Abstract Objectives To determine whether time elapsed since the second injection of the Pfizer-BioNTech BNT162b2 mRNA vaccine was significantly associated with the risk of covid-19 infection after vaccination in people who received two vaccine injections. Design Test negative design study. Setting Electronic health records of a large state mandated healthcare organisation, Israel. Participants Adults aged ≥18 years who had received a reverse transcription polymerase chain reaction (RT-PCR) test between 15 May 2021 and 17 September 2021, at least three weeks after their second vaccine injection, had not received a third vaccine injection, and had no history of covid-19 infection. Main outcome measures Positive result for the RT-PCR test. Individuals who tested positive for SARS-CoV-2 and controls were matched for week of testing, age category, and demographic group (ultra-orthodox Jews, individuals of Arab ancestry, and the general population). Conditional logistic regression was adjusted for age, sex, socioeconomic status, and comorbid conditions. Results 83 057 adults received an RT-PCR test for SARS-CoV-2 during the study period and 9.6% had a positive result. Time elapsed since the vaccine injection was significantly longer in individuals who tested positive (P<0.001). Adjusted odds ratio for infection at time intervals >90 days since vaccination were significantly increased compared with the reference of <90 days: 2.37 (95% confidence interval 1.67 to 3.36) for 90-119 days, 2.66 (1.94 to 3.66) for 120-149 days, 2.82 (2.07 to 3.84) for 150-179 days, and 2.82 (2.07 to 3.85) for ≥180 days (P<0.001 for each 30 day interval). Conclusions In this large population of adults tested for SARS-CoV-2 by RT-PCR after two doses of mRNA BNT162b2 vaccine, a gradual increase in the risk of infection was seen for individuals who received their second vaccine dose after at least 90 days.