Skip to content

Abstract Search

Mental Health

Investigating predictive value and effect estimates between two commonly used depression scales in via secondary data from HPTN 068, a conditional cash transfer trial of adolescent girls and young women in South Africa Valerie Lucas* Valerie Lucas Joanna Asia Maselko Molly Rosenberg Marie Stoner M Luz Reyes McNaughton Kathleen Kahn Audrey Pettifor

Intro

The Child Depression Index (CDI) and Center for Epidemiologic Studies Depression scale (CESD) are scales to measure and screen for depression. CDI is a version of the Beck Depression Inventory simplified for minors, and CESD is generally used for adults. Secondary data from HPTN 068 provided a unique opportunity to assess scale concordance in an adolescent female sample, a time period when minors transition to young adulthood.

Methods

Young women (n = 2530) in rural South Africa were randomized 1:1 to a conditional cash transfer (CCT) or control. Participants were eligible to receive the CCT while they remained unmarried and maintained school attendance. Both the CDI and the CESD were administered at final study visit. Without a gold standard, we compared positive predictive values (PPV) of CDI and CESD for each other across age and intervention assignment and explored how choice of metric impacted CCT effect estimates.

Results

At study conclusion, participants had a median age of 18 (IQR: 17,18). The prevalence of depression was 26% via CDI and 27% via CESD. But PPV for CDI in detecting depression via CESD is only 53%, and the PPV of CESD for CDI is 57%. The PPV is lower for those <18 and higher for those 18+. For the <18 group, the PPV of CDI for CESD is 52%, and the PPV of CESD for CDI is 41%. For the 18+ group, the PPV of CDI for CESD is 57%, and the PPV of CESD for CDI is 62%. Comparing risk difference estimates, the CCT reduced depression by 0.7 percentage points (95% CI: -4.7, 3.4) via CDI and by 2.7 percentage points (95% CI: -0.9, 6.3) via CESD.

Conclusion

The PPV between CDI and CESD are surprisingly low, given the same stated construct and similar prevalence and effect estimates. The PPV of the CESD for CDI is greater than the PPV of CDI for CESD among 18+ participants, and vice versa for <18 participants. These differences suggest that CDI and CESD measure different underlying (sets of) constructs with age-based differences in depression symptoms.