HIV Surveillance and data availability
description
Transcript of HIV Surveillance and data availability
HIV Surveillance and data availability
MTT Winter School, Durban August 2004
Dr Anthony Kinghorn
Controversy and HIV/AIDS
Antenatal Survey
reported that 24%
of pregnant
women are HIV
positive
The HSRC study
reported that
11.4% of people
are HIV positive.
Who is right?
Prevalence of HIV Infection in the Under 20 Year Age Group of Antenatal Clinic Attendees in SA
Year
Prevalence
(%)
95% Confidence
Interval
1996 12.78 11.33-14.23 1997 12.7 11.3-14.2 1998 21.0 18.4-23.8 1999 16.5 14.9-18.1 2000 16.1 14.5-17.7 2001 15.4 13.8-16.9 2002 14.8 13.4-16.1
Source: DOH. National HIV Surveys of Women Attending Public ANC Clinics in SA
HIV INFECTION LEVELS - 15 - 19 year olds
0.00
5.00
10.00
15.00
20.00
25.00
1994 1995 1996 1997 1998 1999 2000 2001 2002
%
Source: National Surveys of Women Attending Antenatal Clinics
Outline of Presentation
– Why measure?
– What can we measure?
• HIV Prevalence
• HIV Incidence
• AIDS Prevalence and incidence
• Mortality
– Using models to understand the epidemic
and it’s impacts
– Second Generation surveillance
Why measure?
– Identify trends in infections and impact
– Identify levels of infection and impact
– Predict future trends and levels of impact
Advocacy and planning
Evaluate interventions for staff and
learners
The Prevalence (Rate)
Definition:Definition:The proportion of a population at risk The proportion of a population at risk
affected by a disease at a specific point in affected by a disease at a specific point in timetime
Prevalence = Prevalence =
No. of people with the disease orNo. of people with the disease orcondition at a specific time condition at a specific time
No. of people at risk in theNo. of people at risk in the population at the specified timepopulation at the specified time
Population at risk for Cancer of the Cervix
All Men All Women
0-25 years
26-69 years
70+ years
Factors influencing the prevalence:
Increased by:•Longer duration of
disease
•Prolonging life but no
cure
•Increase in incidence
•In-migration of cases
•Out-migration of
healthy
•Improved diagnosis/
reporting
Decreased by:•Shorter duration of
disease
•High death rate
•Decrease in incidence
•Out-migration of cases
•In-migration of healthy
•Increased cure rate
The Incidence Rate
This is the rate at which new events This is the rate at which new events occur in a occur in a populationpopulation
= = No. of new cases of a disease in No. of new cases of a disease in aa
specified timespecified time Total number of people at riskTotal number of people at risk
HIV Prevalence
– The main source of HIV Prevalence data is National Surveys of Pregnant Women at Antenatal Clinics
– Other sources include:• Hospital admissions• TB patients• STD clinic attendees• Blood donors• Pre-insurance testing • Workplace and population surveys
– What are the limitations of these sources?
– What are they useful for?
Antenatal HIV Seroprevalence Survey
Source: DOH. National HIV Surveys of Women Attending Public ANC Clinics in SA
0%
5%
10%
15%
20%
25%
30%
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
Limitations of the Antenatal Data
• Usually designed to track trends not national levels • Rising ANC prevalence usually reflects rise in general
population
• May overestimate HIV female and male adult prevalence • Reflects sexually active women, reproductive years, not
using condoms• Over estimates prevalence in teens and high age groups
• But may also underestimate HIV• Excludes women on contraceptives• HIV positive women have a decreased fertility
• Some studies suggest that ANC HIV prevalence is a
reasonable proxy for community adult rate
• Other sampling biases • Rural populations often under-sampled? Other?
