Sage Journals: Discover world-class research

Abstract

Background:

This study shares insights from clinical trialists who have conducted investigator-initiated trials that have linked trial data to administrative data, focusing on the challenges and facilitators of this approach.

Objective:

To provide recommendations for evaluating the feasibility and suitability of using administrative data in clinical trials.

Method:

A convergent parallel mixed-methods study was conducted, surveying Australian clinical trialists and operations staff. Participants could opt-in to in-depth interviews. Survey data were analysed using descriptive statistics, while thematic analysis was applied to interview data, with findings integrated during interpretation.

Results:

Four main themes and 10 sub-themes were identified as critical when evaluating the suitability of administrative data for clinical trials: (i) “trial management considerations” covers operational factors like budgeting, timelines and staffing; (ii) “assessing burdens vs. gains” encourages weighing up the research benefits with the additional operational and consent considerations; (iii) “data preparation and analysis” addresses the processes involved in preparing and analysing data for linkage between trial and administrative datasets; and (iv) “training and support” emphasises the need for researcher support when using linked data.

Conclusion:

Researchers should carefully evaluate the feasibility of using administrative data, considering costs, required skills, timelines and data accuracy. They must also be prepared for delays due to data request processes, participant consent requirements and the mandated use of data access platforms. Early planning can mitigate later complexities.

Implications for health information management practice:

This study highlights the value of health information managers in clinical research, particularly in managing electronic health records and clinical coding. Their expertise in these areas, as well as in data governance and system architecture, can support clinical trials that link to administrative data.

Keywords

clinical trials record linkage electronic health records information systems pragmatic clinical trials health information management real world data administrative data routinely collected data

Introduction

Linking clinical trial data with administrative data, or routinely collected health data, has great potential worldwide for answering research questions. Administrative data refer to information collected by government departments, businesses and organisations for reasons other than research, such as registrations, billing and record keeping (Australian Bureau of Statistics, 2021), and in this paper, we use “linkage” to describe the process of linking clinical trial data to administrative data. The drive to conduct trials more efficiently, reduce costs and lessen the burden on trial sites is a common goal for trial units. Administrative data are often viewed as a cost-effective alternative for clinical trial data collection, and it can provide data, such as health service utilisation and medication adherence, which are often challenging to obtain through traditional methods.

A scoping review of administrative data use in Australian randomised controlled trials (Fahridin et al., 2024) reported a steady increase in publications that linked routinely collected health data to clinical trial data. While the reviewed articles identified several barriers and facilitators to administrative data use, these were limited to the information available in the included studies and did not address the practical use of the datasets or how challenges were overcome. The review found that there were many more studies planning to access administrative data than there were publications, but reasons for this discrepancy were not explored. Guidance around the use of linked administrative data for research purposes is available through state and federal data custodians, or the national data linkage unit network (Flack et al., 2019); however, a knowledge gap remains as to how to best work within the known limitations of administrative data.

In this study, we aimed to share learnings from clinical trialists who have conducted investigator-initiated trials that linked clinical trial data to administrative data, and to understand the challenges and facilitators to its use, while providing suggestions for evaluating its feasibility and suitability for a clinical trial.

Method

Study design and sample selection

A convergent parallel mixed-methods design was used (Katz-Buonincontro, 2024), integrating responses from a survey with interview data from a subsection of the original sample to achieve a comprehensive understanding of administrative data use. The target population were individuals who worked in academic clinical research and who had experience linking clinical trial data with administrative data at any stage of the linkage process. Purposive sampling was utilised to recruit potential participants, through an invitation to complete the anonymous online survey, which was advertised within relevant trial, network and association newsletters, as well as snowball sampling (Braun and Clarke, 2013), where respondents were encouraged to share the survey within their networks. The survey was open for 10 weeks, between March and May 2024. Survey respondents provided consent before completing the survey. (For a copy of the survey, see Figure S3, online Supplemental File). At the conclusion of the survey, all respondents were invited to participate in a single in-depth interview. Those who agreed were asked to provide an email address on a separate form to ensure the anonymity of their survey responses.

Survey data collection

For the survey, a REDCap (Research Electronic Data Capture; Harris et al., 2009, 2019) questionnaire (Supplemental File 3) hosted at the University of Sydney, Australia, was designed to collect responses by mobile phone/tablet or desktop browser. Survey questions collected information via a combination of structured response options (such as yes/no or multiple choice) and open-ended free-text responses regarding participants’ responsibility in the data linkage processes, available support, required skills, storage/accessibility considerations, cost and barriers/enablers to linkage. Participant details such as occupation, employer and experience with data linkage were also collected.

Interview data collection

An information sheet was emailed to the subsample of respondents who had indicated that they were willing to participate in an interview, together with a link to the interview consent form for completion and so they could nominate the times they were available to be interviewed. The interviewer arranged for interviews to take place within 1–3 weeks of the survey completion date. Semi-structured, one-on-one, in-depth interviews were then conducted with this subsample of interview participants. The interviews took place via video call from their usual workplace (university, health service, or home) and lasted between 20 and 64 minutes (average of 45 minutes).

Interviews were based on an interview guide (see Figure S1, online Supplemental File 1), which had been piloted previously with three colleagues who were experienced in clinical trials that linked to administrative data. These pilot interviews were not included in the analysis. With the consent of participants, interviews were audio-recorded using Zoom (2024) and were conducted by one researcher (SF). The interviewer was a health information management graduate and PhD candidate at the time of the study, who also held a leadership role in overseeing data management operations at a large academic clinical trial coordinating centre. This professional context informed the motivation for the research, which was disclosed to the participants at the beginning of each interview. The interviewer had also undergone postgraduate qualitative methods training prior to conducting interviews and was known to some interview participants through existing working relationships. Interview participants were advised that any identifying information in their responses would be anonymised and their responses would not impact their relationship with the interviewer or the organisation at which she (SF) was employed. The interviews focused on the experiences of the participants in requesting, obtaining and using administrative data, specifically, barriers and facilitators to these processes, as well as on the purpose of using the data, support received and overall attitudes about linking trial data to administrative data. The semi-structured interview format allowed participants to speak freely.

