a new method for estimating national and regional art need
DESCRIPTION
A new method for estimating national and regional ART need. Basia Zaba, Raphael Isingo, Alison Wringe, Milly Marston, and Mark Urassa. TAZAMA / NACP seminar Dar-es-Salaam, September 19 th 2008. Outline. Why do we need a new method for estinmating ART need? Explanation of the new method - PowerPoint PPT PresentationTRANSCRIPT
A new method for estimating national and regional ART need
Basia Zaba, Raphael Isingo, Alison Wringe, Milly Marston, and
Mark Urassa
TAZAMA / NACP seminar
Dar-es-Salaam, September 19th 2008
Outline
• Why do we need a new method for estinmating ART need?
• Explanation of the new method• Results for Kisesa• Producing national estimates
Forecasting national ART need using “EPP” and “Spectrum” packages
EPP (no age & sex structure)• ANC surveillance data entered to get prevalence trend• DHS prevalence data added to correct overall level
Spectrum (age, sex and incidence modelling)• Population age structure added to EPP results• Prevalence age and sex patterns modelled by year• Incidence pattern estimated from prevalence changes• Prevalence projected using survival models• Output estimates of new infections, AIDS deaths, new
orphans, treatment need
EPP / Spectrum inputs
National• ANC all age prevalence• DHS all age prevalence• Census population age
and sex distribution
Regional• ANC all age prevalence?• DHS all age prevalence• Census population age
distribution
Model (international)• prevalence age and sex
distribution• deriving incidence from
prevalence• survival post infection
Process• “Black box” with user
friendly inputs and nice graph outputs
TAZAMA method inputs
National• DHS prevalence of HIV
by age and sex• Census population age
and sex distribution
Regional• DHS prevalence of HIV
by age and sex• Census population age
and sex distribution
Model (Mwanza)• age pattern of incidence• age-specific mortality
rates of HIV infected population
Process• Spreadsheet – not yet
user friendly, but has nice graphs …
Assumptions behind new method
• People need to start ART 3 years before they would have died if they didn’t have treatment
• Age-specific patterns of mortality for HIV infected persons not on treatment are the same for both sexes all over the country
• The Kisesa cohort age patterns of incidence can be scaled up or down to represent incidence in different parts of the country
Why we think these assumptions are reasonable
Joint studies in the ALPHA HIV cohort study network showed that:
– from CD4 count of 350 (new UNAIDS treatment start recommendation) people survive for a median period of 3 years without treatment before they die
– age-specific mortality patterns of people infected with HIV in the pre-treatment era did not vary much from one place to another
– HIV incidence has very different levels from one country to another, but the age- and sex- specific patterns are very similar from one place to the next
Example calculation
500 HIV infected people aged 42 in 2004 (before ART available)
HIV+year age death alive expected expected number total
rate at age deaths survivors to start need
2004 42 0.067 500 34 466 99 992005 43 0.070 466 33 433 30 1292006 44 0.073 433 32 401 29 1582007 45 0.076 401 30 371 28 1862008 46 0.079 371 29 342 27 2132009 47 0.082 342 28 314 25 2382010 48 0.085 314 27 287 24 2622011 49 0.088 287 25 2622012 50 0.091 262 24 238
x = - =
sum of previous column
How it all adds up
• The calculation is repeated for infected people at every single year of age in the start year
• We generate expected new infections at each age by multiplying the uninfected by the incidence rate
• We work out expected HIV deaths in those not yet infected in the start year, and their ART need
Cumulated number needing ART by calendar year
0
200
400
600
800
1000
1200
1400
1600
1800
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
incident
prevalent
Annual decline in incidence is 0%
Proportion who are infected by treatment need
0.00
0.02
0.04
0.06
0.08
0.10
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
does not need ART
should be on ART
incidence decline 0 % per year
Impact of incidence decline on treatment need
Cumulated number needing ART by calendar year
0
200
400
600
800
1000
1200
1400
1600
1800
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
incident
prevalent
Annual decline in incidence is 0%
Cumulated number needing ART by calendar year
0
200
400
600
800
1000
1200
1400
