Background Pesticide ingestion is among the most commonly used suicide methods globally. Underestimation of deaths by pesticide self-poisoning is thought to be common but under-researched. We investigated potential misclassification and underestimation of suicide by pesticide poisoning and their impact on trends in pesticide suicide in South Korea.
Methods We compared sociodemographic profiles between deaths certified as suicide, undetermined deaths and injuries by poisoning using pesticides, unspecified chemicals and medicines/alcohol in South Korea (1991–2012), and calculated similarity scores. Joinpoint regression analysis was used to examine time trends.
Results The profiles of deaths classified as pesticide suicides showed generally similar patterns to those for undetermined deaths/injuries by poisoning using pesticides and suicides/undetermined deaths/injuries by poisoning using unspecified chemicals. Over the study period, the number of pesticide suicides (on average 2033 per year; range 623–3530) would increase by 15%, 23% and 31%, when assuming that 20%, 30% and 40% of deaths in alternative cause-of-death categories were misclassified pesticide suicides respectively. Age-standardised rates of pesticide suicide increased more than four times in 1991–2003 but fell by around 50% in 2003–2012. This trend did not change when the assumed proportion of misclassification was 20%, whereas the rapid increase in 2000–2003 would be lessened if misclassification occurred by 30% or more.
Conclusions Pesticide suicides may be underestimated in South Korea as some are possibly misclassified as deaths due to other causes; however, such underestimation would not substantially affect estimated secular trends.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Pesticide ingestion is among the most commonly used suicide methods globally. A systematic review of data for countries around world shows that approximately 30% of global suicides are due to pesticide self-poisoning, and many of these deaths occur in rural areas of South and East Asia.1 However, suicide is commonly under-reported or misclassified as death due to other causes.2 The level of under-reporting/misclassification appears to be more prominent for certain suicide methods such as self-poisoning, in particular pesticide ingestion.3 Nevertheless, the reliability of suicide statistics is understudied and there is a widespread lack of systematic assessment of national data.4 The accuracy of suicide data has important implications for the formulation and evaluation of suicide prevention strategies. Therefore, the WHO has called for more research into the quality of suicide data.5 ,6
In countries where the International Classification of Diseases (ICD) codes (eg, ICD, 10th Revision)7 are used to classify cause of death, suicides (ICD-10 X60–X84) are most commonly misclassified as deaths of undetermined intent (or ‘undetermined death’) (Y10–Y34), and also as injuries (V01–X59), homicides (X85–Y09) or unknown cause (R95–R99).2 In particular, suicide by self-poisoning (including pesticide ingestion) is more likely to be misclassified as undetermined death or injury than suicide from other methods such as hanging and firearm shooting.3 ,8 ,9 However, previous studies mainly used qualitative but not quantitative assessments when determining potential cause-of-death groups that may include misclassified suicides. Few studies have specifically examined pesticide suicide or the impact of underestimation on suicide trends.
South Korea has the highest suicide rate (29.1 per 100 000 population based on 2012 data) among all the Organisation for Economic Co-operation and Development countries10 and a relatively large proportion of suicides from pesticide poisoning compared with other countries.11 In the period 2006–2010, the average number of deaths certified as pesticide suicide per year was about 2700, accounting for 21% of all suicides in South Korea, based on registered death data.12 A recent study showed considerable misclassification of suicide in the South Korean mortality data compared with that of Japan and Hong Kong, and that some deaths certified as undetermined deaths or injuries were likely to be misclassified suicides.13 However, no study so far has conducted a systematic investigation of the misclassification of suicide by pesticide poisoning in South Korea.
The aims of this study are to identify cause-of-death groups that may include misclassified pesticide suicides and to investigate the impact of potential misclassification on the estimated magnitude of and trends in pesticide suicide in South Korea.
