Out of 3,723 records pulled from three research databases, exactly one described the thing the researchers went looking for.
That number sits at the centre of a scoping review published in PLOS One by Jill Vögelin and colleagues at Bern University of Applied Sciences and Vrije Universiteit Brussel. The team wanted to know how medical and health professions educators actually run feedback in assessment situations: not what the textbooks recommend, but what the published studies describe people doing. They searched Embase, ERIC and PubMed for work published between January 2015 and February 2026, screened titles and abstracts in pairs, read 91 full texts, and ended up with 14 studies that met their criteria.
The distinction driving the search is simple enough to explain in a sentence. Feedback looks backward, telling a learner how a past performance went. Feedforward looks ahead, specifying what to do next, how to do it, and under what conditions improvement should show up. In programmatic assessment, the model many health professions schools are moving toward, learning is meant to be stitched together across many low-stakes moments rather than judged in one exam, and feedforward is supposed to be the thread.
In the published literature, that thread is mostly invisible. Vögelin's team found explicit feedforward, with defined goals, strategies, criteria and timelines, described in a single study: a South African paper by Abraham and Singaram on clinical skills laboratories. Three others contained what the reviewers classed as implicit feedforward, where students were nudged to apply feedback to the next attempt without any structured plan. The remaining nine studies did not describe feedforward at all.
What the 14 studies did show
Sorting the interventions produced five families: expert feedback, self-feedback, peer feedback, video-based feedback, and feedforward. Expert feedback, from faculty, clinicians, mentors or standardized patients, was the most common and the most reliably effective. It was structured, tied to explicit criteria, and consistently linked to gains in performance, task efficiency and communication. In head-to-head designs, expert-assisted training beat purely self-directed training, particularly on short-term measures like how long a procedure took and how students scored on a post-test.
Video-based feedback did just as well. Three studies used it, covering asynchronous review of recorded performance and real-time visual navigation during procedural training, and all three reported significant improvements. Asynchronous video beat verbal feedback in a comparison of CPR training. A real-time system used in soft tissue resection cut unnecessary movement. The reviewers suggest the mechanism is unglamorous: video lets you stop, rewind, and break a performance into pieces small enough to analyse.
Peer and self-feedback were the wobbly ones. They supported reflection and gave students more ownership of their learning, but the measured outcomes varied. One Canadian study of a student-led mock clinical exam found no effect on subsequent exam scores at all. What seemed to matter was whether students believed the feedback: credibility, trust, and what the field calls feedback literacy, meaning the skill of interpreting and using comments you receive. The qualitative work turned up an uncomfortable pattern here. Lower-performing students were more likely to doubt the credibility of peer feedback and engage with it less, while higher-performing students engaged more confidently. The learners who stood to gain most got the least out of it, unless the process gave them scaffolding: explicit criteria, worked examples, guided reflection prompts.
The authors are careful about how far any of this reaches. Thirteen of the 14 studies scored moderate to high on the Joanna Briggs Institute appraisal checklists and one scored below half. None of the randomized trials could blind participants or assessors, because you cannot hide from someone which kind of teaching they are getting. Several non-randomized studies lacked comparison groups or follow-up. Every included study sampled medical students, with one adding a subgroup of psychiatry residents, so nothing here has been shown to hold for nursing, physiotherapy or the other health professions the review set out to cover.
Why it matters
The honest reading of this review is that it found an absence rather than an effect. Feedforward is widely advocated in health professions education, but when these reviewers went looking for descriptions of it in a decade of published studies, they found the logic present and the practice rarely named. The authors read that gap two ways: either educators are doing it without labelling it, or the field lacks the vocabulary and indexing to make it findable. Both are problems for anyone trying to build it into a curriculum deliberately.
There is also a quieter point about what gets measured. Most of the included studies tracked things that are easy to count: scores, seconds, error rates. Those are real, and the improvements are real. But the reviewers argue that speed and accuracy leave out the parts of clinical training that take years, including professional identity, judgement, and knowing when your own assessment of your work can be trusted. Their recommendation is not to abandon the countable outcomes but to sequence them, using structured expert guidance early for calibration, then deliberate reflection to build ownership.
With 14 studies, all in medical students, this is a map of a sparse literature rather than a verdict on what works. The authors say as much, calling for longer follow-up, mixed-methods designs, and sampling beyond medicine.