34
Concordance of HIV surveillance and medical record data: What do CD4 and viral loads not tell us about linkage to HIV care? Charu Sabharwal, MD MPH Medical Director Epidemiology and Field Services Program Bureau of HIV/AIDS Prevention and Control NYC Department of Health

Concordance of HIV surveillance and medical record data: What do CD4 and viral loads not tell us about linkage to HIV care? Charu Sabharwal, MD MPH Medical

Embed Size (px)

Citation preview

Concordance of HIV surveillance and medical record data: What do CD4 and viral loads not tell us about

linkage to HIV care?

Charu Sabharwal, MD MPHMedical DirectorEpidemiology and Field Services ProgramBureau of HIV/AIDS Prevention and ControlNYC Department of Health

Acknowledgements• Sarah Braunstein• Rebekkah Robbins• Colin Shepard• HIV Epidemiology and Field Services Program

Background

• NHAS (July, 2010) - first comprehensive roadmap

• A more coordinated response to the HIV epidemic

• Primary Goals for 2015:– Reduce infections– Increase access to care– Reduce health disparities

National HIV/AIDS Strategy

HIV Continuum of Care

Das, Moupali Prevention of HIV Acquisition: Behavioral, Biomedical, and Other Interventions. Medscape 2012

• HIV Care = outpatient HIV visit with provider authorized to prescribe ART1

• Clinical monitoring/treatment guidelines2 – Traditionally, 1st CD4/VL at initial HIV care visit– CD4/VL: every 3-6 months; frequency after ART initiation

• CD4/VLs proxy for HIV care [HIV care visits not reported]

• Since 2004, CSTE encouraged all states (59 jurisdictions) to report all CD4 and VLs3 [New York2005]

• Limited comprehensive evaluation of the validity of surveillance data as proxy of HIV care

Monitoring HIV Care – CD4/VL

1Health Resources and Services Administration. The HIV/AIDS Program: HAB Performance Measures Group 1. In; 2009. 2DHHS Guidelines for the Use of Antiretroviral Agents in HIV-1-Infected Adults and Adolescents. In; 2012. 3 CSTE Position statement 04-ID-07

• Surveillance traditionally measures linkage by a single event: 1st reported CD4/VL on/after HIV diagnosis date

• Accuracy of 1st CD4/VL1,2 drawn prior to referral to HIV care. For example, at the time of– Confirmatory testing after + rapid/point-of-care test – Inpatient diagnosis: CD4 impacts treatment decision

• In New York City: routine medical record (MR) abstraction for linkage to care is not feasible– 3,500 diagnosing providers; 3,000+ HIV cases yearly– Timely linkage – entry into care within 3 months of

diagnosis. Local3 and national measure

Measuring linkage to care

1 Bertolli A. et al The Open AIDS Journal 2012,6:131-141. 2Keller et al. J Acquir Immune Defic Syndr 2013. 3New York City HIV/AIDS Surveillance Slide Sets. http://www.nyc.gov/html/doh/html/data/epi-surveillance.shtml

New York City’s Care Validation Study

• Validate CD4 and VL tests for persons living with HIV (PLWH) in NYC as proxy measure for HIV care in the first year after diagnosis

1° Objective – evaluate the correspondence between a patients 1st CD4/VL on/after HIV diagnosis and linkage HIV care

PurposeValidate 1st lab test (CD4/VL) from the diagnosing facility as measure of timely linkage to HIV care • Hypothesis: early post-diagnostic lab

tests within first 2 weeks are part of diagnostic work-up and not an actual linkage event

Methods

Methods

• Selected high-volume HIV diagnosing sites with co-located care (n=24) –Patients with new, confirmed HIV diagnosis

in 2009 reported the Registry–Patients who had to linked to care at the

same diagnosing facility within 12 months as per the Registry • PLEASE NOTE – Even though Surveillance does not

require linkage to care at the same site of diagnosis, we did in order to conduct this validation study

Study population selection: New York City HIV Registry

3,536 new, confirmed HIV diagnoses among > 13 years in NYC in 2009

1,263 (36%) patients reported from high-volume (> 20 diagnoses) co-located HIV care sites

947 (75%) patients had 1st CD4/VL reported from co-located site within 12 months of diagnosis

eligible for medical record (MR) abstractions

165 (17%) excluded: MR unavailable

Figure 1: Final study population

782 (83%) patients Registry (1st CD/VL) and MR (care visit) data

Data Analysis

Analytic population (n=782)Linkage within 12 months, per Registry

No medical visit group Medical visit group

HIV care visit confirmed by MR

Compared the subgroups based on:• Key demographic characteristics (age, gender, risk)• Proportion concurrently diagnosed with HIV/AIDS

(AIDS within 31 days of HIV diagnosis – local definition)• Proportion diagnosed on inpatient service• Proportion that died within 12 months of diagnosis

YESNO

Timely linkage to HIV care

• Compared the proportion who linked to HIV care within 3 months of diagnosis (timely) by Registry (1st CD/VL) vs. MR (care visit)

Do 1st reported CD4/VLs indicate timely linkage to HIV care?

