The use of continuous electronic prescribing data to infer trends in antimicrobial consumption and estimate the impact of stewardship interventions in hospitalized children

Abstract

Background: Understanding antimicrobial consumption is essential to mitigate the development of antimicrobial resistance, yet robust data in children are sparse and methodologically limited. Electronic prescribing systems provide an important opportunity to analyse and report antimicrobial consumption in detail.

Objectives: We investigated the value of electronic prescribing data from a tertiary children's hospital to report temporal trends in antimicrobial consumption in hospitalized children and compare commonly used metrics of antimicrobial consumption.

Methods: Daily measures of antimicrobial consumption [days of therapy (DOT) and DDDs] were derived from the electronic prescribing system between 2010 and 2018. Autoregressive moving-average models were used to infer trends and the estimates were compared with simulated point prevalence surveys (PPSs).

Results: More than 1.3 million antimicrobial administrations were analysed. There was significant daily and seasonal variation in overall consumption, which reduced annually by 1.77% (95% CI 0.50% to 3.02%). Relative consumption of meropenem decreased by 6.6% annually (95% CI -3.5% to 15.8%) following the expansion of the hospital antimicrobial stewardship programme. DOT and DDDs exhibited similar trends for most antimicrobials, though inconsistencies were observed where changes to dosage guidelines altered consumption calculation by DDDs, but not DOT. PPS simulations resulted in estimates of change over time, which converged on the model estimates, but with much less precision.

Conclusions: Electronic prescribing systems offer significant opportunities to better understand and report antimicrobial consumption in children. This approach to modelling administration data overcomes the limitations of using interval data and dispensary data. It provides substantially more detailed inferences on prescribing patterns and the potential impact of stewardship interventions.

Figure 1.

Daily consumption of all antimicrobials [DOT per 1000 patient-days (PD)]. Consumption in PICU from 2016 is illustrated by yellow triangles and non-PICU consumption by blue circles. The black line indicates fitted values from the AR-MA time-series model. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 2.

(a) Meropenem DOT per 1000 patient-days (PD); fitted AR-MA model. (b) Meropenem DOT per 1000 PD; fitted piecewise AR-MA model, breakpoint at start of expansion of AMS services at the end of 2016. (c) Meropenem DOT per 1000 PD against DDDs per 1000 PD, r = Pearson correlation coefficient with 95% CIs; linear regression line in black. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 3.

Percentage of total monthly consumption of antibacterial agents by ATC group. Black line indicates total monthly DOT per 1000 patient-days (PD). This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 4.

Monthly consumption of antibacterial agents by AWaRe England classification. The black line represents the trend (loess-fitted smoother) in monthly percentage of antimicrobials from the Access group. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 5.

Histogram of the estimated annual change in meropenem consumption from 10 000 simulations of annual PPS data of meropenem consumption. The red line indicates the estimate of the change provided by the continuous model with 95% CIs. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.