Seasonality of urinary tract infections in the United Kingdom in different age groups: longitudinal analysis of The Health Improvement Network (THIN)

Abstract

Evidence regarding the seasonality of urinary tract infection (UTI) consultations in primary care is conflicting and methodologically poor. To our knowledge, this is the first study to determine whether this seasonality exists in the UK, identify the peak months and describe seasonality by age. The monthly number of UTI consultations (N = 992 803) and nitrofurantoin and trimethoprim prescriptions (N = 1 719 416) during 2008-2015 was extracted from The Health Improvement Network (THIN), a large nationally representative UK dataset of electronic patient records. Negative binomial regression models were fitted to these data to investigate seasonal fluctuations by age group (14-17, 18-24, 25-45, 46-69, 70-84, 85+) and by sex, accounting for a change in the rate of UTI over the study period. A September to November peak in UTI consultation incidence was observed for ages 14-69. This seasonality progressively faded in older age groups and no seasonality was found in individuals aged 85+, in whom UTIs were most common. UTIs were rare in males but followed a similar seasonal pattern than in females. We show strong evidence of an autumnal seasonality for UTIs in individuals under 70 years of age and a lack of seasonality in the very old. These findings should provide helpful information when interpreting surveillance reports and the results of interventions against UTI.

Keywords: The Health Improvement Network; Urinary tract infections (UTIs); seasonality.

Fig. 1.

Seasonality in UTI consultations coded in the UK per 100 000 person years by age. The left panels show the rate of UTI consultations by age group. The central red lines represent the fitted predictions of the negative binomial polynomial regression model of degree 2 with the number of patients registered at each of the GP practices on 1 July (mid-year) each year of the study as offset. The central blue lines represent the fitted predictions of the same model but with a seasonal component included. The shaded areas represent the 95% confidence intervals for their respective models. These were calculated using the standard errors from the predict function, which calculates the confidence intervals around the mean. The right panels show the correlograms for the residuals of the regression models without seasonality at lags of 0–12 months for each age group. The September to November period is shaded in grey. The UTI consultations were de-duplicated to one per 30-day period. The y-axes differ between panels.

Fig. 2.

Seasonality in UTI consultations coded in the UK per 100 000 person years by age group and sex. The left panels show the rate of UTI consultations by age group and sex. The central red lines represent the fitted predictions of the negative binomial polynomial regression model of degree 2 with the number of patients registered at each of the GP practices on 1 July (mid-year) each year of the study as offset. The central blue lines represent the fitted predictions of the same model but with a seasonal component included. The shaded areas represent the 95% confidence intervals for their respective models. These were calculated using the standard errors from the predict function, which calculates the confidence intervals around the mean. The right panels show the correlograms for the residuals of the regression models without seasonality at lags of 0–12 months for each age group. The September to November period is shaded in grey. The UTI consultations were de-duplicated to one per 30-day period. The y-axes differ between panels.