Table 2.

Out-of-domain evaluation metrics of PubMedBERT fine-tuned on the full SURUS dataset

Label class	Interventional				Observational
	Precision	Recall	F1	Support	Precision	Recall	F1	Support
Disease	0.99	0.90	0.94	664	0.95	0.87	0.91	302
Drug	0.91	0.85	0.87	4,759	0.81	0.74	0.76	338
Id	1.00	0.98	0.99	341	1.00	1.00	1.00	15
Methodology	0.96	0.89	0.92	3,851	0.91	0.77	0.82	1,627
Parameter	0.83	0.76	0.79	3,003	0.78	0.68	0.73	1,345
Result	0.96	0.96	0.96	5,164	0.93	0.91	0.92	1,976
Therapy	0.97	0.85	0.90	1,273	0.33	0.50	0.40	2
Weighted Mean	0.93	0.88	0.90	19,055	0.88	0.80	0.84	5,605