Mapping of existing data sources

Registers identified through both the literature review and the survey (82% response rate) have been classified by sectors of activity (table 2). This list does not represent an exhaustive list of all institutions collecting injury data in the EU. Not all of the reported injury/accident data collection systems are publicly available. Ideally, registers useful for consumer product safety and injury prevention would contain all of the following information: circumstances of injury (mechanism, intent, location, severity and time of injury, narrative); involved product (product category or more detailed description of product, eg, electrical tool, dispenser); brand of the product (detailed information on product identity and manufacturer) and information on the person injured (age, gender). However, only a very small number of identified registers were designed to intentionally collect such detailed information (eg, research project Susy Safe).

European-centralised database Community Road Accident Database (CARE), Eurostat’s Statistics on Accidents at Work (ESAW) or cosmetovigilance systems are some of the examples of successful data collection practices across Europe which owe their success to both national efforts and well-defined and implemented legal framework that supports them. Many safety policy actions in the past have been driven by evidence from systematic data collection which in return triggered the development of safer products and services, which emphasise the relevance of systematic injury/accidents data collection. The above-mentioned systems (CARE, ESAW, cosmetovigilance) have been excluded from further analyses as they are managed under the umbrella of other policy frameworks and are already well established at the EU level. The Rapid Alert System (RAPEX) is an EU-wide rapid information exchange system on measures taken against consumer products (except for food, pharmaceutical and medical devices) that are identified to pose a risk. The system is not designed to provide information on the actual level of products’ safety in the EU. While notifying authorities are required to report on known incidents and accidents, in the majority of cases such information is not available. It has therefore not been further assessed in this study, due to the limited number of injury cases reported through RAPEX. The focus of this study was further narrowed down to existing data registers that have wide EU coverage, or from sectors whose activities are present in most of the EU Member States and/or could be deployed. By using this approach, the following data sources were analysed in more detail:

• European injury data base (IDB) (full data set (FDS)).

• Online news.

• FRS intervention reporting.

• PC records.

• Fatality statistics.

• Consumer complaints.

• Private economic operators (toy manufacturers, insurance companies).

Potential of EU IDB-FDS for long-term support to consumer product safety

The EU IDB is a network which is running a database that holds standardised data on the external causes of injuries treated in emergency departments across EU.27 IDB was conceptualised in the 1980s for collecting data on home and leisure accidents (European Home and Leisure Accidents Surveillance System (EHLASS)). Its development was triggered by the success of the NEISS of the US CPSC in the 1970s. IDB’s initial purpose was primarily consumer protection, by the use of a well-developed data collection set called FDS to support consumer products safety policy.28 The classification system of the external causes of injuries, including the registration of objects involved, used in IDB is based on EHLASS (V.2000) and ICECI V.1.2 (WHO International Classification of External Causes of Injuries). It allows for a breakdown into a 1000 codes and 20 product groups (eg, furniture, child products, household appliances, items for personal use, sport equipment).10 Variables collected in the FDS are listed in table 3, showing this system’s capacity to code various ways of involvement of an object in an injury event (in bold). It is worth mentioning that the system does not provide information as to whether an injury has been caused by the product’s lack of safety, or its misuse, unless such information is provided in a narrative (as a free text) (table 3).

In addition to the FDS that includes external causes of injuries and codification of objects involved in the incident, the IDB also contains the minimum data set (MDS) that is less costly to collect and contains data relevant for health and national health indicators. Data collection (FDS and MDS) is carried out by national authorities from hospitals/emergency departments who are members of the IDB network, which is coordinated by the European Association for Injury Prevention and Safety Promotion (EuroSafe). Once a year, the data are transferred from network members to a partner at Farr Institute (Swansea University, UK) for the quality check and, consequently, are fed into the database, which is hosted by the European Commission’s Directorate General for Health and Food Safety (SANTE).27 In 2016, according to the report published by EuroSafe, 10 Member States and Turkey uploaded their FDS into the IDB database.29 The number of participating members and hospitals varies slightly from year to year. Despite funding and operational challenges, the data from IDB-FDS are still used by some national authorities (primarily within the Ministries of Health). Over the years the IDB network has been cofunded under various EU funding schemes, such as the EU FP7 project INTEGRIS, JAMIE joint action and the BridgeHealth project, which ended in 2017, thus causing an uncertainty for its future operation.30 Currently, the use of data from IDB-FDS for product standardisation or research purposes is limited by strict data protection policy. Table 4 gives a list of the Strengths, Opportunities, Weaknesses and Threats (SWOT) which should be taken into consideration in the case of assessing the long-term potential of this network to provide data on product/object-related injuries, within its FDS.