Mortality data for South Korea for the period 1991–2012 were obtained from registered death data provided by Statistics Korea (http://kostat.go.kr). The data include information about sex, age, administrative district of residence, education level, occupation, marital status, the cause of death and date of death. Underlying causes of death were coded according to the ICD-10.7 The ICD-10 was adopted in 1995 in South Korea, and Statistics Korea employed the conversion table to recode underlying causes of death originally coded according to ICD-9 for data in 1991–1994. We used data from 1991 as it is the first year for which individual-level mortality data coded according to the ICD-10 are available.
Deaths certified as suicide by pesticide self-poisoning were identified using the ICD-10 code X68. Based on findings from previous studies in South Korea13 and other countries,3 ,9 ,14 the following eight groups of causes of death (ICD-10 codes), which include all poisoning using solid/liquid substances other than homicides, were selected as alternative causes that may potentially include misclassified pesticide suicides (table 1): undetermined death by pesticide poisoning (Y18); injury by pesticide poisoning (X48); suicide (X69), undetermined death (Y19) and injury (X49) by poisoning using unspecified chemicals; and suicide (X60–X65), undetermined death (Y10–Y15) and injury (X40–X45) by poisoning using medicinal/biological substances or alcohol. Only data for deaths of individuals aged 15 years and above were extracted as suicides by individuals aged below 15 years were deemed to be rare and their intent of suicide uncertain. Population data for registered residents by sex, 5-year age group and urbanisation level of residence, for each year in 1991–2012 were obtained from Statistics Korea.
To examine the similarities or differences between certified pesticide suicides and deaths of alternative causes that were potentially misclassified pesticide suicides, we compared sociodemographic characteristics (sex, age, marital status, educational level and occupation), month of death and urbanisation level of residence among deaths of these groups. Rate ratios between sexes, across different age groups and across urbanisation levels were calculated and compared by cause-of-death group. Age categories were 15–49, 50–59, 60–69 and 70+ years and the reference group was 15–49 years; these groups were chosen as trends in certified pesticide suicide within these age bands were similar. Urbanisation levels were identified based on the administrative residential districts provided in the death data. Three categories of urbanisation level—(1) metropolises, which included the capital city, Seoul, and six other major cities; (2) medium-sized and small-sized cities; and (3) rural areas—were determined by governmental administrative divisions where population sizes and rural characteristics were considered. Metropolises were used as the reference group.
We also applied a semiquantitative approach to investigating the similarities or differences in characteristics between certified and possibly misclassified pesticide suicides. A ‘similarity score’ was calculated according to the following procedure. First, the distributions (by percentage) of sociodemographic characteristics, month of death and urbanisation level of residence were compared between certified pesticide suicides and deaths in each of the alternative cause-of-death groups. Second, for each category of each of the characteristics compared, if the relative difference in percentage was less than 10 between certified pesticide suicide and deaths of alternative causes, 1 point divided by the total number of variable categories compared was assigned. Otherwise, a score of 0 was assigned when the relative difference in percentage was above 10. For example, if the relative difference in percentage for the category ‘15–49 years’ of the ‘age’ variable was less than 10 between certified pesticide suicide and undetermined death by pesticide poisoning, 0.25 points (ie, 1 divided by 4 age categories compared) were given under the 10% difference criteria (see table 1). The similarity score can range between 0 and 1 for each characteristic; a score of 1 indicates that for a certain characteristic, the percentages of all variable categories for deaths due to an alternative cause are within a 10% change in those for certified pesticide suicides. Finally, a total score was calculated by summing up values across seven characteristics for each of the eight alternative cause-of-death groups. Sensitivity analyses were conducted using a cut-off of 5% difference (see online supplementary table S1).
We estimated the magnitude of and trends in the age-standardised rate of pesticide suicide, taking into account various levels of misclassification among the selected cause-of-death groups. Age-standardised rates were computed based on the 2000 WHO world standard population.15 Because it is unlikely that all (ie, 100% of) poisoning deaths from the alternative cause-of-death groups are misclassified pesticide suicides, we assumed the proportions of misclassified pesticide suicides to be 20%, 30% or 40% on the basis of findings from previous studies—three studies from the US,16 Finland14 and France17 showed that 21%, 32% and 35% of deaths in the undetermined intent category were misclassified suicides respectively. Joinpoint regression analysis (http://surveillance.cancer.gov/joinpoint) was used to examine secular trends in pesticide suicide and whether trends changed when various degrees of misclassification were allowed. Joinpoint regression analysis is based on linear regression models where time trends are characterised by contiguous linear segments and ‘join points’ (points at which trends change) that fit the data best.
