HLTH 4200 Week 8 Screening Accuracy Write-Up Example

What a Referral From Preschool Vision Screening Establishes: Sensitivity, Specificity and Predictive Values From a Composite County Program in Which Every Child Also Had a Full Eye Examination

Student Name

Walden University

HLTH 4200: Principles of Epidemiology

Instructor Name

Month Day, Year

What a Referral From Preschool Vision Screening Establishes: Sensitivity, Specificity and Predictive Values From a Composite County Program in Which Every Child Also Had a Full Eye Examination

The Program

In the composite Lakeside County program used for this exercise, trained volunteers screened 1,000 four-year-olds at preschool programs using a photoscreener, a handheld device that estimates refractive error and eye alignment from a photograph of the eyes. In this illustration, every screened child, whatever the result, also received a comprehensive eye examination from an optometrist or ophthalmologist, which serves as the reference standard. The target conditions were amblyopia and its risk factors: strabismus, significant refractive error and unequal refractive error between the eyes. Among children younger than 6 years, 1% to 6% have amblyopia or its risk factors (US Preventive Services Task Force, 2017). Because every child had the full examination, the program's results can be laid out in a complete two-by-two table.

The Results

Table 1

Screening Result by Eye Examination Result, 1,000 Children

Counts are composite values chosen to illustrate the calculations; prevalence is 5%, within the range reported for this age group.

Sensitivity and Specificity

Sensitivity asks how many of the affected children the screening catches (Altman & Bland, 1994a). Here, 40 of the 50 children with a target condition screened positive: sensitivity = 40 ÷ 50 = 80%. The other 10 children, one in five of those affected, passed the screening and would not have been referred.

Specificity is the proportion of children without the condition whom the screening correctly passes. Here, 855 of the 950 children without a target condition screened negative: specificity = 855 ÷ 950 = 90%. The other 95 were referred unnecessarily.

These figures are plausible for preschool screening. In a large study comparing tests administered by trained screeners to children enrolled in Head Start, test performance was reported as sensitivity at a fixed specificity of 0.90, and lay and nurse screeners achieved broadly similar sensitivities on most instruments (Vision in Preschoolers Study Group, 2005). Sensitivity and specificity describe the test itself and do not change with how common the condition is in the population screened.

What a Referral Establishes

Sensitivity and specificity answer the questions a test designer asks. Parents and the clinicians receiving referrals ask a different question: if a child is referred, how likely is it that the child actually has a condition? That is the positive predictive value (Altman & Bland, 1994b).

Positive predictive value = true positives ÷ all positives = 40 ÷ 135 = 29.6%.

Of the 135 children referred, only about 3 in 10 had a target condition. The other 7 in 10 were referred and, on examination, found to have no amblyopia or risk factor. A referral, in this program, establishes that a child needs an eye examination; it does not establish that the child has a vision problem.

Negative predictive value = true negatives ÷ all negatives = 855 ÷ 865 = 98.8%.

Of the 865 children who passed, about 99 in 100 had no target condition. A passing result is reassuring but not conclusive: 10 children with a condition were among those who passed.

Why the Positive Predictive Value Is Low

How often a referral is confirmed rises and falls with the condition's frequency in the group screened, not only with the test (Altman & Bland, 1994b). With a prevalence of 5%, children without the condition outnumber children with it 19 to 1. Even a test that passes 90% of unaffected children refers 10% of a very large group, and those false positives, 95 children, outnumber the 40 true positives.

Applied to a group where 10% of children had a target condition, the same test would confirm nearly half of its referrals. Where only 2% did, about one referral in seven would be.

Table 2

Positive Predictive Value at Different Prevalences, Same Test

Sensitivity 80% and specificity 90% are held constant; only prevalence changes.

Why This Illustration Is Unusual

In most screening programs, only children who screen positive receive the full examination. Children who pass are not examined, so false negatives are never found, and sensitivity and specificity cannot be calculated from program data alone. That is why accuracy studies, like the one this illustration imitates, examine every child regardless of result. The illustration assumes that design so that all four cells of the table are known; a real program's records would fill only the top row.

Implications for the Program

Three implications follow. First, letters to parents should say what a referral means: that their child needs a full eye examination, and that most referred children turn out not to have a problem. Overstating the meaning of a referral causes worry; understating it may lead families to skip the examination that the 40 affected children need. Second, the program should track the outcomes of referrals. In a real program, children who pass are not usually examined, so sensitivity and negative predictive value cannot be measured directly, but the positive predictive value can be, by recording the results of follow-up examinations. Third, the 10 children missed by screening are a reminder that a passing result does not replace attention to signs of vision problems at home or at school.

Conclusion

In this composite program, preschool vision screening identified 80% of children with amblyopia or its risk factors and passed 90% of children without them. But because the conditions affected only 5% of children, fewer than a third of referred children were found to have one. A referral establishes the need for an examination, not the presence of a problem, and how much it establishes depends as much on the population screened as on the test.

References

Altman, D. G., & Bland, J. M. (1994a). Diagnostic tests 1: Sensitivity and specificity. BMJ, 308(6943), 1552. https://doi.org/10.1136/bmj.308.6943.1552

Altman, D. G., & Bland, J. M. (1994b). Diagnostic tests 2: Predictive values. BMJ, 309(6947), 102. https://doi.org/10.1136/bmj.309.6947.102

US Preventive Services Task Force. (2017). Vision screening in children aged 6 months to 5 years: US Preventive Services Task Force recommendation statement. JAMA, 318(9), 836-844. https://doi.org/10.1001/jama.2017.11260

Vision in Preschoolers Study Group. (2005). Preschool vision screening tests administered by nurse screeners compared with lay screeners in the Vision in Preschoolers Study. Investigative Ophthalmology & Visual Science, 46(8), 2639-2648. https://doi.org/10.1167/iovs.05-0141