Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
54 8 Discussion The roadmap described in the previous section is divided into steps, which can be thought of as decision points as well. However, the exact path can be different in every state. This section contains additional information and comments on the process that do not fit exactly into the steps in the roadmap, but could be useful in carrying them out. 8.1 Making Progress in Parallel First, even though steps are numbered, some can be done in parallel. In particular, developing and setting up a data warehouse is generally independent of determining how to link databases. For example, a linked dataset could be stored separately until the component databases are added to a data warehouse. Rows that are linked probabilistically can be assigned permanent numeric identifiers in the component databases sometime in the future. Data harmonization can be done at the same time that identifiers are being selected. Probabilistic linkage can even be used while on-scene identifiers are being incorporated into the data collection process. A good system for incorporating person-specific numeric identifiers is less subject to bias than probabilistic linkage. However, putting a process in place and establishing its use across all state agencies is time-consuming, and probabilistic linkage can be used in the interim to improve measurement of serious injuries. Finally, a legal review related to linkage may be time-consuming. This process should start early and can proceed in parallel with the technical development. 8.2 Benefits of a National Standardized Schema and National Datasets Among those we interviewed who were associated with EMS data management and linkage to crash or trauma, a repeated theme was the benefit of having the national NEMSIS database schema for EMS data. Although states have many challenges in getting all EMS units, including volunteer units, to provide data, all vendors that provide software for EMS data collection output data according to the same schema. The Highway Performance Monitoring System is a national dataset that provides data on characteristics of the nationâs highways. States submit data, which are standardized at a national level, and the data are compiled at the national level. This system is analogous to NEMSIS in that the schema is standardized, but the coverage of roads and data elements are more limited in the national sample compared to state roadway datasets. Trauma registries have the NTDS to provide a common structure for their datasets. States are encouraged to move their crash datasets towards the newest MMUCC standard, though compliance is less consistent without a national dataset or XML schema (though one is being developed). The presence of a national standard, encoded in XML, and a national database tends to speed up the process of standardization of datasets in an area. This, in turn, makes change more efficient because states can share developments. Crash datasets and data linkage to crash would benefit from the kind of coordination that a national dataset and standardized XML schema represents. Another benefit of national schemas and datasets in all domains is the ability to share across state borders. Smaller states and those with large cities on the border tend to have many crash cases involving drivers from out of state (for linkage to license files) or occupants who are treated in hospitals across state lines. There is no simple within-state mechanism to handle these linkages.
55 A national schema would mean that if states share data, out-of-state data will look like in- state data and be relatively easy to use. A national dataset would mean that data are available to search for matching records without being arbitrarily restricted by borders. In the meantime, for states that have significant border-related dropout in datasets, the best option is probably to make arrangements with neighboring states to find matching cases. The process for these cases will be different from any in-state system (probabilistic or otherwise), so some level of hand linkage may be necessary. 8.3 Motivators Another theme of interviews was the benefit of having statutes requiring data reporting. These are state requirements, but it provides the state agency in charge of managing the dataset with some leverage to encourage data reporting. For the most part, these statutes are not enforced, but they do provide motivation. Interviewees stressed the need to give information back to the entities providing data. This means emphasizing report generation, not only to meet national requirements, but also to give feedback to the individual organizations that provide data to the state. Timely reporting and timely data upload help those groups benefit from the work they put into the data systems. At the entity level, the motivation for promoting data linkage varies. For departments of transportation (DOTs), MAP-21 requirements motivate the need for crash data to be linked to medical outcome. However, EMS and trauma registries do not generally need crash characteristics for their analyses. For trauma registries, incident location and restraint use are often of greatest interest. In New Jersey, a data linkage project was initiated by a trauma center to capture the incident locations of their crash-related trauma cases. They were interested in targeting educational programs in specific neighborhood schools and centers where certain types of crashes (e.g., pedestrian crashes) are more prevalent. Similarly, knowing when and where crashes happen can help EMS and hospitals better predict future events for resource preparation. There are a number of potential benefits of linking these datasets for all agencies concerned, but it is important to identify and promote these benefits so that agencies are motivated to participate and facilitate. 8.4 Where We Are Now At this point, state roadway, driver license, driver history, crash, EMS, and trauma databases are either in good condition or are being improved in nearly all states. Statewide hospital and ED databases have more variation in coverage and consistency. The exact condition of a stateâs databases matters for near-term progress on linkage, but states are generally moving forward on this front (database completeness and quality). Linkage between EMS and trauma databases is being done in a number of states and tried in others. Linkage between crash and EMS or crash and trauma is planned in many states, but no state that we talked to or surveyed has a fully implemented crash-EMS or crash-hospital linkage that is not probabilistic. The CODES program initiated probabilistic linkage programs in a number of states, and these are being maintained in many cases. Moreover, software developed through CODES, as well as software developed with funds from the CDC are available and help handle some of the technical challenges associated with probabilistic linkage. In principle, probabilistic linkage is not the preferred approach. However, if sufficiently unique identifiers, such as name, are used, probabilistic linkage comes close to linkage using numeric identifiers in performance. In an effort to avoid the problems of MI, states often select
56 only high-probability linkages in their resulting database. However, this introduces bias in analysis that will influence results of performance metrics. Given the strong emphasis in states on probabilistic linkage, it is important to provide states with clear guidance on how to use the approach, how to evaluate the results, and what (if any) versions of a probabilistic linkage system can be considered a final solution to linkage. Otherwise, probabilistic linkage should be considered an intermediate stage on the way to a better approach. Several states are planning and beginning to implement pilot programs to test crash-to- EMS linkage processes that are not probabilistic. As these develop, it will be helpful to learn from statesâ different approaches so that a better-tested âtoolkitâ of approaches is available. Several states have successfully implemented processes to link EMS and hospital data, and these approaches have been described in this document. For the most part, these approaches can be adapted to the crash-EMS linkage process, and indeed, those states are generally considering the possibility.