We used publicly available data without any personal identifiers, and thus, ethical approval was unnecessary.
Figure 1 shows sociodemographic, seasonal and regional profiles by cause-of-death group. Overall, the profiles of certified pesticide suicides (ICD-10 X68) were more similar to those in five cause-of-death groups (undetermined death by pesticide poisoning (Y18); injury by pesticide poisoning (X48); suicide (X69), undetermined death (Y19) and injury (X49) by poisoning using unspecified chemicals) than those in the other three groups (suicide (X60–X65), undetermined death (Y10–Y15) and injury (X40–X45) by poisoning using medicines/alcohol). Compared with the latter three groups, certified pesticide suicides and deaths of the five alternative cause-of-death groups were generally more likely to have a male-to-female rate ratio of 2 or above, involve married individuals, involve individuals with an elementary or lower educational level, involve individuals in the agricultural workforce, occur in summer and have the highest rates in rural areas. All groups had higher mortality rates with increasing age, with undetermined death by pesticide poisoning showing the highest rate ratio in the elderly aged 70+ years and suicide by poisoning using medicines/alcohol showing the lowest rate ratio.
Similarity scores showed a consistent pattern to that observed in figure 1 (table 1): the aforementioned five cause-of-death groups (ICD-10 Y18, X48, X69, Y19 and X49) showed higher scores than the other three groups (X60–X65, Y10–Y15 and X40–X45). The former five cause-of-death groups were thus identified as groups that may contain a considerable number of possible misclassified pesticide suicides. The results using the 5% difference criterion were similar (see online supplementary table S1).
Over the period 1991–2012, the average annual number of certified pesticide suicides was 2033 (ranging from 623 in 1991 to 3530 in 2004); the average age-standardised rate was 5.6 per 100 000 population and the proportion among all poisoning deaths (ICD-10 T36–T65) was 46.6% (see online supplementary table S2). The average annual number of pesticide suicide would increase to 2346, 2504 and 2660 when assuming that the proportions of misclassified pesticide suicides in the five selected alternative cause-of-death groups were 20%, 30% and 40%, respectively, corresponding to an increase of 15.4%, 23.1% and 30.8% compared with the number of certified pesticide suicides only. Online supplementary table S3 provides a detailed overview of the yearly number of deaths by cause-of-death group.
Joinpoint regression analysis showed that pesticide suicide rates, based on certified cases only, increased by 16.4% (95% CI 2.2% to 32.4%) per year from 1991 to 1995, levelled off from 1995 to 2000, rapidly increased by 31.0% (95% CI 2.1% to 67.9%) per year from 2000 to 2003 and then fell between 2003 and 2012 with an average annual percentage change of −7.2% (95% CI −9.0% to −5.4%) (figure 2). Age-standardised rates of pesticide suicide increased by more than four times in 1991–2003 (from 2.09 to 8.75 per 100 000) but fell by around 50% in 2003–2012 (from 8.75 to 4.10 per 100 000). Secular trends in pesticide suicide generally did not change when assuming that misclassified pesticide suicides accounted for 20% of deaths in the five alternative cause-of-death groups. When the misclassification was assumed to be 30% or more, the joinpoint regression analysis only identified a change from an upward trend to a downward trend in 2004, and the rapid increase in 2000–2003 was lessened.
