International Journal of Community Medicine and Public Health

Reddy SD. Int J Community Med Public Health. 2025 Apr;12(4):1873-1876

http://www.ijcmph.com

pISSN 2394-6032 | eISSN 2394-6040

Short Communication

DOI: https://dx.doi.org/10.18203/2394-6040.ijcmph20250938

The role of AI and machine learning in optimizing insulin therapy: a comparative study

Sagam Dinesh Reddy*

Department of Family Medicine/Industrial Health, LMR Hospital, G Konduru, Andhra Pradesh, India

Received: 07 December 2024

Revised: 18 February 2025

Accepted: 19 February 2025

*Correspondence:

Dr. Sagam Dinesh Reddy,

E-mail: dineshsagam143@gmail.com

Copyright: © the author(s), publisher and licensee Medip Academy. This is an open-access article distributed under the terms of the Creative Commons Attribution Non-Commercial License, which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

ABSTRACT

Managing diabetes effectively requires precise insulin dosing. AI and ML have emerged as valuable tools in optimizing insulin therapy. This study compares AI/ML-based insulin optimization with standard therapy to assess its impact on glycemic control and patient satisfaction. A quasi-experimental study was conducted involving 100 patients divided into AI-assisted and standard insulin therapy groups. Primary outcomes measured included HbA1c levels and frequency of hypoglycemic episodes, while secondary outcomes included patient satisfaction and adherence rates. Statistical tests such as paired t-tests, chi-square tests, and ANOVA were applied. Patients in the AI-assisted therapy group exhibited a significant reduction in HbA1c levels (p<0.05), fewer hypoglycemic episodes (p<0.05), and higher satisfaction levels (p<0.05) compared to the standard therapy group. AI and ML-based insulin optimization improve glycemic control, reduce hypoglycemia, and enhance patient satisfaction, making it a valuable addition to diabetes management strategies.

Keywords: Artificial intelligence, Machine learning, Insulin optimization, Diabetes care, Glycemic control, Hypoglycemia

INTRODUCTION

Diabetes mellitus is a chronic metabolic disorder requiring meticulous glycemic control. AI and ML have transformed diabetes management by enabling personalized, real-time insulin dosing adjustments. However, research comparing AI-based insulin optimization with conventional physician-guided insulin therapy remains limited. This study evaluates AI/ML-based insulin dosing’s impact on glycemic control and patient outcomes.1-3

METHODS

Study design

This was a Quasi-experimental study.

Study location

The study was conducted in LMR Hospital, Andhra Pradesh.

Study duration

The study duration was of 1 year from January 2023-December 2023.

Samples size

The sample size taken for study was of 100 patients (50 per group).

Sampling technique

Technique chosen for sampling was purposive sampling.

Inclusion criteria

Adults (≥18 years) with Type 1 or Type 2 diabetes on insulin therapy for ≥6 months

Exclusion criteria

Pregnant women, ESRD patients, those with severe hypoglycemia requiring hospitalization

Intervention

AI-powered CGM-based insulin dosing vs. standard physician-directed insulin therapy.

Outcome measures

Primary

HbA1c change, hypoglycemia frequency

Secondary

Patient satisfaction, adherence rates.

Statistical analysis

Paired and independent t-tests (HbA1c), Chi-square test (hypoglycemia), ANOVA (satisfaction).

RESULTS

AI-assisted therapy group showed 1.2% reduction in HbA1c (p<0.05) vs. 0.6% in standard therapy.4-6 Hypoglycemic episodes reduced in AI-assisted group (p<0.05). Patient satisfaction was significantly higher in AI-assisted group (p<0.05).

