Clustering of opioid prescribing and opioid-related mortality among family physicians in Ontario

METHODS

We performed a population-based cross-sectional analysis among Ontarians aged 15 to 64 on January 1, 2006, who were eligible for prescription drug coverage under the Ontario Public Drug Program, a population in which the prevalence of opioid prescribing is high and in which high-dose prescribing (eg, 200 mg/d of morphine or equivalent) is common.6

We categorized physicians into quintiles according to the rate of their opioid prescribing in 2006. To determine the opioid-prescribing rate for each physician, we divided the number of opioid prescriptions dispensed to public drug plan beneficiaries by the number of public drug plan beneficiaries aged 15 to 64 treated by each physician. We excluded methadone because it is prescribed primarily for opioid addiction rather than pain. We included all opioids prescribed by physicians whose field of practice was either “general practice” or “family practice” in the physician database of the Institute for Clinical Evaluative Sciences. We excluded physicians who practised outpatient palliative care.

We also examined the records of the Office of the Chief Coroner of Ontario to identify deaths related to opioid use.1 The coroner, a licensed physician, ascertains the cause of all deaths that are either sudden and unexpected or unnatural. A postmortem examination, generally including detailed toxicologic testing, is conducted under the authority of the coroner when necessary to determine the cause or manner of death. From these records, we extracted detailed data about deaths in which opioids appeared to play a contributory role. A single abstractor collected data using an encrypted computer database. For each case, the cause and manner of death and the results of toxicologic analysis (including the concentration and specimen type for positive opioid and alcohol results) were recorded. Deaths were deemed to be opioid-related if that was the opinion of the investigating coroner. In almost all instances, this determination was made for 1 of 2 reasons: the toxicologic analysis revealed opioid concentrations sufficiently high to cause death, or the coroner determined that a combination of drugs (including at least 1 opioid present at a clinically significant concentration) collectively resulted in death. If another drug (eg, a cyclic antidepressant) was present at a concentration high enough to cause death and 1 or more opioids were present at levels that could be associated with therapeutic use, the death was not deemed to be related to opioid use. If the abstractor was uncertain whether a death was related to opioid use or not, the file was reviewed by 1 or more physician coinvestigators until consensus was achieved.

For each patient with an opioid-related death in 2006 who received at least 1 prescription for an opioid in the year before death, the final antemortem opioid prescription was assigned to the volume quintile of the prescribing family physician.

To test the hypothesis that the number of deaths increased across prescribing-volume quintiles, we performed a Cochran-Armitage test for trend. We also used bivariate regression to see if particular physician characteristics were associated with higher volumes of opioid prescribing.

This study was approved by the Research Ethics Board of the Sunnybrook Health Sciences Centre in Toronto.