Abstract
This paper employs a causal inference framework to explore two logically distinct forms of preventive effect heterogeneity relevant for studying variation in preventive effect as a basis for developing more personalized interventions. Following VanderWeele (2015), I begin with a discussion of causal interaction involving manipulable moderators that combine to yield more complex nonadditive effects. This is contrasted with effect heterogeneity, which involves variation in causal structure indexed by stable characteristics of populations or contexts. The paper then discusses one particularly promising approach, the baseline target moderated mediation (BTMM) design, which uses theoretically informed baseline target moderators to strengthen causal inference, suggesting methods for using BTMM designs to develop targeting strategies for personalized prevention. It presents examples of recent intervention trials that apply these different forms of moderation, and discusses causal inference and the problem of moderation confounding, reviewing methods for minimizing its impact, including recent advances in the use of propensity score matching.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The causal status of factors that cannot change within individuals, but can vary across individuals, has been the source of some controversy, as witnessed by a recent exchange on race and gender (Glymour and Glymour 2014; Kaufman 2014; VanderWeele and Robinson 2014a, b). Variables such as race can clearly be causes in models where we define them as stimulus conditions for the behavior of others. In that case, they are in fact changeable, since we could vary those conditions and assess the impact of those different conditions. And if the behavior of others in turn has an impact on those who are the “stimuli,” then personal characteristics could have an indirect impact on the person who has that characteristic, mediated through the behavior of others. In this case, variation in race across individuals would provide some index, albeit indirect, of rates of exposure to discrimination by others.
Baseline targets are usually effect moderators rather than part of a causal interaction. The status of a target variable at baseline is historical, that is, when the intervention occurs, that baseline value is in the past, and is unchangeable. The future status of that variable can change as a result of the intervention, and therefore can be considered causal of later changes in more distal outcomes, but the baseline value of that target is immutable when the intervention commences. More complex effects are also possible, although infrequently hypothesized. For example, a target may change dynamically in the period prior to intervention, and prior rates of change (rather than status at baseline) could in principle moderate intervention impact on subsequent rates of change in the target. An intervention might exacerbate that rate of change, or dampen it, and do so in different ways depending on the rate of change at the time the intervention begins. This would reflect a causal interaction, although preintervention rates of change in the target would usually be observed rather than induced.
A reviewer raised concerns as to whether BTMM effects could simply reflect regression to the mean. Values on target variables observed over two or more occasions may change for many reasons, including regression to the mean due to unreliability of measurement, decay of unobserved causes that led to initial increase, or exposure to other unobserved causes that shape change in the target. Unless there is reason to believe that measurement reliability differs for intervention and control groups, leading to differential regression to the mean for a target variable, regression artifacts are unlikely to account for findings of baseline target moderation of intervention impact on change in the target.
A reviewer noted that promotive interventions often seek to build on existing strengths as a way of promoting well-being, suggesting that this is antithetical to the notion that those with baseline deficits will improve the most. This raises some interesting issues. Can a strength be a “latent” protective factor, in that it is present but inactive when risk mechanisms are in operation, and must be activated to become protective? Can strengths in one life arena be activated through intervention to reduce risks in another arena where they are not currently being employed? Both seem plausible, and could easily be incorporated into the BTMM framework. Activation level of a strength could be a reasonable baseline moderator (when the strengths are already activated and are functioning to protect from risk, interventions to activate them will be less necessary and have less impact).
References
Bierman, K. L., Domitrovich, C. E., Nix, R. L., Welsh, J. A., & Gest, S. D. (2014). Integrating evidence-based preschool programs to support social-emotional and cognitive development. In M. Boivin, K. L. Bierman, M. Boivin, & K. L. Bierman (Eds.), Promoting school readiness and early learning: Implications of developmental research for practice (pp. 231–252). New York: Guilford Press.
Brand, J. E., & Thomas, J. S. (2013). Causal effect heterogeneity. In S. L. Morgan (Ed.), Handbook of causal analysis for social research (pp. 189–214). New York: Springer.
Brody, G. H., Chen, Y.-f., Beach, S. R. H., Kogan, S. M., Yu, T., DiClemente, R. J., et al. (2014). Differential sensitivity to prevention programming: A dopaminergic polymorphism-enhanced prevention effect on protective parenting and adolescent substance use. Health Psychology, 33, 182–191. doi:10.1037/a0031253.
Cardemil, E. V., Reivich, K. J., & Seligman, M. E. P. (2002). The prevention of depressive symptoms in low-income minority middle school students. Prevention & Treatment, 5, Article 8. Retrieved from http://psycnet.apa.org/journals/pre/5/8. doi:10.1037/1522-3736.5.1.58a.
Cho, J., Kogan, S. M., & Brody, G. H. (2016). Genetic moderation of transactional relations between parenting practices and child self-regulation. Journal of Family Psychology. doi:10.1037/fam0000228.
