Background and purpose - Several randomized controlled trials have not shown any added benefit of arthroscopy over placebo surgery or physiotherapy in middle-aged patients with knee symptoms without trauma. We studied the characteristics of the knee arthroscopies performed in southern Sweden. Patients and methods - From the orthopedic surgical records from 2007-2009 in the Skåne region of Sweden (with a population of 1.2 million), we retrieved ICD-10 diagnostic codes and selected all 4,096 arthroscopies that were diagnosed peroperatively with code M23.2 (derangement of meniscus due to old tear or injury) or code M17 (knee osteoarthritis). We extracted information on cartilage and meniscus status at arthroscopy, and we also randomly sampled 502 of these patients from the regional archive of radiology and analyzed the preoperative prevalence of radiographic or magnetic resonance imaging (MRI)-defined osteoarthritis. Results - 2,165 (53%) of the 4,096 arthroscopies had the diagnostic code M23.2 or M17. In this subgroup, 1,375 cases (64%) had typical findings consistent with degenerative meniscal tear (i.e. that correspond to a degenerative meniscal tear in at least a third of all arthroscopies). Of the randomly sampled patients, the preoperative prevalence of radiological knee osteoarthritis was 46%. Interpretation - There is a discrepancy between evidence-based medicine treatment guidelines and clinical practice regarding the amount of knee arthroscopies performed in patients with symptoms of degenerative knee disease.