1600
1800
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
incident
prevalent
Annual decline in incidence is 20%
Declining incidence only affects treatment need in those not yet infected in the baseline year
Proportion who are infected by treatment need
0.00
0.02
0.04
0.06
0.08
0.10
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
does not need ART
should be on ART
incidence decline 0 % per year
Proportion who are infected by treatment need
0.00
0.02
0.04
0.06
0.08
0.10
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
does not need ART
should be on ART
incidence decline 20 % per year
Impact of incidence decline on prevalence
Declining incidence means that in future there will be smaller proportion of infected people in the “not yet needing treatment” category
Cumulated MEN needing ART by calendar year
0
100
200
300
400
500
600
700
800
900
1000
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
infected since 2004
infected before 2004
incidence decline 0 % per yearincidence decline 0 % per year
Cumulated WOMEN needing ART by calendar year
0
100
200
300
400
500
600
700
800
900
1000
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
infected since 2004
infected before 2004
incidence decline 0 % per yearincidence decline 0 % per year
Sex differences in number needing treatment
Overall more women than men will need treatment as there are more infected women than men, because of their earlier average age at infection
Proportion MEN infected by treatment need
0.00
0.02
0.04
0.06
0.08
0.10
0.12
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
does not need ART
should be on ART
incidence decline 0 % per year
Proportion WOMEN infected by treatment need
0.00
0.02
0.04
0.06
0.08
0.10
0.12
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
does not need ART
should be on ART
incidence decline 0 % per year
Sex differences in proportion of infected needing treatment
However, women’s earlier age at infection will mean that there will be proportionately more of them in the “not yet needing treatment” category
age in whole pop2004 age distrib infected uninfected
0 929 0 9291 914 0 9142 900 0 9003 885 0 8854 870 0 8705 856 0 8566 841 0 8417 826 0 8268 811 0 8119 796 0 796
10 781 0 78111 766 0 76612 751 0 75113 735 0 73514 720 0 72015 705 0 70516 689 7 68217 674 13 66018 658 19 63919 643 24 61820 627 28 59821 611 32 57922 595 36 56023 579 38 541
estimated numbers
National estimates from 2004 DHS (thousands)
To get national or regional estimates
The only new input needed is the smoothed single year age distribution of the population by HIV infection status. Mortality and incidence age patterns can be taken from the Kisesa cohort, with a suitable scaling factor for incidence
Kisesa situation 1st jan 2004
age in whole pop2004 age distrib infected uninfected
0 1172 0 11721 1138 0 11382 1105 0 11053 1073 0 10734 1041 0 10415 1010 0 10106 980 0 9807 950 0 9508 921 0 9219 892 0 892
10 864 0 86411 837 0 83712 810 7 80313 783 13 77014 758 18 73915 732 23 70916 708 27 68017 683 31 65318 660 34 62619 637 36 60020 614 38 57621 592 40 55222 571 41 53023 550 42 508
estimated numbers
How different is HIV prevalence in Kisesa?
HIV prevalence, both sexes, 2004, Kisesa and National compared
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
15 25 35 45 55 65 75 85
age
pro
po
rtio
n in
fec
ted
Kisesa Kisesa model National National model
HIV Mortality in Kisesa, 1994-2004
0.00
0.05
0.10
0.15
0.20
0.25
15 25 35 45 55 65 75 85
age
ag
e s
pe
cif
ic m
ort
alit
y r
ate single year
five year average
fitted Weibull
HIV incidence patterns in Kisesa, by age and sex
0.000
0.002
0.004
0.006
0.008
0.010
0.012
0.014
0.016
15 25 35 45 55 65
Age
HIV
inc
ide
nc
e r
ate
s
Male hazard rates Female hazard rates
Male incidence model Female incidence model
Cumulated number needing ART by calendar yearNational estimates, thousands
-
200
400
600
800
1,000
1,200
1,400
1,600
1,800
2,000
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
incident
prevalent
Conclusions
The ART need estimation method developed and tested on Kisesa cohort data is easy to adapt for national and regional estimates
It allows us to model various assumptions: – future incidence trends – current prevalence patterns – mortality of those not yet on treatment – years prior to death that treatment should start
It still needs to incorporate:– mortality of those already on treatment– number already receiving treatment in the base year