Programmatic assessment

No single assessment decides anything; the programme does

Programmatic assessment replaces the sequence of pass-or-fail exams with a deliberately designed programme of many low-stakes assessments, each one feeding the learner information to act on. High-stakes decisions about progression are made later, by people, on the accumulated evidence. The idea comes from medical education and is now the default model for competency-based programmes across the health professions.

Programmatic assessment

One-sentence definition: Programmatic assessment is an approach in which assessment is designed as a whole programme — many individual data points, each optimised for learning and low in stakes, aggregated over time so that decisions about progression are proportionate to the evidence behind them.

Cees van der Vleuten, Lambert Schuwirth, and colleagues set out the model (Schuwirth & Van der Vleuten, 2011; Van der Vleuten et al., 2012), and its principles were agreed as an Ottawa¹ consensus statement in 2020 (Heeneman et al., 2021). It starts from an observation about traditional assessment: every exam tries to do two jobs at once, telling the student how they’re doing and deciding whether they pass, and the second job crowds out the first. Students learn to the test, feedback arrives too late to act on, and one bad day carries disproportionate weight.

Programmatic assessment separates the jobs. Any individual assessment — an OSCE² station, a workplace-based observation on a midwifery placement, a written test, a piece of reflective writing — is treated as a single data point. Its purpose is to give the learner rich, usable information, and nothing rides on it. The stakes rise with the number of data points: a mid-course review draws on many of them, and the decision to progress or graduate draws on all of them, made by a committee that looks at the whole picture instead of adding up marks.

Three features follow from this:

• Feedback is the point of each assessment, so it has to be specific and delivered in time to use. A mark alone is a wasted data point.
• Triangulation replaces standardisation. Because no single instrument has to be defensible on its own, the programme can use methods that are individually weak but collectively strong, including narrative judgements from clinical supervisors.
• Decisions are proportional and human. The higher the stakes, the more evidence is required and the more people are involved. Committees make the final call, and their reasoning has to be defensible.

What a programme has to put in place

Van der Vleuten et al.’s Twelve tips (2015) is the usual starting point for a programme lead: a longitudinal portfolio, regular mentor meetings to interpret the accumulating evidence, and a progress committee that meets at fixed points. Ross et al. (2021) note the gap between the people who publish on programmatic assessment and the clinical educators who have to run it, and set out what needs to be in place before it will work. Heeneman et al. (2015) found that the effect on student learning depends on whether students believe the low stakes are real; if a “low-stakes” assessment turns out to count, the programme collapses back into exam preparation.

All of this asks a lot of the institution: mentoring capacity, committee time, an assessment record that can hold narrative as well as numbers, and regulations that let judgement decide progression.

Thin validity evidence, and a question the model wasn’t designed for

Bok et al. (2018) reviewed the validity evidence and found it thinner than the model’s adoption would suggest. The arrival of generative AI raises a further question the model wasn’t designed for: if many of the low-stakes data points are written work, what do they tell you about the learner? The direction of travel is towards evidence gathered by watching students practise and talking with them about it, which programmatic assessment already favours.

Sources

• Bok, H. G. J., de Jong, L. H., O’Neill, T., et al. (2018). Validity evidence for programmatic assessment in competency-based education. Perspectives on Medical Education, 7(6), 362–372. https://doi.org/10.1007/s40037-018-0481-2
• Heeneman, S., de Jong, L. H., Dawson, L. J., et al. (2021). Ottawa 2020 consensus statement for programmatic assessment – 1. Agreement on the principles. Medical Teacher, 43(10), 1139–1148. https://doi.org/10.1080/0142159X.2021.1957088
• Heeneman, S., Oudkerk Pool, A., Schuwirth, L. W. T., et al. (2015). The impact of programmatic assessment on student learning: Theory versus practice. Medical Education, 49(5), 487–498. https://doi.org/10.1111/medu.12645
• Ross, S., Hauer, K. E., Wycliffe-Jones, K., et al. (2021). Key considerations in planning and designing programmatic assessment in competency-based medical education. Medical Teacher, 43(7), 758–764. https://doi.org/10.1080/0142159X.2021.1925099
• Schuwirth, L. W. T., & Van der Vleuten, C. P. M. (2011). Programmatic assessment: From assessment of learning to assessment for learning. Medical Teacher, 33(6), 478–485. https://doi.org/10.3109/0142159X.2011.565828
• Van der Vleuten, C. P. M., Schuwirth, L. W. T., Driessen, E. W., et al. (2012). A model for programmatic assessment fit for purpose. Medical Teacher, 34(3), 205–214. https://doi.org/10.3109/0142159X.2012.652239
• Van der Vleuten, C. P. M., Schuwirth, L. W. T., Driessen, E. W., et al. (2015). Twelve tips for programmatic assessment. Medical Teacher, 37(7), 641–646. https://doi.org/10.3109/0142159X.2014.973388

Notes

• An earlier version of this note (February 2026) described programmatic assessment as automated testing of student-built digital artefacts. That is a different idea with a borrowed name, and it was wrong; the term belongs to the medical education literature above. Rewritten 2026-09-17.
• The model is the natural home for the argument in learning alignment: if assessment is a programme of evidence, student use of AI becomes something to observe and discuss, and not only something to police.