Modeling the types and timing of stress in pregnancy

Abstract

In the current study we examined whether or not stress at critical intervals during pregnancy, or stress consistently experienced across the course of pregnancy was associated with gestational age in a (large) multiethnic sample. After deriving a latent trait-state model of stress, we examined whether or not particular components of stress (i.e., perceived stress, general anxiety, pregnancy-specific anxiety), or stress in general, at specific time points or over time were associated with gestational age. Pregnancy-specific anxiety over the course of pregnancy was associated with shorter gestation after controlling for a number of risk factors, including history of diabetes, smoking, maternal age, and parity. Moreover, these findings suggest that the relation between pregnancy-specific anxiety and gestational age was similar across ethnic groups. The importance of modeling the components and timing of stress with latent variable methodology is discussed.

Keywords:

• Stress
• Modeling
• Latent variables
• Pregnancy
• Gestational age

Acknowledgments

Many thanks to Nancy Collins, Regan Gurung and Chris Rini for their helpful comments on previous drafts of this manuscript. The research was supported by NICHHD grants from the National Institute of Health (R01 HD29553 and R01 HD 28413) to Calvin Hobel, M.D. and Christine Dunkel-Schetter, Ph.D. The first author was supported by a Health Psychology Training Grant from the National Institute of Mental Health (MH 15750). Please address correspondence with regards to this article to either Scott C. Roesch, Ph.D., Department of Psychology, San Diego State University, 5500 Campanile Drive, San Diego, CA 92182-4611, [email protected] or Chris Dunkel Schetter, Ph.D., Department of Psychology, UCLA, Los Angeles, CA 90095-1563.

Notes

The term trait is not used in its traditional sense here. It does not refer to a stable personality characteristic as typically espoused by personality theorists. Rather, trait in this context refers to the cross-situational consistency with which one has responded on a state measure.

See footnote 1.

Maternal age, years of education, and annual family income were entered into the model as continuous variables. Marital status, nulliparity, smoking, and history of diabetes were dummy coded and entered into the model. Smoking was significantly associated with STA-Time 2, r(417)=10, p<0.05) and STA-Time 3, r(417)=0.12, p<0.05, whereas a history of diabetes was significantly related to gestational age, r(417)=−0.19, p<0.05. Because of these relations, we controlled for these two medical risk factors in models predicting gestational age.

In order to completely partial out the impact of the seven control variables, one must also specify associations between the residuals of the stress factors and gestational age. We thank an anonymous reviewer for emphasizing this point.

Life events were measured in this study but were not modeled because of low correlations with the three other measures of stress (all rs <0.15). Moreover, these correlations varied by ethnicity further complicating the development of creating one latent stress factor with the four observed variables.

The Lagrange Multiplier test was used to evaluate the invariance of model parameters across groups. Probability levels greater than the traditional 0.05 were used as indicators that a target parameter was invariant across ethnic groups.

We thank an anonymous reviewer for bringing this important point to our attention.