AbstractThe use of wildlife rehabilitation for conservation is growing, but quantitative criteria are rarely used to guide whether and when to remove animals from the wild. Since 2006, large numbers of African penguin Spheniscus demersus chicks have been abandoned annually when adults enter moult with dependent young still in the nest. As part of conservation initiatives for this endangered species, these chicks were collected and hand reared to fledging age. Post‐release survival has been well documented; in this study we develop models to predict survival of individuals during rehabilitation with the aim of improving hand‐rearing success and guiding the use of scarce resources. For 1455 chicks abandoned between 2008 and 2013, we assessed whether a chick body condition index (BCI) could predict outcome (death or release) and time spent in rearing. In addition, for a subset of 173 chicks in 2012, we assessed whether BCI at admission influenced chick growth rates during rehabilitation and examined whether the use of additional structural measurements and sex provided additional power to predict outcome. Models predicted an 82.9% (95% confidence interval: 73.3–89.5%) release rate for chicks admitted with a BCI >0, the proposed guideline for removal from colonies. This fell below 50% for BCIs < −1.05; 66% of chicks were admitted with BCIs between these thresholds. Adding bill length to BCI improved the relative model fit, but in both cases only ~70% of rehabilitation outcomes were correctly predicted. Chicks that grew more quickly were more likely to be released and, for those that were released, had lower BCI at admission suggesting compensatory growth. Chicks were generally removed at an appropriate time to ensure successful hand‐rearing. However, 32% were admitted in good condition, highlighting the importance of using adaptive management to guide wildlife rehabilitation and the allocation of conservation resources.