. 2016 May 17;24(1):198–208. doi: 10.1093/jamia/ocw042

Table 3.

Areas of improvement

	Challenge	Way to Improve	Rationale
1	Multicenter studies	Use EHRs from multiple sites	Assess portability of models
1	Multicenter studies	Validate Across Sites	Assess portability of models
2	Predictor variables	Incorporate time-varying (longitudinal) factors	Better leverage from the available data
2	Predictor variables	Use larger variable sets	Better leverage from the available data
3	Consideration of biases	Missing data	Assess the robustness of the models
		Loss to follow-up
		Informed presence
4	Evaluation metrics	External validation	Develop an understanding of how models will impact clinical decision making
4	Evaluation metrics	Metrics of Clinical Utility (eg, PPV, Net Benefit)