Using HIV Surveillance Data to Evaluate Outcomes of Site Randomized Interventions in the TLC-Plus Study
1 like449 views
The document describes the design of the HPTN 065 study which tested the efficacy of financial incentives to improve linkage to care and viral suppression outcomes across HIV testing and care sites in the Bronx, New York and Washington D.C. using aggregate HIV surveillance data from local health departments. The study involved randomizing sites to either receive financial incentives or standard of care and then measuring aggregate linkage to care and viral suppression outcomes using surveillance data to compare the two study arms.
Using HIV Surveillance Data to Evaluate Outcomes of Site Randomized Interventions in the TLC-Plus Study
- 1. Using HIV Surveillance Data to Evaluate Outcomes of Site Randomized Interventions in the TLC-Plus Study Deborah Donnell*, Irene Hall, Y Jia, Angelique Griffin, Kathleen Brady, Becky Grigg, Aaron Sayegh, Lucia Torian, Wafaa El Sadr *Vaccine and Infectious Disease Division, Fred Hutchinson Cancer Research Center, Seattle, WA 2011 National HIV Surveillance Conference: Atlanta, GA
- 2. The TLC-Plus Trial (HPTN 065) • Feasibility study of implementing “Test and Treat” for HIV Prevention in the US • Five study components with feasibility outcomes • Two study components testing financial incentives(FI) vs. Standard of care (SOC) – Site randomized outcomes in two intervention cities (Bronx NY, Washington DC) (Session E07, Wednesday 8 am)
- 3. HPTN 065: Study Design Expanded HIV Testing Linkage-to-Care: • Social mobilization Test Randomization of HIV Test Sites • Universal offer of testing in ED/hospital admission FI to link to SOC to link to care care Linkage to Care Prevention for Positives: Individual randomization of HIV+ CARE plus SOC SOC alone Provider & Patient Surveys Initiate Art per current • Knowledge and attitudes regarding ART and FIs guidelines Viral Suppression: Randomization of HIV Care Sites Viral FI for viral SOC for viral suppression suppression suppression
- 4. HPTN 065: Study Design Expanded HIV Testing Linkage-to-Care: • Social mobilization Test Randomization of HIV Test Sites • Universal offer of testing in ED/hospital admission FI to link to SOC to link to care care Linkage to Care Prevention for Positives: Individual randomization of HIV+ CARE plus SOC SOC alone Provider & Patient Surveys Initiate Art per current • Knowledge and attitudes regarding ART and FIs guidelines Viral Suppression: Randomization of HIV Care Sites Viral FI for viral SOC for viral suppression suppression suppression
- 5. Two site randomized components testing efficacy of financial incentives Linkage to Care Viral load suppression • 20 testing sites in Bronx, NY and • 20 care sites in Bronx, NY and Washington DC (40 total) Washington DC (40 total) • 10 sites randomly selected to use • 10 sites randomly selected to use FI coupons FI for achieving low VL • Data from HIV Surveillance for • Data from HIV Surveillance for linkage of newly tested HIV cases viral load of PHWH – Number of newly tested cases in – Number of PLWH in care the previous year – Among HIV-infected people in – Among newly tested cases, care, proportion with last viral proportion linked to care in 3 load < 400 copies/mL months – Cannot assess whether on antiretroviral therapy
- 6. HIV Surveillance used to assess site aggregate data • Baseline data – Site selection – Inform randomization (to achieve balance between arms) – Conduct power calculations for site-randomized trial • Follow-up data – Study outcomes – Monitoring of unintended effects (e.g. site migration)
- 7. HIV Surveillance Information Flow TLC-Plus People with HIV CDC Sources of Reports Local and/or State Health Department Hospital Practitioners 74,353 Private Practitioners Public Clinics Laboratories Active Case Finding HPTN Statistical Center Aggregate surveillance data Also receives: Aggregate testing data Aggregate behavioral data
- 8. To assess site aggregate outcomes HIV case identified as accessing care within jurisdiction People with HIV Testing/Care site identified in laboratory requisition CDC Sources of Reports Local and/or State Health Department Hospital Practitioners 74,353 Private Practitioners Public Clinics Laboratories Active Case Finding Depends on mandatory name Linkage of lab HPTN Statistical Center based reporting of result to case Aggregate surveillance data viral load and CD4 laboratory data Also receives: Aggregate testing data Aggregate behavioral data
- 9. Use of surveillance data for site aggregate measures • HIV Prevention Trials Network (HPTN) study conducted using HIV surveillance data to measure outcomes • Only aggregate site data are released from DoH and CDC as HPTN 065 data – Study conducted under a waiver of informed consent – Strict confidentiality laws for surveillance data
- 10. Issues in compiling aggregate data • Health systems with multiple sites – Not able to separate data unless lab requisitions can be identified by location or provider – Varied completeness of required information • Completeness and consistency of lab data reporting – Mandatory in New York City since 2005 • Mature QC systems between laboratories, state and city – Electronic reporting in Washington DC began 2008 • QC process under development • Data exchange with surrounding states under development • Not all laboratory data reported electronically
