Evaluating the Cost-Effectiveness of Behavior Change

advertisement

Paul Hutchinson, Josh Yukich, Jennifer Wheeler

Tulane University

Background

Since the onset of the HIV/AIDS epidemic, Behaviour Change

Communication interventions using mass media have been standard components of many countries’ HIV/AIDS prevention and treatment programs.

In spite of their ubiquity, such programs often lack rigorous evaluations because of:

Challenges in implementing evaluation designs with strong internal and external validity

Program exposure that is frequently beyond the control of researchers

Mass media which may cover entire countries leaving all but a nonrandom sample of the population unexposed to messages.

Even rarer are rigorous evaluations linking costs to programmatic effects, i.e., economic evaluations

Limited literature on the CE of BCC

Cohen et al (2004): HIV/AIDS mass media, $2 per person, $18,000 per case prevented per year

Hutton et al (2003): mass media and peer group education of highrisk men and young people, $500 per infection prevented

Kumaranayake et al (2004): $359 per infection averted (needle distribution)

Kincaid and Parker (2008): HIV/ AIDS communication campaign -> reduction of > 700,000 HIV cases averted; lifetime savings in treatment costs US$5.6 billion or annual costs of US$260,000,000

Muirhead, Kumaranayake and Watts (2001), who constructed knowledge, intention, and action scales to assess the costeffectiveness of the 4th Soul City Series but did not make a link to health outcomes.

West (2011): $1,625 / infection averted (Lesotho); $591 / case averted

(South Africa)

Pakachere & the OneLove Campaign

Pakachere Institute for Health & Development Communication ran the OneLove campaign starting in 2008/9 focusing principally on multiple concurrent partnerships (MCP)

 Used print media, radio and television education entertainment and other media messaging

Campaign Objectives:

To increase uptake of HIV testing and treatment

To reduce risk behaviors associated with HIV transmission

To reduce HIV-related stigma in the youth and adult populations in

Malawi.

Involved a number of partners, including: faith-based community, the private and public sector, non-governmental organisations

(NGOs)

Part of the Regional Programme of the Southern Africa Social and

Behavior Change Communication, which spanned 8 countries

Source: http://www.pakachere.org/group-item/program/one-love/

Economic Evaluation Objective

To assess the cost-effectiveness – as measured by the cost per Disability Adjusted Life Years Averted - of the OneLove

HIV/AIDS mass media Behavior Change Communication

Programme of the Pakachere Institute for Health and

Development Communication.

Methods: Cost Analysis

Objective:

 To collect program-level financial data to estimate the total resources used in developing and delivering the specific mass media interventions of Pakachere over the period 2008-11.

Details:

 Used a narrow programmatic perspective

 Omitted private household costs / benefits, costs of treatment averted, additional productivity

Overhead costs were apportioned among different interventions (e.g., print media, radio programs and television programs) in proportion to the share of each intervention in total direct costs.

All costs were adjusted for local inflation and calculated in end year

(2011) values.

Converted expenditures from South African Rand to English pounds and US dollars using historical exchange rates gathered from OANDA

Cost Analysis: Included Costs

Program Items

Print Formative audience research, stakeholder workshops; Development costs (writing, translation, editing, testing); Printing costs (design and reproduction, transportation); Distribution

Radio Research and Audience Reception Evaluation; Development costs

(writing, translation, editing); Production costs (recording, sound effects, Production Manager); Post-production (editing, mixing); Air time

Television Piloting, Development and Production; TV Air time; TV Audience reception; dubbing

Overhead Marketing; Evaluation; Operational Costs (office costs; personnel);

Travel and accommodation; Technical Support; Advocacy

Total Costs & Cost Distribution

Intervention Amount % of Total

Print

Television

Radio

Total

$ 2,068,555

$ 247,203

$ 655,040

$ 2,970,798

69.6%

8.3%

22.0%

100.0%

Methods - Effectiveness

Objective

 To ascertain the effects of Pakachere interventions on condom use, HIV testing, and MCP, while controlling for potential confounding influences

Details

 Post-only evaluation comparing outcomes for self-reported exposed and unexposed individuals

 Used data from a nationally representative household survey of 5,149 adults aged 15-49 years.

