Independent Variables of Interest

The measure of social integration is adapted from the Berkman-Syme Social Network Index (SNI) [³⁹], which measures the number of social ties across four domains: marital status, friends or relatives, church attendance, and membership in social organizations. Prior research found the Berkman-Syme SNI to have predictive validity [^{39, 40}] and the index has been adapted for use across numerous epidemiological cohort studies (e.g., National Health and Nutrition Examination Survey [³¹], Framingham Heart Study [²⁹], and Nurses’ Health Study [⁴¹]), including Add Health [²⁸]. The social integration index was measured for this study with the following four items and cut-points consistent with prior research using the Add Health data [²⁸] to facilitate comparisons of findings: currently married/cohabitating, attended church ≥ 12 times in the past 12 months, volunteered once or more in the past 12 months, and reported six or more close friends. One point was given if the participant was affirmative for the item, the items were summed, and the index was standardized.

Data were available in wave 4 of Add Health on the participants’ perceived happiness in their current romantic relationship (regardless of relationship type) as well as their perceived closeness in their relationships with their mother figure and their father figure. As some participants reported not having a romantic partner or a mother or father figure, a multinomial variable was constructed to include those participants who reported not having a relationship of this type. Specifically, for the level of happiness in participants’ current romantic relationship, participants who reported no relationship and those who reported being very happy were compared to the referent category of participants who reported being only fairly happy or not at all happy with their current romantic relationship. Similarly, for perceived closeness to a mother figure and perceived closeness to a father figure, participants who reported having no relationship and those who reported having a somewhat, quite, or very close relationship were compared to the referent category of those participants who reported their relationship was not very close or not close at all. As a strained social relationship may be more detrimental to health than having no relationship, we selected the poor relationship quality as the referent category.