Alaska Blind Child Discovery

Referral Criteria

The ideal screening device would “Refer” everyone with the target disease, but “Pass” everyone who did not have the target disease. Unfortunately there are rarely perfect screeners for real diseases like childhood amblyopia. As a result, even good screeners refer some children who do not actually have the target condition, Amblyopia; this is called a “False Positive” meaning the screener is not perfectly “Specific.” The imperfect screener may fail to refer some children who actually have the target condition leading to “False Negatives” or a screening that is not perfectly “Sensitive.” Issues like “sensitivity” and “specificity” are important aspects of the “Validity” of a screening device or program.
If the screener produces an output with various levels, then the screening program must choose a “referral criteria” or “cut-off” above which children are referred and below which children are passed. Tis means two different kinds of criteria. Higher ‘cut-offs” are less likely to refer normal children (more specific) but then they are likely to miss some children with milder disease (less sensitive). On the other hand, lower referral cut-offs are more sensitive because they are likely not to miss some children with disease, but then more normal children (false positives) are referred making less specific effort. The cost of a specific program is less and the overall cost of a sensitive program can be much more. The relationship between the sensitivity and specificity for a given screening program for various referral cut-offs is graphically demonstrated by a “Receiver-Operator Characteristic (ROC) Curve.”
Analogy: Radar Gun for Catching Speeders: The local speed limit of 55 mph was designed to reduce highway accidents. You then review various technologies and obtain a radar gun that outputs in miles per hour the difference between a traveling object (ie a car) and the radar gun. It costs time and effort to deal with speeders and makes them mad if they are pulled over for minor infractions just above 55. On the other hand, you want to catch every crazy driver going over 100 mph. So you arbitrarily decide to pull over everyone with radar gun readings of 63 MPH and above. Yes, someone traveling at 62 MPH was actually speeding but you didn’t catch them. And someone traveling at 65 MPH was not speeding by very much (you may decide to give them a warning rather than a ticket).
Switching your radar gun cut off to 60 MPH will catch more speeders, but will clog the court system. Switching your radar gun cut-off to 70 MPH will miss more moderate speeders, but will concentrate those caught as having the worst speeding risk.
Pediatric Vision Screeners set a potentially arbitrary cut-off to maximize sensitivity and specificity for the goals of their program. Most parents, pediatricians and busy referral eye doctors prefer a program with high specificity whereas the risk of not referring a mild-moderate case of amblyopia is greater since sensitivity is necessarily reduced. Pediatricians are comfortable with screening tests that are NOT perfectly sensitive as long as a series of age-appropriate screenings take place over the course of development; therefore the AAP Vision Screening Guidelines. Introduction to Validation video.