I had the privilege of attending a conference last week on targeting and tailoring for health interventions convened by the Center for Health Communications & Marketing at the University of Connecticut. Over 40 health communication researchers spent the day in presentations and discussions around the topics of:
- Addressing ethnic and racial health disparities through segmenting, targeting and tailoring
- Approaches to segmenting and targeting
- Meta-analyses of tailored interventions
- Tailoring cases and lessons
Addressing ethnic and racial health disparities through segmenting, targeting and tailoring
Bob Hornik made the point that there have been no studies to systematically evaluate the effectiveness of segmentation strategies in designing and implementing campaigns aimed at specific racial or ethnic audiences as compared with campaigns that do not use a segmentation approach. In the discussion, it was noted that meta-analytic studies of the effectiveness of mass media campaign aimed at specific audiences do find effect sizes similar to those seen among more mainstream campaigns. While the reasons behind using a segmentation strategy may rest more on common-sense than data, there are also political constituencies that exist for racial and ethnic segmentation approaches to public health programs and campaigns (aka eliminating health disparities). [Racial/ethnic disparities and segmentation in communication campaigns.]
One of the points I took away from the presentation of the work of the ONDCP media campaign was their research demonstrating (again) that acculturation among immigrant groups is a salient program design concern with the parents of teens, but not among teens themselves. That is, regardless of their race or ethnicity, teens are more alike than are their parents with respect to KABs about drug use. It was also reported that there is a ‘disconnect’ between what parents say in focus groups about their behavior and communications about drugs and drug use with their kids (including monitoring behaviors) and what their own kids say about their parents’ behaviors. It was mentioned that similar ‘disconnects’ were found in the VERB ™ campaign between parents’ reports of their behaviors and what their children report about the same behaviors.
I was especially interested in data Dr. Perez-Escamilla had is his presentation ‘Selling Health to Latina(o)s.’ He showed an unpublished table from a study where significant changes in fruit and vegetable consumption among the priority audience were related to whether they reported being exposed to campaign messages through 3 or more - or less than 3 - media channels during the campaign. It is one of the few times I have seen this type of analysis that support the idea that it is recall of multiple channel exposures that are related to behavioral changes – not some type of ‘magic bullet.’ He made another point about breast-feeding campaigns among Latinas where the cultural background of the population is an important component for segmentation and tailoring: Mexican Americans come from ‘a breast feeding culture’ whereas Puerto Ricans do not and this factor needs to be taken into account in program development. Another point for the argument that the Hispanic audience is not monolithic.
Matt Kreuter presented a study that is looking at whether media releases for black newspapers (BNs) about cancer prevention and screening topics that are tailored to local demographics and resources will result in greater coverage of cancer topics among newspapers that receive the tailored releases than among those newspapers that do not. The answer is ‘yes,’ but more interesting to me was that mainstream newspapers were still, despite no additional information from the research project, more likely to cover the subject (in one wave of data, 8/12 mainstream, 6/12 ‘tailored BNs’ and 1/12 ‘comparison BNs’ covered news about the new HPV vaccine to prevent cervical cancer). The study underscores the importance of looking at the media as not just a channel for social marketing and communications programs, but also as a potential source or determinant of health inequalities.
Approaches to segmenting and targeting
The presentations and discussion around segmenting and targeting revolved around two issues: (a) the usefulness and feasibility of using behaviors and behavioral attributes, and (b) selecting priority audiences from the various segments identified in the research. [My contribution to this session was a presentation that started with Rediscovering Market Segmentation and ended by touching on the challenges of social media and social networks for segmentation studies]. In the first instance, the consensus seemed to be that behavioral segmentation was a preferred methodology to pursue, but the question is whether a typology of behaviors or attributes could be created to facilitate this process – or would each situation demand a customized approach to the behavior, its determinants and attributes, and the context in which it occurs?
A segmentation study from the commercial sector opened a spirited discussion of the selection of priority audiences. The question boiled down to ‘why do for-profit companies target people with money and other resources and avoid the audiences that we (public health professionals) have to focus on?’ I think we settled that for-profit companies target people who help them achieve their organizational mission (make a profit for shareholders) while public health professionals focus on audiences that help them achieve their organizational (and professional) mission of reducing health disparities and improving overall public health. The philosophy and approach to segmentation between the two is no different, just the ends for which it is employed.
Meta-analyses of tailored interventions
An unpublished (as of the conference) meta-analyses of 61 studies that have looked at tailored print interventions for health behavior change found a significant effect for tailoring materials to audience characteristics when compared to either comparison messages that were not tailored or no treatment controls. Both prevention (smoking cessation, diet and physical activity) and screening (mammography, Pap test) behaviors were equally affected by the tailored interventions. Between 0-9 theoretical constructs were used to inform the tailoring of content in these studies, and the typical one used four of them (such as perceived susceptibility, social norms, processes of change, behavioral intention, social support and self-efficacy). An interesting finding was that using perceived risk as a tailoring construct was less useful than other concepts is leading to successful behavioral outcomes. However, simply using more theoretical concepts did not appear associated with stronger effect sizes.
In a review of worksite health risk appraisal (HRA) interventions the summary was that there was insufficient empirical evidence to determine the effectiveness of just providing health risk information from HRAs to improve health behaviors. When health education components are added to these HRA a variety of positive changes are seen in such health indicators as smoking status, blood cholesterol and blood pressure levels and possibly diet and nutrition. The take-home message was that the ‘industry standard’ of simply assessing, screening and providing feedback of health risks was insufficient for changing risk behaviors and, one assumes, lowering health care utilization and costs in workplaces.
Tailoring cases and lessons
In the cases and lessons sections of the program, the issue of how to scale up and disseminate tailored interventions was presented by Marci Campbell through her group’s work with the VA health system and a web-to-print system. Wayne Velicer presented new work with the Transtheoretical Model in predicting outcomes and mediators of progression through the stages of cigarette smoking adoption by teens. One finding from this work is that peer pressure is not a strong predictor of starting to smoke, but peer selection after starting to smoke (hanging out with other smokers) is a critical social variable in whether the adopted behavior is maintained or not.
In his discussion of the CHESS program with cancer patients, Bret Shaw reminded us that behavioral outcomes are not always the primary ones of interest when working with people with chronic or terminal illnesses: quality-of-life indicators are often the more important considerations among patients, their families and their caregivers. We also talked briefly about how programs such as CHESS might be augmented with social media – an evolution they are actively exploring. Timothy Bickmore gave us a glimpse of health dialog systems as a tailoring device and introduced us to embodied conversational agents (EMA) and his physical activity EMA named “Laura.’ Can’t wait to see her debut in Second Life! One of his research interests is in developing voice systems for health sites as a way to increase engagement and use of these types of sites. [See Health dialog systems for patients and consumers.]