Table 2 –

Accuracy assessment of neoadjuvant chemotherapy data

	Completely automated extraction*	Partially identified by the NLP built dataset**	Completely manual extraction***
Number of patients	265	43	61
Percent of total received neoadjuvant chemotherapy	72%	12%	16%
Accuracy of automated data extraction	Precision: 99% Recall: 99%	Precision: Medication names 99% Dates of receiving medications 80% Recall: Medication names 99% Dates of receiving medications 85%	N/A

^*265 patients received neoadjuvant chemotherapies inside MDA or from our pharmacy (254) for those the information is automatically inserted from the pharmacy system so medication information and dates were accurate. 231 patients had formulated the treatment plan inside MDA and was synoptically reported by the GI medical oncologist so also medications information and dates were accurate.

^**43 patients had neoadjuvant chemotherapy in their clinical Notes identified with dates but received the chemotherapy in an outside facility.

^**61 patients received neoadjuvant chemotherapy in an outside facility and either didn’t have dates with their regimens in the NLP dataset or weren’t included in the NLP dataset.