Figure 1: Sum of final HbA1c (%) by gender and group. | 0 50 100 150 200 250 F M AI-Assisted Therapy Standard Therapy
Figure 1: Sum of final HbA1c (%) by gender and group.
Figure 2: Count of patient ID by gender and group (AI-assisted vs standard therapy). | 0 5 10 15 20 25 30 35 F M AI-Assisted Therapy Standard Therapy
Figure 2: Count of patient ID by gender and group (AI-assisted vs standard therapy).
Figure 3: Sum of change in HbA1c (%) by gender and group (AI-assisted vs standard therapy). | 0 5 10 15 20 25 30 35 F M AI-Assisted Therapy Standard Therapy
Figure 3: Sum of change in HbA1c (%) by gender and group (AI-assisted vs standard therapy).
Figure 4: Sum of age by group and gender (standard vs AI-assisted therapy). | 0 500 1000 1500 AI-Assisted Therapy Standard Therapy M F
Figure 4: Sum of age by group and gender (standard vs AI-assisted therapy).
Figure 5: Sum of initial HbA1c (%) by group and gender. | 0 100 200 300 AI-Assisted Therapy Standard Therapy M F
Figure 5: Sum of initial HbA1c (%) by group and gender.
Figure 6: Sum of adherence rate (%) by group and gender (standard vs. AI-assisted therapy). | 0 1000 2000 3000 4000 AI-Assisted Therapy Standard Therapy M F
Figure 6: Sum of adherence rate (%) by group and gender (standard vs. AI-assisted therapy).
Figure 7: AI workflow for insulin therapy optimization.
Figure 7: AI workflow for insulin therapy optimization.
Table 1: Baseline characteristics.
Table reproduced from source PDF page 3

DISCUSSION

Findings align with studies like Heinemann et al, demonstrating AI’s role in improving glycemic control.7-9 Unlike Bergenstal et al, who found significantly higher adherence rates, our study showed marginal improvement (AI: 88% vs. Standard 82%, p=0.08). While AI-driven therapy shows promise, a larger multi-center study is needed to confirm long-term effectiveness.10-12

Study limitations were small sample size, single-center study, short follow-up duration.

CONCLUSION

AI and ML-based insulin optimization significantly improve glycaemic control, reduce hypoglycemia, and enhance patient satisfaction. These findings support AI’s integration into routine diabetes management.

Funding: No funding sources

Conflict of interest: None declared

Ethical approval: Not required

REFERENCES

  1. 1. Karanth S, Shruthi K, Sheela CN, Ross C. Prevalence of thrombocytopenia in Parturient: Experience in Tertiary Care Center. Indian J Obst and Gynecol Res. 2023;5(1):98-103.
  2. 2. Heinemann L, Deiss D, Siegmund T, Schlüter S, Naudorf M, von Sengbusch S, et al. Glucose measurement and control in patients with type 1 or type 2 diabetes. Experimental and Clinical Endocrinology & Diabetes. 2019;127(1):8-26.
  3. 3. He J, Baxter SL, Xu J, Xu J, Zhou X, Zhang K. The practical implementation of artificial intelligence technologies in medicine. Nature Med. 2019;25(1):30-6.
  4. 4. Shrestha N. Use of artificial intelligence in public health. J of Chitwan Medical College. 2024;14(1):1-4.
  5. 5. Zale A, Mathioudakis N. Machine learning models for inpatient glucose prediction. Current Diabetes Reports. 2022;22(8):353-64.
  6. 6. Bergenstal RM, Tamborlane WV. Benefits of real-time continuous glucose monitoring in diabetes management. JAMA. 2020;2:23-38.
  7. 7. Alam MA, Sohel A, Hasan KM, Islam MA. Machine learning and artificial intelligence in diabetes prediction and management: a comprehensive review of models. J of Next-Gen Engineering Systems. 2024;30:3828.
  8. 8. Giorgini F, Di Dalmazi G, Diciotti S. Artificial intelligence in endocrinology: a comprehensive review. J Endocrinol Invest. 2024;47(5):1067-82.
  9. 9. American Diabetes Association. Standards of Medical Care in Diabetes. Diabetes Care, 2023.
  10. 10. Poonguzhali S, Chakravarthi R. A sensor based intelligent system for classification and assistance of diabetes patients in telemedicine technology. J of Intelligent & Fuzzy Systems. 2021;40(4):6365-74.
  11. 11. Beets B, Newman TP, Howell EL, Bao L, Yang S. Surveying public perceptions of artificial intelligence in health care in the United States: systematic review. J Med Internet Res. 2023;25:40337.
  12. 12. Guan Z, Li H, Liu R, Cai C. Artificial intelligence in diabetes management: advancements, opportunities, and challenges. Cell Reports Medicine. 2023;4(10):53.
  13. 13. Chai SS, Goh KL, Cheah WL, Chang YH, Ng GW. Hypertension prediction in adolescents using anthropometric measurements: do machine learning models perform equally well. Applied Sciences. 2022;12(3):1600.