Table 3

Attribute-level estimates for the panel mixed logit model

AttributeLevelsThe NetherlandsPoland
EstimateSEP sig.EstimateSEP sig.
Precision compared with
finger pricking
Accurate as finger pricking (ref)Mean0.3430.075***0.4570.071***
SD
±0.3 mmol/LMean0.0000.061−0.0810.051
SD0.0360.1010.0470.112
±0.6 mmol/LMean−0.3430.079***−0.3760.073***
SD0.5360.093***0.6550.082***
Average number of finger pricks per day0 times per day (ref)Mean0.3520.059***0.1720.044***
SD
2 times per dayMean−0.3520.059***−0.1720.044***
SD0.4790.070***0.3490.059***
Effort to checkLow (ref)Mean0.1200.039**0.0420.033
SD
ModerateMean−0.1200.039**−0.0420.033
SD
Probability of getting skin irritation or redness5% (ref)Mean0.3360.076***0.3770.064***
SD
20%Mean−0.0590.066−0.0180.059
SD0.0610.1270.0150.166
35%Mean−0.2770.076***−0.3590.066
SD0.4500.097***0.4020.084***
Monthly costsper €1 increaseMean−0.0170.002***−0.0160.001***
SD0.0150.001***0.0190.001***
Glucose informationCurrent glucose level only (ref)Mean−0.1420.063*−0.0560.054
SD
Current glucose level with arrowMean0.0680.0630.0040.055
SD
Current glucose level with daily trendlineMean0.0740.0630.0520.053
SD
AlarmsYes (ref)Mean0.1520.044***0.1480.035***
SD
NoMean−0.1520.044***−0.1480.035***
SD0.2520.076***0.1510.063*
Alternative specific constant for device instead of finger-prick testMean−0.9820.502−2.7700.336***
SD4.5270.386***4.7670.371***
Alternative specific constant indicating left–right biasMean0.3760.085***0.3460.074***
SD0.4460.140***0.5400.098***
  • Higher estimates represent increasing levels of importance for the patient in choosing a device.

  • All attributes were effects-coded, enabling the direct comparison of the estimates. The sum of the effect coded attributes is zero, and therefore the coefficient of the reference category can be easily calculated, and the relative importance of the reference categories of the attributes can be compared with one another, and so that the alternative specific constants have independent interpretation signifying the average utility for that alternative. SDs are given where parameters were found to have a significant random parameter estimate. The significant alternative specific constant indicates that patients tended to choose the alternative on the left side. A normal distribution using 1000 Halton draws was used in model development.

  • *P<0.05, **P<0.01, ***P<0.001.

  • ref, reference level; SD, standard deviation; SE, standard error.