspectives, which are essential for understanding
the acceptability and real-world impact of AI in Ramadan-focused diabetes care. Also,
regional differences in access to AI technologies and health care infrastructure were
not systematically examined, which may limit the generalizability of the results.
Conclusion
This first exploratory survey reveals both optimism and ambivalence among doctors
regarding the role of AI in managing diabetes during RF. Although a majority of respondents
acknowledged AI's potential to enhance glycemic safety—especially through risk-prediction
and real-time monitoring—actual clinical adoption remains limited. A practical roadmap
could include (1) a basic AI literacy module, (2) Ramadan-specific case simulations,
(3) access pathways/cost mitigation, (4) culturally informed messaging templates,
and (5) evaluation metrics.
Future research should investigate patient experiences with AI during RF, assess clinical
outcomes associated with AI-assisted care, and validate AI tools in real-world settings
during Ramadan. Using Delphi methodology as the next step may help achieve more widely
based consensus. Additionally, collaborative efforts are needed to develop training
frameworks and culturally adapted digital interventions, possibly under the guidance
of multidisciplinary bodies such as the DAR International Alliance, Islamic medical
associations, and national endocrinology societies. Larger study populations may enable
deeper exploration of how attitudes differ across geographies, especially between
Muslim-majority versus minority countries, which would be valuable.