Online media text mining using the EMM

Online news-based syndromic surveillance systems and the use of online review articles have already been developed for capturing information on public health threats and providing support to decision-makers.31–33 On this assumption, a pilot study has been carried out to automatically monitor online news articles for 3 months (29 September 2016 to 04 January 2017) with the aim to identify unsafe products, safety concerns or injury events. The EMM, an in-house developed tool, was used for monitoring selected online news outlets. The main operation principle of the EMM is based on setting a number of keywords, which when identified in an online news article classify and ‘capture’ the article as being of potential interest. News articles were classified as potentially being of interest only if at least one keyword from each four groups (see Methodology section) appeared in the online article. The ‘Toy’ product category was designed for the detection of safety concerns or injuries with toys, rather than generic items like ‘consumer products’, since preliminary analyses showed that monitoring of such a vague category as ‘all consumer products’ may result in many of incidents being undetected (due to large number of products for which suitable ‘Product’ keyword should be created). For the ‘Toy’ category, the EMM automatically collected approximately 250 articles in 3 months. Regular assessment and adjustment of the selected keywords finally resulted in a stable percentage of around 50% of relevant results which is likely the best result that was possible to obtain. In the same period, the percentage of false positives (news article is out of scope) dropped to a level of around 20% which is considered normal by the experienced users of EMM. The percentage of non-relevant results (contain all keywords, but not useful to product safety work) continuously decreased with each keyword list refinement, reaching around 30% in the last evaluations. Relevant articles predominantly contained information on product recalls, injury incidents, near-miss events or warnings on safety issues.

In conclusion, our study indicates that EMM has the potential to detect real-time news reporting about safety concerns or incidents related to the specific products, by focusing on a specific product category. If necessary, the system can be customised for other languages. Some limitations need to be taken into consideration:

▪ Despite the tool collects news articles automatically, results must be verified and analysed manually (to avoid misinterpretation or information misuse).

▪ Only the most serious accidents or sensationalised news find their way to the online news sources, therefore this source cannot be used for systematic or reference data collection.

▪ Manual update of new product search keywords may be necessary (eg, ‘fidget spinner’, ‘quadcopter,’).

SWOT analysis of this source to provide information on consumer product-related injuries/accidents is given in table 4.

FRS intervention reporting in support of consumer product safety

FRS are often the first service to arrive at a serious accident, and their national or regional services maintain in-house registers for fire-related interventions. Main objectives of the fire intervention reporting are national statistics (eg, number of firefighters’ interventions per year, resources allocation, insurance claims’ predictions, firefighters safety). Based on the survey information and assessed data registers 39-48 (table 2), it has been observed that the data collection format for fire interventions is not standardised throughout Europe. Differences in intervention reporting can also be found between different regions of one Members State.

Intervention reports usually contain information on the cause of fire and cause of the ignition. If the cause of ignition was the failure of a machine or a device, more details can be provided. In addition, some systems allow ‘free-text’ option for describing the cause of fire, which could include the comprehensive information about the faulty/misused consumer product, however detection of unsafe consumer products is generally of low priority in FRS intervention reporting and is carried out on a voluntary basis (optional) by the responsible fire officer. Some advanced and centralised systems for data collection have been reported to us from the survey respondents. One such system is PRONTO (Finland), a database that collects data from 22 rescue departments, contains detailed accident narratives and may contain information on the products involved, so that the Finish national authority for consumer safety and researchers can already make use of these data. Other interesting examples are Swedish IDA (//ida.msb.se), Danish Online Data Registration and Reporting System (ODIN), Incident Recording System in UK, BRIS system in Norway, to name just a few, which all to a different extent collect information on the product/machine/activity that caused a fire. In some cases, establishing the cause of a fire is determined by the police investigation, which makes completing FRS intervention reports more complex. Our interviews indicated that there is a willingness by certain firefighter associations to promote moving towards more harmonised data collection in the future, which could allow the harnessing of useful information on faulty or high-risk products by many other stakeholders. Unlike some other statistics collected under national authorities, there are no guidelines from Eurostat on the collection of fire incidents data.

The availability and accessibility of the information from FRS registers also varies between countries. Some countries provide part of their registers to the public through online collaborative portals such as Nordstat (www.nordstat.net).

The relevance of FRS intervention reporting for product safety purposes could be increased by harmonising collected information, with a focus on identification of faulty/misused products and circumstances that have caused fire incidents (table 4). Further exploration of the potential and the cost of this sector to provide information on high-risk products involved in fire incidents could be undertaken by means of a pilot project.

