Clinical and socioeconomic predictors of hospital use and emergency department visits among children with medical complexity: A machine learning approach using administrative data

Fig 1. Relative variable importances for the binary submodels. Relative importance is given in parentheses. Clinical and socioeconomic variables are indicated with red and blue, respectively. An asterisk (*) denotes a DA-level measure. Abbreviations: 1Parent, proportion of single-parent families; Alone, proportion of individuals living alone; Cardio, cardiology condition indicator; Class, single or multiple complex chronic conditions; CM, number of chronic medications; Congen, congenital/genetic condition indicator; Employ, employment rate; GI, gastrointestinal condition indicator; H/I, hematology/immunodeficiency condition indicator; MatDep, material deprivation score; Mali, malignancy condition indicator; Meta, metabolic condition indicator; MMHS, mother mental health service use; NI, neurological impairment indicator; NoHS, proportion of individuals without a high school diploma; Renal, renal condition indicator; Resp, respiratory condition indicator; SDW, proportion of individuals who are separated, divorced, or widowed; SocDep, social deprivation score; TA, technology assistance indicator.

Fig 2. Relative variable importances for the conditional submodels. Relative importance is given in parentheses. Clinical and socioeconomic variables are indicated with red and blue, respectively. An asterisk (*) denotes a DA-level measure. Abbreviations: 1Parent, proportion of single-parent families; Alone, proportion of individuals living alone; Cardio, cardiology condition indicator; Class, single or multiple complex chronic conditions; CM, number of chronic medications; Congen, congenital/genetic condition indicator; Employ, employment rate; GI, gastrointestinal condition indicator; H/I, hematology/immunodeficiency condition indicator; MatDep, material deprivation score; Mali, malignancy condition indicator; Meta, metabolic condition indicator; MMHS, mother mental health service use; NI, neurological impairment indicator; NoHS, proportion of individuals without a high school diploma; Renal, renal condition indicator; Resp, respiratory condition indicator; SDW, proportion of individuals who are separated, divorced, or widowed; SocDep, social deprivation score; TA, technology assistance indicator.

Fig 3. Marginal associations (and 95% confidence intervals) for initial LOS in the binary (left) and conditional (right) submodels in the year following initial discharge. Estimates for hospital days are indicated with solid lines and those for ED visits with dotted lines. The rugs on the horizontal axes indicate the first through third quartiles of initial LOS in the analytic sample (given numerically in Table 1). Abbreviations: ED, emergency department; LOS, length of stay.

Fig 4. Marginal associations (and 95% confidence intervals) for age at initial admission in the binary (left) and conditional (right) submodels for hospital days in the year following initial discharge. Estimates for CMCs with multiple CCCs are indicated with solid lines and those for a single CCC with dotted lines. The rugs on the horizontal axes indicate the first through third quartiles of age at initial admission in the analytic sample (given numerically in Table 1). Abbreviations: CCC, complex chronic condition; CMC, child with medical complexity.