So we need to use models to estimate levels of HIV infection in the population and sub-populations
Trends in HIV infection levels in pregnant women
0%
5%
10%
15%
20%
25%
30%Urban
Rural
Adjusted allurban
All pregwomen 15-49
Source: Rwanda HIV Sentinel Sero Surveys and adjustments from population surveys
Limitations of the Antenatal Data
• Increasing difficulties of interpreting ANC data in mature epidemics– Deaths off-setting new infections– Prolonged life due to ARVs– Plateaux due to saturation or behaviour
change?– Etc
Community Prevalence Studies
Community studies more representative of all settings, ages, both sexes
Can link with behavioural surveillance/KAPB/ other data
Big differences from ANC prevalence in the young and old – due to sample bias
Can refine assumptions about community infections used in interpreting ANC data
Results can be surprising or easy to misinterpret
eg. HSRC/NMF study in South Africa HIV prevalence of 11.4% in all > 2 years old 32% prevalence in women aged 25-29
HIV prevalence in Zambia DHS vs Antenatal
0
5
10
15
20
25
30
urban ♀ rural ♀ transitional rural♀
total*
HIV
pre
vale
nce
% (
ages
15-
49)
ANC DHS
* DHS Total = men and women)
ANC vs ZDHS (cont)
• ZDHS: 15.6% prevalence all adults
• ANC 2002: 19% prevalence adult ♀ (15-44yrs)
• ZDHS: 18% prevalence adult ♀ (15-49yrs)
– Similar estimates indicate epidemic still
severe
– Overall, ANC estimates fairly robust?
KDHS versus ANC (2003)
• Adult prevalence – DHS 2003 (women & men): 6.7%
– ANC 2003: 9.4%
→previous over-estimation?
• However, for women 15-49:– ANC 2003: prevalence estimated 9.4%
– DHS 2003: prevalence estimated 8.7%
Age profile of HIV infection levels – Men vs Women
(Zambia DHS 2001)
0
5
10
15
20
25
30
35
15-19 20-24 25-29 30-34 35-39 40-44 45-49
Age group
Prev
alen
ce (%
)
Women
Men
Source: Zambia DHS 2001, Preliminary Report
Age profile of HIV infection levels – Men vs Women(Kenya DHS 2003)
0
2
4
6
8
10
12
14
15-19 20-24 25-29 30-34 35-39 40-44 45-49
Age group
Prev
alen
ce (%
)
Women
Men
Source: Kenya DHS 2003, Preliminary Report
Community Prevalence Studies
Limitations Sample sizes
Especially for sub-groups Biases
Non-Response Other
Expense and complexity Time to establish new time series and trend data Frequency
Probably only repeat every 3-5 years if initial results in line with ANC and expectations
Biological surveillance - workplace sero-prevalence
surveys• Blood or saliva tests for HIV; (STD rates)
• Unlinked anonymous surveys
– VCT usually inadequate for workforce levels
• Advantages
– Accurate refection of risk, including for
employee sub-categories
– Plausible
– Inform projections (still required)
– Track changes and monitor success
Biological surveillance cont.
Challenges• Clear objectives and use of data, including
strategy to communicate results• Limited accuracy if low participation• Employee buy-in
– Credible confidentiality, non-discrimination, programme and response options needed
• Ethics– VCT availability and promotion– Informed consent, anonymity – Ethics committee approval
• Technical and analytical issues – eg. sampling; response rates; stats analysis; lab
• Cost• Limited trend data
HIV prevalence in a service sector workforce (South Africa)
0%
5%
10%
15%
20%
25%
30%
Projected HIV + 2002 National Antenatal Survey Employee HIV salivasurvey
Which data source gave most information and value for money?
HIV Incidence
– Very few sources of data on HIV Incidence
– Usually from large HIV prevention studies
– Main measure of vaccine effectiveness
AIDS Prevalence / Incidence
• Very difficult to measure without
notification
• Only tells us about HIV infections from 5-
10 years ago
• Critical to use a consistent and
recognized classification system!
Other Sources of Data
• Death Registration
– Can be a very useful way to track AIDS trends, as
age related mortality from AIDS is unique
• EMIS; HR and payroll databases; pension funds
• Measuring incidence of opportunistic diseases,
especially TB, is very important for health service
planning
DEATHS by age band 1998 DHS vs Projected(Botswana)
(For previous 24 months)
0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64
1998 DHS
d1
Behavioural surveillance - KAPB
• Standardised questionnaires generate indicators
of Knowledge, Attitudes, Practices, Behaviour eg. – Basic knowledge – Risk e.g. number of non-regular partners; condom use– Views on HIV/AIDS programme
• Can link to blood or saliva tests
• Objectives– Identify target knowledge gaps, behaviour, groups– Identify sources of e.g. information, services– Assess manager and supervisor preparedness– Track levels and trends: baseline and follow-up – Advocacy
KAPB cont.
