Changes in drowning mortality rates and quality of reporting from 2004–2005 to 2014–2015: a comparative study of 61 countries

Background This study assessed international variations in changes in drowning mortality rates and the quality of reporting specific information in death certificates over the past decade. Methods Drowning mortality data of 61 countries were extracted from the World Health Organization Mortality Database. We calculated the percentage change (PC) in age-standardized drowning mortality rates and percentage of drowning deaths reported with unspecified codes between 2004 and 2005 and 2014–2015. Results Of the 61 countries studied, 50 exhibited a reduction in drowning mortality rates from 2004 to 2005 to 2014–2015. Additionally, five countries—Lithuania, Moldova, Kyrgyzstan, Romania, and El Salvador—with a high mortality rate in 2004–2005 (> 40 deaths per 100,000) showed improvement (PC < − 32%). By contrast, four countries—South Africa, Guyana, Morocco, and Guatemala—exhibited a more than twofold increase in mortality rates. Regarding the quality of reporting, 34 countries exhibited a decrease in the percentage of unspecified codes. Additionally, three countries—Paraguay, Serbia, and Croatia—with moderate and high percentages of unspecified codes (> 40%) exhibited a marked reduction (PC < − 60%), whereas three countries—Malaysia, Belgium, and Nicaragua—exhibited a notable increase. Conclusions Large international variations in the extent of changes in drowning mortality rates and the quality of reporting specific information on the death certificate were observed during the study period.

changes in drowning mortality rates. A recent study using the WHO Mortality Database examined the changes in drowning mortality rates across 21 countries meeting the following four criteria: 1) the death registration system covers 70% of the national population; 2) data are available for 10 years or more for the period between 2000 and 2013; 3) at least 100 drowning deaths have been reported for people under 20 years of age for each year; and 4) the International Classification of Disease, Tenth Edition (ICD-10) was used to code deaths for data analysis. The study reported a reduction in drowning mortality rates in 20 of the 21 countries during 2000-2013 (ranging from − 80% to − 13%) [3]. However, this study investigated only decedents under 20 years of age in 21 countries. In 20 among the total 60 countries studied, more than 50% of the drowning deaths were reported among middle-aged and older adults [4]. The circumstances of exposure to different water body types and implementation of prevention measures varied across different age groups and countries. Therefore, we hypothesize that drowning mortality rates vary when considering all age groups rather than considering only the young age group.
The quality of reported mortality data was investigated in the GBD 2016. The quality of morality data was assessed based on the completeness of death registration, fraction of deaths assigned to "garbage" codes, and fraction of deaths assigned to detailed GBD codes [1]. The GBD 2016 did not define "garbage" codes for drowning deaths. Some investigators have suggested the use of the ICD-10 code W74, "unspecified drowning and submersion," as an indicator of the quality of reporting specific information on the death certificate [5][6][7]. A previous study indicated that the percentage of unspecified codes exceeded 50% for 32 of the 69 countries studied [7]. A study revealed an improvement in the quality of reporting specific information in some countries during 2000-2013 but a deterioration in the quality in other countries [3].
In this study, we assessed international variations in drowning mortality rates and the percentage of reporting unspecified codes for drowning deaths among all age groups across countries by using mortality data of the past decade. An international comparison of injuryrelated mortality rates could identify not only the unique health problem of a country but also successful prevention measures [8][9][10][11][12][13]. In particular, we aimed to identify the countries previously having high drowning mortality rates and the percentage of unspecified codes that showed a prominent improvement in the two indicators during the past decade, which could help discern some success stories. Additionally, we aimed to identify countries that exhibited a deterioration in the two indicators; drowning should be considered a high-priority public health concern in these countries. Drowning mortality data for 2004-2005 and 2014-2015  were extracted from the WHO Mortality Database [14]. To ensure stability in comparing the mortality rates between 2004 and 2005 and 2014-2015, only the countries with absolute drowning deaths ≥50 in the same period were included in this study. For some countries, the latest available data might be available for 2012-2013 or 2013-2014. Many countries with high drowning deaths, such as China, India, Bangladesh, Pakistan, Nigeria, and Vietnam, did not submit mortality data to the WHO; therefore, these countries were not included in this study.