 Survey focused on HIV/AIDS-related behaviors and exposure to HIV/AIDS BCC interventions (both Pakachere and other BCC)

Used multivariate regression analysis to attribute observed differences in behaviors to exposure to Pakachere interventions.

Probit models and propensity score matching

Test for non-random (endogenous) exposure

Used marginal effects from multivariate analyses as measures of changes in behavioral outcomes

Used Markov modeling to translate behavior change into Disability Adjusted

Life Years averted

 Only statistically significant parameters included in the Markov model; otherwise, behavioral parameters set to zero (or equal to the no-intervention scenario)

Pakachere Interventions (Exposures)

Measures

OneLove radio: Binary outcome indicating exposure to any OneLove radio shows or

Meet Joe.

One Love booklets: Binary outcome indicating exposure to any of the following:

“Sexual partners and your lives,” “Love and Satisfaction,” “You Haven’t Met Joe,”

Okondedwa athu ndi Moyo wathu, Kukondana ndi Kukwaniritsana

OneLove Television: Binary outcome indicating exposure to any of the following: Love

Stories in the Time of HIV and AIDS, the OneLove talk show, or Untold Stories.

OneLove multimedia: Categorical variable representing intensity of exposure to

OneLove activities. The three levels correspond to the number of media channels through which the respondent was exposed to OneLove interventions, including exposure to OneLove booklets, One Love radio programs, Meet Joe adverts, Love

Stories in the time of HIV, Untold Stories, OneLove talk show. Those coded as low were exposed to one of the following: radio, TV, or booklets. Individuals coded high were exposed to any two media channels.

Markov Model for Estimating the

Probability of HIV-1 Infection

R i

F

E

N

M j

𝑃

P

𝐻𝐼𝑉−1 𝑖𝑗

= Probability of becoming infected with HIV-1 during a given year

= average HIV-1 prevalence among the population of sex partners of the target population

= risk of HIV-1 transmission per act of unprotected sex (infectivity)

= proportion of sex acts in which a condom is used

= efficacy of condoms at prevention of HIV-1 transmission

= number of sex acts per partner

= average number of sexual partners per person per year

= age group

= risk category of the individual

Source: Sweat, M., S. Gregorich, G. Sangiwa, C. Furlonge, D. Balmer, C. Kamenga, O. Gristead,

T. Coates, “Cost-effectiveness of voluntary HIV-1 counseling and testing in reducing sexual transmission

Of HIV-1 in Kenya and Tanzania,” The Lancet, Vol 356, July 8, 2000.

Model Parameters: Probability of

Infection Per Year

Parameter Estimate Source

P (HIV adult 15-49 prevalence)

R (infectivity per sex act)

E (condom efficacy)

M (number of partners per year) (Regular Risk 1)

M (number of partners per year) (Regular Risk 2)

M (number of partners per year) (Regular Risk 3)

M (number of partners per year) (Casual Risk 1)

M (number of partners per year) (Casual Risk 2)

M (number of partners per year) (Casual Risk 3)

M (number of partners per year) (Commercial Risk 1)

M (number of partners per year) (Commercial Risk 2)

M (number of partners per year) (Commercial Risk 3)

N (number of sex acts per partner per year) (Regular)

N (number of sex acts per partner per year) (Casual)

N (number of sex acts per partner per year) (CSW)

Percent of Population (Age Group 1)

Percent of Population (Age Group 2)

Percent of Population (Age Group 3)

Percent of Population (Risk Group 1)

Percent of Population (Risk Group 2)

Percent of Population (Risk Group 3)

0.11

0

0

1

66.5

23.8

39.4

0.43

0.33

0.0034

0.95

1

1

1

1

1

0

0.24

0.90

0.07

0.03

UNAIDS

Boily et al

Sweat et al (2000)