• Challenges– Usually outsourced for expertise and neutrality
• Employee and union buy-in
• Sample size or biases, incl. % responding;
truthfulness
• Survey administration skills
• Ethics
– Cost
– Interpreting, using and communicating results
– Simple or complex questionnaires/ surveys?
– Interfering programmes and influences on KAPB?
– May miss unexpected issues and suggestions
PERCENTAGE OF CHILDREN IN AGE GROUPS WHO WILL BE ORPHANED BY AIDS
0%
5%
10%
15%
20%
25%
30%
35%
0-4'5-9'10-1415-19
Source: Kinghorn et al (2001). The impact of HIV/AIDS on Education in Namibia
What is a Model?
– A model is a hypothesis or theory that tries to explain the real
world
• It gives a framework for design of tools to give answers to questions
about the 'model world'
– A model is only as good as:
• Its underlying assumptions
• Quality of input data
Some use of modeling is probably inescapable to make sense of any
empirical data
Models - Examples
– ASSA 2000/ Doyle/ Metropolitan
• Mix of macro- and micro-model features
• Includes risk groups and geographic differences
– AIM
– US Bureau of Census
– Epimodel
– Other
Projection methodology
Antenatal data – levels and trends in infection
General population projections: age, gender, geographic region
Cross mapping of e.g. educators by age, gender, location, origin
Scenarios; validation/calibration using prevalence, mortality data
Analysis and action
Extrapolation to all women and men
Modifiers
•Mortality data
•(HIV prevalence data)
•(Risk behaviour data)
ASSA 2000 Output
0
5000
10000
15000
20000
25000
30000
35000
40000
45000
5000019
8519
8719
8919
9119
9319
9519
9719
9920
0120
0320
0520
0720
09
To
tal p
op
ula
tio
n
0
1000
2000
3000
4000
5000
6000
7000
Nu
mb
ers
HIV
, AID
S s
ick
an
d H
IV d
eat
hs Total
population
Total HIV
TotalnumberAIDS sick
CumulativeHIV deaths
*Source: Prof R Dorrington, ASSA
Projections - challenges
1. Limitations of all models
2. Demographic data limitations• Population and personnel
• Migration
• Fertility
3. Epidemiological data limitations, particularly• Extrapolation from ANC to general population• Survival time • Fertility impacts - multiple determinants• Epidemic curves for urban/rural, local areas, sub-
groups
Projections - challenges(2)
4. Other enrollment or attrition influences • Policy, other factors – often dominate AIDS
5. Key techniques• Validation – quality of data?• Sensitivity testing • Intervention modeling - Behaviour change; ARVS• Qualitative data
6. Experienced modelers7. Shorter term and more aggregated projections
probably more accurate
Severity of limitations depends on the planning question to be answered
Second Generation Surveillance
• Continue with ante-natal surveys
• Behavioural Surveillance
• Focus on young people
• High-risk sub-groups
• Morbidity and mortality surveillance
BUT – every country is different – needs it’s own
research agenda
HIV prevalence in a large company workforce (South Africa)
Category % HIV+ (95%
CI)
Sexual behaviours
Sex with non-regular partner (last 3
months)
16.8 (14.5 – 19)
No non-regular partner (last 3 months) 6.9 (6.1 – 7.7)
Condom use
Used condom with last non regular partner 14.4 (12 – 16.8)
No condom with last non-regular partner 11.8 (10 – 13.5)
No non-regular partners 4.5 (3.7 – 5.3)
Source: Colvin M Gouws E Kleinschmidt I Dlamini M. The prevalence of HIV in a South African working population. AIDS 2000 Conference poster, Durban 2000
Summary
• Data maybe limited, and the models may be inaccurate, but the main messages in terms of levels and trends are usually clear
• But the epidemic is complex and needs customised responses
• “What is occurring is a collection of epidemics in different stages of increase, stability, and decline” (Sentinel Surveillance of HIV/Syphilis in Zambia, 2003)
• Averages hide variation – much worse or less affected communities
• Don’t contribute to confusion through lack of understanding of HIV/AIDS statistics OR enthusiasm for technical debate