Washington D.C., USA, 22-27 July 2012 Sources of data: estimates of the size of key population groups – mortality data Peter Ghys, UNAIDS.

Washington D.C., USA, 22-27 July 2012www.aids2012.org

Sources of data: estimates of the size of key population groups –

mortality data

Peter Ghys, UNAIDS

Txema Calleja, WHO

Paloma Cuchi, UNITAID

John Stover, Futures Institute


Second Generation Surveillance framework

STISurveillance Second

GenerationSurveillance

HIV andAIDS case &

mortalityreporting

Behavioural or Bio-Behavioural

SurveysSentinel

Surveillance

SizeEstimation

of Risk Groups1


Size matters

Contribution of a subpopulation to the HIV epidemic is determined

by HIV prevalence + risk behaviors + size of subpopulation:– Small population + high HIV incidence + efficient bridge/interactions =

important role to the epidemic– Big population + low prevalence = main contributor HIV epidemic

Use of the SE data:– National estimates: policy, response planning, resource allocation,

advocacy, Understanding HIV surveillance– Local estimates: program planning and management (assessing

commodity, coverage, HIV program evaluation)


Size estimation issues

Few countries with good size estimation of sub-pops at risk • No regular, scientific SE studies & trends• Subpopulation "hidden" and poorly characterized• Not triangulated & validated w/multiple sources• Ad hoc assumptions often made in projection• Point estimates instead of time-varying trends (size change over

time)• Pressure to use “official” estimates & politics

• Low accuracy, large uncertainty of SE estimates and HIV Estimates


Methods

Methods based on data collected from at-risk population:• Census/Enumeration, Capture-

recapture, Multiplier

Methods based on data collected from general population:• Population survey • Network scale-up

Limitations:• Stigmatized populations need to

disclose behaviors (e.g. illegal)

• Geographically limited (1 city, 1 neighborhood) = not nationally representative

• Collect data on 1 population at a time = multiple studies for a full picture


Network Scale Up

• Ideal but not feasible: ask respondents directly about their behaviors (national survey)

• Challenges: stigma, embarrassment, fear

• Ask respondents about acquaintances: national survey, behaviors of others

• Individual’s behaviors are not disclosed

• Each respondent’s personal network contributes to sample

8 countries: Moldova, Ukraine, Kazakhstan, Japan, China, Brazil, Rwanda

Conclusions• On the radar (stigmatized

situations)• Feasible in diverse circumstances

& survey methods • Not for every occasions, needs to

be used appropriately and have data available

• Pending issues


2007 en

Data sources for the size of populations

• Often multiple data sources are available:– Sizes of at-risk populations

• Studies from any of the methods (i.e. Capture-recapture)• Mappings of higher risk sites• Estimates from NGOs (service statistics)• Police arrest records• Security office estimates


Regardless of the method

• National ownership

• Build consensus and agreed on a single estimate

• Use the results

• No harm

• Determine use of SE

• Know what you know

• Use multiple methods to get a better estimate

• Deal with conflicting results

• Repeat study every 2-3 years


Second Generation Surveillance framework

STISurveillance Second

GenerationSurveillance

HIV andAIDS case &

mortalityreporting

Behavioural or Bio-Behavioural

Surveys

SentinelSurveillance

SizeEstimation

of Risk Groups

2


1. Incidence of HIV Infection

3. Mortality from AIDS2. Prevalence of HIV Infection

Underreporting, delays and misclassification to other causes of death in death registration systems

HIV epidemiology


South Africa - overall mortality levels

0

10000

20000

30000

40000

50000

60000

70000

Age0

Age1-4

Age5-9

Age10-14

Age15-19

Age20-24

Age25-29

Age30-34

Age35-39

Age40-44

Age45-49

Age50-54

Age55-59

Age60-64

Age65-69

Age70-74

Age75-79

Age80-84

Age85+

no

of

dea

ths 1993

1995

1996

1998

2000

2005

name South Africa Diseases All Causes

Data

Year

Analyses of the overall mortality can gauge the level of HIV mortality

Analyses for miscoding of AIDS deaths in vital registration data for S Africa, R Fed, Belarus, Ukraine and Thailand (Source and slides “HIV deaths in vital registration data” from Doris Ma Fat, Mortality and Burden of Disease Unit, Department of Health Statistics and Informatics, Dec 2010)


