Research Article

Sequential Pattern Mining to Predict Medical In-Hospital Mortality from Administrative Data: Application to Acute Coronary Syndrome

Table 3

Means of area under the ROC curve (AURC), F-measure, and error rate for the different types of models and similarities in the modeling of ICD-10 code trajectories.

AURCF-measureError rate
ModelSimilarityMean95% CIMean95% CIMean95% CI

NBEdition0.770.68–0.860.700.62–0.820.260.16–0.34
q-gram0.720.64–0.770.640.58–0.700.330.28–0.38
Heuristic0.730.64–0.820.660.60–0.770.320.24–0.39

KNNEdition0.440.38–0.530.580.53–0.630.380.35–0.43
q-gram0.500.45–0.550.570.52–0.610.400.37–0.44
Heuristic0.540.46–0.590.550.52–0.650.410.38–0.46

TreeEdition0.740.66–0.830.660.56–0.790.280.19–0.35
q-gram0.670.62–0.710.630.57–0.700.340.30–0.39
Heuristic0.700.64–0.800.650.57–0.770.310.22–0.38

LREdition0.770.68–0.880.700.62–0.830.270.16–0.35
q-gram0.750.65–0.820.690.62–0.770.290.23–0.38
Heuristic0.740.64–0.820.690.62–0.800.300.21–0.39

SVMEdition0.830.76–0.920.700.61–0.820.250.16–0.33
q-gram0.800.72–0.890.660.60–0.730.310.26–0.37
Heuristic0.840.77–0.920.700.64–0.810.270.20–0.36

ANNEdition0.820.72–0.940.700.59–0.850.250.14–0.33
q-gram0.810.72–0.900.700.62–0.790.280.21–0.37
Heuristic0.830.71–0.960.730.63–0.860.260.14–0.35

CI = confidence interval. Best results are in bold.