Harn-Shen Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Harn-Shen Chen has a total of 118 co-authors. Here are the top co-authors by number of shared publications:
Tzu-En Wu
Department of Ophthalmology, Shin-Kong Wu Ho-Su Memorial Hospital, Taipei, Taiwan, ROC
10 shared publicationsHong-Da Lin
5 shared publicationsTjin-Shing Jap
4 shared publicationsChin-Sung Kuo
Division of Endocrinology and Metabolism, Department of Medicine, Taipei Veterans General Hospital, Taipei, Taiwan, ROC
3 shared publicationsJin-Han Yang
3 shared publicationsRu-Lin Chen
3 shared publications
Latest Publications
- Open Access
- 15 June 2026
- Published by Wolters Kluwer Health in Journal of the Chinese Medical Association
- Vol. 89 (7) , 517-524
Effects of insulin and exenatide therapy on glycemic control and β-cell function in patients newly diagnosed with type 2 diabetes and severe hyperglycemia
Background: To assess the efficacy of short-term insulin therapy compared with exenatide therapy in glycemic control, remission rate, and β-cell function in patients newly diagnosed with type 2 diabetes and severe hyperglycemia. Methods: Patients with newly diagnosed type 2 diabetes and a glycated hemoglobin (HbA1c) level >9% were treated with insulin injections for 10–14 days, after which oral glucose tolerance tests (OGTTs) were performed. The patients were randomized in an open-label design to receive either insulin therapy or exenatide therapy for a further 24 weeks. The OGTT was repeated 6 months after randomization to re-evaluate β-cell function and insulin sensitivity. The participants were followed for another year to evaluate long-term glycemic control. Results: We randomized 40 patients into the insulin group (n=22) and the exenatide group (n=18). One patient in the exenatide group withdrew before the intervention, resulting in 22 and 17 patients in the insulin and exenatide groups, respectively. Body weight was significantly higher in the insulin group than in the exenatide group at baseline (75.8±14.3 vs. 60.5±12.2 kg, p =0.011). The mean HbA1c levels were similar between the insulin and exenatide groups at 6 months (6.74±1.12% vs. 6.88±0.76%, p =0.681) and 12 months after randomization (6.62±0.69% vs. 7.05±0.72%; p =0.119). By 1 year after the intervention, the remission rates were 18.2% (n=4) in the insulin group and 5.9% (n=1) in the exenatide group ( p =0.025). The median time to relapse was 6.5 months in the insulin group and 5.0 months in the exenatide group ( p =0.685). Compared with the exenatide group, the insulin group had better homeostasis model assessment of β-cell function index, insulinogenic index, and total insulin secretion during the OGTT. Conclusion: Insulin therapy may improve β-cell function and reduce remission rates compared with exenatide therapy in patients newly diagnosed with type 2 diabetes and severe hyperglycemia.
- 15 June 2026
- Published by Wolters Kluwer Health in Journal of the Chinese Medical Association
- Vol. 89 (7) , 517-524
Open AccessEffects of insulin and exenatide therapy on glycemic control and β-cell function in patients newly diagnosed with type 2 diabetes and severe hyperglycemia
Background: To assess the efficacy of short-term insulin therapy compared with exenatide therapy in glycemic control, remission rate, and β-cell function in patients newly diagnosed with type 2 diabetes and severe hyperglycemia. Methods: Patients with newly diagnosed type 2 diabetes and a glycated hemoglobin (HbA1c) level >9% were treated with insulin injections for 10–14 days, after which oral glucose tolerance tests (OGTTs) were performed. The patients were randomized in an open-label design to receive either insulin therapy or exenatide therapy for a further 24 weeks. The OGTT was repeated 6 months after randomization to re-evaluate β-cell function and insulin sensitivity. The participants were followed for another year to evaluate long-term glycemic control. Results: We randomized 40 patients into the insulin group (n=22) and the exenatide group (n=18). One patient in the exenatide group withdrew before the intervention, resulting in 22 and 17 patients in the insulin and exenatide groups, respectively. Body weight was significantly higher in the insulin group than in the exenatide group at baseline (75.8±14.3 vs. 60.5±12.2 kg, p =0.011). The mean HbA1c levels were similar between the insulin and exenatide groups at 6 months (6.74±1.12% vs. 6.88±0.76%, p =0.681) and 12 months after randomization (6.62±0.69% vs. 7.05±0.72%; p =0.119). By 1 year after the intervention, the remission rates were 18.2% (n=4) in the insulin group and 5.9% (n=1) in the exenatide group ( p =0.025). The median time to relapse was 6.5 months in the insulin group and 5.0 months in the exenatide group ( p =0.685). Compared with the exenatide group, the insulin group had better homeostasis model assessment of β-cell function index, insulinogenic index, and total insulin secretion during the OGTT. Conclusion: Insulin therapy may improve β-cell function and reduce remission rates compared with exenatide therapy in patients newly diagnosed with type 2 diabetes and severe hyperglycemia.
