Table 1.
Study sample characteristics (N = 68,400).
Fig 1.
Interpreting IRT item characteristic curves (ICC) as defined by medical diagnoses for two example patients from the same subgroup.
This figure illustrates fundamental aspects of IRT models using two hypothetical patients from the same subgroup. Medical conditions and persons are each given estimates which place them on the construct, medical complexity; estimates represent amount of complexity captured by the medical condition or person. Here, the two patients represent low and high amounts of complexity and consequently, which medical conditions they were likely to have given their levels of complexity.
Table 2.
Fit indices of mixture distribution IRT models: 1–7 class solution based on 29 medical conditions (N = 68,400 patients).
Fig 2.
Prevalence of medical conditions by subgroup (prevalence rates <15% not labeled).
This bar chart shows the prevalence of selected medical condition used to help label the subgroups. Not all of these medical conditions were included in the final models because they either were too prevalent (> 95%) or not prevalent enough (< 5%), but were presented here descriptively to help identify the nature of the groups.
Fig 3.
Item characteristic curves showing probability of having a medical condition as a function of amount of medical complexity and patient subgroup.
As previously demonstrated in Fig 1; a medical condition in one group may reside at the lower end of the complexity continuum, while in another group, the same condition represents extreme complexity.
Table 3.
Descriptive demographic and utilization information within subgroups, %(n).
Table 4.
Predicted probabilities (pp), with 95% confidence intervals, indicating the likelihood of the classes being in the listed categories of demographic and service utilization factors, based on multinomial logistic regression analyses (N = 62,579)1.