Patient-reported outcome measures (PROMs): bringing the patient’s voice into clinical follow-up

The French National Authority for Health (HAS) distinguishes two families of indicators. PROMs measure the outcomes of care as patients perceive them. PREMs, patient-reported experience measures, capture the experience of care as patients live it: access, information, coordination, the relationship with care teams. Satisfaction questionnaires, which ask whether care met expectations, form a third category.

Mixing them up has a cost. A PREM will not tell you whether a treatment relieved pain; a symptom score will not tell you whether the pathway was well coordinated. According to the HAS, PROMs work first at the individual level, helping patients express their symptoms and feeding the conversation with the clinician, while PREMs help organizations improve communication, team responsiveness or pain management.

The first question is not “which form should we digitize?” but “which decision should this score inform?” The same PROM can help prepare a consultation, follow a trajectory, spot a deterioration, discuss the patient’s priorities or evaluate the outcomes of a pathway across a department.

Each use comes with its own rhythm, recipient and action threshold. A questionnaire completed before a visit must be readable when the physician opens the record. An institutional performance indicator needs a method for aggregation and risk adjustment. If no one knows what to do with a score, collecting it adds workload without improving follow-up.

The choice depends on the population, the domain measured, the frequency and the intended use. A review published in Health Expectations sums up the trade-off: generic instruments allow comparisons across conditions and suit organization- or system-level use; condition-specific instruments have more credibility with patients and are more responsive to changes linked to a disease or treatment. The authors recommend using both, depending on the level of the health system concerned.

The properties that matter

• validity: does the tool measure what it claims to measure?
• reliability: does it give the same result under the same conditions?
• responsiveness: does it pick up real change?
• interpretability: do we know what a change means?

The same review notes that many PROMs were designed for research rather than clinical practice, with floor or ceiling effects that limit their everyday usefulness. A tool validated in one population does not automatically transfer to another, and each language version needs its own validation. In France, the HAS publishes guides listing PROMs translated and validated in French for several conditions and procedures; equivalent work exists in other countries. Check terms of use and licensing too.

Then there is the burden. Five minutes on paper can become fifteen for someone who is tired, has reading difficulties or is not comfortable with digital tools: test the questionnaire with real patients before rolling it out.

A baseline measurement provides the point of comparison. The rhythm that follows should match the expected pace of change and the team’s capacity to respond: too rare, it misses a change; too frequent, it tires patients and multiplies signals no one can act on. Define the response window and what happens in case of hospitalization, non-response or a change of treatment. Document any change in how data is collected: moving from paper to digital can affect a series.

A PROM becomes useful when it fits into a complete workflow: invitation, response, scoring, display, interpretation, discussion, action and documentation. Each step needs an owner and a time frame. The display shows the current score, the trend, key answers and missing data; a color code is not enough. The clinician should be able to tell whether a change exceeds the minimal clinically important difference, where one has been established for the instrument, or remains within measurement uncertainty.

Answers that call for action

Some questionnaires include sensitive items, such as severe distress. Before launch, decide what triggers an alert, who receives it and how quickly. Tell patients that a questionnaire is not an emergency service and may not be read in real time. An alert with no capacity to respond creates risk: it can be safer to start with use during consultations, without asynchronous alerts. Designing a full alert chain is covered in our guide to remote patient monitoring, and getting scores into the patient record is a matter of healthcare interoperability.

MyTwin for clinicians brings together, between visits, the patient record, measurements and what patients report, including during the regular follow-up call from the AI assistant. The team chooses which modules are active, and interpreting the answers and deciding what to do remain theirs.

The score should not override the patient’s story. In a qualitative study published in BMJ Open, dialysis patients and clinicians stressed one point: results should always be discussed. Filling in a questionnaire with no follow-up leaves patients feeling unheard. They recommend that both sides prepare, and that they choose the three or four priority topics together; patients also asked that results not be presented like a grade.

Show the trajectory, ask what explains a change, and identify with the patient the domain that matters most to them. Overall stability can hide a specific symptom; a deterioration can reflect a passing event. A PROM can justify an assessment, not make a diagnosis on its own. Documenting the discussion and the action makes it possible to check later whether the measurement actually changed the care pathway.

Digital-only collection excludes patients without devices, with disabilities, or with language or reading difficulties. Offer an assisted alternative, test accessibility and track response rates by group: an overall average can hide systematic under-representation.

Explain to patients why the data is collected, who receives it, how long it is kept and what their rights are. Keep care, quality evaluation and research as separate as possible, since they follow different rules. In France, for example, the CNIL, the data protection authority, requires separate formalities before data gathered in a health data warehouse can be reused for research. And collect only what will be used.

1. Pick one care pathway and one specific clinical decision.
2. Design the program with patients, clinicians, and the quality, digital and data protection teams.
3. Choose a validated instrument and write the measurement protocol.
4. Set up the display, alerts, responsibilities and the non-digital alternative.
5. Pilot on a limited scope, measuring workload and use.
6. Decide on scaling up based on documented results.

The first indicators concern implementation: response rate, missing data, time before the score is viewed, documented actions, acceptability. Health outcomes come next, with an analysis plan that avoids attributing every observed change to the questionnaire alone. From 2021 to 2023, the HAS supported four teams, in primary care and in hospitals, that rolled out PROMs in routine practice, and drew from them the obstacles and enablers of implementation.

Four failures come up often: collection without action, where scores pile up unread; overload, when too many tools and reminders lead to drop-out; mechanical reading of a threshold; and unequal access, when only the best-equipped patients respond. A monthly review in the early months, covering incidents, response rates, alerts and team feedback, turns the project into a learning loop rather than a fixed rollout.