CROI 2016 Abstract eBook

Abstract Listing

Poster Abstracts

1035 Differences in HIV Viral Suppression by Frequency and Type of Healthcare Visits Michael A. Horberg 1 ; Jackie Blank 1 ; Kevin B. Rubenstein 1 ; Leo B. Hurley 2 ; Julia L. Marcus 2 ; Daniel B. Klein 3 ; Peter M. Kadlecik 4 ; Michael J. Silverberg 2 1 Mid-Atlantic Permanente Med Group, Rockville, MD, USA; 2 Kaiser Permanente Northern California, Oakland, CA, USA; 3 Kaiser Permanente San Leandro Med Cntr, San Leandro, CA, USA; 4 Mid-Atlantic Permanente Med Group, Washington, DC, USA Background: HIV care typically involves two in-person visits per year, but there is increasing use of alternative patient-provider “visits” (telephone and/or email). We explored differences in HIV viral suppression by frequency and type of visits. Methods: The study population consisted of adult HIV+ patients in Kaiser Permanente Mid-Atlantic States (KP), providing comprehensive HIV care to health plan members in District of Columbia, Maryland, and Virginia. We restricted to members with ≥6 months membership and ≥1 viral load (VL) measurement in 2014. HIV viral suppression (“BLQ”; HIV RNA <200 copies/mL) was measured at last VL measurement during 2014. We compared HIV viral suppression by number of in-person visits to an HIV specialist or primary care (1 or ≥2), and among those with 1 in-person visit by additional use of phone and/or email visits to an HIV specialist (none, phone only, email only, and “both” phone and email). The

reference group was those with ≥2 in-person visits (with or without phone or email). We also compared those with ≥2 in-person visits by visit type, with ≥2 in-person visits only as reference. Adjusted odds ratios (OR) for BLQ by frequency and type of visits were obtained frommultivariable logistic regression adjusting for age, sex, race/ ethnicity, and HIV risk. Results: Among 1958 subjects, there were 1278 (65%) with ≥2 in-person visits and 680 (35%) with 1 in-person visit. Among those with 1 visit, 15% also had a phone visit, 39% had an email visit, 21% had neither, and 26% had both. As shown in the Figure, %BLQ was greater for ≥2 visits compared with 1 visit, except if both phone and email visits were used. In adjusted models, BLQ was reduced compared with ≥2 in-person visits for 1 in-person visit only (OR=0.48 [95% CI: 0.30, 0.74]) or 1 in-person + phone (OR=0.46 [0.28, 0.78]). However, differences were not statistically significant comparing ≥2 in-person visits with 1 in-person + email (OR=0.81 [0.54, 1.21]) or both (OR=1.06 [0.63, 1.80]). Among patients with ≥2 in-person visits only, there was no significant difference in the adjusted odds BLQ among 2 in-person plus email or plus phone or plus both compared with 2 in-person visits only, although there is a trend toward greater odds of BLQ with 2 in-person plus email (OR=1.65 [0.99, 2.76]). Conclusions: One in-person visit per year was not statistically different with respect to HIV viral suppression compared with 2 in-person visits, if supplemented by alternative communications such as email with or without phone.

1036 Community Viral Load: Measure Validation and Public Health Utility Kathryn M. Leifheit 1 ; Christina Schumacher 1 ; Patrick Chaulk 2 ; Carolyn Nganga-Good 2 ; Jacky M. Jennings 1 1 Johns Hopkins Univ Sch of Med, Baltimore, MD, USA; 2 Baltimore City Hlth Dept, Baltimore, MD, USA

Background: Reductions in HIV Community Viral Load (CVL) are associated with decreased HIV infections. In an environment of constrained public health resources, HIV control activities may be best targeted to high CVL areas or CVL “hotspots.” There is a lack of consensus, however, among HIV experts regarding how best to calculate CVL, i.e. how to aggregate individual-level viral loads to community areas in order to best reflect community-level transmission dynamics. The objective of this analysis was to determine the extent to which each of five CVL metrics in the past year predicts new HIV cases in the subsequent year in one mid-Atlantic U.S. city with an established HIV epidemic. The goal of this analysis was to inform HIV control activities in the local jurisdiction. Methods: We calculated 5 different CVL measures at the census tract level, using viral load data between October 2012 and December 2013 in Baltimore, MD via routine surveillance. Data includes point-of-diagnois as well as in-care viral loads. The CVL measures derived from the HIV public health literature included arithmetic mean VL, geometric mean VL, median VL, total VL, and population proportion with unsuppressed VL (VL>400 copies/mL). We then used census tract level rates of new HIV diagnoses in 2014 to gauge the predictive value of each of the CVL measures via inverse variance weighted modeling of the relationships between log-transformed new diagnoses (2014) and log-transformed estimates of CVL (2012-13). Results: The 2012-13 surveillance data was comprised of 2,542 HIV-infected individuals living in 96.5% (193/200) Baltimore City census tracts. The study population was 65.0% male and 84.5% African American, with a mean age of 47.1 years. 9.1% of individuals were newly diagnosed and 33.5% had an unsuppressed viral load. Of the CVL metrics analyzed, proportion with unsuppressed VL was most strongly associated with new cases in the subsequent year (correlation= 0.309; p=0.001). The remaining CVL measures were not significantly associated with new cases (see Figure). Conclusions: Population proportion with unsuppressed VL in the past year was the CVL measure that was most predictive of new HIV diagnoses in the subsequent year. In this one mid-Atlantic U.S. city, population proportion with unsuppressed VL may be the most appropriate measure for utilization in targeted HIV control activities.

Poster Abstracts

443

CROI 2016