the charge to advance theory and improve health outcomes
TRANSCRIPT
C O M M E N T A R Y O N B A R T H O L O M E W A N D M U L L E N jphd_229 36
The charge to advance theory and improve health outcomesSarah Kobrin, PhD, MPH
Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, MD
Keywordstheory; behavior; intervention.
CorrespondenceDr. Sarah Kobrin, Division of Cancer Controland Population Sciences, National CancerInstitute, 6130 Executive Blvd., EPN 4096,Bethesda, MD 20852. Tel.: 3014358662;e-mail: [email protected], [email protected] Kobrin is with the Applied CancerScreening Research Branch, National CancerInstitute.
doi: 10.1111/j.1752-7325.2011.00229.x
Bartholomew and Mullen have written an inspiring call toaction. Better application of behavioral theory in health-related intervention research will improve health outcomesand promote reproducible science. These goals, sometimesperceived as in tension, are both well served by a systematicapproach to intervention design, implementation, evalua-tion, and description.
A coherent model of problem, determinants, and inter-vention is essential. As emphasized by Bartholomew andMullen, this model can contain a single theory – whole orparts – or multiple theories. An effective model is internallyconsistent, based on theoretical and empirical evidenceand the investigator’s experiences with the behavior, popu-lation, and setting. This coherence of the model – not itsderivation from a single theory – produces a cascade ofbenefits. Defining a behavioral outcome within a domainsuggests measureable, proximal outcomes and determi-nants; in their example, the authors focus on “diagnosticdelay” among outcomes related to “late-stage diagnosis.”The determinants of delay imply points of interventionand mediators. Specificity greatly increases the likelihoodof informative findings distinct from interventioneffectiveness.
Accurate and complete reporting of interventions is essen-tial to scientific advancement. As described, publishing onlyintervention success is insufficient to allow others to replicate,adapt, and extend. Publication of sufficient detail of the theo-retical foundations of interventions and analysis of activeingredients is necessary for dissemination. Publishers shouldallow these descriptions, and investigators should make inter-vention materials widely available.
Beyond associated groups of predictors, relations amongposited mediators can be assigned directionality and tempo-rality – increasing X will decrease Y. Describing causal rela-tions allows investigators to pose falsifiable hypotheses.Testing these hypotheses identifies where in the causal chainexpected links did or did not hold – from theory to interven-tion to mediator to outcomes. Furthermore, mediationshould be tested even when overall effect of intervention onoutcome is not significant (1). The intervention may haveinfluenced the mediator, but the theorized relation betweenmediator and outcome did not hold. Or the intervention mayhave failed to influence the mediator, meaning the effect ofmediator on outcome was not actually tested. Tests of mod-eration could show a hypothesized relation held for onegroup but not another, or held differently across groups,“washing out” the overall effect of intervention on outcome.In these tests are found the real contribution to reproduciblescience and improved health outcomes.
One of the most challenging aspects of applying theory isidentifying methods for changing constructs. Bartholomewand Mullen highlight methods drawn largely from communi-cation theories. They make the excellent recommendation ofa table to match objectives with methods and strategies. Manyof their examples of theory-based“methods”are then appliedin mass media campaigns, a strategy that is not always appro-priate (e.g., increasing follow-up to abnormal cancer screen-ing). Unfortunately, most theories of health behavior aremute on how to create change (2). This gap is an opportunityfor future research.
To change norms across health-related interventionresearch, we must all become vectors. NIH funding isawarded through peer-review; the scientific community is thegreatest influence defining acceptable application of healthbehavior theory. In this charge, we should all become leaders.
Conflict of interest
The author has declared no conflict of interest.
References
1. MacKinnon DP. Introduction to statistical mediation analysis.New York: Laurence Erlbaum Assoc.; 2008. p. 48-51.
2. Rothman A. Is there nothing more practical than a goodtheory?: Why innovations and advances in health behaviorchange will arise if interventions are used to test and refinetheory. Int J Behav Nutr Phys Act. 2004;1(11):1-11.
Journal of Public Health Dentistry . ISSN 0022-4006
S36 Journal of Public Health Dentistry 71 (2011) S36 © 2011 American Association of Public Health Dentistry