Clarke, G. N., Hornbrook, M., Lynch, F., Polen, M., Gale, J., Beardslee, W., et al. (2001). A randomized trial of a group cognitive intervention for preventing depression in adolescent offspring of depressed parents. Archives of General Psychiatry, 58, 1127.
Collins, L. M., Murphy, S. A., & Bierman, K. L. (2004). A conceptual framework for adaptive preventive interventions. Prevention Science: The Official Journal Of The Society For Prevention Research, 5, 185–196.
Dumka, L. E., Gonzales, N. A., Bonds, D. D., & Millsap, R. E. (2009). Academic success of Mexican origin adolescent boys and girls: The role of mothers’ and fathers’ parenting and cultural orientation. Sex Roles, 60, 588–599. doi:10.1007/s11199-008-9518-z.
Falloon, I. R. H., Shanahan, W., & Laporta, M. (1992). Prevention of major depressive episodes: Early intervention with family-based stress management. Journal of Mental Health, 1, 53–60. doi:10.3109/09638239209034511.
Ghitza, U. E. (2015). A commentary on ‘A new initiative on precision medicine’. Frontiers in Psychiatry, 6.
Glymour, C., & Glymour, M. R. (2014). Commentary: Race and sex are causes. Epidemiology (Cambridge, Mass.), 25, 488–490. doi:10.1097/EDE.0000000000000122.
Gonzales, N. A., Dumka, L. E., Millsap, R. E., Gottschall, A., McClain, D. B., Wong, J. J., et al. (2012). Randomized trial of a broad preventive intervention for Mexican American adolescents. Journal of Consulting and Clinical Psychology, 80, 1–16. doi:10.1037/a0026063.
Green, K. M., & Stuart, E. A. (2014). Examining moderation analyses in propensity score methods: Application to depression and substance use. Journal of Consulting and Clinical Psychology, 82, 773–783. doi:10.1037/a0036515.
Gunasekara, F. I., Carter, K., & Blakely, T. (2011). Change in income and change in self-rated health: Systematic review of studies using repeated measures to control for confounding bias. Social Science & Medicine (1982), 72, 193–201. doi:10.1016/j.socscimed.2010.10.029.
Harder, V. S., Stuart, E. A., & Anthony, J. C. (2010). Propensity score techniques and the assessment of measured covariate balance to test causal associations in psychological research. Psychological Methods, 15, 234–249. doi:10.1037/a0019623.
Howe, G. W., Reiss, D., & Yuh, J. (2002). Can prevention trials test theories of etiology? Development and Psychopathology, 14, 673–694. doi:10.1017/S0954579402004029.
Howe, G. W., Beach, S., & Brody, G. (2010). Microtrial methods for translating gene-environment dynamics into preventive interventions. Prevention Science, 11, 343–354. doi:10.1007/s11121-010-0177-2.
Howe, G. W., Beach, S. R. H., Brody, G. H., & Wyman, P. A. (2016). Translating genetic research into preventive intervention: The baseline target moderated mediator design. Frontiers In Psychology, 1–9. doi:10.3389/fpsyg.2015.01911.
Josephy, H., Vansteelandt, S., Vanderhasselt, M.-A., & Loeys, T. (2015). Within-subject mediation analysis in AB/BA crossover designs. The International Journal of Biostatistics, 11, 1–22. doi:10.1515/ijb-2014-0057.
Kaufman, J. S. (2014). Commentary: Race: Ritual, regression, and reality. Epidemiology (Cambridge, Mass.), 25, 485–487. doi:10.1097/EDE.0000000000000117.
Kidd, K. K. (1993). Associations of disease with genetic markers: Déjà vu all over again. American Journal of Medical Genetics, 48, 71–73.
Liu, W., Kuramoto, S. J., & Stuart, E. A. (2013). An introduction to sensitivity analysis for unobserved confounding in nonexperimental prevention research. Prevention Science, 14, 570–580. doi:10.1007/s11121-012-0339-5.
MacKinnon, D. P., Taborga, M. P., & Morgan-Lopez, A. A. (2002). Mediation designs for tobacco prevention research. Drug and Alcohol Dependence, 68, S69–S83. doi:10.1016/S0376-8716(02)00216-8.
McClelland, G. H., & Judd, C. M. (1993). Statistical difficulties of detecting interactions and moderator effects. Psychological Bulletin, 114, 376–390. doi:10.1037/0033-2909.114.2.376.
Multisite_Violence_Prevention_Project. (2014). Implementation and process effects on prevention outcomes for middle school students. Journal of Clinical Child and Adolescent Psychology, 43, 473–485. doi:10.1080/15374416.2013.814540.
Nolen-Hoeksema, S. (2001). Gender differences in depression. Current Directions in Psychological Science, 10, 173.
