Data Set Linkages

EMS systems should track patient outcomes into the hospital and beyond. One method for obtaining patient outcome data in EMS is to link together large databases that describe different stages of the continuum of patient care. For example, a statewide EMS database might be linked with a financial dataset that describes inpatient hospital charges, and that database may in turn be linked to a death registry. In theory, such linkages allow researchers to follow a patient from the prehospital phase, through hospitalization, and after discharge. In fact, such linkages are challenging to create. The patient’s name is often stripped out of datasets to preserve confidentiality and other identifying information, such as the patient’s home address, may be missing because those providing care did not have it at the time the records were created. To tackle this problem, enterprising researchers developed a technique called probabilistic matching.⁹³ EMS investigations have used this technique, including examination of the impact of EMS on children with special needs⁹⁴ and linking hospital trauma registry data with prehospital records.⁹⁵

In addition to the technical challenges posed by incomplete data, the best outcome variables may not be recorded in available data sets. Since medical records and database structures are designed independent of specific research questions, key information is often incomplete or simply altogether absent. One temptation is to use the information that is present in the database in an attempt to get as close to the answer as possible. The problem is that this approach can give results that are not meaningful because the most appropriate outcome variables have not been measured.

Another problem is that elements of the health care system may be reluctant to share information. Maintaining patient confidentiality is a major issue. For example, matching a zip code and date of birth in a large database can uniquely identify about 15% of subjects. Some privacy advocates maintain that if a researcher can use a data set to violate the privacy of even one person, then the data should not be collected.

Patient privacy is an important issue in EMS research. Recently the Department of Health and Human Services has developed recommendations to protect against the disclosure of identifiable patient information. The impact of these new privacy regulations on the linking of patient data and its availability for research purposes remains to be seen. These rules may become an additional obstacle to the effective evaluation of prehospital interventions; or they may establish a level of privacy protection that adequately alleviates concerns among the public, thus facilitating advances in clinical research.

One potential solution to the problem of maintaining patient confidentiality is to assign a longitudinal patient identifier. For example, in the State of Washington, trauma patients are given a bracelet with a unique identifying number that remains with the patient throughout the process of care. That number is kept with the medical record but the patient’s name and address is not maintained at the state level, thus preserving confidentiality since the unique number but not patient identifying data moves from the hospital or EMS agency to the state.

Another important regulatory issue that needs to be considered by researchers is a proposed change in the freedom of information law that would allow requests for access to raw research data collected for federally funded research projects. This proposal has several implications. For one, the confidentiality of the study subjects might be compromised. There is also a potential problem with protection of the raw research data from a legal discovery process. If EMS systems and health care providers are going to undertake serious evaluations of their practices in order to improve the care they provide to their patients, they must be assured that the information gathered in that process won’t subsequently be used to support litigation against them.⁹⁶ One possible solution to this problem would be the availability of a “federal certificate of confidentiality” issued by the Office of Management and Budget.

It is useful to link outside data, like law enforcement records about motor vehicle crashes, to EMS and hospital data.⁹⁷ It is also sometimes helpful to link to payment data sets, such as those used in the medical expenditure panel study or the payment databases of health insurance plans. These linkages also raise confidentiality concerns that must be addressed by EMS researchers.