The treatment of Rockwood type III and V acromioclavicular (AC) joint dislocations is controversial, and an individualized treatment algorithm is yet to be developed. The objective of this study was to investigate the association of demographical, clinical, patient-reported and radiological variables with the Western Ontario Shoulder Instability Index (WOSI) score and risk of surgery. Inclusion criteria for this prospective cohort study were patients aged 18-60 with an acute AC joint dislocation with >25% increase in the coracoclavicular distance on bilateral Zanca radiographs. Patients were treated non-surgically with 3 months of home-based training and the option of delayed surgical intervention. The outcomes were the WOSI score and surgery yes/no. Demographical, clinical, patient-reported (WOSI and Shoulder Pain and Disability Index [SPADI]) and radiological variables were collected at baseline and 6 weeks after the injury and investigated for association with the outcomes at 3 months, 6 months and 1 year. Ninety-five patients with Rockwood type III/V AC joint dislocation were included. Pre-injury participation in overhead/collision sports was a risk factor for surgery with an odds ratio of 5 (p = 0.03). Reduced range of motion (ROM) at baseline was associated with reduced WOSI scores and increased risk of surgery. At 6 weeks, reduced ROM, increased SPADI and increased pain during cross-over were associated with the outcomes. Radiological measurements were not correlated with the result. At the 6 weeks follow-up, patients eventually requiring surgery could be detected with a sensitivity of 100% and a specificity of 94% based on a SPADI score of >30 and a ROM ≤ 140° in shoulder flexion or abduction. ROM was the only variable consistently associated with both WOSI and risk of surgery. Six weeks after the injury, it was possible to detect patients in need of surgery based on ROM and SPADI with a sensitivity of 100% and a specificity of 94%. Level II.