- Open Access
- 1 April 2026
- Published by Elsevier in SSRN Electronic Journal
Intensive Glycaemic Control in Adults Aged 80 Years or Older with Type 2 Diabetes: a Randomised Trial Considering Competing Mortality
Background: In adults aged 80 years or older with type 2 diabetes, clinical decision making is complicated by limited life expectancy, multimorbidity, and high competing mortality. Whether intensive glycaemic control provides meaningful vascular benefit in this population remains uncertain. Methods: In this prospective, open-label, randomised trial, ambulatory adults aged 80 years or older with type 2 diabetes were assigned (1:1) to an intensive glycaemic target (HbA1c <7%) or a conservative target (HbA1c <9%). Participants were followed for 5 years. The primary outcomes were composite microvascular and macrovascular events. Analyses were performed according to the intention-to-treat principle. Competing risk analyses were prespecified, with all-cause mortality treated as a competing event. Findings: Among 206 participants (mean age ≥80 years; 102 intensive, 104 conservative), intensive therapy achieved lower HbA1c levels at 5 years (7.42% vs 8.21%; P=0.005) . Microvascular events occurred in 33.3% vs 27.9% (hazard ratio [HR], 1.24; 95% CI, 0.76–2.04), and macrovascular events in 6.9% vs 6.7% (HR, 1.02; 95% CI, 0.36–2.92) . Competing risk analyses showed no reduction in cumulative incidence of vascular outcomes. All-cause mortality occurred in 31.1% and exceeded the incidence of vascular events . Severe hypoglycemia requiring hospitalization was more frequent in the intensive group (7 vs 1 event; HR, 7.31; 95% CI, 1.17–18.77) . Interpretation: In very old adults, the potential benefits of intensive glycaemic control are constrained by competing mortality and limited time to benefit. Treatment strategies should prioritise safety and align glycaemic targets with life expectancy and clinical priorities in ageing populations.
- 1 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessIntensive Glycaemic Control in Adults Aged 80 Years or Older with Type 2 Diabetes: a Randomised Trial Considering Competing Mortality
Background: In adults aged 80 years or older with type 2 diabetes, clinical decision making is complicated by limited life expectancy, multimorbidity, and high competing mortality. Whether intensive glycaemic control provides meaningful vascular benefit in this population remains uncertain. Methods: In this prospective, open-label, randomised trial, ambulatory adults aged 80 years or older with type 2 diabetes were assigned (1:1) to an intensive glycaemic target (HbA1c <7%) or a conservative target (HbA1c <9%). Participants were followed for 5 years. The primary outcomes were composite microvascular and macrovascular events. Analyses were performed according to the intention-to-treat principle. Competing risk analyses were prespecified, with all-cause mortality treated as a competing event. Findings: Among 206 participants (mean age ≥80 years; 102 intensive, 104 conservative), intensive therapy achieved lower HbA1c levels at 5 years (7.42% vs 8.21%; P=0.005) . Microvascular events occurred in 33.3% vs 27.9% (hazard ratio [HR], 1.24; 95% CI, 0.76–2.04), and macrovascular events in 6.9% vs 6.7% (HR, 1.02; 95% CI, 0.36–2.92) . Competing risk analyses showed no reduction in cumulative incidence of vascular outcomes. All-cause mortality occurred in 31.1% and exceeded the incidence of vascular events . Severe hypoglycemia requiring hospitalization was more frequent in the intensive group (7 vs 1 event; HR, 7.31; 95% CI, 1.17–18.77) . Interpretation: In very old adults, the potential benefits of intensive glycaemic control are constrained by competing mortality and limited time to benefit. Treatment strategies should prioritise safety and align glycaemic targets with life expectancy and clinical priorities in ageing populations.
