TL;DR
This paper describes a nationwide pragmatic randomized controlled trial in Switzerland designed to evaluate the effectiveness of personalized, feedback-based interventions for reducing antibiotic prescribing in primary care. By leveraging routinely collected health insurance claims data, the study aims to provide general practitioners (GPs) with insights into their prescribing habits compared to their peers, with the ultimate goal of curbing antibiotic overuse and combating antimicrobial resistance.
What Problem Does This Paper Address?
Antimicrobial resistance (AMR) is a significant global health threat, exacerbated by the overuse of antibiotics. Primary care is a major contributor to antibiotic consumption, particularly for common infections like respiratory and urinary tract issues. While various interventions exist to curb this trend, many are resource-intensive or have shown inconsistent results. This study addresses the need for scalable, low-cost interventions that can be implemented at a system level. Specifically, it builds upon previous pilot work to design a larger, nationwide trial using a feedback mechanism for GPs, aiming to provide them with personalized data on their antibiotic prescribing patterns to encourage more judicious use.
Methods
The study employs a pragmatic, nationwide randomized controlled trial (RCT) design. General practitioners (GPs) in Switzerland who meet specific criteria (board-certified, top 75% antibiotic prescribers, at least 100 patient contacts per year) were randomized in a 1:1 ratio to either an intervention group or a control group.
Intervention Group: GPs received an initial mailing in December 2017 containing evidence-based treatment guidelines, region-specific antibiotic resistance information, and personalized feedback on their antibiotic prescribing. This feedback was delivered quarterly for two years, providing comparisons to peer-physicians and detailed breakdowns of prescribing patterns. Information was also made available on a password-protected website.
Control Group: GPs in the control group were not informed about the study and received no feedback or guidelines.
The core of the intervention relies on personalized antibiotic prescription feedback. This feedback includes:
- The GP's yearly prescription rate (antibiotic prescriptions per 100 consultations) compared to peers.
- Monthly prescription rates for the preceding year, compared to peers.
- Yearly prescription rates stratified by antibiotic type, compared to peers.
- A call to action with tailored advice.
This feedback was generated using pseudonymized, patient-level data from the claims databases of the three largest Swiss health insurers. The study design emphasizes a "Zelen's design" approach for the control group, aiming to reflect actual practice without intervention.
Datasets
The study utilizes pseudonymized, patient-level data from the claims databases of the three largest Swiss health insurers: Sanitas, CSS, and Helsana. This data includes:
- Physician identifiers and demographics.
- Patient identifiers and demographics.
- Consultation data, including the primary outcome data of antibiotic prescribing.
- Claims consultation data with relevant diagnostic test ordering and prescribing of all medications.
- Hospitalization data.
The data is collected through a standardized process, and a six-month delay in processing by health insurers is accounted for in the study timeline. No new benchmarks are introduced; rather, the study leverages existing, routinely collected data.
Key Insights and Ideas
- Leveraging Routinely Collected Data: The study's strength lies in its reliance on comprehensive, real-world claims data. This allows for a nationwide scope and continuous follow-up without the burden of manual data collection, enhancing external validity.
- Personalized Feedback as a Scalable Intervention: The core idea is that providing GPs with personalized, comparative data on their antibiotic prescribing habits can be an effective and low-cost method to influence behavior, addressing the challenge of implementing system-wide changes.
- Focus on Long-Term Effects: By extending the intervention and feedback period to two years, the study aims to assess not just immediate impacts but also the sustained effects of the feedback mechanism on antibiotic prescribing.
- Monitoring Unintended Consequences: The use of patient-level data allows for the investigation of potential negative outcomes, such as increased hospitalizations or reconsultations, which could arise from reduced antibiotic use.
Key Results
The paper describes the design and rationale of the trial, not its results. The primary outcome is the overall antibiotic use (prescribed antibiotics per 100 patient consultations) in the second year of intervention. Secondary outcomes include antibiotic use in the first year and over the entire trial, use of broad-spectrum antibiotics, hospitalization rates (all-cause and infection-related), and antibiotic use in different age groups. The power analysis indicated a median power of 0.93 (95% CI: 0.90 to 0.96) to detect a clinically relevant effect of a 5% reduction in prescription rates.
Limitations and Future Work
Limitations: * Data Latency: A six-month delay in claims data processing means that feedback is always based on past data, and the most recent prescribing patterns may not be immediately reflected. * Potential for Non-Responders: The power analysis accounted for an expected 15% proportion of non-responders to the feedback. * Control Group Blinding: While the control group is fully blinded to the intervention, the paper notes that the "Zelen's design" allows for a comparator that reflects actual practice.
Future Work: * If the feedback intervention proves efficacious, it could be implemented system-wide, allowing for continuous monitoring of antibiotic prescribing at low costs. * The detailed patient-level data will enable secondary analyses of physician practice patterns, antibiotic prescribing, and health economic analyses, including the correlation of prescribing with diagnostic test ordering and specialist referrals. * The study is designed to investigate the impact of the intervention on potential unintended consequences like hospitalization and reconsultations.
Summary updated