Several studies have proposed that the neutrophil–lymphocyte ratio (NLR) is one of the various biomarkers that can be useful in assessing COVID-19 disease-related outcomes. Our systematic review analyzes the relationship between on-admission NLR values and COVID-19 severity and mortality. Six different severity criteria were used. A search of the literature in various databases was conducted from 1 January 2020 to 1 May 2021. We calculated the pooled standardized mean difference (SMD) for the collected NLR values. A meta-regression analysis was performed, looking at the length of hospitalization and other probable confounders, such as age, gender, and comorbidities. A total of sixty-four studies were considered, which included a total of 15,683 patients. The meta-analysis showed an SMD of 3.12 (95% CI: 2.64–3.59) in NLR values between severe and non-severe patients. A difference of 3.93 (95% CI: 2.35–5.50) was found between survivors and non-survivors of the disease. Upon summary receiver operating characteristics analysis, NLR showed 80.2% (95% CI: 74.0–85.2%) sensitivity and 75.8% (95% CI: 71.3–79.9%) specificity for the prediction of severity and 78.8% (95% CI: 73.5–83.2%) sensitivity and 73.0% (95% CI: 68.4–77.1%) specificity for mortality, and was not influenced by age, gender, or co-morbid conditions. Conclusion: On admission, NLR predicts both severity and mortality in COVID-19 patients, and an NLR > 6.5 is associated with significantly greater the odds of mortality.