Article Text
PDF
Abstract
Introduction Imprecise data systems hinder understanding of drowning burden, even in high-income countries like Portugal, that have a well-implemented death certificate system. Consequently, national studies on drowning mortality are scarce. We aimed to explore drowning mortality in Portugal using national data and to compare these to Global Burden of Disease (GBD) estimates.
Methods Data were obtained from the National Institute of Statistics (INE) for 1992–2019, using International Classification of Diseases (ICD)-9 and ICD-10 codes, by sex, age group and cause (unintentional; water transport and intentional). GBD unintentional drowning data were obtained online. Age-standardised drowning rates were calculated and compared.
Results INE data showed 6057 drowning deaths, 4327 classified as unintentional (75.2% male; 36.7% 35–64 years; 31.5% 65+years; 15.2% 0–19 years). Following 2001, an increase in accidental drowning mortality and corresponding decrease in undetermined intent was observed, coincident with Portugal’s ICD-10 implementation. GBD modelled estimates followed a downward trend at an overall rate of decrease of −0.41/decade (95% CI (−0.45 to –0.37); R2 adj=0.94; p<0.05). Conversely, INE data showed an increase in the rate of drowning deaths over the last decade (0.35/decade; 95% CI (−0.18 to 0.89)). GBD estimates were significantly different from the INE dataset (alpha=0.05), either underestimating as much as 0.567*INE in 1996 or overestimating as much as 1.473*INE in 2011.
Conclusions While GBD mortality data estimates are valuable in the absence of routinely collected data, they smooth variations, concealing key advocacy opportunities. Investment in country-level drowning registries enables in-depth analysis of incident circumstances. Such data are essential to informing National Water Safety Plans.
- surveillance
- mortality
- epidemiology
- policy
Data availability statement
Data used in this study are either available in a public, open access repository or available on reasonable request.
Statistics from Altmetric.com
Introduction
With the recent United Nation (UN) declaration recognising drowning as a globally preventable cause of death,1 there is a need for all countries to collect and interrogate drowning data to help develop evidence-based, country-specific prevention strategies. Drowning is defined as the process of experiencing respiratory impairment from submersion/immersion in a liquid, with drowning outcomes classified as death, morbidity and no morbidity.2 Drowning is a leading cause of unintentional-injury death globally,2 and was proclaimed as a public health priority by WHO in 2014,3 however, little is known about the causes of drowning in Portugal.
In Europe, drowning caused the loss of an estimated 17 597 lives in 2019,4 with Western Europe, males and children disproportionately impacted.5 In most European countries, it appears fatal drowning is declining, however, access to data on non-fatal drowning remains a challenge across the region.5
WHO has developed a drowning prevention implementation guide6 which affirms that well-designed research and quality data are key to developing drowning prevention strategies for nations around the world. However, many countries, including high-income countries, still lack a comprehensive, quality drowning surveillance system.3 7 Additionally, multiple organisations collect drowning data, with expectable differences in scope, accuracy and consistency.7 8 The lack of consistent drowning data surveillance systems across agencies hinders full understanding of the drowning burden, causes confusion and limits research, even in high-income countries where death certificate systems are well implemented, such as in Portugal.