Transcripts were produced directly from the audio recordings and were manually checked for accuracy by the researcher (SF). All comments and responses were anonymised during the process of transcription. Transcriptions were emailed to participants to confirm accuracy. As with the survey data, anonymised transcripts of the interviews were stored in the University of Sydney’s Research Data Management System and accessible only to the study team. Data saturation was discussed over two coding sessions by four researchers (SF, KB, CKL and RLM) at which point it was decided that no new themes had been identified and further interviews were not required.

Data processing and analysis

Quantitative survey data were coded and analysed using descriptive statistics. Interview data and qualitative survey data were analysed using the six-phase approach for thematic analyses proposed by Braun and Clarke (2013): (i) familiarising yourself with your data, (ii) generating initial codes, (iii) searching for themes, (iv) reviewing themes, (v) defining and naming themes and (vi) producing the report. The interview transcripts and qualitative survey data from the open-ended, free-text responses were imported into NVivo (Version 14), qualitative analysis software (Lumivero, 2023). Four researchers (SF, KB, CKL and RLM) independently conducted preliminary coding on 20% of the qualitative data and derived themes or patterns by systematically identifying and labelling relevant data characteristics. The coded elements were discussed to identify similarities and differences, and organised into potential themes and sub-themes, and further refined through a review process to generate cohesive and meaningful themes, until consensus was reached. The remaining transcripts were coded and labelled by one researcher (SF). Following the initial thematic analysis, interview participants were emailed a summary of the themes and invited to provide feedback. Their suggestions were considered and incorporated where appropriate to enhance the credibility and accuracy of the findings.

Ethics approval

This research was approved by the University of Sydney Human Research Ethics Committee (Project no. 2023/793). The study has been reported in accordance with the Standards for Reporting Qualitative Research (O’Brien et al., 2014) and Consolidated Criteria for Reporting Qualitative Studies (Tong et al., 2007) (Supplemental File 2).

Results

Sample characteristics

Survey respondents

Seventy survey responses were received, with respondents spread across 18 different Australian institutions, with three-quarters working in an Australian university (n = 53; 76%). Ten universities across five Australian states were represented among these respondents. The majority were investigators, including health economists (n = 29; 41%), project managers (n = 18; 26%), or statisticians/epidemiologists (n = 11; 16%) (Table 1). There was a broad range of data linkage experience, with 66% (n = 46) having worked on between one to four trials involving data linkage, and 29% (n = 20) having worked on at least five trials involving data linkage (Figure 1). Survey respondents were involved in a broad range of duties in the data linkage process. However, of the 10 pre-defined tasks, the majority of respondents played a role in leading and/or assisting processes related to receiving the data (n = 60; 86%), requesting the data (n = 59; 84%) and seeking ethics approval (n = 57; 81%) (Figure 2).

Table 1.

Characteristics of survey respondents and interview participants.

Characteristics	N (%) of survey respondents	N (%) of interview participants
Characteristics	(n = 70)	(n = 22)
Primary role
Investigator/Health economist	29 (41)	6 (27)
Project manager	18 (26)	9 (41)
Statistician/Epidemiologist	11 (16)	4 (18)
Research officer/trial coordinator/assistant	8 (11)	1 (5)
Data manager/systems expert	3 (4)	2 (9)
Medical Monitor and Safety Manager	1 (1)	0
Institution
University	53 (76)	21 (96)
Research Institute	7 (10)	0
Hospital research centre	5 (7)	1 (5)
Government department	3 (4)	0
Unknown	2 (3)	0

Figure 1.

Number of trials with linked administrative data per survey respondent (n = 70).

Figure 2.

Survey respondents’ level of project responsibility in data linkage project/s (n = 70).

Information was collected about 201 trials from the 70 survey respondents. For each trial, respondents could select between one and six different purposes for using administrative data. The most common uses were measuring economic outcomes (n = 111 trials; 55%), tracking health service use (n = 107 trials; 53%), measuring disease or health outcomes (n = 99 trials; 49%) and tracking participant survival status (n = 84 trials; 42%). Fewer trials used administrative data to measure drug adherence (n = 22; 11%) (see Figure 3).

Figure 3.

Purpose of administrative data linkage among survey respondents’ trials (n = 201).*

Interview participants

Twenty-two (31%) of the survey respondents agreed to participate in an in-depth interview and all but one of the interview participants worked at an Australian university (n = 21, 96%).

Themes

The thematic analysis of interview and survey data produced four overarching themes and 10 sub-themes (Table 2): “Trial management considerations” covers operational factors like budgeting, staffing, and timelines; “Assessing burdens vs. gains” evaluates whether the research benefits of administrative data outweigh the additional operational and consent considerations; “Data preparation and analysis” addresses the processes involved in preparing and analysing data for linkage between trial and administrative datasets, while “Training and support” emphasises the need for researcher support in using linked data. The quotes associated with these themes and subthemes are listed in Tables 3 –6.

Table 2.

Outline of themes and sub-themes.

Theme and description	Sub-theme
1. Trial management considerations Operational considerations e.g. documentation, communication, budgeting, staffing, project timelines, when planning a clinical trial involving data linkage.	• Good documentation and communication practices
	• Cost of obtaining data
	• Setting realistic timelines
	• Skilled staff availability
2. Assessing burdens vs. gains Evaluating whether research benefits of administrative data outweigh the additional operational and consent considerations.	• Assessing value added by administrative data
	• Additional participant consent requirements
3. Data preparation and analysis Awareness of the data preparation and analysis processes to link trial data to administrative data.	• Data preparation
	• Mandated data access platforms
4. Training and support Access to training and support to assist researchers.	• Adequate internal training and procedures
	• Access to support

Table 3.

Quotations to illustrate Theme 1 – Trial management considerations.

