Table 1 Summary of literature algorithms for basal insulin attenuation or suspension proposed by academic research groups
Reference paper | Type | Inputs | Prediction method | PH, min | T, mg/dl | Min–max suspension time, min | Assessment type |
|---|---|---|---|---|---|---|---|
Buckingham et al. Diabetes Technol Ther, 2009 [14] | Prediction-based suspension | CGM | Simple linear regression in time, statistical models [15] | 30 | 80 | 90–90 | Inpatient [14] |
Buckingham et al. Diabetes Care, 2010 [16] | Prediction-based suspension | CGM | Voting schema of 5 separate prediction algorithms [17] | 35 | 80 | 30–90 | Inpatient [16] |
Hughes et al. J Diabetes Sci Technol, 2010 [22] | Detection-based attenuation | CGM | – | – | 120 | – | In silico [22] |
Hughes et al. J Diabetes Sci Technol, 2010 [22] | Prediction-based attenuation | CGM, insulin | Kalman filter with metabolic state observer | 15 | 120 | – | In silico [22] |
Patek et al. IEEE Trans Biomed Eng, 2012 [26] | Prediction-based attenuation | CGM, insulin | Simple linear regression in time | 17 | 112.5 | – | In silico [26] |
Cameron et al. J Diabetes Sci Technol, 2012 [19] | Prediction-based suspension | CGM | Kalman filter [18] | 70 | 80 | 0-120 | |
Buckingham et al. Diabetes Technol Ther, 2013 [20] | Prediction-based suspension | CGM | Kalman filter [18] | 30 | 80 | 0–120 | |
Hughes et al. Comput Methods Programs Biomed, 2013 [25] | Prediction-based attenuation | CGM, insulin | Simple linear regression in glucose with parameters from historical data | 30 | 120 or 140 | – | In silico [25] |
Stenerson et al. J Diabetes Sci Technol, 2014 [21] | Prediction-based suspension | CGM, accelerometer, heart rate | Kalman filter [18] | 30 | 80 | 0–120 | In silico [21] |