To address the lack of data and produce consistent results, the Global Burden of Disease (GBD) Study models data from a range of sources to guide policy makers. The GBD is a comprehensive assessment of health associated with risk factors. In 2019 the GBD produced estimates for 369 diseases and injuries, and 87 risk factors, from 204 countries and territories.9 Drowning is one of the 30 mutually exclusive causes of injury related mortality in the GBD study design and is nested within the unintentional injury category of the GBD cause hierarchy.10 The GBD provides a tool to quantify health loss from hundreds of diseases, injuries and risk factors, in order to improve health systems and eliminate inequalities.11
Data, even modelled, are useful for helping decision-makers develop a better understanding of health trends. Where national resources are scarce or national systems are still not designed to efficiently collect and extract drowning data, modelled data may be the only option. To address drowning in their country governments must invest in high-quality, comprehensive data systems which help describe the drowning situation, and identify specific risk factors, thus allowing for targeted interventions to be designed, implemented and evaluated.6
While raw figures of drowning deaths in Portugal are publicly available and can be obtained from official sources,12 studies on drowning mortality are virtually non-existent in Portugal and mostly limited to grey literature using data extracted from media reports or specific to children only.13–15 While media reports can be an important avenue for data on drowning when supplemented by other sources,16 data obtained solely from media has several drawbacks. These include the fact that news coverage is usually more responsive to unexpected deaths17 and increases in death rates than to decreases in death rates.18
There remains no population level study of drowning mortality in Portugal nor an assessment of the GBD model for drowning in this country. Therefore, the aim of this study was to explore the long-term patterns of drowning mortality in Portugal from 1992 to 2019 and compare national drowning data with drowning estimates attributed to Portugal by the GBD study.
Methods
This is a total population retrospective case series study of drowning deaths, between 1992 and 2019, of residents of Portugal. This research observes the Guidelines for Accurate and Transparent Health Estimates Reporting19 and is in compliance with Helsinki Principles and the European Union and National Laws.
Patient and public involvement
The public was not involved in any part of this research.
Data sources
Two data sources were used for this study. Data on drowning deaths in Portugal among the resident population for the period 1992–2019 (inclusive) were sourced from both the National Institute of Statistics (INE) and from the Institute of Health Metrics and Evaluation GBD study20. INE data were obtained after submitting an online request12 and GBD data were obtained using the online GBD results tool after running a direct query on the global health data exchange webpage.4 This information was extracted for the years 1992–2019, for location: Portugal, using cause of death: C.2.2 Drowning, for all age groups and both sexes.
The WHO International Classification of Diseases (ICD) used for the registration of deaths in Portugal according to the cause evolved and was the subject of successive reviews. The year in which Portugal adhered to international agreements to update its classification system did not always coincide with the date on which these systems were actually implemented. The 10th ICD review was approved in 1989 and adopted by the 43rd World Health Assembly, to enter into force on 1 January 1993, but according to information provided by the (Portugal) Directorate General of Health (DGS) by observing the statistical data, deaths have only been coded according to this review since 2002 in Portugal.21 Additionally, individual microdata records of mortality for specific causes have been computerised since 1980, and after 2014 the mortality register input became entirely online.
Drowning deaths of all causes (ie, accidental, undetermined, water transport and intentional) were identified based on the ICD 9th edition from 1992 to 2001 and 10th edition from 2002 to 2019. We obtained and stratified deaths due to drowning where the records showed the primary cause of death as per the codes detailed in table 1 (accidental, water transport, undetermined intent and intentional for INE and unintentional for GBD (due to the redistribution of drowning due to undetermined intent)). From each source, data were obtained as number of drowning deaths overall, and by year, sex and age group.
ICD codes of causes of drowning death extracted from INE and GBD
Statistical analysis
Age was stratified as follows: 0–4, 5–14, 15–19, 20–34, 35–64, 65+. We analysed trends over time, differences by sex and the contribution of each cause and age group to mortality. To describe temporal trends in Portugal, age-specific, sex-specific and cause-specific rates were computed and age-adjusted rates were calculated using the direct method with reference to the Portuguese population 2019 (resident) sourced from the National Population Census.22
For age-specific and sex-specific analyses, a locally smoothed regression (loess) with 95% CIs was used. Regarding cause-specific analysis, INE data were analysed using four categories (accidental, intentional, undetermined and water transport). To compare the trends in rates of official national data retrieved from INE with modelled estimates from GBD, the categories ‘undetermined’ and ‘accidental’ were aggregated within the ‘unintentional’ category. This was undertaken to enable comparison with the GBD unintentional drowning category which includes the corresponding INE categories of accidental and undetermined and due to the observed shift in data classified as undetermined (see the Results section). Unless otherwise stated, unintentional drowning deaths refer to drowning deaths classified as accidental and undetermined, in the national official data from INE. The trend of the rates of drowning deaths was assessed through linear regression. In the case of the official data (INE), a break appeared to be present in the rate of change, while the modelled data (GBD) showed no such break. Thus, a piecewise regression was implemented to the INE data to identify the breakpoint and to characterise each of the regression segments.