Sub-theme	Quotations
Good documentation and communication practices	(On taking over project management of a trial) “So I then had to go, “okay, where is everything up to? Did you ever go through that? Have you ever contacted (custodian)? Do we have the right forms? Did we use the right forms? What was the plan of why we’re collecting this?”. . .if it was well defined, nobody thought to write it down.” (P10, Project Manager). “Usually I’m involved in the planning of the trial. I try to insist on that and tell clinical trialists and clinicians, “don’t come to us at the end of the trial and say we’ve got this wonderful data, can you analyse it please?”” (P22, Health Economist). (On not being involved in the protocol development) “It is when I contacted the chief investigator six months, seven months after, that’s when they sent me the protocol and everything, and I see that this information is completely missing, and I (asked) “how are we going to analyse all this data?”” (P20, Health Economist). “I actually started (documenting) all of the different datasets that we were going to interact with, who the data custodians were, when we needed to get information and apply. . . and I’ve called it a “data linkage protocol” for lack of a better term. . . so that in five years (someone might say), “oh, what am I supposed to do? It’s all here.”” (P17, Project Manager).
Cost of obtaining data	“I don’t think (the fees are) a lot considering I might be doing a $1 million plus budget, and I’m being asked to put in $5,000 to access these databases, I think it’s entirely reasonable.” (P9, Project Manager). “The fact that you only get a quote and the quote can and will change after the linkage is quite difficult.” (P5, Statistician). (On the cost of obtaining linked datasets) “They are what they are. I think the important thing is getting a quote and making sure that we are costing it appropriately and making sure that we’re building it in.” (P12, Project Manager). “A lot of the patients that we see. . . are usually kind of an older demographic, so people are less likely to move interstate. . . It’s not really worth the costs that are required to have (a separate) agreement with (state) births, deaths and marriages.” (P16, Project Manager).
Setting realistic timelines	“It’s a very slow process, very, very slow process.” (P5, Statistician). “If I were an individual academic at the university that doesn’t have a support system to be able to go through all the processes of doing data linkage, I would really think about if it’s worth it, because it is a lot of time, especially if you haven’t done it before.” (P14, Project Manager). “I don’t think the investigators accounted for the data lag between. . . when the outcome is, the data being available at the data custodians. . . and then for us to get it out. . . they were saying it may add up to two years if the timings of the data syncs don’t align, and the grant will be over, and we can’t wait two years to get the outcome data!” (P10, Project Manager). (On advising a researcher new to data linkage) “I tell them to expect it’s going to take a lot longer than what they think and it’s going to be a lot more work than what they think it might be.” (P22, Health Economist). “It’s just a completely moving field, but the problem is (custodians) don’t tell you, and you don’t figure it out until you’ve taken a step and they go, “oh no, you did that wrong, you need to do this form now.” And it happens all the time. . . It’s about the order of events and that order is not straightforward.” (P4, Epidemiologist). “I think the issue for me. . . is this need for (all custodians) to agree on the same protocol. . . (if a change is required), you make that change and now you’re obligated to start back at the beginning and have them reassess.” (P2, Investigator). (On a previously approved long term follow up request) “When we tried to do the linkage, (ethics said the) trial approval is too old, the system has changed, we needed to start from beginning. . . so it’s a bit annoying.” (P13, Project Manager). “Have a really strong handle on the process and try to understand where the hurdles will be before you actually make any submissions, because I find that if you can submit and everyone agrees, then you don’t get it caught in this endless cycle.” (P3, Investigator). “If we are submitting a new project, we typically go to a jurisdiction like (state) first, because they tend to have the most stringent requirements, the most involved application process. So we know if we experience little to no resistance. . . then we’ll take it to other jurisdictions next.” (P7, Epidemiologist).
Skilled staff availability	“I’m really lucky in that (project manager) has done the linkage. . .for the last 15 years, so even though the processes had changed, even though the paperwork and the admin had changed, she knew the philosophy behind it, she knows how it all works.” (P12, Project Manager). (On working with linked data) “So it’s not just the (programming) ability, it is the understanding of what is appropriate and what is not appropriate – that trial experience or epidemiology experience, combined with the statistical and analyst expertise, and increasingly, IT skills to efficiently work with big data.” (P11, Statistician). “We tend to employ a data manager, or a couple of data managers, or biostatisticians with skills in data management. They’re the ones who look at the data and say, “how do you want it?”. . . they’re a great team member.” (P22, Health Economist). “Something that was really difficult for us was to understand how things have been coded, like ICD-10s and procedure codes and things like that. As a non-clinician, who has no access to the coders, it was very difficult to understand how these codes came about. . . That made it really difficult when we were trying to analyse and interpret (the data).” (P5, Statistician).

Table 4.

Quotations to illustrate Theme 2 – Assessing burdens versus gains.

Sub-theme	Quotations
Assessing value added by administrative data	“If I do a trial where people have to consent to participate, then I know all the people that I really want to reach aren’t going to participate in the study. I’ll get all the highly educated health-seeking people, and I don’t really want to do a trial just with them. I want to do a trial with the whole population.” (P6, Investigator and Epidemiologist). “In the area that I work you tend to lose contact with a lot of people [participants] because they’re a fairly chaotic group.” (P1, Investigator). “Having a whole study of linked data is challenging. . .” (P14, Project Manager). “The use of linked data potentially adds more burden on the statistician. . . to simulate a “real trial” database and instead of designing a trial to avoid missing data, you are dealing with more complex analyses to account for missingness. . . you can’t follow up on missing data.” (P11, Statistician). (On the number of patients followed up through administrative data) “Sometimes it just doesn’t seem worth it if it’s not very many patients. I’m not sure what that number is. I feel like maybe it’s a hundred or something like that.” (P14, Project Manager). “Sites are not resourced properly to get (the data) and we don’t pay them enough to resource them to do it, so being able to leverage administrative structures. . . is really great if we can use it.” (P10, Project Manager). “What we’re collecting. . . medication history. . . throughout the life of the study. . . is a huge amount of work for both site staff and our data managers, so in a way, it’s really a much reduced burden on everyone to be able to use (linked data). . . where it’s appropriate.” (P15, Trial Coordinator).
Additional participant consent requirements	“Some of the consent forms. . . are very specific, you can’t just throw it into the consent forms and say, “we’re going to do this,” you have to have their consent form.” (P14, Project Manager). “We ended up only getting maybe a 50% consent rate, even though the idea of the health economic analysis was really important and integrated into the protocol and into the consent forms. There are many factors that contribute to consent rates, such as the way information is presented and the training that staff have around the process and content. These things should be considered to ensure that we optimise giving participants the opportunity to consent.” (P21, Project Manager). “Patients needing to re-consent. . . is a pain. . . there are patients who are now in follow up, they rarely actually come into clinic, in some cases, their visits are done remotely.” (P15, Trial Coordinator). (On trial team needing to reconsent trial participants) “They say that’s too hard, that’s going to be too much work, they’ve done it before and it’s a headache. . . that’s the major barrier.” (P22, Health Economist).

