Objective To investigate the complete recovery rate of renal function, clinical characteristics and risk factors for non-recovery in the elderly with acute kidney injury (AKI), determine the effect of different causes on the short-term prognosis, and investigate the relationship of different stages of AKI with renal function recovery. Methods A retrospective cohort study was carried out on 652 elderly AKI patients (≥75 years old) who hospitalized in our geriatric wards from January 2007 to December 2015. After 90 days’ follow-up, these patients were divided into survival group and death group, and the former group was further assigned into recovery group and non-recovery group according to their renal function. Basic information, clinical data (diagnostic time, mean arterial pressure, urine volume, dialysis, mechanical ventilation, observation indicators and ratio of persistent AKI), AKI stages and causes were collected and analyzed. SPSS statistics 17.0 was used for data processing. Student’s t test, Mann-Whitney U test, Chi-sqaure test or Fisher’s exact test was used for comparison of different data types between groups. Multivariate logistic regression analysis (forward) was adopted to analyze the risk factors of renal function recovery. Results The median age of the 652 patients was 87(84 to 91) years. The mortality rate of the cohort was 33.6%(219/652). Among the 433 survivals, 316 patients (73%,316/433) got complete recovery in renal function, and the other 117 cases (27%, 117/433) did not recover. Infection, hypovolemia, cardiova-scular events, nephrotoxicity, and surgery were the common causes for hospital acquired AKI in the elderly. The incidence of infection (53.0% vs 33.0%; P<0.001) was significantly higher, while the ratio of using nephrotoxic drugs (5.5% vs 15.2%; P=0.001) was obviously lower in the death group than the survival group. For the survived patients, the recovery subgroup had higher ratio of diabetes mellitus, lower baseline serum creatinine (SCr), higher basal estimated glomerular filtration rate (eGFR) than the non-recovery subgroup (P<0.01). Furthermore, the recovery subgroup also had shorter time for diagnosis of AKI, lower peak level and SCr level at diagnosis, higher level of blood urea nitrogen (BUN), and lower ratios of dialysis requirement and persistent AKI (P<0.05). But there was no significant difference in AKI stage between the 2 subgroups (P>0.05). Multivariate logistic regression analysis showed that increased basal eGFR (OR=0.897,5%CI:0.842-0.956; P=0.001) was a protective factor for renal function recovery, while persistent AKI (OR=4.497,5%CI:2.774-7.290; P<0.001) was a risk factor for renal function recovery in elderly patients with AKI at 90 d after diagnosis. Conclusion Clinicians should identify the key causes influencing the prognosis of elderly AKI, that is, infection and application of nephrotoxic drugs, and concern about decreased basal eGFR and persistent AKI. What’s more, early detection and intervention may improve the prognosis of elderly AKI patients.