Just out todayDigital health tech: Ninety-nine nurses, seventy wearable features, and no control groupDigital health tech: Fifty-six trials, 6,154 patients, and headsets beat the threshold patients would noticeSynthetic biology & cell therapy: Roughly one in four patients on these therapies gets a severe infection

Digital health tech/China

Fifty-six trials, 6,154 patients, and headsets beat the threshold patients would notice

A network meta-analysis of 56 trials found extended reality reduced perioperative anxiety above the threshold patients notice.

What the paper found

A Sichuan University team ran a network meta-analysis of 56 randomised trials covering 6,154 patients, using RoB 2 for bias and GRADE for certainty, and judged effects against minimal important differences. Extended reality most effectively reduced perioperative anxiety, SMD 0.60, mean difference 8.05 against a minimal important difference of 6.71, moderate certainty.

Grade 12 reading level6 min read

What happened

A team at the National Clinical Research Center for Geriatrics, West China Hospital of Sichuan University, the Cochrane China Centre and the MAGIC China Centre published in npj Digital Medicine a network meta-analysis of digital health interventions for perioperative patient-reported outcomes.

They searched PubMed, Embase, CENTRAL and Web of Science to 1 March 2025 for randomised trials in adults undergoing elective surgery under general anaesthesia.

Fifty-six trials covering 6,154 patients were included.

Risk of bias was assessed with RoB 2. Certainty was assessed with GRADE. And effects were estimated as standardised mean differences and mean differences with minimal important differences attached.

Extended reality most effectively reduced perioperative anxiety: SMD 0.60, mean difference 8.05 against a minimal important difference of 6.71, at moderate certainty.

The test

The methodological choice that distinguishes this review is the minimal important difference.

A standardised mean difference of 0.60 tells you an effect is moderate in statistical terms. It does not tell you whether a patient would notice. The minimal important difference is the smallest change on a scale that patients consider worthwhile, and comparing an effect against it converts a statistic into a clinical judgement.

Most meta-analyses stop at the standardised effect size. This one reports both, plus GRADE certainty for each comparison, plus RoB 2 for bias.

Fifty-six trials and 6,154 patients is a substantial evidence base. The network structure allows comparisons between interventions that were never tested head to head, which is powerful and rests on the assumption that the trials are similar enough to connect.

The numbers

Outcome and intervention SMD MD vs MID Certainty
Anxiety, extended reality 0.60 (0.37 to 0.84) 8.05 vs 6.71 moderate
Postoperative pain, mobile applications 0.64 (0.32 to 0.95) 1.36 vs 1.0 moderate
Postoperative pain, extended reality 0.51 (0.26 to 0.76) 1.09 vs 1.0 moderate
Trials included 56
Patients 6,154
Bias assessment RoB 2
Certainty assessment GRADE
Search cut-off 1 March 2025

The result

Extended reality reduced anxiety by 8.05 points against a threshold of 6.71. It clears the bar, and the margin is 1.34 points.

For postoperative pain, the margins are narrower still. Mobile applications achieved 1.36 against a threshold of 1.0. Extended reality achieved 1.09 against the same 1.0. The second of those exceeds the minimal important difference by 0.09 points, which is a technical pass.

All three findings carry moderate certainty, not high. GRADE’s moderate rating means the true effect is probably close to the estimate but could differ.

So the accurate summary is that headsets and apps produce effects that patients would probably just notice, with moderate confidence. That is a real finding and a modest one, and reporting it this way is more useful than reporting three statistically significant standardised mean differences.

How this compares to what came before

Perioperative anxiety and pain have attracted a long list of digital interventions: headsets, apps, videos, messaging. The trials are individually small and compare different things, which is exactly the situation a network meta-analysis exists for.

Within this batch’s five evidence syntheses, this is the most carefully reported. A Bayesian network meta-analysis of telerehabilitation pooled 20 trials and 1,854 participants with GRADE per comparison, and rated its AI arm low to very low certainty. A meta-analysis of AI on neck MRI used QUADAS-AI and GRADE and rated its own evidence low. A meta-analysis of infection risk after cell therapies pooled 16 studies and 2,097 patients without a certainty rating. A meta-analysis of brain stimulation for restless legs reported a pooled effect without stating the trial count or a certainty rating.

Ranked by what a reader can independently evaluate, this review comes first, and the minimal important difference reporting is why.

What it means

For perioperative services, a headset that reduces anxiety by an amount patients notice, at moderate certainty, across 56 trials, is a reasonable thing to offer. It is cheap, it is not a drug, and the downside is minimal.

For postoperative pain, the margins over the minimal important difference are small enough that the clinical case is weaker. An effect of 1.09 against a threshold of 1.0 should not be described as pain relief without qualification.