Assumption

Assumption

Assumption

Assumption

Assumption

Assumption

Assumption

Assumption

Assumption

PSI DALY Model

PSI DALY Model

PSI DALY Model

Malawi 2008 Census

Malawi 2008 Census

Malawi 2008 Census

Malawi Data

Malawi Data

Assumed

Model Parameters

Estimate Source Type of

Parameter

Mortality

Probabilities with

HIV

Parameter

Age Group 1

Age Group 2

Age Group 3

Age specific

Mortality Rates

(per year) under 24 (Age Group 1)

25-34 (Age group 2)

35+ (Age Group 3)

Lifetimes with HIV

(years)

15-24 (Age Group 1)

25-34 (Age Group 2)

35+ (Age group 3)

0.058

0.063

0.073

0.0046

0.0091

0.0119

12

11

9.5

Estimated

Estimated

Estimated

Malawi 2008 Census

Malawi 2008 Census

Malawi 2008 Census

Actuarial Society of

South Africa

Actuarial Society of

South Africa

Actuarial Society of

South Africa

Model Parameters: Sexual Behavior

Parameter Estimate Source

Average Number of Sex acts per partner (Regular)

Average Number of Sex acts per partner (Casual)

Average Number of Sex acts per partner (Commercial)

Proportion of sex acts where a condom was used (Risk 1) (F)

Proportion of sex acts where a condom was used (Risk 2)

Proportion of sex acts where a condom was used (Risk 3)

Probability of Condom use with regular partner

Probability of Condom use with casual partner

Probability of Condom use with CSW

Prevalence of HIV-1 among low risk group (Risk 1)

Prevalence of HIV-1 among moderate risk group (Risk 2)

Prevalence of HIV-1 among CSW

P (Risk 1)

P (Risk 2)

P (Risk 3)

E (Condom efficacy)

DALYs Averted per infection averted

Effect of VCT on Condom Use (Odds Ratio)

Baseline VCT

0.11

0.11

0.21

0.95

20

1.69

0.755

66.5

45.15

43.2

0.05

0.20

0.33

0.05

0.62

0.62

0.11

0.11

0.43

Calculated

Calculated

Calculated

Estimated

Estimated

Estimated

Malawi data

Malawi data

Assumed from

Malawi data

UNAIDS

UNAIDS

Est. from UNAIDS and PSI formulation

UNAIDS

UNIADS

Estimated

Sweat et al

Estimated

Denison et al

Malawi Data

Cost per person reached, by media

$1,60

$1,46

$1,40

$1,20

$1,00

$0,80

$0,60

$0,40

$0,20

$-

$0,46

$0,16

$0,72

Print

N=1,432,303

TV

N=631,470

Radio

N=4,026,211

Multimedia

N=4,061,764

Program-attributable effects

a

OneLove

Radio Any TV

Any booklets

Multimedia

0.0000

0.025

0.050

0.000

Increase in VCT

Effect on partnerships per year

Increase in condom use with regular partners

-0.0130

0.000

0.000

0.000

0.0000

0.050

0.046

0.045

Increase in condom use with casual partners 0.0180

0.030

0.000

0.000

Increase in condom use with comm. sex partners

0.0000

0.000

0.000

0.000

a Coefficients are the marginal effects from multivariate regression models controlling for

Endogenous exposure to interventions. Non-zero coefficients represent effects that are statistically

Significant at better than the 5% level.

Cost Per DALY averted

$30,00

$25,00

$20,00

$15,00

$10,00

$5,00

$-

$15,41

$7,79

$4,27

$2,15

Print TV Radio

$24,08

$11,98

Multimedia

5-Year

10-Year

One-way simulations:

$60,00

$50,00

$48,16 $48,30

$40,00

$30,82 $29,74

$30,00

$24,08

$20,00

$10,00

$-

$15,41 $15,73

$8,53

$4,27

$8,24

$4,35

$4,10

$2,02 $2,03

Print TV Radio Multimedia

5-Year Doubling Financial Costs

50% reduction in condom use ME 50% reduction in effects on testing

How do these results compare to non-BCC interventions?

