Advancing Science Through Collaborative Data Sharing and Synthesis

Abstract

The demand for researchers to share their data has increased dramatically in recent years. There is a need to replicate and confirm scientific findings to bolster confidence in many research areas. Data sharing also serves the critical function of allowing synthesis of findings across trials. As innovative statistical methods have helped resolve barriers to synthesis analyses, data sharing and synthesis can help answer research questions that cannot be answered by individual trials alone. However, the sharing of data among researchers remains challenging and infrequent. This article aims to (a) increase support for data sharing and synthesis collaborations among researchers to advance scientific knowledge and (b) provide a model for establishing these collaborations using the example of the ongoing National Institute of Mental Health’s Collaborative Data Synthesis on Adolescent Depression Trials. This study brings together datasets from existing prevention and treatment trials in adolescent depression, as well as researchers and stakeholders, to answer questions about “for whom interventions work” and “by what pathways interventions have their effects.” This is critical to improving interventions, including increasing knowledge about intervention efficacy among minority populations, or what we call “scientific equity.” The collaborative model described is relevant to fields with research questions that can only be addressed by synthesizing individual-level data.

What Drives Academic Data Sharing?

“…Many other authors in our sample examine the issue of competitive misuse [5,22,23,46,48,[102][103][104]. Another issue regarding the recipient's use of data concerns a possible flawed interpretation (e.g., [21,59,89,105,106]). Perrino et al [59], regarding a dataset from psychological studies, state: "The correct interpretation of data has been another concern of investigators.…”

“…Huang et al [35] argue that making data available is no obligation with many funding agencies and that they do not provide sufficient financial compensation for the efforts needed in order to share data. Perrino et al [59] argue that funding policies show varying degrees of enforcement when it comes to data sharing and that binding policies are necessary to convince researchers to share. The National Science Foundation of the US, for instance, has long required data sharing in its grant contracts (see [60]), "but has not enforced the requirements consistently" [61].…”

“…Another issue regarding the recipient's use of data concerns a possible flawed interpretation (e.g., [21,59,89,105,106]). Perrino et al [59], regarding a dataset from psychological studies, state: "The correct interpretation of data has been another concern of investigators. This included the possibility that the [data recipient] might not fully understand assessment measures, interventions, and populations being studied and might misinterpret the effect of the intervention".…”

Does the conclusion follow from the evidence? Recommendations for improving research

“…Similar proposals will have the same effect. These include recent calls for posting raw data (Perrino et al, 2013;Simonsohn, 2013), and greater transparency in disclosing the use and effects of a wide range of research practices (Funder et al, 2014;Sagarin, Ambler, & Lee, 2014;Simmons, Nelson, & Simonsohn, 2012). Journals are increasingly asking researchers to actively state that they have disclosed important research details (Eich, 2014).…”

Open access to research data: Strategic delay and the ambiguous welfare effects of mandatory data disclosure

“…However, it may also be the case that the data has a higher value for R if it remains her private information. For instance, loss of control associated with data sharing can result in data falsi…cation or ‡awed interpretation that may negatively a¤ect R's reputation (Costello, 2009;Perrino et al, 2013). Thus, even though we assume that data disclosure always has a positive e¤ect on the value of the published article that is based on this data, withholding the data after the …nal publication may be the optimal strategy for R in our model.…”