Fig. 3

A, C Illustration of the standardized net benefit of the prognostic model (blue lines) compared with the strategies of monitoring or treating all individuals (black lines) and not providing any follow-up or treatment (red lines) over a range of risk thresholds. B, D The number of persons (out of 1000) who would be categorized at high or low risk for each risk threshold. A The use of the prognostic model to omit persons from follow-up who are at low risk of an urgent care visit or worse. The Y-axis represents the net increase in the proportion of low-risk individuals who avoid unnecessary monitoring (out of a hypothetical maximum achieved when the true negative rate is one and false negative rate is zero) compared with the strategy of enrolling all persons. B Illustration of the expected number of individuals (out of 1000) who would be omitted from monitoring using the prognostic model as a function of low-risk threshold (blue line) and the number of persons who would be omitted and who would never require an urgent in-person evaluation or worse (dashed blue line). C The use of the prognostic model to offer individuals who are at high risk of hospitalization or worse more rigorous follow-up or therapeutic intervention. The Y-axis represents the net increase in the proportion of high-risk individuals who are offered treatment (out of a hypothetical maximum achieved when the true positive rate is one and the false positive rate is zero) compared with the strategy of treating no cases. D The expected number of individuals (out of 1000) who would be offered treatment (blue line) and not be offered treatment (dashed green line) as a function of high-risk threshold. Also shown is the expected number of high-risk individuals (blue line) and those not at high risk (dashed green line) who later would have required hospitalization