Distance elasticity of demand for HIV treatment: sex

DISTANCE ELASTICITY OF DEMAND
FOR HIV TREATMENT: SEX
DISPARITIES IN UTILISATION
Calvin Chiu, Mead Over, Gesine Meyer-Rath,
Frank Tanser, Deenan Pillay, Till Bärnighausen,
Jacob Bor
Background
•
The geographic distribution of HIV treatment services
may be an important determinant of access to care.
•
Efficiency of clinic placement: where should new facilities be
located?
•
Interventions to overcome distance: transport vouchers, home
visits, reduction in number of required visits.
•
We assessed the association between distance to the
nearest HIV clinic and linking to care, initiating ART, and
retention in rural KwaZulu-Natal, South Africa. Analysis
was stratified by sex.
•
ART patients spend 1/3 of median HH income on care-seeking
(Chimbindi et al 2015), primarily due to transport costs.
Distance and the cascade of care
Not yet in
care
Diagnosed
and linked
to care
Initiated
ART
Retained
on ART
Distance and the cascade of care
• Does distance modify
progression through the
cascade of care?
Not yet in
care
Diagnosed
and linked
to care
• Where in the cascade do
access disparities emerge?
Initiated
ART
DISTANCE TO
CLINIC
Retained
on ART
A very simple (rational) model
At each stage, individuals deciding whether to
go to the clinic compare the perceived benefits
and costs of care-seeking.
The time and $$ costs increase in distance.
The perceived benefits are proportional
to a person’s belief that she is HIV+
and in the efficacy of treatment.
Not yet in
care
Diagnosed
and linked
to care
This information becomes
available with care-seeking.
Initiated
ART
DISTANCE TO
CLINIC
Retained
on ART
Data
• 438 km2 surveillance area
• Population >100K, nearly
100% isiZulu-speaking
• Semi-annual household
surveys, 99% response rate
• Linked to public sector ART
• Rural, poor, 29% HIV+
• High ART coverage; by 2011,
13% of all adults had sought
care for HIV; 7% started ART
Tanser et al. 2008, Houlihan et al. 2010
Distance to nearest clinic
Distance to nearest ART clinic was calculated for all
individuals residing in the catchment area of the Africa
Centre’s demographic surveillance, on 31 Dec 2006.
Distribution of individuals' distance to their nearest clinic
1000
0
500
Frequency
1500
2000
•
0
2
4
6
Distance to nearest clinic
8
10
Characteristics of the sample
Variable
Full
Linked to Initiated Retained
Population care
ART
in care
N
29,547
3,383
2,189
1,718
Person-years at risk
126,984
4,865
1,013
-
35
34
36
36
Sex is Male, N(%)
41%
26%
28%
26%
Distance to nearest
clinic, median
2.6
2.3
2.3
2.3
Years of education,
median
11
12
11
11
Age, median
Outcomes and analytic approach
•
Cox Proportional Hazards models to assess associations
between distance to clinic and…
•
linkage to care, ART initiation, and 12 month retention on ART,
measured as time from 31 Dec 2006 to dates of these events
•
Linkage = first CD4 count; 12 month retention on ART = presence of
any lab result 6-18 months after ART initiation.
•
h(t|X) = h0(t)exp(Xβ), X = age, age-squared, HIV prevalence*, socioeconomic characteristics. Models stratified by sex.
•
Regression models
•
Unconditional: same denominator; potentially causal interpretation
•
Conditional models: limited to persons reaching prior stage;
common view of the cascade; may be biased by changing
composition at each stage
Percent of the total adult population that
has been linked, initiated, and retained
14%
12%
10%
8%
Linked
ART
Retained
6%
4%
2%
0%
0-2km
2-5km
5-12km
Distance from the nearest clinic
Distance and the cascade of care
1. Unconditional
cascade
Not yet in
care
Are there
disparities in
access by
distance?
Diagnosed
and linked
to care
Conditional cascade
Where do the disparities
in access emerge?
Initiated
ART
Retained
on ART
Distance is associated with lower careseeking among women
Unconditional cascade
FEMALE
Not yet in
care
Hazard ratios,
* p<.05
** p<.01
0.93***
Diagnosed
and linked
to care
0.94**
Initiated
ART
0.94**
Retained
on ART
Conditional cascade
…but not conditional on linkage
Unconditional cascade
FEMALE
Not yet in
care
0.93***
Hazard ratios,
* p<.05
** p<.01
Diagnosed
and linked
to care
0.94**
0.93***
Initiated
ART
0.95**
0.99
Retained
on ART
Conditional cascade
1.01
No association between distance and
care-seeking among men
MALE
Unconditional cascade
Not yet in
care
1.01
Hazard ratios,
* p<.05
** p<.01
Diagnosed
and linked
to care
1.01
1.01
Initiated
ART
1.04
0.99
Retained
on ART
Conditional cascade
1.01
Results summary
1. Among women, living one additional kilometre away
from the nearest clinic was associated with:
• 6.6% (95%CI 4.9%, 8.3%) reduction in the rate of linkage to care
• 6.5% (95%CI 4.2%, 8.7%) reduction in the rate of ART initiation
• 5.4% (95%CI 3.4%, 7.4%) reduction in rate of retention on ART
2. Conditional on linkage to care, no association between
distance and ART initiation nor retention.
3. No association between distance and care-seeking for
men. Sex*distance interaction significant at all stages.
Conclusions
1. Distance matters, but only for women.
• Further work needed to understand differential patterns by sex.
• E.g., constraints on mobility of women?
Conclusions
1. Distance matters, but only for women.
2. Disparities emerge at testing / linkage, but do not increase
thereafter.
• Causal story: distance is no longer a barrier once patients have
overcome the initial hurdle of seeking care. Information on HIV status
and treatment availability dramatically increase the perceived benefits
of care-seeking, swamping any costs imposed by distance.
• Selection story: among patients who link, those residing further from
the clinic have higher motivation. Distance may still be a barrier to
ART initiation, but this effect could be cancelled out by selection.
Conclusions
1. Distance matters, but only for women.
2. Disparities emerge at linkage; do not increase thereafter.
3. Efforts needed to address barriers to HIV testing and
linkage for women living far from clinics.
4. But distance-specific interventions are unlikely to increase
care-seeking among men. Other approaches are needed
to reach 90-90-90 targets and reduce the prevalence of
unsuppressed HIV in the population.
Acknowledgments
• Colleagues at the Africa Centre for Population Health
• Study participants
• National Department of Health
• Funders
• NIH (NIMH)
• PEPFAR
• USAID
• Wellcome Trust