Methods
We used ICD-10 codes W65-W74 to identify drowning deaths. The age-standardized death rate (deaths per 100, 000 people) of each country was calculated based on the world standard population [15]. The quality of reporting specific information on the death certificate was defined as the percentage of drowning deaths with the ICD-10 code W74, "unspecified drowning and submersion." The higher the percentage, the poorer the reporting quality.
The percentage change (PC) in the age-standardized mortality rate and unspecified codes between 2004 and 2005 and 2014-2015 were calculated to assess the extent of change. Countries were trichotomized as good, fair, and poor in terms of improvement in the two indicators based on the PC. For countries with a drastic increase and decline in drowning mortality rates between 2004 and 2005 and 2014-2015 and countries with a higher ranking during 2014-2015, we further examined the changes in drowning mortality rates by age group.
Data visualization techniques were used to more efficiently illustrate the changes. We used a bubble-scatter plot to contrast the changes in mortality rates and percentage of unspecified codes between 2004 and 2005 (X axis) and 2014-2015 (Y axis), with the size of the bubble indicating the number of drowning deaths in each country during 2014-2015. Subsequently, we used a slopegraph to illustrate the changes in country rankings (the smaller the rank, the lower the mortality rates and percentage of unspecified codes) from 2004 to 2005 to 2014-2015. A downward slope indicates a decline in rankings. The ranking of some countries declined despite the decrease in mortality rates and percentage of unspecified codes. We used different colours to indicate the absolute change in mortality rates and percentages; blue indicates a decrease and orange indicates an increase. The dashboards can be accessed at: https:// public.tableau.com/profile/robert.lu#!/vizhome/Drowningmortality/Story with permission.

Results
The numbers of all and unspecified drowning deaths, age-standardised mortality rate, and percentage of unspecified codes for each studied country during 2004-  Table 1.
Among the 61 countries studied, 50 exhibited a decline in drowning mortality rates from 2004 to 2005 to 2014-2015, which is indicated by their position below the diagonal line in Fig. 1a. To properly interpret the PC (e.g., some high PCs are attributed to small mortality rates at the 2004-2005 baseline), we classified the countries into high, moderate, and low on the basis of mortality rates (Table 2) (Table 2), which can be prominently noted on the right lower side in Fig. 1a, whereas four countries revealed a twofold or more increase in mortality rates, namely South Africa (from 2.0 to 27, + 1195%), Guyana (from 29 to 133, + 368%), Morocco (from 2.8 to 7.6, + 177%), and Guatemala (from 7.8 to 20, + 161%) ( Table 2).
Despite 50 countries exhibiting a decline in drowning mortality rates, only 38 countries were ranked higher during 2014-2015 than during 2004-2005 (Fig. 2). We also found that among the 10 top-ranking (first to tenth) countries in 2014-2015, nine were European countries. Notably, although some countries exhibited a decline in ranking (downward slope) (i.e., Germany [fourth to sixth], Belgium [ninth to tenth], Sweden [thirteenth to fifteenth], Serbia [nineteenth to twenty-first], Uruguay [thirty-fifth to forty-first], and Japan [forty-fourth to forty-sixth]), their mortality rates showed a decreasing trend (blue line).
Despite 34 countries exhibiting a reduction in the percentage of unspecified codes, only 28 countries exhibited higher rankings during 2014-2015 than during 2004-2005 (Fig. 3). Similarly, during 2014-2015, of the 10 top-ranking (first to tenth) countries, seven were European countries. Notably, although some countries exhibited a decline in rankings (downward slope), namely, Slovakia (from sixth to eleventh), Denmark (from eighth to twelfth), Australia (from fourteenth to seventeenth), United States (from fifteenth to sixteenth), Germany (from twenty-second to eleventh), and Kyrgyzstan (from thirty-fifth to thirty-sixth), the percentage of unspecified codes showed a decreasing trend (blue line).
For countries reporting a drastic increase in drowning mortality rates during the study period (South Africa, Guyana, Morocco, and Guatemala), we further analysed

