We have compared the outcome of hemiarthroplasty of the shoulder in three distinct diagnostic groups, using survival analysis as used by the United Kingdom national joint registers, patient-reported outcome measures (PROMs) as recommended by Darzi in the 2008 NHS review, and transition and satisfaction questions. A total of 72 hemiarthroplasties, 19 for primary osteoarthritis (OA) with an intact rotator cuff, 22 for OA with a torn rotator cuff, and 31 for rheumatoid arthritis (RA), were followed up for between three and eight years. All the patients survived, with no revisions or dislocations and no significant radiological evidence of loosening. The mean new Oxford shoulder score (minimum/worst 0, maximum/best 48) improved significantly for all groups (p < 0.001), in the OA group with an intact rotator cuff from 21.4 to 38.8 (effect size 2.9), in the OA group with a torn rotator cuff from 13.3 to 27.2 (effect size 2.1) and in the RA group from 13.7 to 28.0 (effect size 3.1). By this assessment, and for the survival analysis, there was no significant difference between the groups. However, when ratings using the patient satisfaction questions were analysed, eight (29.6%) of the RA group were ‘disappointed’, compared with one (9.1%) of the OA group with cuff intact and one (7.7%) of the OA group with cuff torn. All patients in the OA group with cuff torn indicated that they would undergo the operation again, compared to ten (90.9%) in the OA group with cuff intact and 20 (76.9%) in the RA group. The use of revision rates alone does not fully represent outcome after hemiarthroplasty of the shoulder. Data from PROMs provides more information about change in pain and the ability to undertake activities and perform tasks. The additional use of satisfaction ratings shows that both the rates of revision surgery and PROMs need careful interpretation in the context of patient expectations