Pearl, J. (2009). Causality: Models, reasoning, and inference (2nd ed.). New York: Cambridge University Press.
Pearl, J. (2012). The causal mediation formula—A guide to the assessment of pathways and mechanisms. Prevention Science, 13, 426–436. doi:10.1007/s11121-011-0270-1.
Pelham, W. E., Fabiano, G. A., Waxmonsky, J. G., Greiner, A. R., Gnagy, E. M., Coxe, S., et al. (2016). Treatment sequencing for childhood ADHD: A multiple-randomization study of adaptive medication and behavioral interventions. Journal of Clinical Child & Adolescent Psychology, 45, 396–415. doi:10.1080/15374416.2015.1105138.
Perrino, T., Pantin, H., Prado, G., Huang, S., Brincks, A., Howe, G., et al. (2014). Preventing internalizing symptoms among Hispanic adolescents: A synthesis across Familias Unidas trials. Prevention Science, 15, 917–928. doi:10.1007/s11121-013-0448-9.
Perrino, T., Brincks, A., Howe, G., Brown, C. H., Prado, G., & Pantin, H. (2016). Reducing internalizing symptoms among high-risk, Hispanic adolescents: Mediators of a preventive family intervention. Prevention Science: The Official Journal Of The Society For Prevention Research, 17, 595–605. doi:10.1007/s11121-016-0655-2.
Rassen, J. A., Shelat, A. A., Franklin, J. M., Glynn, R. J., Solomon, D. H., & Schneeweiss, S. (2013). Matching by propensity score in cohort studies with three treatment groups. Epidemiology (Cambridge, Mass.), 24, 401–409. doi:10.1097/EDE.0b013e318289dedf.
Rubin, D. B. (1974). Estimating causal effects of treatments in randomized and nonrandomized studies. Journal of Educational Psychology, 66(5), 688–701. doi:10.1037/h0037350.
Rutter, M. (2007). Proceeding from observed correlation to causal inference: The use of natural experiments. Perspectives on Psychological Science, 2, 377–395. doi:10.1111/j.1745-6916.2007.00050.x.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. New York: Houghton Mifflin Co..
Stormshak, E. A., & Dishion, T. J. (2009). A school-based, family-centered intervention to prevent substance use: The family check-up. The American Journal of Drug and Alcohol Abuse, 35, 227–232. doi:10.1080/00952990903005908.
Teisl, M., Wyman, P. A., Cross, W., West, J., & Sworts, L. (2012). Adaptive intervention to address the needs of children with language delays and behavior problems: Proximal impact on emotion-regulation skill knowledge. Paper presented at the Annual meeting of the Society for Prevention Research, Washington, DC.
VanderWeele, T. J. (2009). On the distinction between interaction and effect modification. Epidemiology (Cambridge, Mass.), 20, 863–871. doi:10.1097/EDE.0b013e3181ba333c.
VanderWeele, T. J. (2015). Explanation in causal inference. Methods for mediation and interaction. Oxford: Oxford University Press.
VanderWeele, T. J., & Robins, J. M. (2007). Four types of effect modification: A classification based on directed acyclic graphs. Epidemiology (Cambridge, Mass.), 18, 561–568.
VanderWeele, T. J., & Robins, J. M. (2008). Empirical and counterfactual conditions for sufficient cause interactions. Biometrika, 95, 49–49.
VanderWeele, T. J., & Robinson, W. R. (2014a). On the causal interpretation of race in regressions adjusting for confounding and mediating variables. Epidemiology (Cambridge, Mass.), 25, 473–484. doi:10.1097/EDE.0000000000000105.
VanderWeele, T. J., & Robinson, W. R. (2014b). Rejoinder: How to reduce racial disparities?: Upon what to intervene? Epidemiology (Cambridge, Mass.), 25, 491–493. doi:10.1097/EDE.0000000000000124.
Weiss, N. S. (2017). What findings are needed to advocate personalized (precision) prevention of disease? American Journal of Public Health, 107, 86–87. doi:10.2105/AJPH.2016.303513.
Yang, S., Imbens, G. W., Cui, Z., Faries, D. E., & Kadziola, Z. (2016). Propensity score matching and subclassification in observational studies with multi-level treatments. Biometrics, 72, 1055–1065.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Funding
This manuscript was supported in part by National Institute of Mental Health grant number R01-MH040859.
Conflicts of Interest
The author declares that he has no conflict of interest.
Ethical Approval
The manuscript does not report any empirical findings; no study was conducted requiring ethical approval.
Informed Consent
The manuscript does not report any empirical findings; no study was conducted requiring informed consent.
Rights and permissions
About this article
Cite this article
Howe, G.W. Preventive Effect Heterogeneity: Causal Inference in Personalized Prevention. Prev Sci 20, 21–29 (2019). https://doi.org/10.1007/s11121-017-0826-9
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11121-017-0826-9