- Open Access
- 1 April 2026
- Published by Wiley in Diabetic Medicine
- Vol. 43 (7) , e70275
Efficacy and safety of switching to iGlarLixi from premixed insulin therapy in patients with type 2 diabetes: A real‐world experience
Aims: Fixed‐ratio combinations such as iGlarLixi are recommended over premixed insulin as first‐line injectable therapy for type 2 diabetes mellitus (T2DM), yet real‐world evidence on transitions from premix to iGlarLixi remains limited. This study aimed to evaluate the efficacy and safety of switching from premixed insulin to iGlarLixi in a real‐world clinical setting. Methods: This retrospective cohort included adults with T2DM who transitioned from premixed insulin to iGlarLixi between July 2020 and July 2023 at Taipei Veterans General Hospital. Glycaemic parameters were collected at baseline, 3 and 6 months after transition. Results: Forty patients (mean age 67.6 ± 10.5 years) were included. Among those with baseline HbA 1c >58 mmol/mol (7.5%, n = 28), HbA 1c decreased from 79 ± 18 (9.4% ± 1.7%) to 67 ± 18 mmol/mol (8.2% ± 1.6%) at 6 months ( p = 0.01) and 23.1% achieved HbA 1c ≤53 mmol/mol (7.0%). The largest HbA 1c reduction was observed in those previously on 40–50 units/day of premix insulin (from 83 ± 21 [9.8% ± 1.9%] to 65 ± 15 mmol/mol [8.1% ± 1.4%], p = 0.037). Total insulin dose and injection frequency significantly decreased ( p < 0.001), while hypoglycaemia incidence dropped from 60.0% to 12.5% ( p < 0.001). Among patients with pre‐existing hypoglycaemia, glycaemic control remained stable despite reductions in insulin dose and injection frequency ( p = 0.001, p = 0.032, respectively). Higher HOMA‐IR was significantly associated with suboptimal glycaemic response in multivariate model ( p = 0.037). Conclusions: Switching from premixed insulin to iGlarLixi is a safe and effective strategy to simplify therapy and improve glycaemic management. Insulin resistance may influence response.