Potential of PC registers across EU to provide data on unsafe consumer products

PC play an important protective role in the use of products containing chemical mixtures. In case of exposure to hazardous chemicals, they provide medical advice to general consumers and physicians. It has been estimated that on average these services at the EU level receive and treat 600 000 calls per year, which is almost 1700 calls per day, mostly related to child exposure. The number of fatalities related to chemical exposure is considered to be over 400 per year. Many Member States operate more than one PC, and in some countries they are organised on a regional level.34

Poisoning incidents are usually documented following a qualitative methodology based on internally standardised documentation of cases. Which data should be collected as a minimum standard has frequently been discussed by members of the European Association of Poison Centres and Clinical Toxicologists, but no formal agreement at the EU level has been reached so far. All PCs typically collect a basic set of information that are essential to do the risk assessment and provide the caller, as soon as possible, with the relevant treatment information. The collected data include some information on the substance or product category involved (eg, home detergents and cleaning products, cosmetics) and its toxicological ingredients/quantities.

Companies placing mixtures on the market have the legal obligation, according to Article 45 of the Regulation (EC) No 1272/2008 (CLP Regulation) to inform appointed bodies (Governmental Authorities or Poison Centres) about the composition and the toxicological properties of mixtures classified as hazardous. On 22 March 2017, the Commission adopted the Regulation (EU) No 2017/542, which amends the CLP Regulation by adding an Annex harmonising information related to emergency health response. The Regulation will require producers and importers of chemical mixtures (such as detergents, paints and household chemicals) to provide information on the product composition, through a Unique Formula Identifier on the label of the product, thus allowing a precise and rapid identification of its specific chemical formulation by an emergency health response operator (eg, PC). Regulation (EU) No 2017/542 is going to fully apply from the 1 January 2020 for chemical mixtures for consumer use.

At present, PC registers may only be used to meet medical demand by formulating curative measures and were requested by the Member State to undertake statistical analysis to identify where improvements to risk management measures may be needed. Confidentiality issues concerning individually registered cases likely concern personal data regulated in both national and European legislation. Accessibility could be explored on condition of anonymisation of the personal information. Considerations on the strengths and weaknesses of this data source to support product safety work are presented in table 4.

In conclusion, data collected at PC contain valuable information on chemical mixtures or liquid consumer products. However, circumstances leading to the intoxication are typically not recorded in a harmonised way and data accessibility is difficult due to confidentiality reasons. Our analyses indicate that direct exploitation of data from PC is not feasible in the short term, without prior attempt for harmonisation among existing data collection methodologies and clarification about the conditions of data accessibility.

Mortality statistics in support of consumer product safety

Mortality statistics is one of the most complete and accurate epidemiological data collection practices in Europe. In all EU countries, the medical certification of death was made mandatory according to the Commission Regulation (EU) No 328/2011 on COD statistics. Member States are required to provide the data specified in this Regulation to the Commission (Eurostat) within 24 months after the end of the reference year. Coding of the COD is performed by the use of a harmonised methodology and coding system (ICD-9 and ICD-10). Most countries in the world and the EU use the WHO international standard form for describing the COD, which allows for the registration of the underlying event or injury that initiated the chain of events causing the fatal outcome (part I of the WHO COD form). The extent of detail on the non-medical circumstances of the injury event (external cause of morbidity and mortality), in the COD, is usually poor and does not allow for simple identification of a potentially involved product or a product category from this type of data source. Trends and inconsistencies in filling the mortality certificates have already been reported and efforts to increase the accuracy of the information recording are ongoing.35–37

For example, based on the Eurostat shortlist of external causes of morbidity and mortality, ICD-10-coded information such as the one listed below could be identified from the death certificates:

W20 ‘Struck by thrown, projected or falling object’,

W21 ‘Striking against or struck by sports equipment’

W36 ‘Explosion or rupture of a gas cylinder’

W87 ‘Exposure to unspecified electric current’, etc

The above listed examples show that the information on the cause of the mortality (by the use of ICD-10) that can currently be recovered from the death certificates is not specific enough to support consumer product safety work. In rare cases, some indication on limited range of product categories (eg, sports equipment) may be retrieved (if the form is properly filled), but no specific details on the product can be recovered. Additional death investigation studies would be necessary for obtaining more details on the circumstances or products involved in the fatal accidents, such as using medical records, exploring online news sources or contacting the family of the deceased. Similar work is being carried out by the US CPSC.38 In Europe, the use of information from the death certificates for research or death investigations is still heavily regulated by national laws on data privacy, therefore the use of this potential data source is not feasible in the short term.39 A SWOT analyses of this potential data source is given in table 4.

