Table 1 Overview of available specification parameters
Parameter | Description of feature |
|---|---|
Patient generation | |
 # of institutions | Total number of institutions to represent in the dataset |
 Provider distribution | Distribution of number of providers within each institution |
 Patient feature set | Distributions and correlations of patient features specified by EHR dataset or user definitions |
 Provider patient volumes | Distribution of the annual number of patients treated by a provider |
 Provider entry | Whether provider entry into case series should be staggered |
Treatment assignment | |
 Treatment prevalence | Proportion of patients receiving each treatment |
 Treatment associations | Associations specifying how patient features influence treatment assignment |
Patient and treatment associated risk | |
 Treatment risk | Difference in the risk of an adverse outcome associated with a novel treatment compared to the reference treatment |
 Outcome risk factors | Associations specifying how patient features influence the risk of an adverse outcome |
 Population adverse event rate | Proportion of the population experiencing an adverse outcome due to patient risk factor and treatment risk |
Provider learning-associated risk | |
 Provider learning – form | Functional form of learning curve (impacts steadiness or steepness of learning) |
 Provider learning – speed | Number of patients receiving the novel treatment before providers reach 95% of asymptotic performance |
 Provider learning—magnitude | Magnitude of learning-associated risk when a provider first starts using the novel treatment |
Institutional learning-associated risk | |
 Institutional learning – form | Functional form of learning curve (impacts steadiness or steepness of learning) |
 Institutional learning – speed | Number of patients receiving the novel treatment before institutions reach 95% of asymptotic performance |
 Institutional learning – magnitude | Magnitude of learning-associated risk when an institution first starts using the novel treatment |
Data finalization | |
 Missingness | Proportion of missing values |
 Noise | Additional random noise in outcome generation |
 Omitted variables | Patient features to be excluded from final data |