Whether host genetic factors determine clinical severity of influenza virus infections is unclear. This study examined the effects of single nucleotide polymorphisms (SNPs) of IFITM3, TLR3, CD55, and TLR4 genes on outcomes of avian influenza A (H7N9) and pandemic influenza A (H1N1)pdm09 in Chinese patients. This multicentre cohort study took place in four centres in Hong Kong, Beijing, Guangdong, and Shanghai and included adults prospectively diagnosed with H7N9 and H1N1pdm09 infections during three respective seasonal outbreaks (H7N9, 2013–15; H1N1pdm09, 2011–14). The inclusion criteria were PCR-confirmed H7N9 or H1N1pdm09 virus infection, age 18 years or older, and Chinese ethnicity. Host DNA was extracted from diagnostic respiratory samples, and IFITM3(rs12252), TLR3(rs5743313), CD55(rs2564978), and TLR4(rs4986790/4986791) were genotyped by Sanger sequencing. The data were tested to see whether they deviated from the Hardy-Weinberg equilibrium, and were compared with the 1000 Genomes data. The primary outcome for analysis was mortality. The effects of the SNPs were examined under different genetic models (recessive, additive, and dominant). The joint effect of significant loci was tested by computing the genetic risk score. All identifying patient information was removed from the dataset, and only anonymous data were used for analysis; individual consent was not required. Ethics approvals were obtained from the institutional review boards of all participating institutes. Our cohort consisted of 275 adults with H7N9 and H1N1pdm09 infections, of whom 33 died. In patients who died, we found over-representation of the homozygous IFITM3 CC genotype (18 54% of 33 patients with available data who died vs 70 33% of 211 with available data who survived; p=0·017) and the TLR3 CC genotype (28 93% of 30 vs 160 77% of 208; p=0·039). Recessive genetic models, adjusted for age, comorbidity, and antiviral treatment, showed that these genotypes had a significant association with increased risk of death (IFITM3 CC genotype, adjusted hazard ratio HR 2·78 95% CI 1·29–6·02; and TLR3 CC genotype, adjusted HR 4·85 1·11–21·06). Cumulative effects were found (adjusted HR 3·53 95% CI 1·64–7·59 per risk genotype; adjusted HR 9·99 1·27–78·59 with both genotypes). The case-fatality rate in patients with a genetic risk score of 2, 1, and 0 was 23%, 11%, and 3%, respectively. Results were similar for each influenza subtype and other severity indicators such as development of acute respiratory failure. The population-attributable risk for mortality was 35·5% (95% CI 9·9–69·8) for IFITM3 CC and 74·6% (21·7–100) for TLR3 CC (combined effect, 83·6% 52·3–100), owing to their high prevalence among Chinese people. A SNP of CD55 (genotype TT) was linked to severity of influenza as indicated by an increased risk for hospital admission (adjusted odds ratio OR 2·77, 95% CI 1·21–6·36; p=0·02); TLR4 was shown to be non-polymorphic in this cohort. Host genetic factors might influence clinical outcomes of avian and pandemic influenza. Our findings could have important implications on public health and health-care planning, patient care, and design of clinical trials. Health and Medical Research Fund (RRG-09) of the Hong Kong Special Administrative Region, China.