Lead is a potent neurotoxin, with young children especially vulnerable to its toxic effects. There is renewed interest in childhood lead poisoning, including updated estimates of its costs. Administrative claims data are a popular source of information on healthcare costs and utilization, but billing diagnosis codes may have limited accuracy. We used linked claims and laboratory test results to assess the accuracy of using ICD-9 and ICD-10 diagnosis codes for lead toxicity in claims to identify children with elevated blood lead (BPb).
We identified privately insured children ages <72 months who had a BPb test result of 0-200 μg/dL from 1/1/2013 – 6/30/2017 and were continuously enrolled for 179 days after the test using BPb test LOINC codes (5671-3, 77307-7, and 10368-9) in the IBM MarketScan Laboratory database. We dichotomized results into “elevated” (5-200 μg/dL) and “not-elevated” (0-4.9 μg/dL). We calculated the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of lead toxicity diagnosis codes (ICD-9: 984.x, E861.5, E866.0; ICD-10: R78.71, T56.0X1x, and T56.0X4x). We used an 89-day follow-up period for confirmatory testing of children with an initial elevated test. We considered two elevated test results as confirmatory of elevated BPb. We searched for lead toxicity diagnosis codes within 179 days of the most recent test.
Among 24,652 children age <72 months tested, 195 were continuously enrolled for 179 days and re-tested within 89 days. Of 79 children with an elevated BPb result, 33 (41.8%) had a second elevated result. The PPV of lead toxicity diagnosis codes was 80.0% (11/12), sensitivity was 33.3% (11/33), specificity was 99.4% (161/162), and the NPV was 88.0% (161/183).
Relying on stand-alone claims data may under-identify children with elevated BPb. Linking medical records to claims data, e.g., state Medicaid data, may yield more accurate results.
© 2020 Published by Elsevier Inc.