Holly Hedegaard

Improving National Data Systems for Surveillance of Suicide-related Events

BACKGROUND Describing the characteristics and patterns of suicidal behavior is an essential component in developing successful prevention efforts. The Data and Surveillance Task Force (DSTF) of the National Action Alliance for Suicide Prevention was charged with making recommendations for improving national data systems for public health surveillance of suicide-related problems, including suicidal thoughts, suicide attempts, and deaths due to suicide. PURPOSE Data from the national systems can be used to draw attention to the magnitude of the problem and are useful for establishing national health priorities. National data can also be used to examine differences in rates across groups (e.g., sex, racial/ethnic, and age groups) and geographic regions, and are useful in identifying patterns in the mechanism of suicide, including those that rarely occur. METHODS Using evaluation criteria from the CDC, WHO, and the U.S.A.-based Safe States Alliance, the DSTF reviewed 28 national data systems for feasibility of use in the surveillance of suicidal behavior, including deaths, nonfatal attempts, and suicidal thoughts. The review criteria included attributes such as the aspects of the suicide-related spectrum (e.g., thoughts, attempts, deaths) covered by the system; how the data are collected (e.g., census, sample, survey, administrative data files, self-report, reporting by care providers); and the strengths and limitations of the survey or data system. RESULTS The DSTF identified common strengths and challenges among the data systems based on the underlying data source (e.g., death records, healthcare provider records, population-based surveys, health insurance claims). From these findings, the DSTF proposed several recommendations for improving existing data systems, such as using standard language and definitions, adding new variables to existing surveys, expanding the geographic scope of surveys to include areas where data are not currently collected, oversampling of underrepresented groups, and improving the completeness and quality of information on death certificates. CONCLUSIONS Some of the DSTF recommendations are potentially achievable in the short term (<1-3 years) within existing data systems, whereas others involve more extensive changes and will require longer-term efforts (4-10 years). Implementing these recommendations would assist in the development of a national coordinated program of fatal and nonfatal suicide surveillance to facilitate evidence-based action to reduce the incidence of suicide and suicidal behavior in all populations.

PA 20-2-1227 Using small area estimation techniques to examine geographic trends in suicide rates in the united states

Background Since 2008, suicide has ranked as the 10th leading cause of death in the United States (U.S.), with nearly 45 000 suicides occurring in 2016. Suicide rates vary by state, with higher rates in the West. To target prevention efforts, a detailed understanding of geographic variation is needed, however interpreting county-level suicide rates can be challenging because rates based on small numbers can be unstable and highly variable year-to-year. Objective To use small area estimation methods to generate stable estimates of annual county-level suicide rates for 2005 through 2015 and to examine variation across counties by geography and over time. Methods Suicides were identified from National Vital Statistics System Mortality Data using ICD-10 underlying cause codes U03, X60-X84, Y87.0. Hierarchical Bayesian models were used to estimate suicide rates for 3140 counties. Models included time-varying county-level covariates representing risk factors for suicide (demographic, socioeconomic and health- and crime-related characteristics) and included terms to account for spatial and temporal dependence. Model-based estimates were mapped to explore geographic and temporal patterns and examine urban-rural differences. Findings From 2005 to 2015, model-based county-level suicide rates increased by at least 10% for 99% of counties, with 87% of counties showing increases of 20% or more. Counties with the highest model-based rates were consistently located across the western and northwestern U.S.; geographic patterns did not change over time. Rural counties had higher estimated rates than urban counties, and saw the largest increases in rates during the study period. Conclusion Small area estimation methods can be used to overcome many of the challenges associated with examining geographic variation in suicide rates at a more granular level. Policy implications Maps of model-based estimates can help target prevention efforts both within and across state boundaries, and inform research on community-level risk and protective factors for suicide mortality.

27 Development and testing of surveillance case definitions and reporting frameworks to standardise the use of ICD-10-CM coded data for injury surveillance, epidemiology and research

Purpose The National Centre for Health Statistics and the National Centre for Injury Prevention and Control have proposed surveillance case definitions and reporting frameworks for analysis of injury data coded using the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM). This presentation describes recent collaborative efforts to test the proposed definitions and the framework for reporting external cause (injury mechanism by intent). Methods 1) Analysis of administrative data coded in ICD-10-CM, with comparison to historic results from data coded in ICD-9-CM, and 2) analysis of dual coded data (i.e., same medical record coded in both ICD-9-CM and ICD-10-CM). Results Using dual-coded data, the proposed ICD-10-CM definition identified more injury hospitalizations than the ICD-9-CM definition. 95% of the hospitalizations identified using the ICD-10-CM definition but not the ICD-9-CM definition were ‘subsequent’ encounters. Comparison of results from the proposed ICD-10-CM external cause matrix to historic results using the ICD-9-CM matrix showed a decrease in the number of cases assigned to Unintentional Fall, Motor Vehicle Traffic (MVT)- Pedestrian, Other Pedalcyclist, Other Transportation, Overexertion and Undetermined Intent, while increases were seen in the number of cases assigned to Unintentional MVT-Pedalcyclist, Other Pedestrian and Unspecified Mechanism. Conclusions Analysis results will be used to finalise the ICD-10-CM surveillance case definitions and the external cause matrix by deciding whether to: 1) include subsequent encounters, 2) limit case inclusion to a principal diagnosis of injury only or to include records with mentions of injury in other diagnosis fields, 3) include records based on presence of an external cause code even if the principal diagnosis is not an injury, and 4) whether re-assignment of codes to specific cells of the external cause matrix is needed. Significance Use of finalised ICD-10-CM injury surveillance case definitions and reporting frameworks will help standardise the comparison of results across jurisdictions and time.

378 International comparisons of drug-related deaths

Background In the past decade, many countries have seen a significant increase in their drug-related mortality rates. In the United States (US), drug-related deaths now outnumber deaths from any other injury cause. To better understand international differences, this study compared drug-related deaths in the US, England/Wales, Scotland and Australia, examining trends, demographic characteristics and differences in the drugs involved. Methods Drug-related deaths were identified using public access data and reports from each country’s statistical office. Cases were selected based on ICD–10 underlying cause codes of F11–16, F18–F19 (Drug abuse), X40–X44 (Accidental poisoning), X60–X64 (Intentional self-poisoning), X85 (Assault by drugs) and Y10–14 (Drug poisoning of undetermined intent). Results In 2013, the rate of drug-related deaths in the US (146 per million population) was 1.5 times the rate in Scotland, twice the rate in Australia and more than 3 times the rate in England/Wales. In all countries, rates were higher for males than for females, with the greatest gender difference seen in Scotland. For underlying cause, in all countries, the majority of the deaths were accidental, however in England/Wales a higher percent were intentional (33%), in Australia a higher percent had a mental/behavioural cause (15%) and in Scotland a higher percent were categorised as undetermined intent (17%). In all countries, opioids including morphine, heroin and methadone were implicated in a high percent of the deaths, although drug-specific comparisons were limited due to variation by country in the completeness of the information on specific drugs. Conclusions While similar patterns in drug-related deaths were identified, differences were also seen. The extent to which these differences are true or due to variation in death investigation, reporting and coding is unclear. Further work is needed to enhance the international comparability of mortality data on drug-related deaths.