A secondary analysis of retention
across the PMTCT cascade:
Rwanda
Sam Kalibala, HIVCore/Population Council
Godfrey Woelk, HIVCore/Elizabeth Glaser
Pediatric AIDS Foundation
17 September 2015
IATT Webinar
HIVCore background
• Improve the efficiency, effectiveness, scale, and quality of
HIV treatment, care, and support, and PMTCT programs by:
– Conducting operations research and program evaluations
– Promoting use of research and program results to enhance decisionmaking
– Building local capacity to conduct operations research.
• 5-year project (Oct 2011–Sep 2016), 19 studies in portfolio
• Funded by USAID
• Led by Population Council in partnership with the Elizabeth
Glaser Pediatric AIDS Foundation, Palladium, and University
of Washington
HIVCore’s contribution in PMTCT
• Enhancing PMTCT program effectiveness
– Identifying gaps and successes in retention and adherence
of PMTCT mothers and HIV-exposed infants (HEI)
• Example: Identifying reasons for delays and loss to follow-up
among each step of the PMTCT cascade in Cote d’Ivoire
– Measuring MTCT rates
• Example: National PMTCT Evaluation in Zambia
• Testing new approaches for improving HEI testing
and treatment
– Example: Testing combined m-health and transport
reimbursement approaches in Tanzania
Introduction
• Global consensus that pediatric AIDS can be
eliminated.
• Goal of elimination (transmission rate <5%) by 2015.
• Retention in care is critical in achieving this goal.
• In 2012, 62% of HIV+ pregnant women received the
most efficacious regimes.
• Paucity of knowledge on retention.
• Monitoring challenges.
• 4 country case studies: Kenya (Option A), Malawi
(Option B+), Rwanda (Option B), Swaziland (Option A).
– Rwanda (Option B) findings presented here.
Research questions
• What are the levels of and factors related to retention
among women and infants in PMTCT care?
(retention = rates of program [clinic] attendance)
– What are the levels of program attendance 30 days after
entry into the PMTCT program at: delivery; 6 weeks; and 2,
3, 6, and 12 months post-delivery?
– How do the levels of program attendance vary by
demographic, clinical and facility characteristics?
– What program characteristics are associated with retention?
Methodology
• Retrospective chart (registers) review and patient
records: (ANC, PMTCT, Labor and delivery, Child welfare, post natal care
[PNC], early infant diagnosis, pharmacy and laboratory).
• Retrospective cohort constructed of HIV+ women
attended ANC from 2010–2011.
• In-charge interviews to obtain health facility data.
• Sampled from EGPAF-supported sites.
• Stratified by type and location (urban/rural) of facility.
• Logistics such as number of HIV+ pregnant women,
distance, quality of records considered.
Sampling and sample size (Rwanda)
• 5 sites (2 urban and 3 rural) selected—minimum of
40 HIV+ pregnant women per site/year.
• Sampled proportional to expected number of women
HIV positive at ANC.
• Expected 50% to attend 12 month visit; proposed
sample 474.
• Sampled all HIV+ women attending ANC (Option B).
• Interviewed approximately 5–10 in-charges, speaking
to at least one at each facility.
Derivation of endpoint
• Retention estimated: from the registration date to the
end of the study, this period was divided into 6 nonoverlapping time segments
Delivery
3 months
12 months
Registration
30 days after
registration
6 weeks
6 months
• For each time interval, if there is an indication that the
mother/child made a visit to health center or
clinic/hospital or pharmacy to pick up drugs, they got
1 for the visit and 0 otherwise.
Derivation of endpoint (cont.)
• Retention 30 days after registration: From the
registration date, investigated the 1st drug receipt
date. If date was at 30 days or within 2 weeks after
the 30 days = retained.
• Retention at delivery: Used all possible record
indications that delivery had occurred: place of
delivery, delivery date (where available), child date of
birth (if available). If any available, concluded mother
was retained at delivery.
Endpoint derivation (cont.)
• Retention after delivery: At 2-4 months, 5-7
months, and 11-13 months, any indication of ART
receipt.
• Retention of mother: count outcome indicating how
many visits were accomplished out of a maximum of
6 visits that they were expected to attend. If a mother
made > 1 visit during a given time period, only
indicated that she was retained (did not count
multiple visits during the same period).
Analysis
• Conducted bivariate and multivariate analyses using
Generalized Estimating Equations (GEE) to
determine factors associated with retention.
• For mothers’ retention: visits 1 and 2 have the total
number of those who registered as denominator.
The remaining visits have the number of those who
made it to visit 2 (those delivered) as the
denominator.
Findings
Proposed versus actual sample size
Type of site*
Site #1: Urban
Site #2: Urban
Site #3: Rural
Site #4: Rural
Site #5: Rural
Total
*All
Proposed (women)
N (%)
65 (13.7)
211 (44.5)
44 (9.3)
111 (23.4)
43 (9.1)
474 (100.0)
the sites are health centers
are more infants than mothers due to multiple births.