Our data showed that pesticide suicides may be misclassified as deaths in some cause-of-death groups (ie, undetermined deaths/injuries by poisoning using pesticides and suicides/undetermined deaths/injuries by poisoning using unspecified chemicals) and that pesticide suicide might be underestimated by 15%–31% in South Korea in 1991–2012. However, the possible misclassification has little impact on trends in pesticide suicide. Since suicide is an important public health issue and pesticide ingestion is among the most frequently used methods of suicide worldwide, providing a more reliable estimate of the burden of and trends in pesticide suicide has important implications for strategies to prevent pesticide suicide.
Our finding of 15%–31% possible underestimation of pesticide suicide is similar to those from previous studies that showed that officially recorded mortality statistics underestimated the true burden of suicides by 10%, 20% and 31% in Finland,14 England18 and Taiwan,3 respectively. However, these previous studies did not specifically investigate pesticide suicides and mainly focused on undetermined deaths as potentially misclassified suicides, while the current study systemically examined not only undetermined deaths but also suicides and incidental deaths by poisoning with various agents that may include misclassified pesticide suicides. Furthermore, our estimates were based on the assumption of 20%–40% misclassification in the alternative cause-of-death groups, which we feel is more reasonable than the approach adopted by previous research which treated all (100%) of the deaths in the alternative groups as misclassified suicides.
In the present study, deaths from poisoning by pesticides and by unspecified chemicals showed sociodemographic profiles and regional and seasonal characteristics similar to those of pesticide suicides, while deaths from medicines/alcohol poisoning had relatively low similarity. It is likely that some deaths that were certified as poisoning by unspecified chemicals were actually suicides by ingesting pesticides in South Korea, while poisoning deaths where the causing agents were identified as medicines or alcohol were unlikely to be pesticide suicides. Previously, incidental pesticide poisoning and certified suicides by poisoning using solids/liquids substances (mainly pesticides) also showed similar patterns in demographic profiles from registered death data in Taiwan,3 and the majority of poisoning death cases involving pesticide were thought to be suicides.19 Our findings suggest that in South Korea, poisoning by unspecified chemicals may have served as a convenient category for certifiers or coders to classify possible pesticide suicides when a full investigation or autopsy was not conducted to identify the underlying causing agents. Additional research is needed to better understand certifying or coding practices in relation to poisoning deaths in South Korea and whether these practices have changed over time.
Our study showed that the overall impact of misclassification on time trends in pesticide suicide was limited. In other words, taking into account misclassification would considerably increase the estimate of the burden but not substantially change time trends in pesticide suicide. Previous studies in 15 European Union member countries,20 France17 and Taiwan3 reported that suicide misclassification contributed to only a minor proportion of the variation in suicide rates or did not considerably influence their time trends. In contrast, other studies in Portugal21 and Malaysia22 showed that national trends in suicide would change substantially when suicide misclassification was taken into account. Of note, our data showed a marked increase in pesticide suicide in South Korea between 1991 and 2003–2004 even after considering the potential impact of misclassification, in keeping with findings of previous studies based on certified suicides only.23 The rapid increase in pesticide suicide in 2000–2003 when the assumed proportion of misclassification was 30% or more may have disappeared because the data contained a substantial portion of possibly misclassified pesticide suicide from certified poisoning using unspecified chemicals in 1997–1998 when the suicide rate increased rapidly under the economic crisis,24 so the trend in pesticide suicide changed to a gradual increase in 1991–2004. However, the weak impact of misclassification on downward trends in pesticide suicide after 2003 is probably due to the steadily improving classification of manner of death in South Korea.13
The accuracy of cause-of-death data was reported to have improved since 2000 in South Korea because Statistics Korea (the South Korean governmental agency in charge of death registration and publishing vital statistics) has initiated the linkage of the registered death data to a number of administrative datasets, that is, National Health Insurance data, National Cancer Center data, the police data and autopsy data from the National Institute of Scientific Investigation.25 Improved data reliability may have contributed to some of the increase in certified suicide mortality during 2000–2003.