- 11. Results: Site selection • Site identification began in 2008, randomization occurred in 2010/11 – Linkage to care – newly diagnosed cases • Bronx NY: 2007 data for selection, 2008 for randomization • Washington DC: 2008 data for selection, 2009 for randomization – Viral load suppression – most recent viral load at site • Bronx: 2008 data for selection and randomization • Washington DC: 2008 data for selection, 2009 for randomization
- 12. HPTN 065: Test Site Selection Bronx NY Washington D.C. Total number of test sites Total number of test sites identified by DOH: 27 identified by DOH: 31 Number of sites approached: Number of sites approached: 27 28 2 Not seeing HIV-infected 3 Did not respond 3 Declined 3 Did not respond Number of sites that signed Number of sites that signed LoI: 19 LoI: 25 1 Combined 2 Combined 4 Test volume too low Number of sites selected for Number of sites selected for study participation: 18 study participation: 19 Randomized Randomized Randomized Randomized to FI: 9 to SOC: 9 to FI: 10 to SOC: 9
- 13. HPTN 065: Care Site Selection Bronx NY Washington D.C. Total number of care sites Total number of care sites identified by DOH: 36 identified by DOH: 32 Number of sites approached: Number of sites approached: 36 27 2 Not seeing HIV-infected 1 Declined 3 Declined 3 Did not respond 4 Did not respond 2 Other reason Number of sites that signed Number of sites that signed LoI: 25 LoI: 23 2 Combined 3 Combined 3 Low volume of patients 1 Declined Number of sites selected for Number of sites selected for study participation: 20 study participation: 19 Randomized Randomized Randomized Randomized to FI: 10 to SOC: 10 to FI: 10 to SOC: 9
- 14. Calculating power for a site randomized study Linkage to Care Viral suppression (Outcome: Newly tested linked (Outcome: VL <400 copies/mL) w/i 3 months) • Total number of sites • Total number of sites • Mean number of HIV • Mean number of cases in positive cases per site care per site • Baseline probability of • Baseline proportion of viral linkage to care suppression • Intracluster correlation • Intracluster correlation coefficient for linkage coefficient for low viral load Variability in Variability in VL < linkage probability 400 copies/mL across sites across sites
- 15. Study design: Linkage to Care Bronx Bronx Washington Washington (2007) (2008) DC (2008) DC (2009) Number newly diagnosed cases Median 13 13 24 20 (Q1, Q3) (9-41) (3-44) (13-60) (3-44) Mean 22 28 40 38 Proportion linked to care in 3 months Median 75% 69% 77% 54% (Q1, Q3) (49%-86%) (50%-86%) (57%-87%) (33%-71%) ICC* 0.27 0.42 0.31 0.64 *Intracluster correlation coefficient
- 16. Study design: Viral load < 400 copies/mL Bronx Bronx Washington Washington (2008) (2008) DC (2008) DC (2009) Number of cases assessed at care site Median 174 251 100 153 (Q1, Q3) (121-310) (130-806) (48-229) (50-348) Mean 692 625 245 311 Proportion with HIV viral load suppression Median 57% 57% 37% 64% (Q1, Q3) (39%-60%) (54%-61%) (27%-50%) (56%-72%) ICC* 0.07 0.04 0.11 0.18 *Intracluster correlation coefficient
- 17. Power of site randomized studies Linkage to Care Viral suppression • 40 sites (37 sites) • 40 sites (39 sites) • 54 linkage cases per • 180 cases in care per site (mean 33 per year) site (mean 481 per site) • ICC of 0.27 • ICC of 0.11 • 80% power to detect • 80% power to detect increase from 67% to increase from 60% to 80% linkage to care 66% VL <400 copies/mL
- 18. Randomization Strategy • Restricted randomization – Small number of sites – Protect against imbalance in factors predicting outcome – Volume of site; baseline outcome measure • Randomization index: – Sites divided into R1, R2 t statistic for t statistic for difference in difference in site volume baseline outcomes
- 19. Data for randomization Baseline linkage to care Viral load suppression 160 6000 140 5000 Number of new diagnoses 120 Number of patients 4000 100 80 3000 60 2000 40 1000 20 0 0 0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100% 120% Proportion linked to care in 3 mos. Proportion with VL < 400 cp/mL
- 20. Issues in conducting randomization • Test sites could not start until care sites had initiated study • Additional restriction added to ensure balance for highest volume sites • Randomization of sites after IRB approvals required different start times – added blocks – Washington DC • Test: two blocks (Feb and March 2011) • Care: three blocks (October 2010, Jan and March 2011) – Bronx • Test: one block (Feb 2011) • Care: one block (Jan 2011)
- 21. Summary • HIV surveillance data were aggregated in selected sites to inform study design and randomization for HPTN 065 (TLC-Plus) • Linkage to care at baseline – Levels of linkage to care were similar in Bronx, NY and Washington DC. – More cases were being identified in Washington DC. • Suppressed VL at baseline – Levels of viral suppression were modest and similar in Bronx and Washington DC – Includes patients not on ART. – More PLWH were in care in the Bronx
- 22. Implications • HIV surveillance data has the potential to provide information for assessment of site level outcomes – an opportunity for conducting rigorous implementation science • Additional resources provided at DoH to facilitate obtaining timely information – especially needed for site identification • Upload of complete lab data (CD4 and viral load) into eHARs facilitates uniform assessment of outcomes across jurisdictions
- 23. Acknowledgments • Special thanks to HIV surveillance staff in New York City, Washington DC. • HPTN 065 is sponsored by the NIAID and NIMH under Cooperative Agreement #UM1 AI068619 and #UM1 AI068617, by the CDC, National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention, via an interagency agreement. • The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.