Exhibiting caution with use of big data: The case of amphetamine in Iceland's prescription registry

Data from large electronic databases are increasingly used in epidemiological research, but golden standards for database validation remain elusive. The Prescription Registry (IPR) and the National Health Service (NHS) databases in Iceland have not undergone formal validation, and gross errors have repeatedly been found in Icelandic statistics on pharmaceuticals. In 2015, new amphetamine tablets entered the Icelandic market, but were withdrawn half a year later due to being substandard. Return of unused stocks provided knowledge of the exact number of tablets used and hence a case where quality of the data could be assessed.

A case study of the quality of statistics in a national database on pharmaceuticals.

Data on the sales of the substandard amphetamine were obtained from the Prescription Registry and the pharmaceuticals statistics database. Upon the revelation of discrepancies, explanations were sought from the respective institutions, the producer, and dose dispensing companies.

The substandard amphetamine was available from 1.9.2015 until 15.3.2016. According to NHS, 73990 tablets were sold to consumers in that period, whereas IPR initially stated 82860 tablets to have been sold, correcting to 74796 upon being notified about errors. The producer stated 72811 tablets to have been sold, and agreed with the dose dispensing companies on sales to those. The producer's numbers were confirmed by the Medicines Agency.

Over-registration in the IPR was 13.8% before correction, 2.7% after correction, and 1.6% in the NHS. This case provided a unique opportunity for external validation of sales data for pharmaceuticals in Iceland, revealing enormous quality problems. The case has implications regarding database integrity beyond Iceland.