Previous studies from other countries have shown variations in certification behaviours by the medical examiner or coroner26 and the referral of unnatural deaths to coroners across different hospitals;27 such variations may contribute to geographic variations in the quality of suicide data. In South Korea, if the cause of death is suspected or known to be unnatural, it is required that the death should be referred to and investigated by the prosecutors. However, a small number of coroners or medical examiners are responsible for the postmortem examination of deaths referred for autopsy.28 Therefore, the certification practice of individual coroners or medical examiners may substantially influence data quality for unnatural deaths, including suicides, in the area where the coroners or medical examiners work, and variations in the certification practice may contribute to variations in suicide data quality across areas in South Korea. Furthermore, there was an indication that referral to autopsy was less common in rural than non-rural areas in South Korea before 2000.29 Therefore, the underreporting of suicide may be more marked in rural than other regions.
This study has important limitations. First, this study was based on information available in death certificates. We did not have access to other information sources such as psychiatric treatment records or forensic reports, which may contain more detailed information needed to fully investigate the reliability of the coding of pesticide suicide and alternative causes of death. Second, we investigated only selected causes of death—poisoning by solid/liquid substances (pesticides, medicines/alcohol, and unspecified chemicals)—as potential alternative codes used to classify pesticide suicides. Thus, the possibility of pesticide suicides being misclassified as deaths due to other causes cannot be completely ruled out. However, suicides where the mechanism of injury involves low energy (eg, poisoning) are more likely to be misclassified as undetermined deaths than suicides that involve more active means (eg, suffocation and firearm shooting).8 Third, the choice of 20%, 30% and 40% as the possible levels of misclassification was based on research findings from other countries as we could not identify any previous South Korean studies that can inform our choice. In addition, the level of misclassification may vary by cause of death, calendar years or individual characteristics such as sex, age and marital status. Further research considering these factors would produce more accurate estimates of the magnitude of pesticide suicide underreporting.
In conclusion, this study is the first systematic investigation of possible cause-of-death groups that may contain misclassified pesticide suicides and of the impact of such misclassification on trends in pesticide suicides in South Korea. Our data show that some pesticide suicide cases may be misclassified as deaths of undetermined intent and injuries by pesticide poisoning as well as poisoning deaths by unspecified chemicals. However, such possible misclassification would not have substantially influenced the estimate of secular trends in pesticide suicide.
What is already known on the subject
Suicides by self-poisoning are more likely to be misclassified and underestimated than suicides by other methods.
Suicides by pesticide ingestion are prevalent in South Korea, but no previous study has examined the effect of suicide misclassification on secular trends in pesticide suicide.
What this study adds
Pesticide suicides may be misclassified as deaths of undetermined intent and injuries by pesticide poisoning as well as poisoning deaths by unspecified chemicals.
However, the misclassification has little impact on trends in pesticide suicide.
Distraction from digital billboards
In the old days we used to fuss that billboards were an eyesore but rarely thought that they distracted drivers. Now it is becoming evident that digital billboards (DBB) do so. An American traffic safety consultant concluded that there is ‘persuasive evidence that DBB take drivers’ eyes off the road for dangerously long gazes’. A study along high-speed roadways in Alabama and Florida found 25%–29% higher crash rates near DBBs than at control sites. The danger seems greater for novice and older drivers and the number of DBBs is increasing. Comment: not surprisingly, the more effective the DBB is in attracting attention, the more dangerous it appears to be.
Dementia linked to head injuries
Although we now accept that serious head injuries can have lasting effects, a new study suggests that repeated minor head trauma may also lead to early dementia. A Master's student in New Zealand, using a case–control design, showed that ‘the long term effects of repeated mild blows to the head were as exceedingly common as they were unreported and as damaging as a single episode of a higher grade brain trauma’. Equally unexpected was the finding that dementia also may be related to injury to other parts of the body.
Contributors ESC was responsible for analysis and interpretation of the data and drafting of the manuscript. S-SC provided advice on analysis of data, contributed to interpretation of the data and critically revised the manuscript. WJL was responsible for the study design, interpretation of the data and revision of the manuscript.
Funding This work was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2014R1A1A2054120, NRF-2014R1A6A3A01055217).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.