Abstract
OBJECTIVE: To assess the validity of the International Classification of Diseases, 10th Revision (ICD-10) healthcare database diagnosis codes for lithium toxicity at hospital admission in Ontario, Canada. DESIGN: Population-based retrospective validation study. SETTING: A total of 152 hospitals linked to a provincial laboratory database in Ontario, Canada, from 2007 to 2023. PARTICIPANTS: Patients 50 years of age or older taking lithium with hospital-based serum lithium laboratory measurements during admission to the hospital (n=2804). OUTCOME MEASURES: Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) comparing an ICD-10 diagnostic coding algorithm for lithium toxicity to a serum lithium concentration of 1.5 mmol/L or more. The codes used in the algorithm were T568, T435, Y495, X41 and X49. Serum lithium values and changes in the concentration of serum lithium from baseline levels in patients with and without a diagnosis code for lithium toxicity (code-positive and code-negative, respectively). RESULTS: The sensitivity of the ICD-10 coding algorithm for identifying a serum lithium level≥1.5 mmol/L was 84% (95% CI 81% to 87%). The specificity and the NPV were over 88%, and the PPV was 63% (95% CI 60% to 66%). The median (IQR) serum lithium measurement in code-positive patients was 1.7 (1.2 to 2.2) mmol/L, and it was 0.6 (0.4 to 0.9) mmol/L in code-negative patients. The median (IQR) increase in serum lithium concentration compared with the most recent prehospital baseline values was 0.7 (0.2 to 1.3) mmol/L in code-positive patients and 0.0 (-0.2 to 0.2) mmol/L in code-negative patients. CONCLUSION: In Ontario, the sensitivity of the ICD-10 coding algorithms was moderate for identifying a serum lithium level≥1.5 mmol/L at hospital admission. The presence or absence of the ICD-10 codes for lithium toxicity at hospital admission successfully differentiated two groups of patients with distinct serum lithium measurements.