Hylke Salverda

39 2 Automated oxygen control in preterm infants, how does it work and what to expect in the Sophie ventilator (Stephan GmbH, Gackenbach, Germany). Information on the workings of the algorithm is limited.34 The control loop iterates every two seconds, conveniently using the SpO2 measured by an existing bedside monitor. Separate PID controllers update the set FiO2 and the baseFiO2. The primary controller adjusting set FiO2 unconventionally uses variable PID coefficients which depend on the range in which SpO2 falls and the speed and direction of the change. In an attempt to account for the shape of the oxygendissociation curve, the proportional term is exponentially weighted. The second PID loop updates the baseFiO2 every 5 minutes by comparing recent values for set FiO2 to the previous baseFiO2. Clinical effect Although the data was only recently published34 the algorithm was clinically tested in 2014 in a randomised crossover trial (Table 1). The algorithm proved superior to manual titration in twelve preterm infants on non-invasive respiratory support (manual titration: 68.5% within TR; SPO2C controller: 77.8%; p<0.001). VDL 1.121 How it works The VDL 1.1 algorithm, an adaptive PID algorithm, is available as the Oxygenie option on the SLE6000 ventilator (SLE, Croydon, United Kingdom). Like other algorithms it uses the baseline oxygen requirement, known as reference FiO2, which is calculated every 30 minutes using the preceding 60 minutes of data. Combining this value and the P, I and D terms the algorithm calculates a new value for the set FiO2 each second, rounded to 0.5%. Measurements with low signal IQ are labelled as missing, in which case the last set value for FiO2 is used. The algorithm does not use a lock-out period. To account for known pathophysiological idiosyncrasies, as well as the limitations of pulse oximeters, each of the PID terms is modified in different ways. The proportional term is adapted to the degree of lung dysfunction by multiplying with 0.5-1, corresponding to a reference FiO2 in the range of 21%-40%. Furthermore, the error in the proportional term is attenuated whilst SpO2 is in TR to minimise adjustments during good control.21, 36, 41 Finally, to correct for the increasing imprecision of pulse oximetry at SpO2 values below 80%42, the negative error is limited to 13%. During protracted hypoxia, the integrand (sum of past errors) will rapidly increase in