Potentially, Eurostat’s shortlist of external causes of morbidity and mortality could be amended to satisfy desired granularity or focus on specific type of injury that is involving objects/consumer products. This would require collaboration with a wide range of stakeholders from the Member States. The development of electronic death registration systems may provide more opportunities for data usage from mortality statistics in the future. In addition, the use of the new coding system ICD-11, that is in preparation under the umbrella of WHO, may allow improved coding of external causes of injuries (by linking the ICECI coding system with the ICD module for external causes) and therefore higher disaggregation of collected data.9

Consumer complaints in support of consumer product safety

According to the GPSD (2001/95/EC), Article 9,2 Member States’ authorities are required to provide opportunities for submitting consumer’s complaints on product safety. The survey with CSN was focused on finding out how product safety complaints from consumers are collected and analysed. Eighteen Member States CSN representatives replied to the survey (56% response). The responses indicated that all institutions allow for electronic submission of complaints in various forms (downloadable forms, online form, photos, videos), and also by phone or fax. It has been observed that responsible jurisdictions have different ways of analysing data, categorising injuries and products. Some CSN members reported the use of classifications by in-house coding systems for products which contain elements of RAPEX and international standardisation codes or reported using LOTUS system for registering emails and performing prioritisations based on the severity risk. The number of product-related consumer complaints reported to the responsible authority varied considerably per country. The survey indicated different views on the subject of sharing data on complaints among the members of the CSN. A third of the survey respondents expressed that they would share their data on complaints on unsafe products for the purpose of research or policymaking. The observed strengths and weaknesses of the existing system of consumer complaints is provided in table 4.

An interesting IT system has been reported by a Slovakian member of CSN on teaching citizens how to be proactive in product safety and market surveillance: ECHO system–consumer information on dangerous products (available at soi.sk). An information system called ECHO was launched in 2007 with the objective to gather information on accidents caused by the use of any non-food products in the household, in leisure and in sport activities. The system has an educational element in itself to promote more active communication with the consumers and the market surveillance body.

Results of the survey suggest that national authorities should continue to operate and innovate their systems for collecting consumer safety complaints. The quality of product-related injury/accident data could be increased by harmonising the procedures/methodologies for receiving safety complaints from consumers and then potentially forwarding details of the incident and product concerned to an EU-wide integration platform. The obstacles for anonymised data sharing should be explored.

Can the private sector contribute to data on product safety?

Data needs

When developing a new data infrastructure, preliminary analyses of the stakeholders data needs are necessary in order to understand if target objectives by different stakeholders can be met (eg, increasing safety of a product, injury prevention, risk reduction, monitoring of legislation implementation, standardisation work).

Detailed analyses of methodological approaches and pitfalls to consider before setting up systematic collection of statistics for the safety of services have been previously published in a report commissioned by the European Commission’s DG SANCO.42 Based on this report, some considerations on the data/information needs by various stakeholders have been indicated in table 5. Defining the minimum, but sufficient number of parameters and quality of data, that are fit for purpose, is essential for rationalising data collection efforts. This is equally valid for developing new data infrastructure or deployment of the existing databases.

A platform for remote data access

Based on the priority data sources that have been identified and assessed here as being of potential interest to the consumer product safety work, we propose to put in place an IT platform for remote access of data in a modular-based fashion. A successful example of a distributed remote data access via IT platform, in the area of environmental chemical occurrence data, is the European Commission’s Information Platform for Chemical Monitoring (IPCHEM). IPCHEM has been recently recognised by the Parliament as an initiative reducing the knowledge gap on the environmental data, related to the implementation of the 7th Environment Action Programme.43 IPCHEM allows remote access to data from various providers such as European Agencies, Members State authorities and research institutions who committed to support data sharing policy, in the spirit of collaborative initiative called Shared Environmental Information System (SEIS).44 Similar to IPCHEM, the possibility of modular interlinking of structured, semi-structured and unstructured data, which can be accessed remotely, while taking advantage of IT technologies, could be applied in the field of consumer product safety. This would imply working with data coming from various data sources, such as fatality statistics and consumer complaints, complemented with relevant online news articles/customer reviews (eg, EMM-retrieved information) for detecting emerging risks and trends. In the long term, other sources such as harmonised fire and rescue service records/PC or information supplied from the private sector could also be integrated, if required. Basic data flow and options for remote data access or the use of online and offline data sources delivered by data providers are visualised in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Proposed data flow diagram and options for data access by remote linking of potential data sources onto a common information platform

Furthermore, novel IT paradigms of interlinking different data sources such as the Semantic Web using a Linked Data approach could be explored.45 However, it should be stressed that such option would require major collaborative efforts (technical, organisational and in terms of governance) across participating data infrastructures and information systems.

In our estimate, the functional use of data from assessed sources, apart from EMM, cannot be achieved in the short term (ie, less than 5 years). Due to limited time and resources available for this project, it was not possible to perform a cost-effectiveness analysis of the various options. However, such analysis is necessary to be able to conclude on the feasibility of the proposed options.