Table 5.

Event and attribute classification performance with gold standard medication mentions.

Model	Metric	MedSingleTask		MedMultiTask		MedSpan		MedIdentifiers		MedIDTyped
		Macro	Micro	Macro	Micro	Macro	Micro	Macro	Micro	Macro	Micro
event	P	0.771	0.884	0.774	0.894	0.830	0.917	0.848	0.931	0.798	0.913
	R	0.697	0.884	0.739	0.894	0.818	0.917	0.844	0.931	0.851	0.913
	F1	0.729	0.884	0.755	0.894	0.824	0.917†	0.846	0.931 ‡	0.815	0.913†
action	P	0.704	0.728	0.789	0.741	0.856	0.808	0.825	0.797	0.884	0.821
	R	0.482	0.479	0.570	0.616	0.671	0.726	0.685	0.739	0.706	0.762
	F1	0.568	0.578	0.646	0.673δ	0.739	0.765†	0.742	0.767†	0.775	0.791 ‡
actor	P	0.614	0.761	0.675	0.833	0.711	0.857	0.755	0.865	0.721	0.876
	R	0.513	0.707	0.528	0.684	0.552	0.782	0.623	0.811	0.564	0.805
	F1	0.554	0.733	0.592	0.751	0.611	0.818†	0.677	0.837 †	0.622	0.839†
temporality	P	0.707	0.729	0.768	0.802	0.727	0.785	0.724	0.804	0.743	0.812
	R	0.570	0.622	0.592	0.645	0.631	0.691	0.655	0.749	0.651	0.746
	F1	0.629	0.671	0.665	0.715δ	0.675	0.735δ	0.687	0.776†	0.691	0.778 †
certainty	P	0.580	0.730	0.670	0.806	0.748	0.846	0.760	0.856	0.737	0.851
	R	0.656	0.713	0.555	0.635	0.684	0.749	0.782	0.795	0.718	0.779
	F1	0.611	0.722	0.598	0.710	0.701	0.795†	0.766	0.824 †	0.711	0.813†
Overall*	P	0.675	0.766	0.735	0.814	0.774	0.843	0.782	0.851	0.777	0.855
	R	0.584	0.681	0.597	0.694	0.671	0.773	0.718	0.805	0.698	0.801
	F1	0.618	0.718	0.651	0.748δ	0.710	0.806†	0.744	0.827 †	0.723	0.827 †

^*Overall score is an unweighted average of the event and attribute scores.

^‡indicates performance significance $\left(p<0.05\right)$ compared across all models.

^†indicates performance significance $\left(p<0.05\right)$ compared against MedSingleTask and MedMultiTask.

^δindicates performance significance over MedSingleTask.