South Africa 2004: Further analyses of trends and patterns are necessary to identify potentially misclassified HIV deaths

ALRI - M

0500

10001500

2000250030003500

A0M

A51

0M

A15

19M

A25

29M

A35

39M

A45

49M

A55

59M

A65

69M

A75

79M

A85

00M

199319962004

Meningitis - M

0

100

200

300

400

500

600

700

A0

M

A5

10

M

A1

51

9M

A2

52

9M

A3

53

9M

A4

54

9M

A5

55

9M

A6

56

9M

A7

57

9M

A8

50

0M

1993

1996

2004

other infectious - F

0

500

1000

1500

2000

2500

A0M

A51

0M

A15

19M

A25

29M

A35

39M

A45

49M

A55

59M

A65

69M

A75

79M

A85

00M

1993

1996

2004

ENDOCRINE - F

0

500

1000

1500

2000

2500

A0M

A51

0M

A15

19M

A25

29M

A35

39M

A45

49M

A55

59M

A65

69M

A75

79M

A85

00M

1993

1996

2004

Diarhoea - F

0500

1000150020002500300035004000

A0M

A51

0M

A15

19M

A25

29M

A35

39M

A45

49M

A55

59M

A65

69M

A75

79M

A85

00M

1993

1996

2004

Ill defined - M

0500

10001500

2000250030003500

A0M

A10

14M

A25

29M

A40

44M

A55

59M

A70

74M

A85

00M

199319962004

Acute lower resp. inf -male Other infectious dis - maleMeningitis - male

Diarrhoea - female Endocrine disorders - female Ill-defined injuries - male


AIDS-related mortality

• HIV has a significant impact on mortality• Measuring HIV mortality to evaluate the impact of NAP’s

efforts • One of the clearest indicators of success is a decrease

in HIV mortality • Two of the 10 MDG require mortality data• Provide evidence of equity in distribution of health

services

In many cases this information is not available


Data sources HIV-related mortality

•Civil registration systems - gold standard

• Verbal autopsy – most common

– Nationally representative sample vital registration with

verbal autopsy (SAVVY)

• Facility-based mortality surveillance (e.g., HIV treatment

and care facilities, hospitals, prisons, drug treatment

facilities, morgues)


Data sources HIV-related mortality

• Burial systems with verbal autopsy (cadaver autopsy,)

• Surveys & research

– Population-based surveys with verbal autopsy (VA)

(e.g., DHS and post-census mortality surveys) -

retrospective

– Prospective Demographic Surveillance Systems

(DSS) with verbal autopsy (ALPHA Analyzing

Longitudinal Population-based HIV/AIDS data on Africa)

linkages between DSS participants and HIV prevention,

treatment and care services


5 data considerations

1. Data identification: multiple data sources need to

identify all. Organization, creativity, ongoing

2. Data quality and completeness: evaluate all potential

sources (strengths, weaknesses). Underestimation,

quality (cause, date, sex, age…. )

3. Data management: different sources & ways to

collect data, duplications, use

4. Data Analysis: limitations in the analysis of mortality

data


5 data considerations

5. Future data issues: strengthening current systems, data collection, sharing systems and collaboration

• Short-term goals: obtaining measures of HIV mortality

• Longer-term goals: identify opportunities and advocacy strategies for health systems strengthening and creating strong civil registration systems


GUIDELINES

Guidelines available on UNAIDS and WHO website

WWW.UNAIDS.ORG WWW.WHO.INT

Washington D.C., USA, 22-27 July 2012 Sources of data: estimates of the size of key population groups – mortality data Peter Ghys, UNAIDS.

Documents

hiv estimates washington

se data

national estimates

use of se

general population

representative collect

picture washington

multiple studies