†There
Actual (women)
N (%)
60 (13.1)
207 (45.3)
43 (9.4)
109 (23.9)
38 (8.3)
457 (96.4)
Actual (infants)†
N (%)
61 (13.2
210 (45.5)
44 (9.5)
109 (23.6)
38 (8.2)
462 (97.5)
Description of HIV+ women attending ANC 20102012 (N=457)
Variable
Age: median, (Q1, Q3)
Parity: median, (Q1, Q3)
Marital status (N=455)
Married
Living as married
Single
Divorced/separated
Widowed
Education—highest level completed (N=421)
None
Primary
Secondary
Employment status (N=420)
Farmer
Housewife
Trader
Other
Referred to ART initiating site (N=206)
Yes
No
Number of ANC visits for this pregnancy (N=369)
1
2
3
4 or more
%
28.5 (23.5, 32.9)
2.0 (1.0, 4.0)
203
192
36
22
2
44.6
42.2
7.9
4.8
0.4
60
310
51
14.3
73.6
12.1
222
123
53
22
52.8
29.3
12.6
5.21
100
106
48.5
51.5
75
101
125
68
20.3
27.4
33.9
18.4
Description of the women (N=457) (con’t)
Variable
Known HIV positive at first ANC (N=457)
Yes
No
CD4 tested this pregnancy (N=456)
Yes
No
WHO staging at baseline (N=324)
Stage 1
Stage 2
Stage 3 or 4
Type of regimen (N=364)
TDF/3TC/EFV
TDF/3TC/NVP
AZT/3TC/EFV
Other
Toxicities/side effects (N=247)
No
N
%
272
185
59.5
40.5
339
117
74.3
25.6
291
20
13
89.8
6.2
4.0
196
84
55
28
53.8
23.0
15.1
7.7
243
98.4
Description of the infants (n=462)
Variable
Place of delivery (N=290)
Health center
Gender (N=289)
Male
Female
Mode of delivery (N=284)
Normal vaginal delivery
C-section
Other
Infant DNA PCR (N=292)
Yes
Co-trimoxazole (N=294)
Yes
Mode of feeding (N=285)
Exclusive breast feeding
HIV antibody testing at 18 months (N=254)
Yes
No
Not applicable
N
%
288
99.3
155
134
53.6
46.4
244
34
6
85.9
12.0
2.1
286
97.9
283
96.3
279
97.8
111
6
137
43.7
2.4
53.9
Description of the health facilities (N=5)
Variable*
No. of ANC clients
No. of HIV+ pregnant
women
No. of deliveries
Ratio of health staff:
ANC clients
Ratio of HIV trained nurses:
HIV+ pregnant women
No. of doctor visits† last quarter
CD4 test TAT‡
Follow up starts <week
*Data
Site 1
Urban
880
Site 2
Urban
1,124
Site 3
Urban
2,254
Site 4
Rural
1,885
Site 5
Rural
1,350
31
34
156
81
50
1,498.6
(562.4)
70.4 (51.8)
393
403
1,086
713
640
647.0 (283.4)
1:49
1:47
1:59
1:54
1:43
1:51
1:3
1:3
1:16
1:10
1:4
1:7
0
7
Yes
8
2
Yes
—
14
No
12
3
No
9
7
Yes
for 2011.
normally visit health facilities for consultation, supervision, and support purposes.
‡Turn-around-time (days): This is the amount of time taken from receipt of samples at the health facility to the
availability of the test results at the same health facility.
†Doctors
Mean (sd)
7.3 (5.1)
7
NA
Women’s and infant’s retention
Women’s retention by specified time intervals*
Time period
30 days
Delivery
6 wks
2–4 m
5–7 m
11–13 m
N expected
348
348
202
202
202
202
N observed
175
204
151
169
158
136
Prop.
0.51
0.59
0.75
0.84
0.78
0.67
95% CIs
0.46, 0.56
0.54, 0.64
0.69, 0.81
0.79, 0.89
0.73, 0.84
0.61, 0.74
*The
retention analysis for the women is from 4 instead of the 5 selected health facilities
as it was discovered that one of the facilities (Site #4) was not an ART site during the
records review period (2011). This facility referred HIV-positive women to a nearby
hospital for follow up. However, the facility retained the infants for follow-up.
Infants’ retention at specified time intervals
Time period
N expected
N observed
Prop.
