Evaluating Anti-Smoking Media Campaigns Using Non-experimental Data: An Analysis using Propensity Score Matching and Simultaneous Equations Models with Data from Egypt and Jordan

APHA Scientific Session and Event Listing

Paul Hutchinson, PhD, International Health and Development, Tulane University, 1440 Canal Street, Suite 2200, New Orleans, LA 70118, 504 988-6078, phutchin@tulane.edu

In numerous countries, anti-smoking campaigns have been highly effective in disseminating information related to the health risks of smoking, in altering norms related to smoking behavior, and ultimately in reducing smoking prevalence and encouraging smoke-free environments. However, high rates of smoking continue, and in fact are increasing, in many countries, particularly in the Middle East. This study examines the effectiveness of nationwide smoking prevention and cessation campaigns in Egypt and Jordan on smoking knowledge, intentions and behavior of young adults aged 15 to 24. Two issues, common to such evaluations, complicate the analysis. First, measures of exposure to the campaigns are derived from self-reports in population-based surveys. Such measures of exposure may in fact lead to biased estimates of campaign effectiveness because they are likely to be non-random, reflecting a variety of measured characteristics (e.g. access to media, attitudes, education) and unmeasured characteristics of respondents (e.g. motivations, health conscientiousness, supply-side variables). Second, the anti-smoking campaigns were national in coverage, precluding the randomization of respondents into intervention (exposed to the communication campaign) and control (non-exposed) groups. This study compares the performance of two quasi-experimental analytical methods to correct for the problems of non-random assignment: (1) propensity score matching to compare differences in outcomes across groups with similar likelihoods of exposure and (2) simultaneous equations models to purge estimates of exposure of common unmeasured effects across recall and behaviors. These methods are used with data collected in 2004 and 2005 by the Health Communication Partnership.

Learning Objectives:

By the end of the session, participants will be able

1. To identify and discuss the problems of non-random reporting of exposure to mass media messages in evaluations of communication programs
2. To identify factors associated with the differential effectiveness of anti-smoking campaigns on different population sub-groups in Egypt and Jordan
3. To describe the differing conditions under which propensity score matching and simultaneous equations models are applicable to evaluations of health communication campaigns

Keywords: International Public Health, Evaluation

Presenting author's disclosure statement:

Any relevant financial relationships? No

Recorded presentation

Communicate Effectively: Campaigns and Evaluation Techniques

The 134th Annual Meeting & Exposition (November 4-8, 2006) of APHA