Your Constant score is 14 points more optimistic than your patient

Design

single-centre retrospective study with Bland-Altman agreement analysis and Firth penalised logistic regression

Population

74 shoulders in 72 patients (mean age 56 years, 56% male) after locking plate fixation of displaced proximal humerus fractures, minimum 1 year follow-up

Primary outcome

agreement between Constant-Murley score and shoulder subjective value, and predictors of poor self-evaluated outcome

Effect

mean difference -14.21 points (95% CI -17.25 to -11.18); pain independently associated with poor subjective value (coefficient 3.5, 95% CI 1.53 to 6.47)

Outcome after proximal humerus fracture fixation is usually reported with the Constant-Murley score, which a clinician measures, and increasingly alongside the shoulder subjective value, which the patient states as a percentage of a normal shoulder. This study of 74 shoulders in 72 patients, all having had locking plate fixation of a displaced fracture between 2015 and 2024 with at least a year of follow-up, asked how far apart the two run.

On Bland-Altman analysis the mean difference was -14.21 points (95% CI -17.25 to -11.18), an interval that does not include zero, and Pitman's test found the variances differed too (p=0.029). So the physician-measured score is systematically higher, and the disagreement is not uniform across patients. On penalised logistic regression, the only factor independently associated with a poor self-evaluated outcome was the presence of shoulder pain (coefficient 3.5, 95% CI 1.53 to 6.47). Limitations of elevation, external rotation and internal rotation, and reduced strength, were not.

That is the finding worth carrying into clinic. Range of motion is what gets measured, recorded and discussed at follow-up, and it is not what determines whether the patient thinks their shoulder is any good - pain is. In the 66 shoulders without avascular necrosis, malreduction (p=0.009) and an absent postoperative medial hinge (p=0.032) were both associated with more frequent pain, which connects the intraoperative decision to the outcome the patient actually reports.

• Ask for the patient's own percentage rating alongside whatever score you measure - they are answering different questions.
• Treat persistent pain, not range of motion, as the marker of a poor outcome from the patient's point of view.
• Protect the medial hinge at fixation; its absence was associated with more frequent postoperative pain.
• Review reduction quality on the postoperative film with the same seriousness as the construct.
• Small single-centre series of 74 shoulders - the direction of the disagreement is more reliable than the 14-point figure.

The statistics, in plain English

Bland-Altman analysis asks whether two measurements agree, not whether they correlate, and a mean difference of -14.21 with an interval of -17.25 to -11.18 says they systematically do not. The authors are careful to say they are not testing substitutability, because the two instruments measure different constructs - one is an examination, the other an opinion. The threshold defining a poor subjective outcome (69.4) was derived from the same dataset, which risks circularity; the sensitivity analysis using a pre-specified threshold of 75 reaching the same conclusion is what makes the pain finding credible. With 74 shoulders, the associations with malreduction and medial hinge are exploratory bivariate comparisons, not adjusted analyses.

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.