T-tests with pooled error variance were used to compare the slopes of the regression lines, within INE data and to compare INE with GBD. All statistical tests were carried out in the statistical programme R (V.4.0.3; http://www.R-project.org) and using the ‘segmented’ R package (V.1.3–323).
Results
From the INE data, a total of 6057 deaths from drowning were reported among which 4327 (71.4%) were unintentional (accidental +undetermined), 81 (1.3%) were water transport and 1649 (27.2%) were intentional. Overall, males represented 69% (n=4165) of the deaths and most drowning deaths were aged 65 years and over (n=2309, 38%) (table 2).
Drowning deaths in Portugal between 1992 and 2019 retrieved from national official source (INE) by cause of drowning
Of the 4327 unintentional deaths, 1882 (43.5%) were accidental and 2445 (56.5%) were undetermined. Males accounted for 77.1% of all unintentional deaths (n=3334). The most common age groups were 35–64 years (n=1648, 38.1%), followed by adults aged 65 years and over (n=1372, 31.7%) (table 2). Children and adolescents (0–19 years) accounted for 16.9% (n=733) of all unintentional drowning deaths, with 5.2% (n=224) aged 0–4 years.
An annual average of 2.9 water transport related drowning deaths were identified, ranging from 0 to 14 deaths. Intentional drowning deaths represented an average of 58.9 deaths per year, 27.2% of the total deaths, ranging from 32 in 1997 (12.1% of that year) and 1998, to 98 in 2014 (45.4% of that year).
Mortality rates of drowning deaths for INE and GBD by age group and sex are shown in figure 1. While GBD modelled estimates tended to mimic the INE data trend for younger age groups, substantial differences were apparent for 35 years and older and stronger for males compared with females. In particular, an increase was observed in INE data for the 20–34 years age group and was particularly evident among males aged 65 years old and over, suggesting that these age groups are determining the overall trend of drowning deaths in Portugal. Age-specific unintentional drowning mortality rates for children under 5 years of age and adults aged between 35 and 64 years were, on average, the same, 1.47/100 000 residents (figure 1).
Age-specific and sex-specific drowning death rates in Portugal per 100 000 residents 1992–2019. Points show the yearly rate of drowning derived from each dataset and lines show locally smoothed data (loess regression) with 95% CIs (shades). Blue lines represent GBD data and the red lines INE data. The vertical black line in 2008 marks the breakpoint identified in the INE full dataset by way of piecewise regression. GBD, Global Burden of Disease; INE, National Institute of Statistics.
The age-standardised rate of unintentional drowning deaths has varied from a peak of 3.18 drowning deaths per 100 000 residents in 1992 to 0.70 in 2011 with 1.09 more recently in 2019 (figures 2A and 3 and online supplemental table S1). Considering the INE dataset, the rate of unintentional drowning deaths has been changing by an average of −0.76 per decade (95% CI (−0.96 to –0.56); R2adj=0.69; p<0.05). However, piecewise regression analysis confirmed the presence of the apparent increase in the rate of such drowning deaths over the last decade (R2adj=0.85), shifting from a decline of −1.32 per decade until 2008 (the identified inflection year; 95% CI (−1.59 to –0.04); p<0.05) to an increase by 0.35 per decade from 2009 onwards (95% CI (−0.18 to 0.89); p>0.05).