- 1 April 2026
- Published by Wiley in Diabetic Medicine
- Vol. 43 (7) , e70275
Open AccessEfficacy and safety of switching to iGlarLixi from premixed insulin therapy in patients with type 2 diabetes: A real‐world experience
Aims: Fixed‐ratio combinations such as iGlarLixi are recommended over premixed insulin as first‐line injectable therapy for type 2 diabetes mellitus (T2DM), yet real‐world evidence on transitions from premix to iGlarLixi remains limited. This study aimed to evaluate the efficacy and safety of switching from premixed insulin to iGlarLixi in a real‐world clinical setting. Methods: This retrospective cohort included adults with T2DM who transitioned from premixed insulin to iGlarLixi between July 2020 and July 2023 at Taipei Veterans General Hospital. Glycaemic parameters were collected at baseline, 3 and 6 months after transition. Results: Forty patients (mean age 67.6 ± 10.5 years) were included. Among those with baseline HbA 1c >58 mmol/mol (7.5%, n = 28), HbA 1c decreased from 79 ± 18 (9.4% ± 1.7%) to 67 ± 18 mmol/mol (8.2% ± 1.6%) at 6 months ( p = 0.01) and 23.1% achieved HbA 1c ≤53 mmol/mol (7.0%). The largest HbA 1c reduction was observed in those previously on 40–50 units/day of premix insulin (from 83 ± 21 [9.8% ± 1.9%] to 65 ± 15 mmol/mol [8.1% ± 1.4%], p = 0.037). Total insulin dose and injection frequency significantly decreased ( p < 0.001), while hypoglycaemia incidence dropped from 60.0% to 12.5% ( p < 0.001). Among patients with pre‐existing hypoglycaemia, glycaemic control remained stable despite reductions in insulin dose and injection frequency ( p = 0.001, p = 0.032, respectively). Higher HOMA‐IR was significantly associated with suboptimal glycaemic response in multivariate model ( p = 0.037). Conclusions: Switching from premixed insulin to iGlarLixi is a safe and effective strategy to simplify therapy and improve glycaemic management. Insulin resistance may influence response.
- 29 March 2026
- Published by SAGE Publications in Journal of Diabetes Science and Technology
A Deep Learning Prognostic Model for Diabetes Patients Using Bilateral Fundus Imaging
Aims: Diabetes mellitus shortens life expectancy, driven primarily by premature mortality from vascular complications. Mortality models for intensive care units are well established, whereas outpatient mortality prediction remains challenging. To empower proactive, life-extending care in aging societies, we developed a prognostic model that utilizes routine bilateral color fundus photography for identifying high-mortality-risk patients years before irreversible complications, enabling timely interventions that can substantially improve survival. Methods: We analyzed 19 029 persons with diabetes who received routine color fundoscopy examinations. We adapted deep learning architectures to align with Cox proportional hazards modeling of mortality events, as opposed to simple classification. Results: Our image-trained mortality model stratified the testing cohort into quartiles exhibiting escalating hazard ratios (HRs) of 1.67 (95% confidence interval [CI]: 0.80-3.47), 3.02 (1.55-5.88), and 7.19 (3.91-13.25) relative to the lowest-risk group. The five-year survival rate of the highest-risk quartile was lower than 80%. For comparison, we evaluated a regulatory-approved commercial ophthalmic system, which generated HRs of 1.48, 2.02, and 3.11 across its strata. When evaluated in strata of equivalent sizes to those used by the commercial system, our system demonstrated stronger separation than the commercial system, yielding HRs of 3.64 (95% CI: 3.18-4.16), 4.98 (4.08-6.07), and 6.89 (6.01-7.89). The top 8% patients identified by our model exhibited a five-year survival rate <70%. Conclusion: We developed a survival-aware, image-based model that can predict mortality risk directly from routine fundus photography, achieving prognostic discrimination which facilitates diabetes management.
- 29 March 2026
- Published by SAGE Publications in Journal of Diabetes Science and Technology
A Deep Learning Prognostic Model for Diabetes Patients Using Bilateral Fundus Imaging
Aims: Diabetes mellitus shortens life expectancy, driven primarily by premature mortality from vascular complications. Mortality models for intensive care units are well established, whereas outpatient mortality prediction remains challenging. To empower proactive, life-extending care in aging societies, we developed a prognostic model that utilizes routine bilateral color fundus photography for identifying high-mortality-risk patients years before irreversible complications, enabling timely interventions that can substantially improve survival. Methods: We analyzed 19 029 persons with diabetes who received routine color fundoscopy examinations. We adapted deep learning architectures to align with Cox proportional hazards modeling of mortality events, as opposed to simple classification. Results: Our image-trained mortality model stratified the testing cohort into quartiles exhibiting escalating hazard ratios (HRs) of 1.67 (95% confidence interval [CI]: 0.80-3.47), 3.02 (1.55-5.88), and 7.19 (3.91-13.25) relative to the lowest-risk group. The five-year survival rate of the highest-risk quartile was lower than 80%. For comparison, we evaluated a regulatory-approved commercial ophthalmic system, which generated HRs of 1.48, 2.02, and 3.11 across its strata. When evaluated in strata of equivalent sizes to those used by the commercial system, our system demonstrated stronger separation than the commercial system, yielding HRs of 3.64 (95% CI: 3.18-4.16), 4.98 (4.08-6.07), and 6.89 (6.01-7.89). The top 8% patients identified by our model exhibited a five-year survival rate <70%. Conclusion: We developed a survival-aware, image-based model that can predict mortality risk directly from routine fundus photography, achieving prognostic discrimination which facilitates diabetes management.
