From human review to Straight-Through Compliance
[see also part 4 of this series]
The previous articles argued that clinical research may be approaching a challenge similar to the one international banking encountered several decades ago. The industry has largely digitised its documentation, established common filing structures, and invested heavily in document management platforms. Yet quality control, reconciliation, and inspection readiness remain heavily dependent on human interpretation.
This dependence on human review should not be viewed as a failure. Clinical trials are complex, highly regulated, and involve significant judgement. Human expertise will always remain essential. The more interesting question is whether human expertise is currently being applied where it creates the most value.
Consider a typical quality-control activity. A reviewer receives an artefact, reads it, determines what it demonstrates, compares those conclusions with other artefacts, evaluates whether policies and regulations have been satisfied, and ultimately decides whether an issue exists. The process is repeated across thousands of artefacts and across countless studies.
The industry has become highly efficient at moving artefacts between participants. It has become less efficient at evaluating them consistently.
This observation is not unique to clinical research. During the 1980s and 1990s, the banking industry faced a similar challenge. SWIFT had already established a global communication network and standardised message formats. Financial institutions could exchange information reliably. Yet significant manual processing remained. Different institutions interpreted messages differently, applied different operational practices, and introduced different exceptions. Connectivity had improved substantially. Consistency had not.
The industry’s response became known as Straight-Through Processing, or STP. Contrary to popular belief, STP was not primarily an automation initiative. It was a rulebook initiative. The objective was to ensure that participants evaluated transactions according to a common set of rules and expectations so that fewer transactions required manual intervention and repair. This distinction is important because automation is often viewed as the objective. In reality, automation is merely a consequence. The banking industry did not automate first and standardise later. It standardised first and automated what became unambiguous.
Once those rules became sufficiently clear, a powerful shift occurred. Transactions satisfying the agreed conditions could proceed without manual intervention. Human expertise became focused on exceptions, ambiguities, and genuinely unusual situations. The result was not the elimination of human judgement. It was the more effective use of human judgement.
Looking back at the STP initiative during the 1990s, one observation stands out. Most transactions were routine. A small percentage consumed a disproportionate amount of operational attention because information was incomplete, contradictory, or unusual. The objective was never to eliminate human judgement. The objective was to focus that judgement where it created the most value.
When I look at clinical research today, I see a remarkably similar pattern. Most artefacts are routine. A small percentage consume a disproportionate amount of attention because the evidence is incomplete, contradictory, or unusual. Experienced reviewers already understand this intuitively. The challenge is not to eliminate review. The challenge is to ensure that human expertise is applied where it contributes most.
Consider a delegation log. Today, two reviewers may independently determine whether all required signatures are present, whether responsibilities are assigned appropriately, and whether authorisations were valid during the relevant period. The activity is routine, but it is repeated thousands of times across studies, sponsors, CROs, and sites. Similar assertions exist across informed consent forms, training records, temperature logs, investigator qualifications, equipment records, and countless other artefacts.
The question is not whether these checks are important. The question is whether every reviewer should have to derive the same conclusions independently.
Today, these evaluations are performed repeatedly by humans. Different organisations may apply slightly different interpretations. Different reviewers may reach different conclusions. Significant effort is therefore spent ensuring consistency across studies, sites, sponsors, and CROs.
An alternative approach would be to make the evaluation criteria themselves more explicit.
Rather than asking reviewers to repeatedly derive the same conclusions from the same categories of artefacts, the industry could define common evaluation rules. Human reviewers would remain responsible for exceptions, ambiguities, contextual judgement, and regulatory interpretation. Routine evaluations, however, could be performed consistently across studies and participants.
Such an approach would not replace quality control. It would transform its focus.
Instead of spending large amounts of time determining whether common conditions have been satisfied, reviewers could concentrate on situations where the evidence is incomplete, contradictory, unusual, or genuinely difficult to interpret. Human expertise would move closer to where it creates the greatest value.
This is ultimately what Straight-Through Processing achieved in banking. The industry did not eliminate human oversight. It reduced the amount of routine interpretation that required human involvement. It created a common framework for evaluating transactions and allowed participants to focus on exceptions.
Clinical research may eventually follow a similar path.
One might call this Straight-Through Compliance. Not because compliance becomes fully automated, and certainly not because human judgement disappears. Rather, because routine evaluations are performed consistently, allowing human expertise to focus on exceptions, ambiguities, and genuinely difficult situations.
This distinction is important. Straight-Through Processing did not emerge because banks automated payments. Banks automated payments because they first agreed how those payments should be evaluated. Automation was the consequence of standardisation, not the other way around.
Clinical research may face a similar challenge. Before routine compliance activities can be performed consistently, participants must first agree on what evidence should demonstrate, how common assertions should be evaluated, and where human judgement remains essential.
The next stage of maturity may therefore not be another repository, another workflow engine, or another document-management platform. It may be the gradual emergence of common evaluation practices that allow independent participants to arrive at similar conclusions when presented with similar evidence.
If the previous decades were about digitising documentation, the next decade may be about standardising how we evaluate what documentation is supposed to prove.
One Reply to “”