How would you write an ethics approval for a study conducted on human data, when the dataset is publicly available?

Question

The dataset was not collected by the authors, but the study was conducted on a publicly available dataset. The source of the data is cited, but the source does not have an ethics statement itself. The data is heartbeat time series.

Answer 1

The following assumes that the analysis is being done as research (i.e., as a systematic investigation intended to contribute to generalizable knowledge) and that the intention is to publish the results of the analysis in a scientific journal.

There are two separate questions here.

What kind of ethics review is required when analyzing publicly available data about humans?

How should the ethical review of data about the analysis of publicly available data about humans be described when submitting a manuscript that reports the results of the analysis?

The answer to each question is complex. The answer to the first question would be modified by the specific requirements of the author’s institution as well as the local (country) context. The answer to the second question would depend on the specific requirements of any journal in which an attempt is made to publish the results of the analysis of the data.

You should clarify your institution’s requirements for review of publicly available data with an expert on these matters at your institution. They are likely to want to know a lot more about the data: when it was collected, how it was collected, whether the humans whose data are contained in the dataset consented to the data collection and to the use of the data for research, and, importantly in secondary analysis, whether there are elements of the data that could be used to identify or potentially identify a specific living human being.

Medical journals vary widely in what they require in an ethics statement. You should think about what journals might be interested in the results of your analysis and review their requirements for an ethics statement. A review of two or three journals that might be interested would give you an idea of the range of requirements.

A “fail-safe” method for assuring that an analysis of publicly available data about humans will not raise concerns about the ethics of use of the data at the journal submission stage (or worse, after publication) is to submit the plan for the analysis to the institutional entity charged with ethics review and have this entity determine what kind of review is required or deem that the analysis is not subject to (further) review. This sounds like (and is) quite bureaucratic but it also protects the analyst(s) from allegations that ethics were breached in analyzing the data.

Answer 2

You should not need ethics clearance at all. There is no ethical impediment to making inferences from publicly available data and reporting your results publicly.

Answer 3

In general you wouldn't need ethics approval to use this data set; it would be covered by the ethics statement from the original experiment. Presumably the original data collectors had one, even if it's not referenced; if you want to be extra-careful, you should try to find the peer-reviewed publication associated with the data set, and look there for the ethics statement.

More generally, it's hard to see how you could be causing harm by using a public, already collected data set; you didn't have any interaction at all with the subjects! The exception to this would be if you had reason to believe that the original data set might not have been ethically collected, in which there would be an argument for not using the data even though it has already been collected. Practically speaking, though, if these are data collected in the last few decades, it's likely that the researchers were required to undergo ethics approval before data collection.

There are lots of edge cases here:

• data that were unethically collected, either decades ago (Tuskegee/Nazis/Stanford Prison study/Milgrom obedience studies etc.) or more recent studies that slipped through the cracks of the modern research-ethics framework
• data that are insufficiently anonymized
• data that are OK in the jurisdiction where they were originally collected but not in yours

If possible, you should be careful and follow @DianaPetitti's advice by checking with your university's research ethics board; hopefully they will tell you you're good to go. (Although I absolutely understand why it is so, ethics boards are often cautious to the point of [researcher] frustration even when the probability of harm is almost vanishingly small.)

Disclaimer: I am not an ethicist and have not been involved in dealing with ethics statements with the exception of a few surveys and student evaluation exercises.

---

Tracking the source of the data back we get to Greenwald, S.D. (1990). Improved detection and classification of arrhythmias in noise-corrupted electrocardiograms using contextual information. Thesis. Massachusetts Institute of Technology. https://dspace.mit.edu/handle/1721.1/29206

We established a test database of 35 half-hour ECG records ... [Some data] are from the MIT/BIH Arrhythmia Database ...
[Some other data] are from the AHA Arrhythmia Database ...
Tapes in the 800 series were collected for this research from long-term Holter records archived in the Beth Israel Hospital Arrhythmia Laboratory ...

This is admittedly a little disappointing; there's no evidence that informed consent was obtained, although from a practical standpoint it seems incredibly unlikely that these records could either be personally identified, or that any significant harm was caused in their collection. (This describes the data that are first presented in the thesis — for complete diligence you'd have to go back farther to the sources of the other two arrhythmia databases listed here and see what they say about sample collection ..)

Answer 4

I suggest to start your research for answers from expert in the field of ethics and data usage. For example:

https://www.ama-assn.org/delivering-care/ethics/release-data-unethical-experiments

Please avoid expecting a meaningful response for such kind of delicate questions.