- Open Access
- 17 March 2026
- Published by Wolters Kluwer Health in Journal of the Chinese Medical Association
- Vol. 89 (6) , 407-414
Ocular complications from glucagon-like peptide-1 receptor agonists: Clinical evidence, potential mechanisms, and clinical recommendations
Glucagon-like peptide-1 receptor agonists (GLP-1 RAs)—central to type 2 diabetes mellitus (T2DM) management owing to their comprehensive benefits (glycemic control, weight reduction, cardiovascular benefits, and renal protection)—exhibit a concerning ocular safety profile. Clinical trial data, notably from SUSTAIN-6, revealed a higher incidence of diabetic retinopathy (DR) complications with semaglutide, particularly in patients with preexisting DR and rapid HbA1c reduction, reflecting early worsening. Conversely, the REWIND and LEADER trials reported no significant increase in DR risk, suggesting variability among agents and the potential influence of glycemic trajectory. Observational studies and meta-analyses provide mixed findings: some suggest increased DR progression risk in vulnerable populations, while others indicate a lower risk than insulin therapy. Beyond DR, recent pharmacoepidemiologic studies and pharmacovigilance reports have implicated GLP-1 RAs—especially semaglutide—in nonarteritic anterior ischemic optic neuropathy (NAION), leading the European Medicines Agency to classify NAION as a very rare adverse effect and prompting regulatory scrutiny. Conversely, real-world data suggest possible protective associations with other retinal disorders, including reduced incidence of neovascular age-related macular degeneration and diabetic macular edema, although findings remain inconsistent. The mechanistic pathways involve rapid metabolic shifts, vascular dysregulation, and potentially direct effects on retinal or optic nerve tissue. While GLP-1 RAs confer substantial systemic benefits, their ocular risks appear concentrated in high-risk subgroups. Individualized prescribing, baseline ophthalmic assessment, and interdisciplinary monitoring are essential until ongoing prospective trials clarify their long-term ocular impact.
- 17 March 2026
- Published by Wolters Kluwer Health in Journal of the Chinese Medical Association
- Vol. 89 (6) , 407-414
Open AccessOcular complications from glucagon-like peptide-1 receptor agonists: Clinical evidence, potential mechanisms, and clinical recommendations
Glucagon-like peptide-1 receptor agonists (GLP-1 RAs)—central to type 2 diabetes mellitus (T2DM) management owing to their comprehensive benefits (glycemic control, weight reduction, cardiovascular benefits, and renal protection)—exhibit a concerning ocular safety profile. Clinical trial data, notably from SUSTAIN-6, revealed a higher incidence of diabetic retinopathy (DR) complications with semaglutide, particularly in patients with preexisting DR and rapid HbA1c reduction, reflecting early worsening. Conversely, the REWIND and LEADER trials reported no significant increase in DR risk, suggesting variability among agents and the potential influence of glycemic trajectory. Observational studies and meta-analyses provide mixed findings: some suggest increased DR progression risk in vulnerable populations, while others indicate a lower risk than insulin therapy. Beyond DR, recent pharmacoepidemiologic studies and pharmacovigilance reports have implicated GLP-1 RAs—especially semaglutide—in nonarteritic anterior ischemic optic neuropathy (NAION), leading the European Medicines Agency to classify NAION as a very rare adverse effect and prompting regulatory scrutiny. Conversely, real-world data suggest possible protective associations with other retinal disorders, including reduced incidence of neovascular age-related macular degeneration and diabetic macular edema, although findings remain inconsistent. The mechanistic pathways involve rapid metabolic shifts, vascular dysregulation, and potentially direct effects on retinal or optic nerve tissue. While GLP-1 RAs confer substantial systemic benefits, their ocular risks appear concentrated in high-risk subgroups. Individualized prescribing, baseline ophthalmic assessment, and interdisciplinary monitoring are essential until ongoing prospective trials clarify their long-term ocular impact.