Sub-theme

Quotations

Assessing value added by administrative data

“If I do a trial where people have to consent to participate, then I know all the people that I really want to reach aren’t going to participate in the study. I’ll get all the highly educated health-seeking people, and I don’t really want to do a trial just with them. I want to do a trial with the whole population.” (P6, Investigator and Epidemiologist).
“In the area that I work you tend to lose contact with a lot of people [participants] because they’re a fairly chaotic group.” (P1, Investigator).
“Having a whole study of linked data is challenging. . .” (P14, Project Manager).
“The use of linked data potentially adds more burden on the statistician. . . to simulate a “real trial” database and instead of designing a trial to avoid missing data, you are dealing with more complex analyses to account for missingness. . . you can’t follow up on missing data.” (P11, Statistician).
(On the number of patients followed up through administrative data) “Sometimes it just doesn’t seem worth it if it’s not very many patients. I’m not sure what that number is. I feel like maybe it’s a hundred or something like that.” (P14, Project Manager).
“Sites are not resourced properly to get (the data) and we don’t pay them enough to resource them to do it, so being able to leverage administrative structures. . . is really great if we can use it.” (P10, Project Manager).
“What we’re collecting. . . medication history. . . throughout the life of the study. . . is a huge amount of work for both site staff and our data managers, so in a way, it’s really a much reduced burden on everyone to be able to use (linked data). . . where it’s appropriate.” (P15, Trial Coordinator).

Additional participant consent requirements

“Some of the consent forms. . . are very specific, you can’t just throw it into the consent forms and say, “we’re going to do this,” you have to have their consent form.” (P14, Project Manager).
“We ended up only getting maybe a 50% consent rate, even though the idea of the health economic analysis was really important and integrated into the protocol and into the consent forms. There are many factors that contribute to consent rates, such as the way information is presented and the training that staff have around the process and content. These things should be considered to ensure that we optimise giving participants the opportunity to consent.” (P21, Project Manager).
“Patients needing to re-consent. . . is a pain. . . there are patients who are now in follow up, they rarely actually come into clinic, in some cases, their visits are done remotely.” (P15, Trial Coordinator).
(On trial team needing to reconsent trial participants) “They say that’s too hard, that’s going to be too much work, they’ve done it before and it’s a headache. . . that’s the major barrier.” (P22, Health Economist).

Table 5.

Quotations to illustrate Theme 3 – Data preparation and analysis.

Sub-theme	Quotations
Data preparation	“(Data custodians) generally have. . .a data dictionary to describe the data you’ve been given, and that is usually very well documented and easy to follow. Everything has been coded and formatted to work with.” (P11, Statistician). “Some of these datasets. . . are huge. . . In these increasingly large datasets, the analyses, even simple analyses take a lot of time to run.” (P11, Statistician). “(Data custodians) don’t know much about the content of the data, so it’s difficult to ask anyone from the data linkage unit about how to analyse things.” (P5, Statistician). (On learning to work with administrative data) “It was a really, really steep learning curve. Luckily, we have collaborated with people at [university] who work specifically on linked data so we had some help from them in terms of how to clean the data or manipulate the data so it’s analysis ready, but it was just all trial and error. . . I have also attended some workshops. . . but you know, it’s just you learn as you go kind of a thing.” (P5, Statistician). “I would do things like clean up the dataset, create the right variables, chop up the cohort, that sort of thing, and then the focus would change and I would have to do it all over again.” (P8, Statistical Programmer).
Mandated data access platforms	(On data curation platforms) “If you want to get graphics or tables of analyses. . . it goes through a special portal and it is checked before it comes out. . . but it’s not like you put it in the portal and in five minutes you take it out. You put it in there and a couple of days, weeks later, you get it out. . . I can understand why they want that level of security, but it is extremely burdensome and time consuming.” (P11, Statistician). “You pay for the computing power that you have in the cloud, and it’s an annual subscription fee. I think we have the basic one, and it is very slow to be running R code or SAS code in. I’ve dealt with analysing large datasets before and this is a very slow system.” (P11, Statistician).

Sub-theme

Quotations

Data preparation

“(Data custodians) generally have. . .a data dictionary to describe the data you’ve been given, and that is usually very well documented and easy to follow. Everything has been coded and formatted to work with.” (P11, Statistician).
“Some of these datasets. . . are huge. . . In these increasingly large datasets, the analyses, even simple analyses take a lot of time to run.” (P11, Statistician).
“(Data custodians) don’t know much about the content of the data, so it’s difficult to ask anyone from the data linkage unit about how to analyse things.” (P5, Statistician).
(On learning to work with administrative data) “It was a really, really steep learning curve. Luckily, we have collaborated with people at [university] who work specifically on linked data so we had some help from them in terms of how to clean the data or manipulate the data so it’s analysis ready, but it was just all trial and error. . . I have also attended some workshops. . . but you know, it’s just you learn as you go kind of a thing.” (P5, Statistician).
“I would do things like clean up the dataset, create the right variables, chop up the cohort, that sort of thing, and then the focus would change and I would have to do it all over again.” (P8, Statistical Programmer).

Mandated data access platforms

(On data curation platforms) “If you want to get graphics or tables of analyses. . . it goes through a special portal and it is checked before it comes out. . . but it’s not like you put it in the portal and in five minutes you take it out. You put it in there and a couple of days, weeks later, you get it out. . . I can understand why they want that level of security, but it is extremely burdensome and time consuming.” (P11, Statistician).
“You pay for the computing power that you have in the cloud, and it’s an annual subscription fee. I think we have the basic one, and it is very slow to be running R code or SAS code in. I’ve dealt with analysing large datasets before and this is a very slow system.” (P11, Statistician).

Table 6.

Quotations to illustrate Theme 4 – Training and support.

