Comparing Complex Chronic Conditions Versions to Identify Children With Medical Complexity

The identification of children with medical complexity (CMC) from administrative data sources is essential to understanding illness severity, morbidity, mortality, and use patterns in this population.¹ CMC have severe chronic conditions that require multiple subspecialties, lead to functional and high-impact family limitations, and are often associated with medical technology dependence.² Identifying CMC using operational definitions is important given their contribution to pediatric expenditures.³ The open-source complex chronic condition (CCC) classification system, published in 2000, and updated to version 2 in 2014, is frequently used for this purpose.^1,4 The CCC system uses diagnoses and procedure codes to identify children with medical conditions that are expected to at least 12 months and affect multiple organ systems or one system severely enough to require specialty pediatric care and likely hospitalization in a tertiary care center.^1,4 Another revision to the CCC (version 3) incorporated new, missing, and retired codes, and built a separate classification for technology dependence.⁵ 

The report that published version 3 indicated that a similar proportion of children met versions 2 and 3 CCC criteria in the Pediatric Health Information System (PHIS) and the Medicaid Merative MarketScan Research Databases but did not describe their overlap or investigate how versions may compare when CMC receive care in emergency departments (EDs) or intensive care units (ICUs).⁵ Understanding this overlap has implications for research interpretation and health system planning, including in ensuring applicability of research findings to present-day contexts and for resource allocation. We evaluated differences in case identification of the CCC versions 2 and 3 among children receiving care in children’s hospitals.

We performed a retrospective study using PHIS, a multi-institutional pediatric data repository that provides anonymized clinical and administrative information from participating US children’s hospitals. We included all ED and inpatient encounters in 2023 for patients aged under 21 years from 41 hospitals that contributed consistently during this time. We evaluated the percentage of encounters meeting versions 2 and 3 criteria for CCC and their overlap among ED, inpatient, and intensive care unit (ICU) encounters. We compared demographics, costs, and individual CCC categories for each criterion.⁶ This study was approved by our institutional review board.

The ED, inpatient, pediatric, and neonatal intensive care unit samples respectively included 3 385 937, 561 744, 105 434, and 41 833 encounters. The percentages of encounters meeting version 3 criteria were smaller than those meeting version 2 criteria. The overlap between versions 2 and 3 ranged from 86% to 95% (Figure 1). Among children with at least 1 CCC, individual categories of CCC occurred in similar proportions between versions but with varying overlap (Supplementary Figure 1; Table 1). Children meeting both version 2 and 3 criteria had higher mortality than those only meeting only one criterion. Among ED encounters only meeting CCC version 3 criteria, the neuromuscular (45.7%) category occurred most frequently, and 13.5% had concurrent technology dependence. Median costs were usually higher among encounters meeting both criteria compared with those meeting one criterion (Supplementary Tables 1–4).

We demonstrate substantial overlap between the CCC versions 2 and 3 across multiple settings. Our findings, which compare different versions of CCC, differ from prior work that has compared CCC rubrics with other operational rubrics.⁷ 

Our findings expand upon prior work by specifically evaluating the overlap of criteria and their individual categories within different settings. Feinstein et al reported that similar percentages of patients met CCC criteria among patients admitted to the hospital (38.3% in version 3 vs 40.1% in version 2)⁵ and added information for how to interpret patient samples across CCC versions. We add to their work by demonstrating that the degree of overlap within individual CCC categories differs widely. Version 3 assigns technology assistance codes when patients also have a diagnosis for another organ system CCC group (eg, respiratory, cardiovascular) rather than having technology assistance assigned as a separate grouping. This approach ensures that technology dependence alone does not automatically assign a CCC status and provides information regarding the organ system the technology supports not evident in prior versions.

Limitations of the present analysis include the use of retrospective data and the use of data from only children’s hospitals, which may be less generalizable. Nevertheless, these findings have implications in interpreting and conducting research involving CMC presenting to acute care settings.