Validation of Stevens-Johnson syndrome or toxic epidermal necrolysis diagnoses in the Clinical Practice Research Datalink.
Frey NBasel Pharmacoepidemiology Unit, Division of Clinical Pharmacy and Epidemiology, Department of Pharmaceutical Sciences, University of Basel, Basel, Switzerland.
Bircher AAllergology, University Hospital Basel, Basel, Switzerland.
Jick SSBoston Collaborative Drug Surveillance Program, Boston University School of Public Health, Lexington, MA, USA.
Meier CRBasel Pharmacoepidemiology Unit, Division of Clinical Pharmacy and Epidemiology, Department of Pharmaceutical Sciences, University of Basel, Basel, Switzerland.
Spoendlin JBasel Pharmacoepidemiology Unit, Division of Clinical Pharmacy and Epidemiology, Department of Pharmaceutical Sciences, University of Basel, Basel, Switzerland.
English
PURPOSE To evaluate the validity of recorded diagnoses of Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) in the Clinical Practice Research Datalink (CPRD).
METHODS We identified patients with a diagnosis of SJS or TEN between 1995 and 2013 in the CPRD. We reviewed information from patient records, free text, and hospital episode statistics (HES) data, and excluded patients with no indication of a secondary care referral. Remaining patients were classified as probable, possible, or unlikely cases of SJS/TEN by two specialised clinicians or based on pre-defined classification criteria. We quantified positive predictive values (PPV) for all SJS/TEN patients and for patients categorised as 'probable/possible' cases of SJS/TEN, based on a representative subsample of 118 patients for whom we had unequivocal information (original discharge letters or HES data).
RESULTS We identified 1324 patients with a diagnosis of SJS/TEN, among whom 638 had a secondary care referral recorded. Of those, 565 were classified as probable or possible cases after expert review. We calculated a PPV of 0.79 (95% CI, 0.71-0.86) for all SJS/TEN patients with a recorded secondary care referral, and a PPV of 0.87 (95% CI, 0.81-0.93) for probable/possible cases. After excluding 14 false positive patients, our study population consisted of 551 SJS/TEN patients.