6 wks
2–4 m
5–7 m
11–13 m
348
348
348
348
140
310
303
265
0.40
0.89
0.87
0.76
95% CIs
0.34, 0.46
0.85, 0.93
0.83, 0.91
0.71, 0.81
Retention, demographic, and clinical characteristics
Marital status by levels of retention
30 days
Delivery
6 wks
2–4 m
5–7 m
11–13 m
D/S
0.27
0.73
0.40
0.33
0.33
0.47
95% CIs
0.14, 0.54
0.56, 0.94
0.22, 0.74
0.16, 0.68
0.16, 0.68
0,27, 0.80
LM
0.52
0.57
0.70
0.82
0.78
0.64
95% CIs
0.47, 0.60
0.51, 0.65
0.62, 0.79
0.75, 0.90
0.70, 0.86
0.56, 0.74
M
0.29
0.72
0.44
0.48
0.46
0.39
95% CIs
0.23, 0.36
0.66, 0.79
0.36, 0.53
0.40, 0.57
0.38, 0.55
0.32, 0.48
S
0.42
0.44
0.53
0.60
0.05
0.40
95% CIs
0.28, 0.61
0.31, 0.64
0.33, 0.86
0.40, 0.91
0.27, 0.80
0.22, 0.74
D/S: Divorced/separated; LM: Living as married; M: Married; S: Single
Retention by employment status
30 days
Delivery
6 wks
2–4 m
5–7 m
11–13 m
Farmer
0.29
0.73
0.43
0.45
0.44
0.35
95% CIs
0.23, 0.36
0.68, 0.79
0.36, 0.51
0.38, 0.53
0.37, 0.52
0.29, 0.43
Housewife
0.49
0.47
0.69
0.86
0.76
0.73
95% CIs
0.41, 0.59
0.39, 0.57
0.58, 0.82
0.78, 0.96
0.66, 0.88
0.62, 0.85
Other
0.51
0.52
0.71
0.76
0.74
0.63
95% CIs
0.41, 0.63
0.42, 0.65
0.58, 0.87
0.64, 0.91
0.61, 0.89
0.50, 0.81
Referral to ART initiating site by retention levels
30 days
Delivery
6 wks
2–4 m
5–7 m
11–13 m
Yes
0.05
0.80
0.05
0.07
0.09
0.07
95% CIs
0.02, 0.12
0.72, 0.88
0.02, 0.12
0.03, 0.16
0.04, 0.18
0.03, 0.16
No
0.51
0.71
0.81
0.90
0.84
0.65
95% CIs
0.43, 0.62
0.626, 0.80
0.731, 0.91
0.827, 0.97
0.762, 0.93
0.555, 0.77
Factors associated with retention among women
Variable
Odds ratio
95% CIs for odds ratio
p-value
Facility location (rural)
29.37
2.079, 4.682
<0.001
ANC clients 2011
0.996
0.995, 0.998
<0.001
Deliveries 2011
1.001
1.0004, 1.0025
0.007
Doctor visits
0.999
0.9998, 0.9999
<0.001
• Rural location positively associated with retention with an odds
ratio of 29.37
• Deliveries in 2011 (per each delivery) also positively associated
with retention
• ANC clients (per client) and doctor visits (per visit) in 2011
inversely related to retention.
Summary
• Women: 67% retained at 12 months postpartum.
• Infants: 76% at 12 months of age.
• Characteristics associated with retention:
– Facility location (rural) and number of deliveries (positively
associated).
– ANC clients and doctor visits (inversely associated).
Discussion
• 67% retention similar to 70% reported by MOH.
• Retention improvement after delivery; women may
deliver elsewhere; implications for tracking and
tracing.
• Better retention in rural areas; greater mobility in
urban.
• Inverse association with ANC clients and doctor
visits; association with larger facilities where women
come to deliver from other areas, more doctor visits
where fewer doctors.
Conclusion
• Patient files helped in assessing retention.
– Yet much missing data.
• Facility characteristics important.
– Inverse association of ANC clients, but positive association
with number of deliveries, suggests association with larger
and more comprehensive facilities, where women come to
deliver from other areas.
– Retention associated with fewer doctor visits. Sites may
have more doctor visits where there is insufficient staff and
doctors, and these sites may tend to be smaller.
Factors associated with retention overall
•
•
•
•
Rural location
Lower workload
Faith-based facilities
Provision of comprehensive services may be also
associated positively with retention
• Active follow-up
• Availability of doctors
Acknowledgements
•
•
•
•
•
•
•
•
Martha Mukaminega
Dieudonne Ndatimana
Epiphanie Nyirabahizi
Heather Hoffman
Placidie Mugwaneza
Muhayimpundu Ribakare
Ministry of Health Rwanda
Elizabeth Glaser Pediatric AIDS Foundation Rwanda
Thank you!
Full report is
available
at hivcore.org
Analysis(2)
• Computed pairwise correlation between variables.
For any 2 pairwise correlation of above 0.8, one
variable was dropped based on what was judged as
clinically important.
• Variable selection was based on fitting a sequence of
models beginning with a simple model with only an
intercept term, and then includes 1 additional
explanatory variable in each successive model.
• The importance of the additional explanatory variable
was assessed by the difference in deviances
between successive models.