Integrating the Life Course Perspective Into Health Disparities Intervention Research
Over the past several decades there has been considerable research focused on uncovering the causes of risk behaviors during childhood and adolescence, with a primary focus on factors involved in the development of health risk behaviors (e.g., Capaldi, Kerr, Eddy, & Tiberio, 2016). There have been calls for more dynamic intervention models bridging strong theoretical models with advanced methodology (e.g., Lerner, 2012). One notable shortcoming of most research on preventive intervention programs targeting adolescents' health risk behaviors is their reliance on the assumption of a homogeneous sample population. That is, existing research on intervention effectiveness has used analysis techniques that treat the sample as a homogeneous group when assessing how interventions affect adolescents'behavioral outcomes (Brincks, Perrino, Howe, Pantin, Prado, Huang, et al., 2018).
As introduced previously, the life course perspective emphasizes the possibility of distinct developmental patterns of health risk behaviors (i.e., heterogeneity in health behavior trajectories). Consonant with this perspective, prevention scientists have begun to focus on identifying effective prevention programs for heterogeneous groups of youths (Connell, Stormshak, Dishion, Fosco, & Van Ryzin, 2018). For example, Brincks, and colleagues (2018) reported differential intervention effects among Hispanic youth with different trajectories of health problems (i.e., internalizing problems) using GMM. They identified three distinct trajectories of health problems: 1) low internalizing symptoms (60% of the sample), 2) moderate internalizing symptoms (27% of the sample), and 3) high internalizing symptoms (13% of the sample). They found that intervention was most beneficial for the youth with high internalizing symptom levels. That is, within this class, youth receiving the intervention experienced reduced trajectories of internalizing symptoms over time, whereas youth in the control condition experienced increased internalizing symptoms. This type of person-centered approach allows health researchers to identify which adolescents are most likely to benefit from the intervention. Therefore, the results can help direct the intervention to youth who are most likely to gain from it and identify youth who are not likely to benefit, for whom other interventions may be more effective.
In a GMM, when a predictor variable is the intervention indicator (i.e., a dichotomous variable indicating whether the individual received the intervention or not), the model specification is slightly different from the previously introduced GMM, where predictors are entered as independent variables that are thought to explain the latent class variable (see Panel a of Figure 19.2). Class membership should be independent of condition assignment so that the discrete groups (or classes) defined by the trajectory are not influenced by whether the participants are assigned to intervention or control (Brincks, et al., 2018). Instead, the intervention condition may interact with the class membership trajectories to explain the differential effects of the intervention depending on class membership. Therefore, a latent class variable is used to estimate interaction effects between class membership and intervention condition on trajectories of health risk behaviors (see a dotted line in Panel c of Figure 19.2; Brincks, et al., 2018).
An Example Study: Integrating the Life Course Perspective Into Health Disparities Intervention Research
A study by Liu, Hedeker, Segawa, and Flay (2010) is an example of this analytic approach. The study employed a GMM to identify classes of substance use behavior trajectories with a sample of African American male adolescents, and they identified differential intervention effects among these subgroups of individuals. The study used data from the Aban Aya Youth Project (AAYP), which is a longitudinal preventive intervention trial targeting health-compromising behaviors among adolescents in disadvantaged areas of Chicago. An ordinal scale was used to assess substance use behavior as the outcome of interest. Self-reported data were collected from the participating youth at the beginning of fifth grade and at the end of grades five, six, seven, and eight (» = 668 adolescents, M = 10.8 years in fifth grade).
First, a GMM indicated two classes of adolescent drug use trajectories; Class 1 (44.7% of the sample) reported a relatively higher initial level of drug use (fifth grade) and had a small increase over time; Class 2 had a lower baseline level and a greater increase over time. Next, the intervention condition was entered into the GMM to predict the slope factor. The results showed an interaction effect between youth in Class 2 and intervention condition (b = —0.912, p < .05). That is, in Class 2 (the group with the lower baseline drug use but greater increase), the adolescents in the intervention group showed a slower increase in drug use trajectories compared to adolescents in the control group, while the drug use trajectory pattern for adolescents in Class 1 did not differ significantly between the intervention and control groups. These findings suggest that interventions are more (or less) effective for certain individuals based on their trajectory pattern. Consequently, these heterogeneous intervention effects indicate the need for varied, targeted interventions based on trajectory patterns. For example, based on this study, the intervention was not helpful for youth who displayed higher levels of drug use at the first measurement occasion (fifth grade). Instead, the intervention was appropriate for youth with lower levels of early drug use. Together, these findings could signal the need for more targeted intervention for youth who exhibit early initiation of drug use behavior.
Strengths and Limitations
The major advantage of the life course perspective for health research is in the area of epidemiology of health and illness. Researchers are applying the life course perspective to understand group differences in health and how health behaviors develop and persist across time. The life course perspective emphasizes the accumulation of threats for individuals exposed to persistent disadvantage over time and how these threats can negatively affect health. This focus provides etiologic insights about the developmental processes that generate health disparities.
However, many existing studies have statistical limitations. Despite their methodological sophistication and public health relevance, many observational (or quasi-experimental) studies have been limited by their inability to make causal conclusions about the associations between contextual risk factors and health behaviors because of possible confounder effects. To address this limitation, many studies incorporate a broad array of possible confounder variables into the statistical models as predictors (i.e., control variables) using traditional regression, which may result in limited statistical power.
Under such circumstances, inverse propensity weight (IPIV) can be used. Conceptually, IPIV is similar to using survey weights in an analysis. IPIV is the probability of assignment to, or membership in, a specific group, conditional on a set of observed confounder variables (Rosenbaum & Rubin, 1983). For example, in IPIV, individuals with a low probability of having a reported level of contextual risk factors, given their levels of confounders, are up-weighted, and individuals with a high probability of having a reported level of contextual risk factors, given their levels of confounders, are down-weighted. Thus, the weighted sample mimics a randomized sample where individuals are randomly assigned to levels of contextual risk factors and confounders are evenly distributed between individuals with different levels of contextual risk factors. IPW allows researchers to adjust the data for confounding variables in the absence of randomization, assuming that all possible confounders are measured (Rosenbaum & Rubin, 1983). In addition, IPIV allows researchers to control fora larger, more diverse array of potential confounders than traditional regression methods, and the ability to more accurately control for these confounders increases the confidence that can be placed in the findings and enhances the ability to draw causal inferences.
Recommendations for Future Research
Several important extensions to the analytical approaches presented earlier are important to emphasize for future research. First, more research should investigate the existence of underlying global risk, such as youth health risk lifestyle (Lawrence, Mollborn, & Hummer, 2017; Wickrama, Conger, Wallace, & Elder, 1999), rather than focusing on specific health behaviors. Such global factors can be captured under latent constructs defined by multiple health risk behaviors as observed indicators. Second, because socioeconomic context can change over time, it is important to investigate the effects of time-varying socioeconomic context on health behavior trajectories. For example, trajectories of family economic hardship may influence adolescent health behavior trajectories. Such investigations can uncover parallel processes between elements of the socioeconomic context and health risk behaviors. Similarly, more research is needed to identify parallel processes between specific health risk behaviors (e.g., substance use and eating a poor diet). Such investigations can identify the extent of longitudinal comorbidity between health risk behaviors. Third, continued research is needed focused on the onset timing of health risk behaviors, such as substance use and risky sex. In order to investigate these event outcomes, categorical regression can be utilized predicting the onset of these risky health behaviors as logistic coefficients or odds ratios. More importantly, the onset timing of health risk behaviors may be influenced by socioeconomic factors, and this possibility can be investigated using survival models.