Sub-theme	Quotations
Adequate internal training and procedures	“I would say that our organisation has room for improvement when it comes to data linkage processes and training.” (P21, Project Manager). “I had no idea of the background of the collection of this additional data, and I think that would have been really good, to have some sort of training . . . because I felt like I went in kind of blindly.” (P18, Data Manager). (On publicly available custodian guidance) “I would doubt that what’s on (a) central website would be everything that I need to give me confidence that what I was writing in a (participant information consent form) was the wording that would be acceptable to my centre.” (P19, Project Manager). “Because I hadn’t done it before, I wrote a series of steps because it was so confusing. And I thought, if someone needs this. . . this could be helpful to them.” (P4, Epidemiologist). “It usually takes us between six months to a year to train a staff member on how these processes work in order for them to be independent.” (P7, Epidemiologist).
Access to support	“What would be great is if there was a point person. . . that you go to right at the start. . . and they scope it and loop in the people you need. . . someone who. . . has a lot of experience in it and can co-design a bit of a plan with you, because it impacts so many other things. It will impact your monitoring plan, your data plan and your budget. . . and I don’t think it’s realistic that all the project managers will get (to) that level of knowledge, so you need to have a subject matter expert that you can talk to.” (P10, Project Manager). “It is useful to have someone that you could go to and get all that information. If the university had someone with that expertise within the central research team or the faculty research team, I think that would be valuable.” (P6, Investigator and Epidemiologist). “We have a monthly meeting just to chat about “how’s the data going? Does anyone have any issues?”” (P4, Epidemiologist). (On data custodians) “I think the most useful thing. . . has been, if you have a good contact. . . we had someone really good and it just meant every time we needed to do something we could go, “what do we need to do about this?” Because it’s not written down anywhere on their website. I know they try. . . but they don’t cover everything. . . There are so many questions you need to ask.” (P4, Epidemiologist). “I think the easiest way is to find someone else, probably in your institution or some institution that you’ve had some connection with, or cold calling. . . to call up someone that published something with the data that looks like something that you would want to do and ask them how they made that happen. . . that’s how it sort of starts and then it unfolds slowly and you start to understand it.” (P3, Investigator).

Sub-theme

Quotations

Adequate internal training and procedures

“I would say that our organisation has room for improvement when it comes to data linkage processes and training.” (P21, Project Manager).
“I had no idea of the background of the collection of this additional data, and I think that would have been really good, to have some sort of training . . . because I felt like I went in kind of blindly.” (P18, Data Manager).
(On publicly available custodian guidance) “I would doubt that what’s on (a) central website would be everything that I need to give me confidence that what I was writing in a (participant information consent form) was the wording that would be acceptable to my centre.” (P19, Project Manager).
“Because I hadn’t done it before, I wrote a series of steps because it was so confusing. And I thought, if someone needs this. . . this could be helpful to them.” (P4, Epidemiologist).
“It usually takes us between six months to a year to train a staff member on how these processes work in order for them to be independent.” (P7, Epidemiologist).

Access to support

“What would be great is if there was a point person. . . that you go to right at the start. . . and they scope it and loop in the people you need. . . someone who. . . has a lot of experience in it and can co-design a bit of a plan with you, because it impacts so many other things. It will impact your monitoring plan, your data plan and your budget. . . and I don’t think it’s realistic that all the project managers will get (to) that level of knowledge, so you need to have a subject matter expert that you can talk to.” (P10, Project Manager).
“It is useful to have someone that you could go to and get all that information. If the university had someone with that expertise within the central research team or the faculty research team, I think that would be valuable.” (P6, Investigator and Epidemiologist).
“We have a monthly meeting just to chat about “how’s the data going? Does anyone have any issues?”” (P4, Epidemiologist).
(On data custodians) “I think the most useful thing. . . has been, if you have a good contact. . . we had someone really good and it just meant every time we needed to do something we could go, “what do we need to do about this?” Because it’s not written down anywhere on their website. I know they try. . . but they don’t cover everything. . . There are so many questions you need to ask.” (P4, Epidemiologist).
“I think the easiest way is to find someone else, probably in your institution or some institution that you’ve had some connection with, or cold calling. . . to call up someone that published something with the data that looks like something that you would want to do and ask them how they made that happen. . . that’s how it sort of starts and then it unfolds slowly and you start to understand it.” (P3, Investigator).

Theme 1 – Trial management considerations

The use of administrative data involves key operational considerations that researchers must navigate. Of the 70 survey participants, almost half (n = 31; 44%) felt that they would do something differently the next time they are involved in a data linkage process, including better planning or having clearer processes (n = 12), engaging data custodians earlier (n = 7) or allowing for more money and time (n = 5). Some participants said they would reconsider using administrative datasets altogether (n = 8) due to the added complexities. These findings highlight the operational complexities that can significantly impact the feasibility and success of administrative data projects.

Good documentation and communication practices

Participants commented that maintaining detailed records of correspondence and decisions was a valuable practice as the duration between initial discussions and requesting of data can be years. This may include discussions with data custodians and internal decisions, such as when and why datasets were required, who was responsible for each process, such as preparing applications, managing consent forms and receiving and storing the data. Good communication was seen as particularly important where trials involved multiple parties, especially if they were from different institutions. Participants agreed that connecting early and regularly with the full research team would help to agree on plans for the data; however, some project managers mentioned the challenge of balancing investigator expectations and project demands to keep the trial on time and under budget, as changes to project scope or expectations typically result in increased time and cost. Health economists, in particular, expressed frustration that they were consulted too late in the planning or were not kept updated on the progress of the trial.

Cost of obtaining data

Interview participants stressed the importance of evaluating the benefits of obtaining the data in relation to the fees charged and noted the substantial differences in data access fees among various data custodians. The cost of obtaining data influenced study design, such as whether an additional jurisdiction would be included in the participant sample, or whether an endpoint could be assessed. There were varying opinions on whether the cost of obtaining data was considered reasonable, with interview participants acknowledging that research is not the custodians’ core business, and some citing considerable fee increases over time. However, participants generally agreed that the most important factor was ensuring a quote is obtained from the data custodians and factored into grant applications. Challenges raised by the participants were the unpredictability of the final fee for the data, as often quotes are received years before data are required and are subject to change when the dataset is finally generated, suggesting a buffer should be added to all quoted costs when developing a budget. Further, some were unaware they would need to add the cost of the data storage and curation platform, which is increasingly becoming a mandatory requirement when working with administrative data (see subtheme “Mandated data access platforms”).

