Premium
Comparison of dispensed medications and forensic‐toxicological findings to assess pharmacotherapy in the Swedish population 2006 to 2013
Author(s) -
Forsman Jonas,
Taipale Heidi,
Masterman Thomas,
Tiihonen Jari,
Tanskanen Antti
Publication year - 2018
Publication title -
pharmacoepidemiology and drug safety
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.023
H-Index - 96
eISSN - 1099-1557
pISSN - 1053-8569
DOI - 10.1002/pds.4426
Subject(s) - medicine , medical prescription , pharmacoepidemiology , drug , concordance , population , forensic science , kappa , confidence interval , pharmacotherapy , emergency medicine , psychiatry , pharmacology , environmental health , linguistics , philosophy , veterinary medicine
Purpose To investigate person‐level agreement between medication exposure as predicted using the PRE2DUP (a prescription‐based design to estimate continuous drug use) method and postmortem toxicological findings, in the Swedish population during the years 2006 to 2013. Methods Using the Swedish National Board of Forensic Medicine's toxicology database and the Swedish National Board of Health and Welfare's registries on causes of death, dispensed medications, and in‐patient care, forensic‐toxicological findings were compared with prescription‐based estimates of drug use for 27 medications. We modeled expected drug‐use periods with the PRE2DUP using an algorithm of demonstrated high validity that evaluates personal drug‐purchasing patterns with consideration to possible stockpiling of drugs and package information. Excluding criteria included self‐inflicted death and recent in‐patient care. Results In data from 18 627 performed autopsies, as well as 10 160 instances of dispensed drug use, the agreement between PRE2DUP drug‐use periods and forensic toxicology was, overall, moderate (Cohen's kappa: 0.56 [95% confidence interval {CI}: 0.55‐0.57]) with a positive predictive value, or predicted adherence rate, of 46.0%. The group‐level predicted adherence and agreement were highest for antidepressants, at 71.0% (Cohen's kappa: 0.74 [CI: 0.73‐0.76]), and lowest for cardiovascular drugs, at 21.5% (Cohen's kappa: 0.33 [CI: 0.31‐0.36]). Predicted recreational use (negative predictive value) was low for all investigated drugs (0.0%‐1.4%). The biological half‐life explained 29% ( P = 0.003) of the variability of the false‐positive rate. Conclusions Measured agreement between PRE2DUP‐based drug‐use estimates and forensic‐toxicological findings is dependent upon a number of factors, including true continuous drug use and postmortem detectability of the investigated drugs, as well as the occurrence of unconventional dosing and true non‐adherence.