Scholarlink

Background <p>Young and middle-aged type 2 diabetes mellitus (T2DM) patients exhibit high readmission rates, yet research on their self-management behaviors remains insufficient. This study investigates the impact of medication and follow-up adherence on 30-day readmission risk to provide a quantitative foundation that could inform the future development of a knowledge graph-based early warning system.</p> Methods <p>A retrospective cohort study enrolled young and middle-aged (18–59 years) T2DM patients discharged between 2020 and 2024. Exposure variables were medication adherence (medication possession ratio ≥ 80%) and follow-up adherence (≥ 1 outpatient visit in prior 6 months). The outcome was 30-day all-cause readmission. Multivariable logistic regression identified independent predictors, with robustness verified via Bootstrap sampling and sensitivity analyses.</p> Results <p>A total of 950 patients were included, with a readmission rate of 10.9%. Multivariable analysis identified medication non-adherence (aOR = 4.37, 95% CI: 2.79–6.82), lack of follow-up (aOR = 4.02, 95% CI: 2.53–6.38), HbA1c (per 1% increase) (aOR = 1.15, 95% CI: 1.03–1.29), and insulin use during hospitalization (aOR = 1.76, 95% CI: 1.13–2.75) as independent risk factors. The prediction model demonstrated excellent discrimination (AUC = 0.897) and good calibration (Hosmer-Lemeshow test <i>P</i> = 0.609). Bootstrap validation confirmed model robustness. An individualized risk calculation tool was developed with a 12% stratification threshold.</p> Conclusion <p>Medication and follow-up non-adherence are strong independent predictors of 30-day readmission in young and middle-aged T2DM patients. The high-performance prediction model provides key feature variables, precise quantitative weights, and a reliable risk stratification tool. These outputs constitute a crucial quantitative foundation that may contribute to the future development of a knowledge graph-based clinical warning system, which could ultimately facilitate a shift in diabetes management towards more proactive prevention.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predictors of 30-day readmission in young and middle-aged T2DM patients: a retrospective cohort study

  • 期刊论文

  • 发表日期：

    2026年5月2日

查看全文
  • Xiaoqing Hu,
  • Xiaohong Su,
  • Siyu Chen,
  • Qin Wang,
  • Fan He,
  • Rong Zhang

摘要

Background

Young and middle-aged type 2 diabetes mellitus (T2DM) patients exhibit high readmission rates, yet research on their self-management behaviors remains insufficient. This study investigates the impact of medication and follow-up adherence on 30-day readmission risk to provide a quantitative foundation that could inform the future development of a knowledge graph-based early warning system.

Methods

A retrospective cohort study enrolled young and middle-aged (18–59 years) T2DM patients discharged between 2020 and 2024. Exposure variables were medication adherence (medication possession ratio ≥ 80%) and follow-up adherence (≥ 1 outpatient visit in prior 6 months). The outcome was 30-day all-cause readmission. Multivariable logistic regression identified independent predictors, with robustness verified via Bootstrap sampling and sensitivity analyses.

Results

A total of 950 patients were included, with a readmission rate of 10.9%. Multivariable analysis identified medication non-adherence (aOR = 4.37, 95% CI: 2.79–6.82), lack of follow-up (aOR = 4.02, 95% CI: 2.53–6.38), HbA1c (per 1% increase) (aOR = 1.15, 95% CI: 1.03–1.29), and insulin use during hospitalization (aOR = 1.76, 95% CI: 1.13–2.75) as independent risk factors. The prediction model demonstrated excellent discrimination (AUC = 0.897) and good calibration (Hosmer-Lemeshow test P = 0.609). Bootstrap validation confirmed model robustness. An individualized risk calculation tool was developed with a 12% stratification threshold.

Conclusion

Medication and follow-up non-adherence are strong independent predictors of 30-day readmission in young and middle-aged T2DM patients. The high-performance prediction model provides key feature variables, precise quantitative weights, and a reliable risk stratification tool. These outputs constitute a crucial quantitative foundation that may contribute to the future development of a knowledge graph-based clinical warning system, which could ultimately facilitate a shift in diabetes management towards more proactive prevention.