Discussion
The findings of this study indicate that most of the studied countries (50/61) showed a decline in drowning mortality rates during the past decade. However, only half of the studied countries (34/61) exhibited an improvement in the quality of reporting specific information on the death certificate. Some countries demonstrated a larger PC in the two indicators, which could provide insights for preventing drowning deaths. One possible explanation for the decline in drowning mortality rates in many countries is the proposal and implementation of several evidence-based drowning prevention interventions during the past decade [2,[16][17][18][19]. However, Towner et al. indicated that countries with the widest range of injury prevention legislations and strongest enforcements have not made relative progress in reducing child injury deaths since the 1970s [11]. In a historical study, Staines and Ozanne-Smith highlighted other key factors of the reduction in drowning mortality rates (i.e., infrastructural development [e.g., the development of a piped water supply system eliminated the need for reliance on hazardous water supplies or storage systems], urbanization, and mandatory schooling [thereby reducing the amount of time that children were unsupervised and in hazardous environments]) [20]. European countries showed high ranking in both the indicators, for example, nine of the ten highest ranked (lowest mortality rates) countries during 2014-2015 were European countries (Fig. 2). Most of the decline in drowning mortality rates from 2004 to 2005 to 2014-2015 in these countries were attributed to the reductions in child drowning mortality rates and possibly due to the Tools to Address Childhood Trauma, Injury and Children's Safety funding and partnership coordinated by the European Child Safety Alliance [21]. On July 14, 2008, the European Child Safety Alliance, EuroSafe, launched a new resource titled "Protecting children and youth in water recreation: safety guidelines for service providers" [22]. This water safety resource is particularly optimized for people working in the water recreation and tourism sectors to assist them in offering safe water-related activities and services for children and families throughout Europe.
We found that four countries (South Africa, Guyana, Morocco, and Guatemala) exhibited the highest increases in drowning mortality rates from 2004 to 2005 to 2014-2015, in which an increase occurred in each age group. According to the report of The Global Facility for Disaster Reduction and Recovery, we found tsunami and flood were the possible causes of the increase in the drowning mortality rates in these four countries [23].
Regarding the improvement in the reporting quality of Paraguay, Serbia, and Croatia, we are expecting that some studies may be performed to understand if there were any interventions or efforts implemented during the past decade. The coding for drowning deaths as unspecified may result from various factors, such as lack of specific information on the circumstances that led to the drowning, inadequate collection of primary data owing to insufficient police and medico-legal investigations, and incompleteness or errors during the death certification and coding processes [7]. Information on the concerned water body is paramount in designing relevant  prevention programs. For example, in Japan, the main water body for drowning deaths among older adults is the bathtub [24], for which the prevention measures would be quite different from measures required for preventing children from drowning in swimming pools. Several limitations should be considered while interpreting the findings of this study. First, some countries with high drowning death rates, such as China, India, Bangladesh, Pakistan, Nigeria, and Vietnam, were not included in this study because of the unavailability of recent data or details in the WHO Mortality Database. Second, only two time points (2004-2005 and 2014-2015) were used to examine the changes, and some countries with a relatively small number of drowning deaths might exhibit large fluctuations in mortality rates. To avoid the improper interpretation of the changes, we further analysed the yearly mortality rates to examine whether the pattern of the mortality trend was linear or nonlinear. Third, we did not have detailed contextual information on drowning mortality rates and the reporting quality in each studied country, which impeded a reliable explanation of the changes.

Conclusions
The study findings indicate large international variations in the extent of changes in drowning mortality rates and the quality of reporting specific information on the death certificate during the past decade. Additional studies are required to thoroughly understand the reasons for the drastic improvement or deterioration in both the indicators.