Table 1.

Contributions of the twenty included papers in response to the five requirements for deep phenotyping.

Reference	First author	Summary of Contributions
Requirement 1: Natural Language Processing
[14]	Datta, S	A systematic review of NLP on cancer notes
[15]	Liu, Q	Symptom extraction for patient stratification
[16]	Lyudovyk, O	NLP on pathology notes for subtyping
[17]	Liu, C	Ensemble of NLP for better portability
Requirement 2: Standardization
[18]	Hong, N	A FHIR-based EHR phenotyping framework
[19]	Shang, N	An empirical study of “making phenotyping work visible” that demonstrates the need for standardized processes
[20]	Hripcsak, G	Demonstrate OMOP’s value in improving phenotyping algorithms’ portability
[21]	Ostropolets, A	Adapting EHR phenotypes to claims data using OMOP Common Data Model
[22]	Reps, J	OMOP CDM-based probabilistic phenotyping algorithms using self-reported data
[23]	Swerdel, J	OMOP CDM-based standardized phenotype evaluation algorithms
[24]	Warner, J	Expansion of OMOP CDM to cancer phenotypes
[25]	Shen, F	Extension of HPO using embedding of phenotype knowledge resources
Requirement 3: Novel Data for Phenotyping
[26]	Trace, JM	Using voice to diagnose Parkinson’s disease
Requirement 4: Temporal Phenotyping and Subtyping via Similarity Metrics
[27]	Mate, S	A graphical model of temporal constraints
[28]	Meng, W	Temporal phenotyping of cancer treatment pathways
[29]	Zhao, J	Temporal phenotyping via tensor factorization
[30]	Chen, X	Phenotypic similarity for rare diseases
[31]	Xu, Z	Subtyping for acute kidney injury
Requirement 5: Scalability
[32]	Zhang, L	Automated grouping of medical codes
[33]	Chen, P	Deep representation learning for phenotyping