The structural limitation is blinding. A patient knows whether they are wearing a headset, and anxiety and pain are both self-reported. Expectation is present in every included trial, and no analytic technique removes it. For an intervention with no pharmacological action, a substantial part of the effect may simply be expectation, which is not worthless and is not what a device claim usually implies.

Business ideas from this paper

Perioperative extended reality as a service for day surgery units, priced against premedication and nursing time rather than against drug costs.

Minimal important difference benchmarking as a standard in digital health evidence packs. This review shows what it reveals, and most vendors’ evidence would not survive it.

Anxiety-specific rather than pain-specific products, since the anxiety margin over the threshold is ten times larger than the pain margin for extended reality.

What could be wrong with it

Blinding is essentially impossible in these trials, so expectation is baked into every included result and into the pooled estimate.

The margins over the minimal important difference are small: 1.34 points for anxiety, 0.36 and 0.09 for pain.

Certainty is moderate, not high, for the main findings.

Network meta-analysis estimates comparisons that were never made directly, and with 56 heterogeneous trials the transitivity assumption that licenses those estimates is substantial.

Data are stated to be included in the published article, which is not a public deposit, so the network cannot be refitted independently.

The interventions are grouped into categories such as extended reality and mobile applications, which contain very different products.

How sure can you be?

Confident that the methods are the most rigorous among this batch’s evidence syntheses: 56 trials, RoB 2, GRADE per comparison, and minimal important difference benchmarking.

Confident that extended reality reduced perioperative anxiety by more than the minimal important difference, at moderate certainty.

Confident that the pain effects exceeded their threshold by small margins.

Less confident that the effects reflect the interventions rather than expectation, since blinding is infeasible and all outcomes are self-reported.

Less confident about the indirect comparisons in the network, which rest on transitivity across 56 varied trials.

Not confident about which specific products within each intervention category produce the effect.

Our rubric places this in the upper part of the range for this batch: peer-reviewed, several institutions, a synthesis of randomised trials with formal bias and certainty assessment, and data in the article rather than deposited.

Do this today

Ask for the minimal important difference alongside any effect size. This review shows effects can be statistically clear and clinically marginal, and the difference is visible only when the threshold is stated.

If you run day surgery, extended reality for preoperative anxiety has the strongest case here. For postoperative pain the margin is thin enough to be cautious about.

If you are assembling an evidence pack for a digital product, include the minimal important difference. Vendors who do will stand out, and those who do not will eventually be asked.

Source: Digital health interventions for perioperative patient-reported outcomes: a network meta-analysis, npj Digital Medicine, 2026-02-03. PMID 41634301 · doi.org ().

Just Out Tech explains new research in plain language. This article was drafted with AI assistance and checked by a human against the original source.

What to remember
  • Network meta-analysis of 56 randomised controlled trials covering 6,154 adults undergoing elective surgery under general anaesthesia.
  • Risk of bias assessed with RoB 2, certainty with GRADE, and effects judged against minimal important differences.
  • Extended reality most effectively reduced perioperative anxiety: SMD 0.60 (95% CI 0.37 to 0.84), mean difference 8.05 against a minimal important difference of 6.71, moderate certainty.
  • For postoperative pain, mobile applications (SMD 0.64, MD 1.36, MID 1.0) and extended reality (SMD 0.51, MD 1.09, MID 1.0) were probably effective at moderate certainty.
Where this really is

Network meta-analysis of 56 randomised trials with RoB 2, GRADE certainty and minimal important difference benchmarking; blinding is infeasible in the included trials.

Questions people ask

What is a minimal important difference?

The smallest change on a scale that a patient would notice and consider worthwhile. Comparing an effect against it is how you tell a statistically significant result from a meaningful one.

Why does that make this review unusual?

Because most meta-analyses report statistical significance and standardised effect sizes, which say nothing about whether a patient would feel the difference.

How large is the margin?

For anxiety, 8.05 against a threshold of 6.71. It clears the bar and not by a wide margin. For pain, 1.36 and 1.09 against a threshold of 1.0, which is tighter still.

Can these trials be blinded?

Essentially not. A patient knows whether they are wearing a headset, so expectation is present in every included trial.

About the author

Mark Alex

Mark Alex is the founder and Managing Director of Real Biz Digital, a technology company operating out of Nairobi since 2018. He works in agentic AI and the Model Context Protocol, AI governance, enterprise software architecture and cybersecurity. He holds an MSc in Mechatronical Engineering from Obuda University in Budapest and a BSc in IT, Forensic Technology and Cybercrime, from USIU-Africa in Nairobi, and has published IEEE conference research on an AI-powered digital twin for greenhouse systems. He is the author of seven books. Between 2020 and 2024 he mentored more than 200 university students and interns in Nairobi. He writes every Just Out Tech article from the original research paper.