- 25 February 2026
- Published by Elsevier in Growth Hormone & IGF Research
- Vol. 84, 101686
Prevalence and clinical correlates of pre-operative glucose homeostasis in patients with acromegaly
- 25 February 2026
- Published by Elsevier in Growth Hormone & IGF Research
- Vol. 84, 101686
Prevalence and clinical correlates of pre-operative glucose homeostasis in patients with acromegaly
- Open Access
- 11 February 2026
- Published by Elsevier in Diabetes & Metabolism
- Vol. 52 (2) , 101740
Glucagon-like peptide-1 receptor agonists and the risk of nonarteritic anterior ischemic optic neuropathy: Evidence from a global real-world cohort
- 11 February 2026
- Published by Elsevier in Diabetes & Metabolism
- Vol. 52 (2) , 101740
Open AccessGlucagon-like peptide-1 receptor agonists and the risk of nonarteritic anterior ischemic optic neuropathy: Evidence from a global real-world cohort
- Open Access
- 14 August 2025
- Published by Springer Nature in Pituitary
- Vol. 28 (5) , 91
Evaluating pituitary tumor management: aligning with pituitary tumor centers of excellence criteria
Background: The Pituitary Society established criteria to develop Pituitary Tumor Centers of Excellence (PTCOE) for optimization of patient care. These criteria were later transformed into quantitative standards based on real-life data from accredited pituitary centers worldwide. The aim of this study was to evaluate the pituitary tumor care capacity at our institute and compare it with the PTCOE criteria. Method: We retrospectively reviewed the data of patients who underwent sellar, suprasellar, or parasellar surgery during 2021–2023 at Taipei Veterans General Hospital. Adults older than 18 years who were diagnosed with pituitary tumors were included. We collected data regarding baseline patient and tumor characteristics, the surgical method, complications, and activity volumes across involved departments. The relevant data were compared with those of the PTCOE standards. Results: In total, 182 patients with pituitary tumors underwent surgery via the endoscopic endonasal approach during 2021–2023. Their median age was 51 (range, 20–91) years. Among them, 90.7% had macroadenomas. Functional remission rates were 55.6% for acromegaly, 69.2% for prolactinoma, and 93.3% for Cushing’s disease. Our institute met the acceptable PTCOE criteria for the number of pituitary interventions, postoperative readmissions for complications, dynamic endocrine tests, neuroradiologists, and neuro-oncologists. Further, we met the preferred PTCOE criteria for mortality rate and the numbers of dedicated surgeons, endocrinologists, trained nurses, neuropathologists, and neuro-ophthalmologists. Conclusion: Most indicators in our study met the acceptable standards for a PTCOE. During the study period, the multidisciplinary team at our institute collaborated closely to provide comprehensive care for patients with pituitary adenomas.
