Dual-mobility choice is a property of the surgeon, not of the patient

Design

Retrospective cohort with cross-classified multilevel logistic regression, National Joint Registry linked to Hospital Episode Statistics

Population

238,455 primary total hip arthroplasties in England, 2018-2022, of which 7,032 dual-mobility

Primary outcome

Proportion of variation in dual-mobility use attributable to patient, surgeon and institution

Effect

Surgeon 32.4 per cent, institution 22.9 per cent, patient about 10 per cent (marginal R-squared 0.10; conditional R-squared 0.598)

All 238,455 primary total hip arthroplasties recorded in the National Joint Registry for England between January 2018 and December 2022 were linked to Hospital Episode Statistics, and cross-classified multilevel logistic regression was used to partition where the variation in dual-mobility implant use actually sits. Seven thousand and thirty-two hips received a dual-mobility component; those patients were older, frailer and more often being treated for a fractured neck of femur.

That clinical patterning is real but small. Patient-level characteristics explained about 10 per cent of the variance (marginal R-squared 0.10). Surgeon-level factors explained 32.4 per cent and institutional factors 22.9 per cent, with the full model accounting for 59.8 per cent. Restricting the analysis to high-volume surgeons did not change the picture.

The uncomfortable reading is that whether a given patient receives a dual-mobility cup depends more on which list they land on than on their dislocation risk. This is a registry study and cannot say which surgeons are right - it is entirely possible that the high users are correctly anticipating instability that the recorded variables do not capture. But unwarranted variation of this size is the standard argument for a prognostic tool and a targeted randomised trial, and the authors make it. In the meantime the practical action is local: know your own unit's dual-mobility rate and the reasoning behind it, because if it differs sharply from the unit next door, that difference is currently unexplained by the patients.

• Find out your unit's dual-mobility rate and how it compares with peers; this is auditable data.
• Record the specific instability risk factors that led to the choice, so the decision is reconstructable.
• Treat frailty and neck-of-femur fracture as the recognised indications; they drove use here.
• Do not read high use as evidence of better practice - the registry cannot judge appropriateness.
• Where a trial is offered, this is the selection question it needs to answer.

The statistics, in plain English

Marginal R-squared measures how much the fixed patient variables explain; conditional R-squared adds the surgeon and hospital random effects. The jump from 0.10 to 0.598 is the whole finding: most of what predicts a dual-mobility cup is clustering by who and where, not by patient. Registry data cannot capture unmeasured clinical judgement, so some of that surgeon variance may be appropriate selection this dataset cannot see.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.