Intervention

VCT

VMMC

Print

Television

Radio

Multimedia

Cost per person

$7-$11

$74-$90

$1.46

$0.40

$0.16

$0.72

Cost per infection averted

$1,315 a

$306-$1,216; b

$176-$3,554 c

$0.77

$0.21

$0.10

$1.20

Cost per

DALY averted

$65.75

$15.30-$60.80

$8.80-$177.70

$15.41

$4.27

$2.02

$24.08

a Behrman and Kohler, (2011); b Njeuhmeli, Forsythe, et al (2011); c Gray, Kigozi et al (2007)

Key Findings

High levels of reach mean that even modest effects on behaviours will avert sufficient numbers of infections and DALYs to make Pakachere interventions costeffective

Programs do not need to affect all behaviors; just a few.

But modesty should prevail; the research design does not ensure that effects are causal.

Acknowledgements

We are grateful to Esca Scheepers for her comments, suggestions, and guidance with the conceptualization, design, evaluation and analysis. We thank Soul City

Institute colleagues Sue Goldstein, Bongiwe Ndondo, Renay Weiner, and Michael

Jana, who contributed their regional experience to the evaluation. We also thank

Katia Da Silva and Eugene Strydom for compiling the information on the costs of the Soul City program. Data collection for the household survey was led by Susan

Cotts Watkins and Invest in Knowledge Initiative. We would like to thank Susan

Watkins, Senior Research Scientist at the California Center for Population

Research, Peter Fleming, Executive Director of IKI, Mike Dalious, Research

Director of IKI and James Mkandawire, Assistant Research Director. We would also like to thank Stephane Helleringer for designing the sampling scheme and

Kingsley Manda of the National Statistics Office, Malawi for assistance in selecting the sample. Finally, we would like to thank DFID for their support of this evaluation, particularly Petra Nahmias, Anna Seymour, Dirk Mueller, Liz Harmer, and Max Gasteen. Useful comments on earlier drafts have been provided by

Lesong Conteh, Alan Whiteside, Dirk Mueller, Anna Seymour and Jane Bertrand.

Bibliography

Creese, A., K. Floyd, et al. (2002). "Cost-effectiveness of HIV/AIDS interventions in Africa: a systematic review of the evidence." Lancet 359(9318): 1635-1643.

Denison, J. A., K. R. O'Reilly, et al. (2008). "HIV voluntary counseling and testing and behavioral risk reduction in developing countries: a meta-analysis, 1990--2005." AIDS Behav 12(3): 363-373.

Drummond, M. F., M. J. Sculpher, et al. (2005). Methods for the Economic Evaluation of Health Care

Programmes. Oxford, United Kingdom, Oxford University Press.

Hutton, G., K. Wyss, et al. (2003). "Prioritization of prevention activities to combat the spread of

HIV/AIDS in resource constrained settings: a cost-effectiveness analysis from Chad, Central Africa."

Int J Health Plann Manage 18(2): 117-136.

Kincaid, D. L. and W. Parker (2008). National AIDS Communication Programmes, HIV Prevention

Behaviour, and HIV Infections Averted in South Africa, 2005 Pretoria, South Africa, JHHESA.

Muirhead, D., L. Kumaranayake, et al. (2001). Economically Evaluating the 4th Soul City Series: Costs and Impact on HIV/AIDS and Violence Against Women. Johannesburg, South Africa, Centre for

Health Policy, University of Witswatersrand and Health Policy Unit, London School of Hygiene and

Tropical Medicine.

Sweat, M., S. Gregorich, G. Sangiwa, C. Furlonge, D. Balmer, C. Kamenga, O. Gristead, T. Coates,

“Cost-effectiveness of voluntary HIV-1 counseling and testing in reducing sexual transmission of HIV-

1 in Kenya and Tanzania,” The Lancet , Vol 356, July 8, 2000.

Walker, D. (2003). "Cost and cost-effectiveness of HIV/AIDS prevention strategies in developing countries: is there an evidence base?" Health Policy Plan 18(1): 4-17.

Watkins, S. C., Invest in Knowledge, et al. (2012). External Evaluation of the Southern African Regional Social and

Behaviour Change Communication Programme, as Implemented in Malawi. Zomba, Malawi, Invest in Knowledge.

West, P. (2010). Cost-Effectiveness of Regional HIV and AIDS Behaviour Change Communication (BCC) Programme.

Download