Investigating the Performance of FDA Databases in Terms of Reporting Medical Device Problems 1 - The
Medical device reporting (MDR) is part of the Food and Drug Administration (FDA)s strategy for post-market surveillance of medical devices. There are four main databases which have medical device reports of adverse events: the Manufacturer and User Facility Device Experience (MAUDE) database, the MDR database, the Medical Device Recalls database, and the Total Product Life Cycle (TPLC) database. This investigation into the utility of the FDA databases focuses on these four databases (See the following flowchart). The MAUDE database houses also data from now discontinued MDR database, so in this article, we only study the MAUDE database, the Medical Device Recalls database, and the TPLC database.
The different databases were compared by looking at 30 medical devices through the case studies method and analysed by the content analysis method. Data validation was achieved by triangulation and member checking.
Before discussing the utility problems, there are two subtle points that should be emphasised. Firstly, there is limited information about the specific users of these databases, making it difficult to identify the target users. However, in many papers, these databases, especially the MAUDE database, are widely used by researchers to identify specific devices problems. It is likely that users are either manufacturers, researchers, or medical device designers. Researchers, or even medical device designers, who may not have a great deal of experience in manufacturing, tend to focus on searching for devices problems rather than manufacturer information. However, the MAUDE database is not ideal for these users. Secondly, compatibility problems are found through comparing search queries, which should give identical or similar results as they can be selected to access the same data. Therefore, when comparing, we did not use the search fields unique to each database; these are fields such as the ‘Brand Name’, the ‘Manufacturer’, the ‘510K Number’, the ‘PMA Number’, the ‘Event Type’ in the MAUDE database; the ‘Regulation Number’ in the TPLC database; the ‘Recall Class’, the ‘Recall Number’, the ‘Recalling Firm’, and the ‘Reason for Recall’ in the Medical Device Recalls database. What is more, due to the Medical Device Recalls database only containing data on recalling of devices, its data is limited, and it cannot be compared to the MAUDE database or the TPLC database. Thus, we focused on the analysis of the MAUDE database and TPLC database. As these two databases have the same medical device data, they should provide the same adverse event reports when queried.
1. Search compatibility problems between the databases
Compatibility problems were investigated by searching data with varying search queries through three exploration studies. In the first exploration, there were two search queries used to compare the result - a specific device name and the specific date: using the ‘Product Class’ search box in the MAUDE database; and the ‘Device’ search box in the TPLC database. The report received year is since 01/01/2004 till now. By comparing the results from the MAUDE database and the TPLC database, it was found that typically the number of results do not match. This should not be the case as the data only came from the MAUDE in the both cases. The issue of non-matching search results also happens when other medical devices were searched. What is more, from using different search queries in the two databases, we can see the MAUDE database results focus more on the manufacturer, and the TPLC database results focus more on device problems, even if they sample the same adverse events of medical devices reports data.
In the second exploration, the FDA product code is an important entry when searching data, because they identify the generic category of medical devices. That means the 'Product Codes' search box can be used for the both databases to find out problems of medical devices in a specific device category. So, the FDA product code and the specific date (the report received year is since 01/01/2004 till now) were used to compare the results. However, the results from the MAUDE database and TPLC database still did not match each other.
In the third exploration, by considering the unsatisfactory search results, two or more than two search queries were used to explore the relationships between the MAUDE database and the TPLC database. We could find that there were same results that could be found from these two databases.
Through the three search explorations, it was found that there are several approaches which can be used when searching the TPLC database. The first approach is likely used by people who do not have a lot of knowledge or experience concerning category information on medical devices. In this case, users just need to type the device that they want into the ‘Device’ search field. Another approach is where users also type the name of the category into the ‘Device’ search field; for example, ‘Atrial Defibrillator’ could be used to find out device problems, which could be used as the second search approach. However, the ‘Device’ field can only hold 23 characters, so it is not suitable for medical devices with a long name or long category names. The third approach is only to type a product code into the ‘Product Code’ search field. The last approach is to use search fields concurrently. All four search approaches can be used in TPLC give the same final data.
This is opposed checking matching data between the MAUDE database and the TPLC database, where the exploration studies failed to give matching search results. As the TPLC database always provides the same results with different search approaches, we can deduce that when failing to get the same data from the MAUDE database, the TPLC database is incompletely searching the MAUDE database. That may suggest that if users, particularly users who do not have enough knowledge and experience about manufacture, who want to get accurate data on adverse events with a medical device from the MAUDE should fill in as many fields as possible.
2. Interaction interface problems between the FDA databases
The interaction interface problems have two main aspects: problems with consistency between the interfaces in different databases, and problems with the search interface in the MAUDE database.
Firstly, different words are used for the same meaning for the search fields in MAUDE, TPLC, and Medical Device Recall databases. For example, in the MAUDE database, ‘leak’ can be found from the ‘Product Problem’ field, but it in the TPLC is found under ‘Device Problems’ in the second search stage. In the MAUDE database, ‘hemodialyzer, re-use, low flux’ is entered into the ‘Product Class’ field, but in the TPLC database, it must be entered in the ‘Device’ field. Furthermore, in the Medical Device Recalls database, the main field, ‘Recall Class’, has a totally different role to the ‘Product Class’ in the MAUDE database, but has the same role as ‘Device Class’ in the TPLC database. The only unified search query is ‘Product Code’. The different labelling of search fields clearly obstructs the utility of the databases, and sometimes it may lead to mistakes. Therefore, it is recommended the search fields should have unified language to improve their utility.
Secondly, in the MAUDE database search interface, there are a great many device categories listed under ‘Product Class’ field, and there are also a great many product problems listed in the ‘Product Problem’ field. Users need to select a device category from the ‘Product Class’ to match a product problem from the ‘Product Problem’. For manufacturers, if they have the knowledge and experience on a specific product problem, a specific product category, and specific manufacturers’ information, they would likely find results very quickly. However, some researchers, designers, or other users may not have the experience to match a device category to a device problem. If they want to find relevant data for the adverse events of medical devices in the MAUDE database, it is likely they would waste a lot of time finding what they were looking for.
Compatibility issues were found to search consistency problems and interaction interface problems. Although the causes of these problems were attempted to be investigated in this study, the FDA database is a complicated and large system, and the reasons for these problems are difficult to answer. It was assessed that these problems might be caused by the uncompleted search fields, limited data or some data being missing, or caused by data being from a different source.