Setting realistic timelines

A recurring theme was the significant time required to prepare data linkage applications, delays in receiving approval for data requests and delays in receiving data once the final request is submitted to data custodians. Participants observed that it was common for less experienced researchers to underestimate the work required to prepare data linkage applications, and this was further exacerbated by the fact that linkage requirements are becoming increasingly complicated. Some said it was worthwhile developing a strategy for applying for access to administrative data, especially in multi-jurisdictional trials, where the order of events seemed to impact the outcome. Participants employed different strategies to deal with this. Some decided to approach the custodians who had very specific requirements first, whereas others would approach these custodians last to ensure that they were satisfied with all the changes made following applications to prior custodians. It was clear that applying to multiple custodians increased the workload significantly, as applications were not streamlined. Reusing responses from prior approved applications was one way to avoid unnecessary rework; however, custodian requirements could change between applications. Appreciating custodians have limited resources to deal with research requests; delays of up to 2 years were not uncommon and depended on the custodian and the complexity of the linkage, in particular, the number of jurisdictions and datasets involved, as these added layers of complexity in terms of multiple ethics submissions and approvals. Often the frustration stemmed from a lack of or unreliable alternatives, such as hospital episodes of care for a health economic analysis, where the alternative to the administrative data is patient recall.

Skilled staff availability

Many participants, regardless of their position, emphasised that prior experience in data linkage or access to mentoring is crucial, as the skills required are unique. Of the 70 survey participants, 60 (86%) agreed that specific skills are required. Given trials linking to administrative data are already prone to delays, having less experienced staff can lead to further setbacks. Participants felt that resourcing additional time for trial operations staff was important, particularly at the beginning of the trial, to account for increased administrative requirements, and this should increase accordingly with each additional custodian data set request. Several participants also emphasised the importance of having access to staff with strong data manipulation and programming skills to prepare the data, whether it be a statistical programmer, biostatistician or data manager (see subtheme “Data preparation”).

Theme 2 – Assessing burdens versus gains

The burden of additional operational processes and resources, as well as the administrative burden of added consent processes, means that researchers should weigh up the value added by linking to administrative data with the burden of including it in the trial design.

Assessing value added by administrative data

Participants emphasised the trial design, such as the trial population, planned length of follow-up or the number of trial participants, can help decide whether administrative data are preferable over traditional case report form (CRF) data collection. In general, participants agreed that while there are known limitations to administrative data, such as missing data that cannot be followed up through sites, there are instances where it is a preferred data source, such as where the alternative is patient recall, or where researchers seek straightforward data, such as the death status. Some interview participants felt that administrative data can hold great potential for including priority populations who may otherwise not take part in trials due to barriers such as their regional/remote geographical location, language and socioeconomic status. Trial participants who are less likely to attend follow-up appointments are also better suited to be followed up through administrative data.

From a trial coordination perspective, an advantage for sites can be reduced data entry into trial databases for data which will be obtained administratively, with one participant commenting that CRF data collection has its own challenges, such as obtaining timely and complete data. Another interview participant suggested that administrative data should supplement, not replace, CRF data, as some data sources are more difficult to obtain, such as educational and general practice data. Researchers are also influenced by the data custodians’ level of experience in working with researchers, which impacts on the choice of data source. Establishing which data are available through linkage is vital, with more than half of participants mentioning direct discussions with data custodians were the most effective way to gather this information.

Additional participant consent requirements

Although trials involving data linkage often add complexities to the participant consent process, several factors can contribute to a more streamlined consent process. Establishing a consent process for data linkage at the beginning of the trial was viewed as essential to the success of administrative data related outcomes. Challenges included the struggle to have consent forms approved before commencing recruitment, particularly where there are competing interests and a pressure to meet trial milestones. There were instances where trial participant consent to access administrative data was collected as a contingency, despite the uncertainty of funding, particularly where the data would not be required until years later, such as a health economic analysis or long-term follow-up. Overall, it was felt that it was better to wait for the consent form to be approved and finalised as the reconsenting process created a high administrative burden on sites and coordinating staff, as well as confusion for trial participants. There was some frustration with specific or lengthy mandated text in consent forms, as well as poorly designed forms resulting in the need for additional site monitoring to ensure that the consent form information was correct. Programming the consent form into the clinical trial database for site data entry seemed to help with reducing potential errors. Training site staff as well as monitors in how to administer the consent forms, as well as the coordinating centre project team on what to expect from sites, was viewed as crucial as missing information may hinder linkage. Reviewing consent forms early to identify potential errors or low response rates could help to identify sites requiring additional training or processes. While there is increasing uptake of e-consent in trials, acceptability of e-consent amongst data custodians for administrative data access remained unclear. Without e-consent, it was difficult to obtain consent if trial visits were performed remotely.

Theme 3 – Data preparation and analysis

Participants highlighted the technical complexities involved in preparing and analysing administrative datasets, as well as accessing them where it had to be done through a curated data access platform.

Data preparation

Participants responsible for preparing administrative datasets for analysis said it was a skill mostly learnt through trial and error, as datasets are unique, and contained certain intricacies only learnt over time. While data dictionaries supplied by custodians are essential to understanding the datasets, they can vary considerably in terms of quality and detail between custodians. Deciphering classification and coding systems, such as the International Classification of Diseases and Diagnostic Related Groups, were particularly challenging for researchers, as knowledge in classification and coding is a niche skill and researchers did not always have access to an experienced clinical coder, skilled in understanding historical changes to the coding of diagnoses and procedures when analysing longitudinal data. The large size of the datasets and lengthy computer processing times for analyses was something those working with the data needed to be prepared for, even with the help of a high-performance computer and the highest speeds available through virtual trusted research environments.

Mandated data access platforms

Participants’ experiences with receiving data from custodians varied considerably depending on the custodian and their requirements. Some custodians permitted data sharing via a secure download, while others only allowed access through a curated data platform. Curated platforms enable secure sharing and analysis of sensitive data and are managed independently by universities or research institutes rather than the custodians themselves; therefore, researchers are required to enter into separate agreements with the platform provider. Besides the cost (see subtheme “Cost of obtaining data”), issues around using curated data platforms ranged from challenges in setting it up, the speed of running a program and the difficulty in retrieving results. Participants who were aware of plans of having a single national linked data system, with all data stored centrally, said they hoped it would make the process more seamless.

Theme 4 – Training and support

Those involved in operationalising a trial with linked administrative data, such as project managers and data managers, wanted to be better equipped to respond to investigator requests for data linkage in their trials and felt that they would have benefited from better training and support.

