Implementation of antiretroviral therapy improved between 2001 and 2010 in the US

Bora Youn, MS, MPH; Yoojin Lee, MS, MPH; Theresa Shireman, PhD, RPh; Omar Galárraga, PhD; Aadia Rana, MD; Ira Wilson, MD

11th International Conference on HIV Treatment and Prevention AdherenceMay 9-11, 2016

Study background

Complete implementation of anti-retroviral therapy (ART) is a critical component of the Care Continuum

Improved ART adherence in some individual HIV clinics and academic-based cohorts

Limited nationally representative data on time trends or sociodemographic predictors

Generalizable estimates can help identify areas for targeted interventions

Objectives

(1) To examine the changes in ART implementation in a representative U.S population with Medicaid between 2001 and 2010

(2) To determine the factors associated with ART implementation in a real-world setting

Persistence vs. Implementation

PERSISTENCE: duration of use without exceeding the permissible gap

IMPLEMENTATION: % of doses taken as prescribed during the corresponding period of persistence

StartMedication

PermissibleGap

Ends Medication

Non-persistent periods

Data

Medicaid Analytic Extract (MAX) file, 2001-2010

Medicaid is the single largest source of care for HIV patients

14 states with the highest HIV prevalence (75% of US cases)

Individuals with HIV based on ICD-9 diagnosis codes and ART fill records

n=397,836

Individuals with complete ART regimen episoden=227,531

Individuals fully observable in the Medicaid FFS system n=44,456

Study Inclusion Criteria

Individuals who initiated ART (no ART fill records six month prior to initiation & continuously eligible)

n=91,741

Outcome Measurement

Month-level implementation rate:

number of days of medication supplied for each calendar month / number of days in each month during the corresponding period of persistence.

n=808,682 months of observations from 44,456 patients (unit of analysis)

Dichotomized monthly implementation rate for adjusted model: ≥90% vs. <90%

Study Variables

Main Independent Variable

Calendar year of treatment(2001-2003, 2004-2006, 2007-2010)

Covariates

Age group Gender Race/Ethnicity State Initial ART regimen type

(integrase inhibitor based, NRTI based, NNRTI based, PI based, multiple classes)

Initial NRTI backbone (TDF/ABC, AZT, ddl/d4T, others)

Single tablet regimen use

Statistical Analysis

Population-level monthly implementation rates obtained by averaging the rates across all persistent patients

Generalized estimating equations (GEE) model with autoregressive correlation used to evaluate the factors associated with complete implementation, adjusting for covariates

Sensitivity Analysis: 80% and 95% cutoff, exchangeable correlation, and logistic regression model.

Descriptive Statistics

0

100000

200000

300000

400000

2001-2003 2004-2006 2007-2010

Calendar Year of Treatment

n=167,094

n=281,361

n=360,227

0

10

20

30

40

50

<25 25-34 35-44 45-54 55+

Age group

2001-2003 2004-2006 2007-2010

0

10

20

30

40

50

60

70

Male Female

Gender

2001-2003 2004-2006 2007-2010

0

10

20

30

40

50

60

Black White Hispanic Asian/PI/NA Multi/unknown

Race/Ethnicity

2001-2003 2004-2006 2007-2010

%

% %

All p<.0001

Descriptive Statistics

0

10

20

30

40

50

CA FL GA IL LA MA MD NC NJ NY OH PA TX VA

State

2001-2003 2004-2006 2007-2010

0

20

40

60

IntegraseInhibitor Based

NNRTI based PI based NRTI based Multiple classes

Regimen type

2001-2003 2004-2006 2007-2010

0

20

40

60

80

TDF/ABC AZT ddl/d4T others

NRTI backbone

2001-2003 2004-2006 2007-2010

0

20

40

60

80

100

Yes

Single Tablet Regimen Use

2001-2003 2004-2006 2007-2010

All p<.0001

Monthly Implementation Rate by Calendar Year (unadjusted)

Distribution of monthly implementation rate

0

0.2

0.4

0.6

0.8

1

2001200220032004200520062007200820092010

>80% implementation cutoff

>90% implementation cutoff

>95% implementation cutoff

Proportion of months with complete implementation

0

0.2

0.4

0.6

0.8

1

2001200220032004200520062007200820092010

Monthly Implementation Rate by State and Race/Ethnicity

Distribution of monthly implementation rate by State

0.50

0.60

0.70

0.80

0.90

1.00

2001-2003 2004-2006 2007-2010

Black White

Hispanic Asian/PI/NA

Multiracial/Unknown

Mean monthly implementation rate by race/ethnicity and calendar year

0

0.2

0.4

0.6

0.8

1

CA NY IL NJ PA MA FL NC OH VA GA LA MD TX

Generalized Estimating Equations Model

Binary outcome: >90% implementation (yes/no)

OR>1: more likely to completely implement ART

The following factors were associated with higher odds of achieving complete implementation: older age, male, non-black, new ART regimen, recent calendar year, and living in NY.

aOR 95% CI p-value

Calendar year (ref=2001-2003)

2004-2006 1.29 (1.25, 1.32) <.0001

2007-2010 1.60 (1.54, 1.66) <.0001

Generalized Estimating Equations Model

0.0

0.2

0.4

0.6

0.8

1.0

1.2

CA IL NJ OH PA FL NC MA LA VA GA MD TX

Adjusted OR of Complete Implementation by States (ref=NY)

All p<.0001, except CA

Generalized Estimating Equations Model

aOR 95% CI p-value

Gender (ref=female) Male 1.04 (1.01, 1.08) 0.01

Race/Ethnicity (ref=Black)

Asian/PI/NA 1.41 (1.19, 1.66) <.0001

Hispanic 1.22 (1.17, 1.28) <.0001

Multiracial/Unknown 1.22 (1.14, 1.30) <.0001

White 1.16 (1.11, 1.21) <.0001

Regimen Type(ref=PI based)

Integrase Inhibitor Based

1.17 (1.03, 1.32) 0.02

NNRTI based 1.20 (1.16, 1.24) <.0001

NRTI based 1.03 (0.98, 1.07) 0.28

Multiple Classes 0.98 (0.93, 1.04) 0.56

NRTI backbone (ref=TDF/ABC)

AZT 0.87 (0.84, 0.90) <.0001

ddl/d4T 0.84 (0.81, 0.88) <.0001

others 0.95 (0.90, 1.01) 0.10

Single tablet regimen use (ref=no)

Yes 0.98 (0.93, 1.03) 0.39

Conclusions

Marked improvement in ART implementation between 2001 and 2010

Adherence support programs as a potential explanatory factor

Disparities for blacks remain

State differences are concerning, may relate to Medicaid generosity, and merit further study

Limitations

No viral loads or CD4 counts

Not generalizable to the uninsured, commercially insured, and Medicare population

Not all states were included

Implications

National, population-based data that can be generalized to HIV patients in the U.S with Medicaid

Can help identify areas for targeted interventions

Differences between the results of persistence and implementation analysis

Acknowledgements

Team Members: Yoojin Lee, Theresa Shireman, Omar Galárraga, Aadia Rana, and Ira Wilson

NIMH 1R01MH102202

Providence/Boston Center for AIDS Research (Providence/Boston CFAR NIH/NIAID grant P30AI042853)

Thank you