- 14 August 2025
- Published by Springer Nature in Pituitary
- Vol. 28 (5) , 91
Open AccessEvaluating pituitary tumor management: aligning with pituitary tumor centers of excellence criteria
Background: The Pituitary Society established criteria to develop Pituitary Tumor Centers of Excellence (PTCOE) for optimization of patient care. These criteria were later transformed into quantitative standards based on real-life data from accredited pituitary centers worldwide. The aim of this study was to evaluate the pituitary tumor care capacity at our institute and compare it with the PTCOE criteria. Method: We retrospectively reviewed the data of patients who underwent sellar, suprasellar, or parasellar surgery during 2021–2023 at Taipei Veterans General Hospital. Adults older than 18 years who were diagnosed with pituitary tumors were included. We collected data regarding baseline patient and tumor characteristics, the surgical method, complications, and activity volumes across involved departments. The relevant data were compared with those of the PTCOE standards. Results: In total, 182 patients with pituitary tumors underwent surgery via the endoscopic endonasal approach during 2021–2023. Their median age was 51 (range, 20–91) years. Among them, 90.7% had macroadenomas. Functional remission rates were 55.6% for acromegaly, 69.2% for prolactinoma, and 93.3% for Cushing’s disease. Our institute met the acceptable PTCOE criteria for the number of pituitary interventions, postoperative readmissions for complications, dynamic endocrine tests, neuroradiologists, and neuro-oncologists. Further, we met the preferred PTCOE criteria for mortality rate and the numbers of dedicated surgeons, endocrinologists, trained nurses, neuropathologists, and neuro-ophthalmologists. Conclusion: Most indicators in our study met the acceptable standards for a PTCOE. During the study period, the multidisciplinary team at our institute collaborated closely to provide comprehensive care for patients with pituitary adenomas.
- Open Access
- 30 May 2025
- Published by Wolters Kluwer Health in Journal of the Chinese Medical Association
- Vol. 88 (7) , 513-519
Application of machine learning algorithms to identify risk factors for depression in type 2 diabetes mellitus patients: A Taiwan diabetes registry study
Background: We analyzed variables reported during routine clinical practice using a registrational database to estimate risk factors for depression in people with type 2 diabetes mellitus. Methods: A Patient Health Questionnaire (PHQ-9) score of 15 was selected as the cut-off for clinically meaningful depression. Missing data was either filled in with a median value, the k-nearest neighbors’ method, or the entire variable was removed. Logistic regression, random forest, and decision tree machine learning models were used to decide which factors were most relevant to depression. The accuracy of each algorithm was evaluated with a testing set. Results: When all variables were included in the logistic regression model, the area under the receiver operating characteristic curve was 0.81. In the random forest model, the most important factor was quality of life (QoL). Upon removing QoL-related variables, bloating, and autoimmune disease became the greatest contributing factors. Model accuracy was 83.1%. In the decision tree model, QoL was also observed as the most decisive factor. Upon removing QoL variables, bloating was the first node. Model accuracy was 82.5%. Conclusion: QoL, bloating, and autoimmune disease were the most important factors associated with depression in type 2 diabetes mellitus patients.
- 30 May 2025
- Published by Wolters Kluwer Health in Journal of the Chinese Medical Association
- Vol. 88 (7) , 513-519
Open AccessApplication of machine learning algorithms to identify risk factors for depression in type 2 diabetes mellitus patients: A Taiwan diabetes registry study
Background: We analyzed variables reported during routine clinical practice using a registrational database to estimate risk factors for depression in people with type 2 diabetes mellitus. Methods: A Patient Health Questionnaire (PHQ-9) score of 15 was selected as the cut-off for clinically meaningful depression. Missing data was either filled in with a median value, the k-nearest neighbors’ method, or the entire variable was removed. Logistic regression, random forest, and decision tree machine learning models were used to decide which factors were most relevant to depression. The accuracy of each algorithm was evaluated with a testing set. Results: When all variables were included in the logistic regression model, the area under the receiver operating characteristic curve was 0.81. In the random forest model, the most important factor was quality of life (QoL). Upon removing QoL-related variables, bloating, and autoimmune disease became the greatest contributing factors. Model accuracy was 83.1%. In the decision tree model, QoL was also observed as the most decisive factor. Upon removing QoL variables, bloating was the first node. Model accuracy was 82.5%. Conclusion: QoL, bloating, and autoimmune disease were the most important factors associated with depression in type 2 diabetes mellitus patients.