Adequate internal training and procedures

Of the 70 survey participants, 51 (73%) reported following some form of written guidance; however, 39 (77%) of these also faced challenges or felt that processes could be improved. Of the 19 (27%) participants who did not follow written guidance, 15 (79%) felt that guidance would have been helpful. Of the 51 survey participants who followed written guidance, 40 (78%) used procedures or guidance directly provided by the data custodian and 17 (33%) used their own institutions procedures. Resources such as standard operating procedures, templates and checklists were seen as ideal tools to have access to and could be broadly written to account for the variability across custodians and data sets. Participants felt that understanding the terminology around data linkage would be particularly helpful, and some participants developed their own guidelines which they shared with colleagues when requested. Interview responses varied when asked whether they understood the purpose of data linkage, and unsurprisingly, it was those who were involved in the operations of the trial who felt that they had the most limited understanding, despite being responsible for overseeing the process to ensure it functioned well.

Access to support

Participants, particularly in larger research centres, felt that resourcing a local subject matter expert would be beneficial as they could be consulted during concept and protocol development and could provide a framework for how to apply data linkage processes within their institutions. Some participants had access to support networks within their institutions, of whom questions could be asked and resources shared. A fixed contact within the custodian unit was seen as helpful for obtaining timely information about the data request process, with 45 of the 70 survey participants (64%) receiving some form of support from a data custodian.

Discussion

This study contributes to enhancing the use of administrative data in trials by offering insights into the experiences of clinical trialists and trial operations staff across diverse roles, expertise and therapeutic areas. The issues identified in the Australian context broadly align with those reported in previously published studies from international jurisdictions (Toader et al., 2024; Williams et al., 2023). Some challenges are specific to Australia. A large proportion of health data are collected by individual Australian states, and there is limited uniformity between state collections. Further, Australia does not have a centralised health record, and the movement towards this (Australian Digital Health Agency, 2024) has been met with mixed reactions from Australians (Knaus, 2019), which has been further impacted by data breaches across the country (Department of Home Affairs, 2024). Internationally, there are initiatives pushing for more standardised planning for trials using linked administrative data, including the SPIRIT-ROUTINE extension for protocols of trials conducted using administrative data (McCarthy et al., 2021), CONSORT-ROUTINE for the reporting of randomised controlled trials using cohorts and administrative data (Kwakkenbos et al., 2021), and RECORD for reporting of observational studies using administrative data (Benchimol et al., 2015). These guidelines aim to assist the transparency and quality of trial reporting.

Progress has been made within Australia in recent years to streamline access to administrative data for research purposes. The National Health Data Hub (Australian Institute of Health and Welfare, 2024) and the National One Stop Shop (Australian Commission on Safety and Quality in Health Care, 2024) both aim to eliminate redundant processes through centralisation of datasets, and the introduction of The Data Availability and Transparency (DAT) Act 2022 (Australian Government, 2023) aimed to create and support transparency and data sharing in Australia and provides a searchable government data catalogue for policy makers and researchers. This Act also introduced the role of the National Data Commissioner. The formation of a network of researchers with the stated goal of offering a coordinated approach for researchers to engage with the National Data Commissioner and contribute to shaping the national data sharing and release framework (University of New South Wales, 2018) represents another initiative that has empowered researchers by involving them in addressing a shared challenge, and funding was recently awarded to create better services for clinical trials using administrative data (University of Western Australia, 2023). Although these initiatives were viewed as positive steps forward by some interview participants, they noted that they were yet to observe significant changes in practice. The Australian Research Data Commons (ARDC, 2025), who were involved in discussions with the Office of the National Data Commissioner to implement the DAT Scheme, recently commented that data access continues to be a challenge, and that the Act has not been successful in its goal of increasing access for researchers.

The themes identified in this study were consistent across survey and interview participants, despite the broad range of researcher roles and experiences. Overall, participants agreed the process of obtaining data, while cumbersome, was worth repeating where the data are crucial to analysing trial outcomes; however, they expressed a strong desire for a more streamlined and accessible process.

Strengths and limitations

A key strength of this study was the in-depth one-on-one interviews, which enabled researchers to speak candidly and allowed the interviewer to delve deeper into responses. Further, surveying researchers allowed the capture of a larger number of participant views.

A limitation of this study was that clinical trials were the focus; however, many of the themes could be applicable to other study designs that utilise administrative data, such as prospective cohorts or implementation trials. An additional limitation may be the generalisability of study findings, as it is possible that those surveyed and interviewed represented individuals who faced more difficulties and were keen to share their experiences, compared with researchers who found the process straightforward. While the interviewer’s background provided valuable insight into the topic, during interviews it may have introduced a potential bias toward promoting efficient and positive data management practices. The use of snowball sampling excluded the possibility of assessing response rates.

Implications for practice and future research

Researchers can facilitate a streamlined outcome by establishing a well-planned process for submitting applications, which should involve regular and early communication with data custodians, while maintaining a detailed record of decisions and conversations, given the potential delays in between decision making and obtaining the data. Additionally, institutions are well-equipped to manage trials linking to administrative data if they have access to local expertise and provide comprehensive training and procedures.

Conclusion

This study presents candid experiences of researchers in administrative data use, and practical information for Australian researchers when considering applicability of administrative data to their own clinical trials. Primarily, researchers should assess whether obtaining administrative data is worthwhile, considering the fees and charges, availability of skills, time to obtain and analyse data and whether the data are more accurately captured through administrative data or through traditional data collection. Further, researchers should be aware of the additional processes required to access administrative data, including lengthy application processes, extra requirements for participant consent and the mandated use of data access platforms. Consideration of these factors will enable researchers to make pragmatic decisions around the inclusion of jurisdictions and datasets in their trial. Considerations are best made during the planning stages of the trial, so that complexities, such as reconsenting trial participants and overspending on trial budgets, are avoided.

