Table 1 Common limitations of randomized controlled trials and whether they can be amended by routinely collected health data
From: Routinely collected data for randomized trials: promises, barriers, and implications
Limitations of RCTs [10] | What using RCD for RCTs can offer | Challenges | Potential of RCD to improve RCTs |
|---|---|---|---|
Generalizability and real-world relevance | No specific data collection processes (follow-up visits, measurements) outside routine care, avoiding artificial situations | Random allocation of interventions may still require some deviation from routine care processes (e.g., obtaining informed consent). | Very high |
Costs and resources | No costs to the trial for data collection processes and related activities (study site setup, study staff salary, monitoring and auditing activities, training costs) | Potential costs for obtaining the RCD (if the collecting entity does not provide it for free; e.g., data brokers); additional costs for data management, processing, merging, cleaning, and so forth | Very high |
Specific conditions/subgroup effects | Larger sample sizes that are less influenced by resource constraints and feasibility issues may provide sufficient power for evaluating subgroups. | More opportunities for exploratory analyses with spurious findings | High |
Late outcomes | RCD can provide long-term outcome data without actively following patients and often reducing the number of patients lost to follow-up | Patients moving away from RCD infrastructure will be lost and may still require active contact, highly dependent on RCD infrastructure | High |
Speed | No cumbersome outcome ascertainment (follow-up contacts, data recording and collection) and no need for setting up the data collection infrastructure, thus results can be obtained faster | Management, processing, merging, and “cleaning” of large datasets may be time-consuming. Reporting of specific adverse events may be delayed. | High to moderate |
Conflicts of interest/sponsorship bias | Collection of RCD is more objective and less easily manipulated to obtain a desired result. | Data may still be analyzed and reported nonobjectively to convey preferred conclusions. | Moderate |
Understudied healthcare questions | Providing information on routine care allows researchers to address understudied healthcare questions because more resources are spared or different outcomes are collected. | Not all desired endpoints might be available; funding may not be the sole barrier | Moderate |
Regulations | Obtaining approval for intervention imposes several bureaucratic loopholes; RCD are already available and might require different ethical clearance. | RCD still require approval in terms of data protection and confidentiality. | Moderate |
Rare or uncommon conditions | Recruiting an appropriate sample size may be hard with rare diseases; larger samples with RCD and easier EHR or registry recruitment can reduce these difficulties. | Only possible if RCD resources are extensive, highly dependent on RCD infrastructure | Moderate |