Table 2.

Potential registration levels for epidemiologic research

1No registrationCurrent predominant paradigm; may continue to be common, but novel published results from such studies should be seen primarily as exploratory analyses requiring confirmation
2Dataset registrationShould be feasible to achieve in large-scale; each dataset registers the variables that it has collected and their definitions; this would allow knowing how many studies with how many participants who have measured variables or markers of interest, instead of guessing what data are available on that marker beyond what has been published
3Availability of detailed dataIndividual-level (raw) data are made available; this practice may be subject to policy/consent/privacy constraints for past studies and their data; easier to anticipate and encourage in the design of future studies
4Availability of data, protocols, and analyses codesOptimal ability to evaluate the reproducibility of analyses, to maximize the integration of information across diverse studies, and to allow improvements on future studies based on exact knowledge of what was done in previous studies
5Live streaming of analysesInvestigators not only post all their data and protocols online, but analyses are done and shown in realtime to the wider community as they happen. Live streaming can be coupled with crowdsourcing of analyses across large communities of analysts