Supplemental Material

sj-docx-1-him-10.1177_18333583251413652 – Supplemental material for Using linked administrative data: Insights and tips from academic clinical trialists

Supplemental material, sj-docx-1-him-10.1177_18333583251413652 for Using linked administrative data: Insights and tips from academic clinical trialists by Salma Fahridin, Karen Bracken, Chi Kin Law and Rachael L Morton in Health Information Management Journal

Supplemental Material

sj-docx-2-him-10.1177_18333583251413652 – Supplemental material for Using linked administrative data: Insights and tips from academic clinical trialists

Supplemental material, sj-docx-2-him-10.1177_18333583251413652 for Using linked administrative data: Insights and tips from academic clinical trialists by Salma Fahridin, Karen Bracken, Chi Kin Law and Rachael L Morton in Health Information Management Journal

Supplemental Material

sj-docx-3-him-10.1177_18333583251413652 – Supplemental material for Using linked administrative data: Insights and tips from academic clinical trialists

Supplemental material, sj-docx-3-him-10.1177_18333583251413652 for Using linked administrative data: Insights and tips from academic clinical trialists by Salma Fahridin, Karen Bracken, Chi Kin Law and Rachael L Morton in Health Information Management Journal

Footnotes

Acknowledgements

The authors thank all of the survey and interview participants for their time and thoughtful responses and contribution to the article.

Accepted for publication December 20, 2025.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

ORCID iDs

Salma Fahridin, BAppSc (HIM), MHSc (CDM)

Karen Bracken, BEc, MPH, PhD

Chi Kin Law, BFin, PhD

Rachael L Morton, BAppSc, Grad Dip (Mgmt), MScMed(Clin Epi)(Hons), PhD

Supplemental material

Supplemental material for this article is available online.

References

Australian Bureau of Statistics (2021) Enhancing the 2021 Census Using Administrative Data. Available at: https://www.abs.gov.au/census/about-census/2021-census-overview/administrative-data (accessed 3 September 2024).

Australian Commission on Safety and Quality in Health Care (2024) The National One Stop Shop - A National Cross Government Platform for Health-Related Human Research. Available at: https://www.safetyandquality.gov.au/our-work/health-and-human-research/national-one-stop-shop-national-platform-health-related-human-research (accessed 19 August 2024).

Australian Digital Health Agency (2024) My Health Record. Available at: https://www.digitalhealth.gov.au/initiatives-and-programs/my-health-record (accessed 19 August 2024).

Australian Government (2023) Data Availability and Transparency Act 2022. Canberra.

Australian Institute of Health and Welfare (2024) National Health Data Hub. Available at: https://www.aihw.gov.au/reports-data/nhdh/about (accessed 19 August 2024).

Australian Research Data Commons (ARDC) (2025) Submission: Statutory Review of the DAT Act - Issues Paper. Available at: https://www.finance.gov.au/sites/default/files/2025-07/Australian-Research-Data-Commons.pdf (accessed 16 June 2025).

Benchimol

Smeeth

Guttmann

, et al. (2015) The reporting of studies conducted using observational routinely-collected health data (RECORD) statement. PLoS Medicine 12(10): e1001885.

Braun

Clarke

(2013) Successful Qualitative Research: A Practical Guide for Beginners. Los Angeles, CA: SAGE.

Department of Home Affairs (2024) MediSecure cyber security incident. Available at: https://www.homeaffairs.gov.au/about-us/our-portfolios/cyber-security/cyber-coordinator/medisecure-cyber-security-incident (accessed 19 August 2024).

10.

Fahridin

Agarwal

Bracken

, et al. (2024) The use of linked administrative data in Australian randomised controlled trials: A scoping review. Clinical Trials 21(4): 516–525.

11.

Flack

Kemp Casey

Wray

(2019) Australian Clinical Trials Alliance, Using Linked Administrative Data in Clinical Trials: A Guide for Clinical Trialists and Researchers. Available at: https://clinicaltrialsalliance.org.au/wp-content/uploads/2019/09/Linked-data-guidance_final.pdf (accessed 3 September 2024).

12.

Harris

Taylor

Minor

, et al. (2019) The REDCap consortium: Building an international community of software platform partners. Journal of Biomedical Informatics 95: 103208.

13.

Harris

Taylor

Thielke

, et al. (2009) Research electronic data capture (REDCap)–a metadata-driven methodology and workflow process for providing translational research informatics support. Journal of Biomedical Informatics 42(2): 377–381.

14.

Katz-Buonincontro

(2024) How to Mix Methods: A Guide to Sequential, Convergent, and Experimental Research Designs. Washington, D.C.: American Psychological Association.

15.

Knaus

(2019) More than 2.5 million people have opted out of My Health Record. The Guardian, 20 February 2019. Available at: https://www.theguardian.com/australia-news/2019/feb/20/more-than-25-million-people-have-opted-out-of-my-health-record (accessed 19 August 2024).

16.

Kwakkenbos

Imran

McCall

, et al. (2021) CONSORT extension for the reporting of randomised controlled trials conducted using cohorts and routinely collected data (CONSORT-ROUTINE): Checklist with explanation and elaboration. BMJ 373: n857.

17.

Lumivero (2023) NVivo 14. Available at: https://www.lumivero.com (accessed 19 August 2024).

18.

McCarthy

O’Keeffe

Williamson

, et al. (2021) A study protocol for the development of a SPIRIT extension for trials conducted using cohorts and routinely collected data (SPIRIT-ROUTINE). HRB Open Research 4: 82.

19.

O’Brien

Harris

Beckman

, et al. (2014) Standards for reporting qualitative research: A synthesis of recommendations. Academic Medicine 89(9): 1245–1251.

20.

Toader

Campbell

Quint

, et al. (2024) Using healthcare systems data for outcomes in clinical trials: Issues to consider at the design stage. Trials 25(1): 94.

21.

Tong

Sainsbury

Craig

(2007) Consolidated criteria for reporting qualitative research (COREQ): A 32-item checklist for interviews and focus groups. International Journal for Quality in Health Care 19(6): 349–357.

22.

University of New South Wales (2018) Australian and New Zealand Real-World Data Network (RADiANT). Available at: https://radiant.med.unsw.edu.au/ (accessed 19 August 2024).

23.

University of Western Australia (2023) Population Health Research Network wins $10.24M funding boost. Available at: https://www.uwa.edu.au/news/article/2023/october/population-health-research-network-wins-10-24m-funding-boost (accessed 19 August 2024).

24.

Williams

ADN

Davies

Farrin

, et al. (2023) A DELPHI study priority setting the remaining challenges for the use of routinely collected data in trials: COMORANT-UK. Trials 24(1): 243.

25.

Zoom (2024) Zoom Workplace (Version 6.5.12). Zoom Communications. Available at: https://www.zoom.com/en/about/ (accessed March-May 2024).