1. Are we there yet? Spatial-temporal trend of mother to child HIV transmission in western Kenya,
Anthony Waruru, Thomas Achia, Hellen Muttai, Lucy Ng’ang’a, Abraham Katana, Peter Young, Jim Tobias, Peter Juma, Thorkild Tylleskär
26th July 2017
Ministry of Health

2. No conflicts of interest to declare.
Conflict of Interest
No conflicts of interest to declare.

3. Background (1)
Elimination of mother to child transmission of HIV (e-MTCT) can be achieved through PMTCT efforts to reduce HIV transmission.
PMTCT services are often prioritized by district-level service planning.
Early infant diagnostic testing (EID) is a strategy of identifying HIV status for infants and assess PMTCT impact.

4. Background (2) Measuring e-MTCT progress can be done through:
Measuring proportion of infected infants [<5%]
Measuring case rates / 100,000 births [<50]
Putting these measures in a spatial context is important.

5. Aims Assess district level trends and factors associated with MTCT.
Model MTCT rates over time and space.

6. methods

7. Study setting

8. Data, laboratory procedures, analyses and mapping
Dried blood spot (DBS) samples collected accompanied by a submission form and sent to a regional laboratory.
HIV testing performed using PCR.
Analysis and mapping done using:
Stata v.14.
R – Integrated Nested Laplace Approximation (INLA)
Quantum GIS (QGIS) to map fitted MTCT rates

9. All infant and children samples
Analysis data set
Over 1 year old
n=5,186 (5.1%)*
Missing age
n=1715 (1.7%)
Included in analyses
n=95,215 (93.2%)
All records
N=102,116
Infants ≤12 months old
Exclusions
All infant and children samples

10. RESULTS

11. Raw MTCT rates and early infant diagnosis
Testing by 8 weeks/2 months considered “early”

12. Factors associated with MTCT
Characteristic
Total (n)
Positive, n (%)
Adjusted
aOR
[95% CI]
Total
95,215
10,095
Age at diagnosis
Under/= 8 weeks
52,504
3,307 (6.3%)
ref.
Over 8 weeks
42,711
6,788 (15.9%)
1.17
(1.08,1.26)
Maternal regimen
SdNVP only
2,763
279 (10.1%)
2.51
(2.32,2.72)
AZT+NVP+3TC | short course
11,634
871 (7.5%)
1.51
(1.33,1.72)
ART for prophylaxis
4,551
328 (7.2%)
1.71
(1.49,1.97)
ART for treatment
22,389
1,171 (5.2%)
Covariates: year of diagnosis, sex, infant’s age, age at diagnosis, maternal regimen, breastfeeding, and mother ARV status

13. Models comparison
Model type
DIC
Effective parameters
Model choice
Model 1:- A generalized linear model (non-spatial)
1,153
4.0
Fourth
Model 2:- Spatial model without covariates
1,319
11.8
Fifth
Model 3:- Spatial-temporal model without covariates
306
59.7
Second
Model 4:- Spatial non-temporal model with covariates
325
62.3
Third
Model 5:- Spatial-temporal model with covariates
305
58.8
First*
The best fitting model was spatial-temporal model with covariates (age at diagnosis, breastfeeding, sdNVP use, infant’s age)

14. Spatial-temporal MTCT trend

15. Case rates /100,000 births District Estimated live births in 2013*
Women tested for HIV in 2013†
HIV+ women in 2013
Infants tested in 2013
Absolute transmission (number infected)
Transmission rates per 100,000 live births‡
Rank (low to high)
All
275,169
203,069
15,136
17,129
1,231
447 -
Bondo
13,262
9,925
1,372
1,739
116
875 11
Kisii
36,841
25,143
622
701 46
125 3
Gucha
17,231
17,316
375
293 20 2
Homa Bay
43,423
13,159
1,257
1,968
163 5
Kisumu
24,931
29,599
2,882
2,469
167
670 9
Kuria
11,696
13,774
214
473 35
299 4
Migori
30,193
26,391
2,503
2,582
190
629 7
Nyamira
26,640
15,827
354
445 22 83 1
Nyando
19,063
10,307
1,208
1,286
102
535 6
Rachuonyo
17,243
12,658
1,451
1,457
121
702 10
Siaya
24,984
21,589
1,998
2,276
652 8
Suba
9,662
7,381
900
1,440 86
890 12
* Kenya population estimates
† PEPFAR annual progress report (APR 2013) data
‡ Transmission rate per 100,000 live births = Absolute transmission in 2013 Estimated live births in x 100,000

16. Summary of findings Early testing rate has improved over time.
Significant drop in mother to child transmission of HIV in 7-year period.
Case rate per 100,000 live births is still high.
Spatial-temporal model with covariates was best in explaining MTCT geographical variation.

17. CONCLUSION

18. Does spatial-temporal modeling help us tell the story?
Limitations
Routine data from programs are often incomplete.
Did not take into account the underlying population.
Strengths
May be better than other models.
Offers a visual tool to help program planners focus efforts.

19. Are we there yet?
Improvement in uptake of infant testing and reduction of MTCT rates ~ growth of the PMTCT program.
Overall, the PMTCT program coverage has contributed to reduction in MTCT rates in western Kenya.
Geographical disparities may signify gaps in distribution of e-MTCT efforts.
More spatial and spatial temporal analyses should be considered as additional tools for planning.

20. Thank you Acknowledgements Ministry of Health
Kenya medical research institute (KEMRI) – field work & laboratory
U.S. Centers for Disease Control and Prevention (CDC)/PEPFAR - funding
University of Washington/University of Nairobi – GIS training
Attribution of Support: This evaluation was supported by the President’s Emergency Plan for AIDS Relief (PEPFAR) through U.S. Centers for Disease Control and Prevention, Division of Global HIV/TB (CoAg # GH000041).
Disclaimer: The findings and conclusions in this presentation are those of the authors and do not necessarily represent the official position of the funding agencies.