Supplemental material
(A) annual age-standardised rate of fatal drowning from ine (accidental and undetermined) and GBD (unintentional), linear regression with 95% CIs, Portugal, 1992–2019; (B) annual age-standardised rate of fatal drowning from ine (accidental only) and GBD (unintentional), linear regression with 95% CIs, Portugal, 1992–2019; (C) the contribution of drowning deaths with undetermined intent to the INE drowning dataset, expressed as a percentage of the yearly total (undetermined / (accidental +undetermined)), Portugal 1992–2019. Note: points show the yearly rate of drowning derived from each dataset. Lines show the linear regression of the data, and the shade and dashed lines depict the 95% CIs of the slopes. Blue lines represent GBD data and the red lines INE data.The vertical black line in 2008 marks the breakpoint in the ine dataset. GBD, Global Burden of Disease; INE, National Institute of Statistics.
Age-standardised drowning mortality rate in Portugal, 1992–2019, with data from INE and GBD and the percentage of GBD data matching INE data. (A) INE data considering undetermined and accidental drowning deaths; (B) INE data considering just accidental drowning deaths. Blue lines represent GBD data and the red lines INE data;Grey area denotes %GBD/INE. GBD, Global Burden of Disease; INE, National Institute of Statistics.
Modelled data (GBD), on the other hand, ranged from 2.11 drowning deaths per 100.000 residents in 1992 to 0.89 in 2019, with a minimum rate of 0.88 recorded in 2017. Importantly, GBD data had considerably lower variance, with a SD of 0.35, much lower than the SD of 0.75 found in the INE dataset. This is reflected in the unchanging and steady decline in the rate of drowning evidenced in the GBD dataset, with an overall drop of −0.40 per decade (95% CI (−0.45 to –0.397 R2adj=0.94; p<0.05). The slope of the linear regression of the GBD data was significantly different (alpha=0.05) from the slope of the linear regression of the full INE dataset, as well as the slopes of the piecewise regression segments before and after 2008. While the trend of both INE and GBD datasets seems more similar when only drowning deaths coded as accidental in INE’s dataset are compared with GBD (figure 2B), the adjusted R-squared is much lower (-0.02 vs 0.87), indicating a worse fit.
The contribution of drowning deaths with undetermined intent to the overall picture of drowning in Portugal (figure 2C), peaked in the year 2000 at 86.7% (6.52 times higher than the rate of accidental drowning deaths), falling to 17.8% in 2003 (0.22 times the rate of accidental drowning deaths).
Discussion
This study aimed to address a research gap in Portugal by reporting country level, all-age fatal drowning using data from a national source and then comparing this to estimates derived from the GBD study. Our study identifies drowning deaths have been decreasing, a finding that is consistent between INE and GBD. However, we note an increase in drowning identified in the INE data late in the study period (after 2008) not shown in the GBD data. Changes in coding from ICD-9 to ICD-10 appear to impact drowning coding and this will need to be monitored when mortality surveillance systems globally move from ICD-10 to ICD-11.
While mortality rates for drowning are declining overall in Portugal, this trend conceals high-risk groups. First, much like every other country in the world, fatal drowning impacts males more than females.24–26 This is seen across all age groups, where females always and at all time points have had lower rates of fatal drowning. Similarly to many other countries,27 28 drowning in Portugal occurs mainly among the youngest (under 5 years), although this study identified increasing rates for those of over 35 years of age. However, more research exploring age-based risk differences in drowning in Portugal is required.
While child drowning has been consistently declining over the last decades, the trajectory for the 65+ age group has been strikingly different. There has been a marked increase in drowning rates among the older population, especially among—but not restricted to—males. Importantly, this increase takes place in spite of a growing elderly population,29 30 and may potentially be due to increased physical activity of this age group. Healthy ageing is important for reducing the burden on the health system and providing a better quality of life,31 and aquatic activity is both fun and often reduces weight bearing burden32 making it perfect for an older cohort. However, encouraging aquatic based exercise for older population in Portugal must also consider drowning risk. There is a need to conduct further research into causal factors behind rising rates of drowning among older people both in Portugal and globally, given the world-wide issue of ageing populations.33 34
Interestingly the issue of drowning among older populations impacts modelled drowning estimates. Our study has found that it is the 65+ years age group that is most poorly modelled, with the GBD dataset largely underestimating the magnitude of drowning in older people of both sexes when compared with national data (INE). This finding emphasises the risks of relying on modelled data to identify at-risk groups and the importance of having a high-quality national drowning registry to bridge that gap.