• Compared subgroups:– Median time to 1st lab per the Registry– Proportion of 1st labs in 0-7 days and 0-14 days

• Calculated sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) of Registry data in correctly classifying patients’ true timely linkage to care status based on the 1st CD/VL within:– 0-91 days (no labs excluded: National standard)– 8-91 days (excluded labs from 0-7 days)– 15-91 days (excluded labs from 0-14 days)

RESULTS

Figure 2: Linkage to care (n=782)Registry vs. MR

Registry MR0%

20%

40%

60%

80%

100%

% L

inke

d to

car

e w

ithin

12

mon

ths

of d

iagn

osis

Medical visit80% (n=625)

1st CD4/VL100% (n=782)

No Medical visit20% (n=157)

No medical visit

N=157

Medical visit

N=625 P value

Age at HIV diagnosis (median, range) 42 (16-80) 37 (15-78) 0.001

Male Gender (%) 69.4 74.2 0.220

Race/ethnicity (%)

Black 54.1 48.0 0.370

Hispanic 36.9 37.4

White 7.6 11.0

Transmission risk (%)

Men who have sex with men 18.5 42.9 <0.001

Injection drug use 8.9 4.2

Heterosexual 33.8 29.3

No identified risk 38.9 23.7

Concurrent AIDS diagnosis 61.8 37.0 <0.001

Table 1: Demographics/clinical outcomes

Figure 3: Inpatient diagnoses

Diagnosed in acute setting0%

20%

40%

60%

80%

100%

73%

32%

% S

tudy

pop

ulati

on

p<0.001

No medical visit Medical visit

Figure 4: Mortality outcomes:Deaths within 12 months of HIV diagnosis

No medical visit Medical visit0

5

10

15

20

18%

1 %

% D

ied

with

in 1

2 m

onth

s of

HIV

di-

agno

sis

p<0.001

Timely Linkage to Care

Figure 5: Timely linkage to careRegistry vs. MR

Registry MR0%

20%

40%

60%

80%

100%

% L

inke

d to

car

e w

ithin

12

mon

ths

of d

iagn

osis

97%

1st CD4/VL(proxy measure):

0-91 days

75%

True linkage event(HIV care visit):

0-91 days

Timely linkageAre labs within the early

post-diagnostic period indicative of timely linkage to care?

Figure 6: Median time (days) to linkage based on 1st CD4/VL, by subgroups

MR0%

20%

40%

60%

80%

100%

% L

inke

d to

car

e w

ithin

12

mon

ths

of d

iagn

osis

8 days (IQR 0-20days)

1 day (IQR 0-5 days)No medical visit

Medical visit

p <0.001

Figure 7: Proportion of 1st labs in the early post-diagnostic period, by subgroups

0 - 7 days 0 - 14 days0%

20%

40%

60%

80%

100%

80% 85%

49%

66%

% P

atien

ts w

ith fi

rst l

ab o

ccur

ing

in

time

peri

od

No medical

visit

No medical

visit Medical

visit Medical

visit

p <0.001p <0.001

31%

19%

Sensitivity Specificity PPV NPV0%

20%

40%

60%

80%

100%

84%

24%

76%

33%

73%

23%

73%

22%

11%

77%

96%99%

0-91 days 8-91 days 15-91 days

Figure 8: Performance of Registry data

Refinement of NYC’s timely linkage to care indicator

Figure 9: Final study population: Refining timely linkage to care

Study population0

20

40

60

80

100 97%

81%75%

% L

inke

d tim

ely

(with

in 3

m

onth

s) to

car

e Gold standard: care visit

Lag applied

No lag applied

Figure 10: New York City’s refined Timely linkage to care indicator

All new 2009 diagnoses All new 2010 diagnoses0

20

40

60

80

100

75% 73%66% 66%

% L

inke

d tim

ely

to c

are

No lagLag

No lagLag

Conclusions• First population-based study to validate the use of HIV

Surveillance’s proxy measure of timely linkage to care• Substantial misclassification of timely linkage in the

early post-diagnostic period • NYC DOHMH implemented a refined definition of

timely linkage to care (labs 8-91 days after diagnosis)– HIV labs in 1st 7 days not indicative of linkage

• Surveillance data overestimated linkage for older persons, non-traditional HIV risk transmission, and those who died soon after diagnosis

Limitations• Selection of provider– A portion had a CD4/VL at an alternate provider which may

be the linkage to care visit –DID NOT validate if these patients EVER linked

– Oversampled the acute care setting

• Selection of study population – Due to the complexities of HIV laboratory reporting, the

1st lab may have been misclassified to the incorrect provider

Future directions• Exploration of surveillance-based retention in

care measures vs. medical abstraction data– All care visits at diagnosing provider during first 12

months immediately following diagnosis

• In depth exploration of mortality within 12 months of HIV diagnosis

[email protected]

Thank you!