Qatar Metabolic Institute, Qatar.
Qatar Computer Research Institute (QCRI), Doha, Qatar.
Diabetes Res Clin Pract. 2020 Nov;169:108388. doi: 10.1016/j.diabres.2020.108388. Epub 2020 Aug 26.
To develop a machine-based algorithm from clinical and demographic data, physical activity and glucose variability to predict hyperglycaemic and hypoglycaemic excursions in patients with type 2 diabetes on multiple glucose lowering therapies who fast during Ramadan.
Thirteen patients (10 males and three females) with type 2 diabetes on 3 or more anti-diabetic medications were studied with a Fitbit-2 pedometer device and Freestyle Libre (Abbott Diagnostics) 2 weeks before and 2 weeks during Ramadan. Several machine learning techniques were trained to predict blood glucose levels in a regression framework utilising physical activity and contemporaneous blood glucose levels, comparing Ramadan to non-Ramadan days.
The median age of participants was 51 years (IQR 49-52); median BMI was 33.2 kg/m (IQR 33.0-35.9) and median HbA1c was 7.3% (IQR 6.7-7.8). The optimal model using physical activity achieved an R of 0.548 and a mean absolute error (MAE) of 30.30. The addition of electronic health record (ehr) information increased R to 0.636 and reduced MAE to 26.89 and the time of the day feature further increased R to 0.768 and reduced MAE to 20.55. Combining all the features together resulted in an optimal XGBoost model with an R of 0.836 and MAE of 17.47. This model accurately estimated normal glucose levels in 2584/2715 (95.2%) readings and hyperglycaemic events in 852/1031 (82.6%) readings, but fewer hypoglycaemic events (48/172 (27.9%)). The optimal XGBoost model prioritized age, gender, BMI and HbA1c followed by glucose levels and physical activity. Interestingly, the blood glucose level prediction by our model was influenced by use of SGLT2i.
XGBoost, a machine learning AI algorithm achieves high predictive performance for normal and hyperglycaemic excursions, but has limited predictive value for hypoglycaemia in patients on multiple therapies who fast during Ramadan.
从临床和人口统计学数据、身体活动和血糖变异性中开发出一种基于机器的算法，以预测在接受多种降血糖治疗且在斋月期间禁食的 2 型糖尿病患者中出现高血糖和低血糖波动。
对 13 名（10 名男性和 3 名女性）接受 3 种或更多种抗糖尿病药物治疗的 2 型糖尿病患者，使用 Fitbit-2 计步器设备和 Freestyle Libre（Abbott Diagnostics）在斋月前 2 周和斋月期间 2 周进行研究。利用身体活动和同期血糖水平，在回归框架中训练了几种机器学习技术，以比较斋月和非斋月的血糖水平。
参与者的中位年龄为 51 岁（IQR 49-52）；中位 BMI 为 33.2kg/m（IQR 33.0-35.9），中位 HbA1c 为 7.3%（IQR 6.7-7.8）。使用身体活动的最佳模型的 R 为 0.548，平均绝对误差（MAE）为 30.30。添加电子健康记录（ehr）信息后，R 增加到 0.636，MAE 降低到 26.89，时间特征进一步增加到 0.768，MAE 降低到 20.55。将所有特征结合在一起，得到了最佳的 XGBoost 模型，R 为 0.836，MAE 为 17.47。该模型在 2715 次读数中的 2584 次（95.2%）中准确估计了正常血糖水平，在 1031 次读数中的 852 次（82.6%）中准确估计了高血糖事件，但在 172 次读数中的 48 次（27.9%）中低血糖事件较少。最优的 XGBoost 模型优先考虑年龄、性别、BMI 和 HbA1c，其次是血糖水平和身体活动。有趣的是，我们模型的血糖预测受到 SGLT2i 应用的影响。
XGBoost 是一种机器学习人工智能算法，在预测正常和高血糖波动方面具有较高的预测性能，但在接受多种治疗且在斋月期间禁食的患者中预测低血糖的能力有限。