When exploring temporal trends for drowning mortality, one should not ignore the possible effect from the revision of ICD codes during the study period, as has been seen in other studies.35–37 Our study identified significantly higher coding of drowning deaths as undetermined intent between 1992 and 2002, resulting in a rate that, in the year 2000, was six times higher than the cases coded as accidental drowning. Following the year 2001, a substantial increase in accidental drowning mortality was observed and drowning deaths classified as being of undetermined intent saw a sharp decrease. This is most likely the result of the changes in coding introduced in 2002 with the implementation of ICD-10 in Portugal, in parallel with a verbal autopsy process that took place from 2001 to 2004 targeting medical doctors responsible for filling in the death certificates for external causes of death.21 38 This mechanism is consistent with the slow but upward trend in the contribution of undetermined drowning deaths since the mid 2000s to about 50% in 2019. As a coding system ages, it appears to increase the number of deaths classified as of undetermined intent. While it is unclear why this occurred, it may be due to a reduction in zeal around the coding, a loss of skill from people who were originally trained moving to other jobs, a lack of priority given to good coding, an increase in workload or some other mechanism. This finding is valuable as it shows that regular refreshment efforts targeted at those responsible for coding mortality data should be implemented to increase quality and accuracy of data reported in death certificates.39
Distinguishing between drowning due to undetermined intent and intentional drowning can be quite challenging to determine.40 Intentional drowning deaths make up a significant proportion of the overall drowning toll in Portugal, reaching a high of 45% of all drowning deaths of Portuguese residents in 2014.
On the other hand, although variable, overall undetermined intent accounted for 37% of all drowning deaths. The overuse of the undetermined intent code can lead to a poorer understanding of drowning and its prevention. Additionally, noisy data have previously been identified as an issue when studying suicides in Portugal.21 Since former procedural changes for data collection and coding may have secondarily impacted the fidelity of drowning death rates in Portugal, it is important to identify periods in which there have been a greater/less consistency in the registration of intentional and undetermined drowning deaths, in order to detect positive or negative coding practices and apply measures to improve.
Drowning represents a global challenge to public health3 and, while mortality data estimates such as those produced by the GBD study are a useful source of information where national level data are missing, this study shows that it removes the variability seen in national data. This lack of variability can hide areas of concern, such as increasing drowning rates among older age groups as identified in this study. This may result in missed opportunities for advocacy around topical issues. For example, a summer with high drowning deaths could be used as an impetus for change.41
For countries such as Portugal that rely on a National Death Certificate system, even one that has since 2014 been fully online42 there is still a need for more detailed analysis of the data and preferably a detailed drowning deaths registry. There remains a need to improve data collection accuracy and quality as well as data accessibility. There are many data items outside of what is collected in death certificate system around drowning that are required for development of a national water safety plan for Portugal and more targeted drowning prevention interventions such as aquatic location, activity prior to drowning, and risk factors such as alcohol consumption and pre-existing medical conditions.6
Moreover, a far-reaching analysis of all drowning events, including non-fatal, is important to fully understand the burden of drowning morbidity and how drowning mortality rates are being modulated throughout time. Adequate, sustainable and long-term funding is needed to establish and maintain accurate and high-quality interagency drowning data collection and linkage systems.6 The recent UN resolution on drowning prevention1 may prove to be a useful tool for drowning prevention researchers and advocates to use to generate political support and resourcing for drowning data collection, research and interventions, not only in Portugal, but in all nations which lack quality drowning data.
GBD enables a standardised process to compare countries and accounts for differences in completeness of data and quality of coding. This is both a strength and a limitation of the GBD data and this study is the first to compare GBD to country-level data for drowning. We acknowledge the need for the GBD study to help strengthen our global understanding of the burden of diseases and injury, however we also highlight the need for country level studies to improve the modelling within the GBD and to inform local level action.
