BACKGROUND: Although Marin County consistently ranks as the healthiest county in California, it has higher drug poisoning mortality rates than most California counties. In response to this growing epidemic, Marin County developed a drug poisoning mortality surveillance system to inform local programs and policies. We evaluated this surveillance system in Marin County, California from 2007-2013 to determine if the system was adequately describing the burden of drug poisoning deaths given existing data sources.
METHODS: The surveillance system was evaluated according to the CDC’s 2001 Updated Guidelines for Evaluating Surveillance Systems. Local data were analyzed from: Death Statistical Master Files (DSMF), containing primary cause and manner of death (intentional, unintentional, or undetermined), and Multiple Cause of Death (MCOD) Files, containing contributing causes of death and associated pharmaceuticals. Data quality was assessed by comparing death counts between DSMF and MCOD data. Key stakeholders were interviewed via structured questionnaires to assess the system’s simplicity, timeliness, and acceptability. Stability was assessed by evaluating how manner of death changed over time. California counties were queried through a state-wide data managers group to compare case definitions for drug overdose deaths.
RESULTS: The number of reported deaths varied between data sources; from 2007-2011, DSMF data contained 190 drug poisoning deaths compared to 198 deaths identified through MCOD files. The stakeholder survey (n=9) revealed: 78% believed existing data did not adequately monitor trends to inform policies, 67% believed data were not easy to interpret, and 33% believed the system released data in a timely way. The proportion of drug poisonings with an undetermined manner of death decreased from 40% in 2008-2010 to 1% in 2011-2013, corresponding to a change in personnel involved in the death classification process. Five out of eight counties who responded to the case definition query were conducting drug poisoning mortality surveillance; of those, 80% used the same case definition as Marin to identify drug poisoning deaths.
CONCLUSIONS: The system meets basic objectives, however, has several weaknesses reflective of an emerging surveillance system. Stakeholders’ subjective concerns regarding accuracy corresponded with our objective findings. Data quality might be enhanced by routine cross-validation between data sources. The system would gain stability through a standardized process of classifying deaths applied uniformly across time and personnel. A future direction may be evaluating a new mortality data system implemented in 2015 for improved timeliness, data quality, and completeness.