Background:  HIV behavioral studies often rely on patient recall of CD4+ counts and viral loads (VL). The recall accuracy of such data has not been established. We report data from two large HIV clinics comparing recall with chart review.

Methods:  202 patients at DeKalb and Cobb County Ryan White Clinics were enrolled in a CDC-sponsored prevention in care project. Patients completed Audio Computer-Assisted Self-Interviews (ACASI), including questions about CD4+ and VL values. Chart abstraction ascertained actual CD4+ and VL values.  Relationship between recalled and actual values was examined by Pearson correlation. Predictors of correlation were evaluated by ANOVA and multiple regression analysis. 

Results: Pearson correlations between recalled and actual CD4+ / VL values were as follows:

	Correlation	p value
Most recent VL	0.823	0.01
Most recent CD4+	0.762	0.01
Lowest CD4+	0.456	0.01

Race, gender, sexual identity, age, income, education, time in care, and ART use did not predict recall accuracy for CD4+ or VL.  Site of care was the only other predictor, for perceived CD4 count.  Those with undetectable VL demonstrated highest correlation between recalled and actual CD4+ (0.85, p=0.01). Those with CD4+ > 200 demonstrated highest correlation between recalled and actual VL (0.90, p=0.01). 

Conclusions:  Patients had high correlation between recalled and actual values for recent CD4+ and VL, particularly for those with undetectable VL or CD4+>200 cells/µl. Further analyses are underway to examine other predictors of recall accuracy and associations with sexual risk behaviors.