Changed frequency of enterovirus hospitalisations after change of polio vaccination scheme: a nation-wide cohort study from Denmark.

INTRODUCTION: This study examined whether ending oral poliovirus vaccination (OPV) increased enterovirus (EV) associated hospitalizations in children. Inactivated poliovirus vaccination (IPV) protects against disease caused by poliovirus. In contrast to OPV, it does not provoke a mucosal defence. Therefore, it is hypothesised that IPV does not offer cross-protection against other EV infections. METHODS: The prevalence and incidence of EV cases were estimated in three cohorts of Danish children. To address the historically sparse diagnosis and registration of EV, we trained supervised machine-learning models in 2017-2018 admission data, validated them against the 2015-2016 cohort and applied the final model to the period 1996-2014. The study was conducted using nation-wide Danish registries on hospital admissions in 1996-2018. Three cohorts of all children living in Denmark born between 1 January 1996 and 31 December 1998 (201 460 live births), 1 January 2000 and 31 December 2004 (325 825 live births) and between 1 January 2005 and 31 December 2007 (194 303 live births) were followed in hospital registers until age 12. The outcomes were the prevalence and incidence of EV cases in the three cohorts and the ratio of estimates between the OPV and IPV cohorts. RESULTS: Among 201 460 live births in the OPV cohort, 178 EV cases were identified, corresponding to an incidence rate of 6.48 per 100 000 risk years. The incidence rates per 100 000 risk years were 10.64 and 13.81 in the IPV cohorts, with 460 and 326 EV cases, respectively. The incidence rate ratio was 1.64 (95% CI 1.38 to 1.95) and 2.13 (1.78 to 2.56) in the two IPV cohorts compared with the OPV cohort. CONCLUSION: Children born in Denmark after 2001 had significantly higher rates of EV-related hospitalisations, suggesting potential cross-protective effects of OPV for EV-related hospitalisations. While supporting the hypothesis, the results should not be seen as risk attribution estimates, as the residual impact of diagnostic improvements is unmeasured.