Pediatric cancer in the United States: A comparison of three data sources

We used data from the SEER–17 registries, which encompass 26.5% of the US population; from the ConcertAI (CAI) RWD360® dataset, a rw dataset of > 6 million oncology pts from across the US from electronic health record (EHR) systems; and Optum’s de-identified Market Clarity Data (Optum® Market Clarity) (OMC), a US EHR and linked claims data source. For each source, demographics, clinical characteristics, and survival were evaluated among children and adolescents (0-19 yrs) diagnosed between 2015-2020 (inclusive) with any of 3 relatively more common pediatric cancer types: acute lymphocytic leukemia (ALL), acute myeloid leukemia (AML), and central nervous system tumors (CNS-T). Descriptive statistics and Kaplan-Meier 1-year overall survival were calculated.

Total diagnosed pediatric cases for ALL, AML, and CNS-T were 4,886, 1,249, and 4,016 from SEER; 349, 56, and 332 from CAI; and 1,204, 221, and 6,488 from OMC, respectively. Higher proportions of younger children (0-4 yrs) were observed in SEER; 16-21% of CAI pts and 26-31% of OMC pts had unknown/other information on race. Compared to SEER, CAI and OMC included a lower proportion of Asian pts (4-5% and 3% vs 8-13%) and a similar proportion of Black pts (5-21% and 6-10% vs 7-15%). Geographic location by US Census regions also varied with overrepresentation of the West in SEER (60%) and Midwest in CAI (29%) and OMC (40%) and underrepresentation of the Midwest in SEER (4%) and Northeast in CAI (11%) and West in OMC (13%). Survival at 1-yr from dx for ALL, AML, and CNS-T was 95.6% (95% CI: 94.9-96.1%), 81.6% (79.2-83.7%), and 87.1% (85.9, 88.1%) in SEER; 98.5% (96.5-99.4%), 90.9% (79.6-96.1%), and 94.5% (91.4-96.5%) in CAI; and 98.1% (97.1-98.7%), 92.4% (87.9-95.3%), and 95.7% (95.2-96.2%) in OMC, respectively. In SEER, ALL pts had progressively worse survival as age group at dx increased (p-trend < 0.01).

SEER-17 captured > 10 times as many diagnosed cases as CAI in each of the 3 pediatric cancers while SEER had > 4 times as many ALL and AML cases but fewer cases of CNS tumors than OMC suggesting an error in how CNS tumors are diagnosed in OMC data. Although the depth of clinical data available from CAI and OMC was greater, patient race and ethnicity were not always available. Across the three sources, 1-yr survival was higher for all cancers in CAI and OMC compared to SEER. Choice of data source may be determined by the specific research question of interest including the need for relatively large patient numbers versus in-depth clinical/treatment information.