Fig 1.
Flow chart: Transforming the database of the Acute Pain Service (APS) into nine data sets.
For each of the three postoperative days, three data sets are created: one for patient’s opinion (PO), one for nurses’ observation (NO) and one for the combined variable (PONO). *Before this point multiple records are possible per patient but after this point the number of records equals the number of patients. mv = missing values.
Table 1.
Name, abbreviation, values and coding of variables used in the logistic regression models to estimate the relationships between four explanatory variables and each of three response variables.
Table 2.
Numbers and characteristics of patients.
Fig 2.
Relative frequencies for observations of patients’ opinion (A), nurses’ observation (B), and the measure combining patient’s opinion and nurses’ observation (C) against NRS-MEP scores. The observations in all patients gathered during the first three postoperative days were pooled.
Fig 3.
Estimated proportion (solid curve with its 95% confidence band) of patients that accept the pain (A-C), move appropriately (D-F) or accept the pain and move appropriately (G-I) as a function of NRS-MEP for day 1, 2 and 3 after surgery. The open circles represent the observed proportions of patients at each of the eleven points of the NRS. For each of the nine data sets, one estimated proportion is computed and shown at NRS-MEP = 7.
Fig 4.
ROC curves for the dependent variables PO, NO and PONO for the three first postoperative days.
The dashed line is the line of identity where the AUC = 0.5. Open circles are the points where Youden’s J-statistic is maximal for PONO. These points are, by definition, the ‘optimal’ cut-off points.
Table 3.
Cut-off points obtained from the logistic regression model using the 11-point Numerical Rating Scale for movement-evoked pain as explanatory variable for each of the three dependent variables PO, NO, and PONO.
Shown are the optimal cut-off points with their associated sensitivities and specificities, as well as the areas under the ROC curves (AUC).