Strengths and limitations
This is the first national study exploring drowning mortality in Portugal. Similarly, this is the first study to compare GBD modelled data for drowning to country-level records which has important implications for policy markers relying on modelled data. However, there are some limitations associated with this study. First, as with all studies using death certificates, this study relies on the quality and homogeneity in the manner and cause-of-death reporting by the personnel involved in the death registration procedures, which includes police officers, doctors, coders and government agencies. We have not considered X-38 (flood related deaths), so it is possible that we may be missing some flood-related drowning cases. We did not include this code in the analysis, because flooding events resulting in fatal drowning are not common in Portugal43 and due to methods used by GBD we would have been unable to compare all drowning between INE and GBD — as a result, water transport and intentional drowning cases could only be analysed using INE data. The cases that could have been coded as undetermined in GBD might have been redistributed into the drowning category. However, it is not clear exactly how many of all the cases would be expected to be redistributed to intentional drowning. As such less than 100% would be distributed to the accidental drowning category. Similarly, we did not factor in any political or socioeconomic conditions that might have impacted the observed trend. Furthermore, we are comparing two data sources with unknown accuracy and quality. We have not included the 95% uncertainty intervals provided with GBD estimates and it is possible that these uncertainty intervals may overlap in some degree with INE dataset. Additionally, distinguishing injury intent has intersectoral relevance, from clinical, to medicolegal and public health professionals and coding rules of ICD-10 prioritise intent of injury over injury mechanism.44 Distinguishing intentional from unintentional injury is important for injury surveillance, as prevention, treatment and rehabilitation differ.45
Conclusion
This is the first study to analyse trends of fatal drowning in Portugal and the first to compare country level data with GBD data. The two datasets show some homogeneity, but with differences among age groups and individual year of drowning death. This highlights the need for the development of country-specific drowning registries to enable the development of plans, research and advocacy opportunities. Unintentional drowning in Portugal disproportionately affects the youngest and mature populations. A comprehensive analysis of all drowning events, including non-fatal, is required to fully understand the burden of drowning morbidity and how drowning mortality rates change throughout time. Improving data collection systems and furthering understanding of the risk factors of drowning events, as well as the development, implementation and evaluation of prevention programmes, are essential to reduce the risk of drowning in Portugal.
What is already known on the subject
Drowning is a key public health issue, however, there are no previous population level studies in Portugal.
Global Burden of Disease (GBD) modelled estimates are an important source of data for decision making on drowning, especially where national level data are absent or not routinely collected, as well as allowing for a standardised comparison among countries.
No previous study has compared modelled GBD data to country level data for drowning.
What this study adds
From 1992 to 2019, 6075 people fatally drowned in Portugal, an average of 217 per year, with an average age-standardised rate of −0.76/decade (95% CI (−0.96 to –0.56); R2 adj=0,69; p<0.05).
Modelled estimates from GBD followed a downward trend between 1992 and 2019, while National Institute of Statistics (INE) data showed initially a decrease and then an increase in the rate of drowning deaths over the last decade (0.35 per decade; (95% CI (−0.18 to 0.89)).
Modelled estimates from GBD were significantly different from the INE dataset (alpha=0.05), varying from 59.2% to 147.5% and had considerably lower variance (SD=0.35) than the variance found in INE data (SD=0.75).
While mortality data estimates can provide invaluable appraisals of global national trends, investment in country-level, high-quality and detailed data collection systems are essential for good planning and evaluation.
Abstract translation
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.Data availability statement
Data used in this study are either available in a public, open access repository or available on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
It is exempted from ethical approval because it uses population-level aggregated and irreversibly anonymised data from a dataset made public by a state entity.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @Franklin_R_C, @amyepeden
Contributors All authors have made substantial contributions to all of the following (1) drafting the article or revising it critically for important intellectual content, (2) final approval of the version to be submitted. ACQ and RS were responsible also for data analysis. ACQ is the guarantor for the content of this article.
Funding The author RS was funded through national funds CEECIND/01424/2017.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.