Determination of Date of First HAART Initiation Among All Patients in VACS

1
Determination of Date of First HAART Initiation
Among All Patients in VACS

• Neel Gandhi, MD
• VACS 1st HAART Work Group
• Albert Einstein College of Medicine
• 14 October 2008

2
Background

• Advent of Highly-Active Antiretroviral Therapy
  (HAART) transformed HIV infection from
  universally fatal illness to chronic disease
• HAART most effective if initiated before the
  onset of AIDS (CD4 count lt200 cell/mm3)
• HIV-infected veterans roughly half have CD4lt200
  when initiating HIV care at VA
• Study patients entering VA HIV care from
  1998-2002 in VACS Virtual Cohort, a large
  administrative dataset

3
Date of HAART initiation

• Difficult to determine whether patients were
  diagnosed with HIV earlier and received HAART
  outside VA
• Determination of date of first HAART important
• for assessing delayed diagnosis or treatment
  initiation
• for determining impact of interventions and
  comorbidities

4
Use of Large Datasets

• Increasingly, large datasets of HIV patients are
  used to characterize impact of interventions and
  comorbidities
• Administrative or claims data
• Pharmacy databases
• Pooling of smaller cohorts
• No clear method available to establish patients
  who previously received HAART at entry into
  database
• Heterogeneous cohort
• Difficult to draw conclusions about area of
  interest

5
Objectives

• To determine the date of first HAART initiation
  for all patients in VACS cohort
• To determine if first HAART within VA or outside
• Will allow assessment of impact of our
  interventions
• To devise an algorithm for predicting HAART naïve
  status in a large database
• By comparing to pharmacy and laboratory data

6
Objectives

• To determine the date of first HAART initiation
  for all patients in VACS cohort
• To determine if first HAART within VA or outside
• Will allow assessment of impact of our
  interventions
• To devise an algorithm for predicting HAART naïve
  status in a large database
• By comparing to pharmacy and laboratory data
• To test accuracy of pharmacy database to identify
  first HAART regimen received within VA

7
Methods

• Retrospective chart review of all VACS patients
• Performed at each of 8 sites by coordinators and
  PIs
• Determine if patients ever received
  antiretroviral therapy (either HAART or
  pre-HAART)
• Date of first HAART prescription
• Site of first HAART received within VA or
  outside
• Regimen received

8
Definitions

• HAART defined as
• Concurrent use of 3 antiretroviral medications
• Irrespective of classes of individual medications
• Ritonavir used at low doses not counted
• Mono- or dual-therapy defined as pre-HAART

9
Comparison with Virtual Cohort

• VACS Virtual Cohort
• Large database of all HIV-infected veterans at
  VAs nationwide identified using ICD-9 codes
  (N33,420)
• Utilization, pharmacy and laboratory data
  available
• All VACS patients are included in Virtual Cohort
• Create and test algorithm for prediction of
  HAART-naïve status at entry into VA
• Use chart review as gold standard
• Devise algorithm based others used in literature

10
Proposed Algorithm

• Algorithm to identify HAART-naïve patients
• Patients should have a CD4 viral load before
  receiving a HAART prescription
• Viral load should be greater than 400 copies/ml
• Sensitivity analysis performed using differing
  thresholds of viral load up to 5000 copies/ml

11
Outcomes

• Percent of patients who are HAART-naïve at time
  of VA presentation
• Sensitivity and positive predictive value of
  proposed algorithm for identifying HAART naïve
  patients in Virtual Cohort

12
Preliminary Results
3227 HIV-infected VACS 8 pts
2918 Charts reviewed
13
Preliminary Results Pre-HAART data
3227 HIV-infected VACS 8 pts
2918 Charts reviewed
345 Unable to determine if received a pre-HAART
regimen
2573 Charts analyzed
1239 (48) recd pre-HAART
1334 (52) no pre-HAART
14
Preliminary HAART data
3227 HIV-infected VACS 8 pts
2918 Charts reviewed
421 Baltimore charts
2482 Charts analyzed
2335 (94) received HAART
147 (6) never received HAART
15
HAART-Naïve at Entry to VA
2482 Charts analyzed
2335 (94) received HAART
147 (6) never received HAART
16
HAART-Naïve at Entry to VA
2482 Charts analyzed
2335 (94) received HAART
147 (6) never received HAART
169 HAART initiation Site unknown
418 (19) initiated HAART outside VA
1748 (81) initiated HAART within VA
17
HAART-Naïve at Entry to VA
2482 Charts analyzed
2335 (94) received HAART
147 (6) never received HAART
169 HAART initiation Site unknown
82 HAART Naïve at VA entry
418 (19) initiated HAART outside VA
1748 (81) initiated HAART within VA
18
Preliminary Results
19
Preliminary Results
20
Preliminary Results
21
HAART Naïve VACS Site
22
Year of Presentation to VA
23
HAART Initiation Date
24
Components of Initial HAART Regimen

• Data collected during chart review
• Still being cleanedhas not been analyzed yet

25
Attempts at a HAART-naive Algorithm
26
Attempts at a HAART-naive Algorithm
27
Attempts at a HAART-naive Algorithm
28
Attempts at a HAART-naive Algorithm
29
Preliminary Findings

• Although nearly all patients received HAART, more
  than 80 were HAART-naïve at time of presentation
  to the VA
• HAART-naïve patients were more likely to be
  African-American or Latino and have lower CD4
  counts
• Majority of patients entered care in VA before
  1996
• Algorithm was able to reach nearly 90 positive
  predictive value

30
Next Steps

• Finish cleaning and analysis of these data
• Validation of Pharmacy database comparison of
  the first HAART regimen identified in Virtual
  Cohort with the actual regimen from chart review
• Establish a variable reflecting date of first
  HAART initiation for each VACS patient which can
  be used for all future analyses of VACS data