AI-assisted pain checks found pain in one in five assessments—but did care improve?
A Canadian long-term-care home recorded 3,826 PainChek assessments and pain in 764 of them. Nurses generally accepted the tool, but the uncontrolled case study did not show whether residents received better treatment or outcomes.
The short version
- What happened
- PainChek was introduced across four units with 160 beds after 18 training sessions. Sixty-five nurses were surveyed before implementation and 12 weeks later during a June-to-October 2024 quality-improvement project.
- Who it affects
- Residents who cannot explain pain may benefit when staff have another structured prompt to look for it. Nurses still make the assessment and treatment decision, and a facial-analysis score should not override clinical context, family knowledge or changes that require urgent medical review.
- What remains unproven
- This was a quality-improvement case study in one Canadian long-term-care home, without randomisation or a concurrent control group.
Keep in mind
This was a quality-improvement case study in one Canadian long-term-care home, without randomisation or a concurrent control group.
The paper reports 3,826 assessments but not the number of unique residents assessed; repeated observations are therefore not independent people.
It did not measure treatment changes, pain relief, adverse effects, responsive behaviour with a validated outcome, hospital use or family experience.
Nurse experience was self-reported, and the 12-week post-implementation window cannot establish sustained use or long-term benefit.
Source analysis · 1 linked source. Primary analysed source date: 31 October 2026.
Keep following the evidence
Join The Impact Brief founding list for selected AI reports, their effect on people and the evidence limits. Read a sample first.
Read the full report · 4 min The evidence, methods, implications and limitations
The tool combined a facial scan with staff observation
Pain can be missed when residents with dementia or other cognitive impairment cannot describe it reliably. PainChek uses automated facial analysis alongside staff-entered observations of voice, movement, behaviour, activity and bodily signs. A Canadian long-term-care home introduced the commercial app across four units with 160 beds, running 18 training sessions before a 12-week evaluation. Sixty-five nurses completed surveys before implementation and again afterwards.
During the project, staff logged 3,826 assessments and recorded mild-to-severe pain in 764 of them, or 20%. The share of assessments followed by another check increased from 69% in the first month to 78% in the third, while zero-pain scores fell from 14% to 6.2%. Nurses generally rated the tool easy to use and aligned with existing work. Tablet shortages were reported by 20% and the time required by 18% as barriers.[1]
More recorded pain is not yet evidence of better treatment
The study did not find significant changes in nurses' confidence in their own assessments, existing pain-assessment practices or perceived workflow effects. Confidence in the technology's reliability decreased slightly after use, a change reported with p=0.009. That could reflect more informed judgement after hands-on experience, but the design cannot establish why it changed. Nurses were also uncertain that the tool reduced responsive behaviours.
The denominator matters. Four units had 160 beds, but the paper does not report how many unique residents received the 3,826 assessments, so repeated checks cannot be interpreted as 764 people in pain. There was no control unit, random assignment or comparison with care before deployment. The project did not report analgesic changes, comfort, adverse effects, hospital transfers or family experience. CABHI funded the implementation and evaluation but was reported to have no role in the study; the authors declared no competing interests. The next useful evidence is a controlled evaluation linking assessments to treatment decisions and resident outcomes.[1]
The distinction that matters
What the project measured—and what it did not
Assessment activity increased, but treatment and resident outcomes were outside the reported evidence.
Implementation
Train staff and introduce the app across four units
Nurses reported generally high usability and satisfaction
Recorded assessments
Complete 3,826 checks during the evaluation
764 assessments recorded mild-to-severe pain
Care outcome
Link detection to treatment and resident wellbeing
Not measured with a control or patient-outcome comparison
Watch the background, then check the evidence
Illustrative photo · not a video frame
The Royal Society · 44 min 28 sec · Background explainer
Is this the AI future we want? | The Royal Society
YouTube loads only when you choose; playback starts when you press play.
A Royal Society discussion with Madhumita Murgia and biomedical engineer Alison Noble on AI’s potential in healthcare, science and culture, recorded at the 30 June 2026 event.
Evidence check: A discussion of opportunities and risks, including future possibilities. It is not evidence that every application described is available or clinically proven.
Published 2026-07-01. Source checked 14 October 2026. Check captions on the publisher’s video page.
Check the healthcare evidencePhoto: Daniel Sone / National Cancer Institute / Unsplash · Licence
Continue the story
Read next
Continue exploring computer vision: compare the finding and its evidence limits.
Can AI identify a medicine safely outside the lab?
A review of 91 studies finds rapid progress in pill detection, classification and imprint reading, but little external, prospective or clinical validation. High benchmark accuracy is not evidence that an app prevents medication harm.
Read the next reportExplore the full source and evidence record
What this means for people
- Residents who cannot explain pain may benefit when staff have another structured prompt to look for it. Nurses still make the assessment and treatment decision, and a facial-analysis score should not override clinical context, family knowledge or changes that require urgent medical review.
Global context
AI-assisted pain tools are being considered in aged care because pain is often under-recognised in people with advanced cognitive impairment. Deployment studies can reveal workflow barriers that validation studies miss, but systems should not be judged by assessment volume alone. The important question is whether detection leads to safer, more appropriate relief without over-treatment or missed alternative diagnoses.
What the evidence does not yet show
- This was a quality-improvement case study in one Canadian long-term-care home, without randomisation or a concurrent control group.
- The paper reports 3,826 assessments but not the number of unique residents assessed; repeated observations are therefore not independent people.
- It did not measure treatment changes, pain relief, adverse effects, responsive behaviour with a validated outcome, hospital use or family experience.
- Nurse experience was self-reported, and the 12-week post-implementation window cannot establish sustained use or long-term benefit.
What to watch next
- Unique resident and repeat-assessment counts
- Controlled comparisons with usual care
- Treatment and pain-relief outcomes
- Performance across skin tones and facial conditions
- Staff override and escalation patterns
- Sustained use, cost and tablet availability
Living evidence record
Impact record IAI-1KVG2B9
Evidence stage
Studied
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Present
Record status
Monitoring
Last checked
31 October 2026
Source trail
1 direct source across 1 source type.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Single-source reporting disclosure
This record analyses one direct source. It can establish what BMC Health Services Research published or reported, but it is not independent corroboration of every performance claim or predicted outcome. The confidence label will change only when broader evidence is added.
Evidence trail
Sources used for this report
Links checked 31 October 2026
Research question, topics and regions
Topics: Long-term care · Dementia · Pain assessment · Nursing · Clinical implementation · Computer vision
Regions: North America · Canada
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Reader commentary
What does this mean where you are?
Tell us what you have seen at work, in your community or in everyday life. A question is welcome too. No account is needed; we review notes before publishing them.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.