METHODS
The New York City Poison Control Center provides toxicology consultation services to a population of approximately 12 million people. It maintains an electronic database of all cases dating back to Jan 1, 2000. Cases are coded for basic demographics, acuity, clinical effects, the treatments recommended and provided, and medical outcome, all using standardized fields in Toxicall (Computer Automated Systems, Aurora Colorado) and in accordance with case definitions created by the American Association of Poison Control Centers. Case records also include a free-text narrative that describes the clinical course and additional diagnostic studies. By routine, these narratives are updated until the patient’s clinical course plateaus, the patient is discharged, or the patient is transferred to psychiatric care. We performed a Structured Query Language (SQL) search for all cases of human lithium poisoning in which hemodialysis was coded as either recommended and/or performed. It is important to note that if the poison center was consulted regarding the indications for hemodialysis, the decision was made by the on-call consultant. After 2015, the EXTRIP criteria may or may not have been applied by the individual consultants but were never formally adopted as a poison center protocol.
The study protocol was preliminarily reviewed by our IRB and deemed exempt from comprehensive review due to absence of potential harm to research subjects. The database was searched from its inception (1/1/2000) through 5/24/20. Cases were then manually reviewed by one of two authors (PD, EF), and data were extracted using a predetermined form. Data were collected and managed using REDCap electronic data capture tools.[8],[9] The data collection form was piloted and reviewed, coding rules for subjective variables were agreed upon by abstractors a priori. Abstractors were not blinded to study objectives. Cases met inclusion criteria if there was a documented Li concentration >1.2 mmol/L and new neurological symptoms were recorded.
For each case record, the following information was collected: date; age; sex; initial and peak lithium concentrations ([Li]) if multiple concentrations were provided; acuity; coingestions; presence or absence of neurological symptoms, dysrhythmias, seizures, coma, or confusion; creatinine if provided; whether there was a either a documented [Li] >1 mmol/L after 36 hours or a documented [Li] <1 mmol/L before 36 hours; hemodialysis recommendations and additional treatments or GI contamination including IV fluids, activated charcoal, and whole bowel irrigation; whether or not of hemodialysis was performed; and medical outcome, graded as resolved, improved, unchanged, worsened, or not reassessed. For cases in which the coded data conflicted with the narrative account, the narrative account was taken as authoritative.
Cases were then evaluated to determine if they met criteria established by the EXTRIP workgroup sufficient to “suggest” or “recommend” hemodialysis and whether or not they met Paris criteria (Tables 1). Calculations of the eGFR by the CKD-EPI formula were not adjusted for race, as race is not routinely collected in our database. Predicted lithium concentration at 36 hours after presentation was determined to be a subjective variable not coded or routinely collected in our database, and thus was not used to make determinations of EXTRIP criteria applicability.