Table 7.

Event and attribute classification performance with predicted medication mentions on the withheld CMED test set. We used the Clinical BERT model to extract the medications.

Model	Metric	MedSingleTask		MedMultiTask		MedSpan		MedIdentifiers		MedIDTyped
		Macro	Micro	Macro	Micro	Macro	Micro	Macro	Micro	Macro	Micro
event	P	0.733	0.850	0.736	0.860	0.795	0.884	0.808	0.896	0.761	0.877
	R	0.664	0.850	0.703	0.860	0.784	0.884	0.803	0.896	0.807	0.877
	F1	0.693	0.850	0.718	0.860	0.789	0.884	0.805	0.896	0.776	0.877
action	P	0.686	0.708	0.774	0.718	0.844	0.790	0.809	0.772	0.872	0.800
	R	0.472	0.466	0.558	0.596	0.661	0.710	0.669	0.717	0.694	0.743
	F1	0.555	0.562	0.632	0.651	0.728	0.748	0.727	0.743	0.763	0.770
actor	P	0.600	0.733	0.665	0.813	0.702	0.839	0.741	0.837	0.708	0.851
	R	0.499	0.681	0.520	0.668	0.544	0.765	0.609	0.785	0.552	0.782
	F1	0.540	0.706	0.582	0.733	0.602	0.801	0.663	0.810	0.610	0.815
temporality	P	0.687	0.702	0.745	0.773	0.716	0.770	0.704	0.776	0.727	0.791
	R	0.553	0.599	0.575	0.622	0.621	0.678	0.636	0.723	0.636	0.726
	F1	0.611	0.647	0.645	0.690	0.665	0.721	0.667	0.749	0.676	0.757
certainty	P	0.572	0.710	0.662	0.785	0.742	0.831	0.748	0.828	0.730	0.833
	R	0.648	0.694	0.548	0.619	0.679	0.736	0.772	0.769	0.711	0.762
	F1	0.603	0.702	0.591	0.692	0.696	0.781	0.755	0.797	0.704	0.796
Overall	P	0.655	0.741	0.716	0.790	0.760	0.823	0.762	0.822	0.760	0.830
	R	0.567	0.658	0.581	0.673	0.658	0.755	0.698	0.778	0.680	0.778
	F1	0.600	0.693	0.634	0.725	0.696	0.787	0.723	0.799	0.706	0.803