When epidemiologists are looking at the health experience of a group of people, they often use available computer registries, such as vital records registries or cancer registries. Registries provide information about specific events such as births in a certain state, for example. Registries are records of events (births, deaths, disease, marriages, etc.), or exposures (such as persons exposed to asbestos or heavy metals). State laws require, for example, that those persons who perform marriages report the event to the state’s marriage registry. Similarly, hospitals and health care workers report births to the birth registry and cancer to the cancer registry; laboratories report elevated mercury levels to the heavy metals registry.
Remember from the last newsletter insert on descriptive statistics when we were talking about the occurrence of ear infections in Oz and in Kansas? Let’s say that neither Oz nor Kansas has a registry to record ear infections. We’d have to count cases without a registry, which is a lengthy and costly job. Researchers need to find all Oz children within the defined age limits and years of the study, find out how many and when they had ear infections, verify the information through their doctors, and record all the information. If we had a registry, all this information would have been already reported and available to match its records against a list of Oz residents or to search the registry for Oz children. The registry also provides data to use for comparison, or "background" rates of ear infection.
There will be many predictable errors in an ear infection study done without the benefit of a registry - especially if we are studying several past years (which is desirable from an epidemiologist’s point of view) to get complete data over time. Over the years, children move, grow up and marry, change names, switch doctors, and move out of the area. Doctors also move, retire and lose or destroy their records. Children and parents alike will not remember ear infections with accuracy. The same limitations are present when looking at ear infections in the comparison population of Kansas. Non-registry studies happen and are needed, and can be interpreted reasonably well as long as the limitations are clearly stated.
However, if a registry of disease incidence is available, using it makes good sense since the data are easily accessible, it takes less time to obtain the data, and there are likely fewer errors. Another advantage of using a registry is that minimal or no action is needed on the part of the study subjects. Yet another is that diagnoses and criteria for inclusion and exclusion in a registry are consistent. This lessens the chance of false findings due to things like individual physician’s practices or conventions being inconsistent, or regional variations in how patients’ health care is diagnosed, reported, treated or managed.
Even though using a registry offers more complete observations for less time and money than not using a registry, there are some trade-offs. An obvious one is that you need to choose health effects to study which are reported and maintained in a registry, which limits the choices considerably. Another is that the data quality can never exceed that of the registry, and so it is dependent upon the completeness of reporting and how well and how promptly the registry maintains its data. For example, there may be errors in how an event like cause of death is recorded on a death certificate. Such errors may make it harder to identify health effects associated with certain exposures, but are inevitable when registries are used. Another limitation in the use of a disease registry is that real matches might be missed due to errors in recording identifiers such as social security number or date of birth. If the date of birth were entered wrong, the match might never be noticed.
We are using registry matches in the Love Canal study to obtain the most complete information we can about former residents. We will also take steps to reduce some of the limitations. We requested "loose matches" from the registries instead of exact matches on all identifiers. This way we received many records to look at more closely, for things like incorrect but close social security numbers or dates of birth. If enough information matched so we are confident that the person was the same as the previous Love Canal resident, we would include that event which would otherwise have been missed.
The Love Canal Medical Fund Board urged us to find ways to increase our ability to study reproductive experiences of Love Canal parents, especially births before the Congenital Malformations Registry was established in 1983. We are looking at reports of children with birth defects made by parents in 1978 and verified by Health Department staff through medical records. The medical records were reviewed, but the classifications used were not consistent with those used today. Therefore, we asked staff who now code birth defects for the Congenital Malformations Registry to code the old events as they would currently. This maintains data consistency and objectivity. Another suggested improvement being used is to search the marriage license data base and identify marriage names for Love Canal mothers and wives’ names of Love Canal fathers to get the most complete reproductive information possible.
A related problem associated with using registry searches to capture health events is the fact that many former Love Canal residents have moved outside of New York State for periods of time. Registries are maintained by other states, but unless we know in which states former Love Canal residents lived, we will not know to look for information from those states. Each state has its own criteria for inclusion in the registry, how the data are coded or accessed, etc. Some states do not even have the registries we seek, which further lowers our chances of capturing these events.
Registry matches provide valuable information about the group being studied and others against which to compare our findings. Knowing the strengths and limitations helps interested readers to better understand the results of registry studies.
|Send questions or comments to: firstname.lastname@example.org|
|Revised: July 2001|