To develop and validate the novel measure, we conducted our analysis in three steps: (1) derived the marginal effect of financial burden using multivariable logistic regression predicting unmet need with bivariate splines for income and OOPC, adjusting for covariates; (2) performed recursive partitioning on the marginal effect of financial burden to identify cut-points in income and OOPC; and (3) assessed the performance of the novel measure for predicting unmet need, compared to the standard 10 percent cut-point.
Using a bivariate tensor product spline, which modeled a joint function of OOPC and income, allowed the relationship between OOPC, income, and unmet need to vary separately by the value of OOPC and income, similar to real-world conditions (e.
continuous variable output from the spline models) could be categorized by distinct thresholds in income and relative OOPC (i.
Models with a bivariate tensor product spline for income and OOPC supported the presence of effect dependence between income and OOPC (i.
First, measures of financial burden that rely on one-dimensional assessment of burden (such as absolute OOPC alone) may fail to account for the various components that contribute to the latent construct of financial burden.
In addition, most existing research has relied, almost entirely, on arbitrary OOPC cut-points for financial burden without any empirical evidence that these cut-points can adequately capture the true underlying threshold effect (Merlis 2002; Banthin and Bernard 2006; Banthin, Cunningham, and Bernard 2008; Collins et al.
The proposed novel measures may be a promising new tool for capturing a more nuanced definition of financial burden for families with children because it relies only on total family income and total OOPC and can be calculated independent of insurance status.
OOPC are a modifiable aspect of the health care system and, as such, policies that address excessive OOPC will be a key component of targeted efforts to improve health care access for families.
Ultimately, our novel measure of financial burden could be used to identify families who are at risk of delayed or forgone care, and assist policy makers in devising and applying policies to ensure that families of varying income do not incur OOPC that could lead to unmet health care need, or adverse health outcomes.
Further, OOPC were imputed and subject to potential misclassification; however, the robustness of the observed association suggests that our results may be invariant to this potential misclassification.