Providing Antibiotic Prescribing Feedback to Primary Care Physicians: The Ontario Program To Improve AntiMicrobial USE (OPTIMISE)
|The safety and scientific validity of this study is the responsibility of the study sponsor and investigators. Listing a study does not mean it has been evaluated by the U.S. Federal Government. Read our disclaimer for details.|
|ClinicalTrials.gov Identifier: NCT03776383|
Recruitment Status : Active, not recruiting
First Posted : December 14, 2018
Last Update Posted : April 2, 2019
|Condition or disease||Intervention/treatment||Phase|
|Antibiotic Prescribing Audit and Feedback||Other: Antibiotic use feedback letter||Not Applicable|
Hide Detailed Description
Background: Antibiotic overuse is the major driver of rising antimicrobial resistance rates. Unnecessary antibiotic use is associated with adverse medication side effects and rising rates of resistance. Mailing letters to high antibiotic prescribers in other countries have shown variables effects on reducing antibiotic prescribing.1-3
Objective: To evaluate the impact of a letter to the 25% of primary care physicians who prescribe the most antibiotics in Ontario on increasing enrollment for HQO primary care practice reports.
Secondary objectives: To test the impact of different formats of a notification letter on antibiotic prescribing. Our primary hypothesis is that notifying physicians that antibiotic use is being monitored will decrease use of antibiotics. The investigators further hypothesize that targeted change ideas on antibiotic initiation or duration will have varying impacts on these different prescribing behaviours highlighting the importance of targeted messaging.
Methods: The effect of letter mail outs to the top 25% prescribing primary care physicians will be evaluated using a randomized controlled trial (RCT) design. The investigators will allocate physicians to intervention versus control arms in a 3:3:1 ratio to maximize the number of physicians exposed to the intervention. The investigators are using the IQVIA Xponent database from March 2017 to February 2018 to identify the top 25% of primary care physician antibiotic prescribers in Ontario (family physicians or general practitioners) defined by total numbers of antibiotic prescriptions. They will be mailed a letter to indicate that they prescribe more antibiotics than 75% of Ontario primary care physicians, based on total numbers of antibiotic prescriptions per year. The letter will encourage them to sign up for HQO My Practice primary care reports which in the future may include antibiotic indicators.
The letter was designed using proven theoretical frameworks of behaviour change to improve antibiotic prescribing by physicians. Most physicians are aware of the global public health threat of antimicrobial resistance, however many underestimate its importance locally. Furthermore, unnecessary antibiotic prescribing is frequently attributed to external factors out of their control, such as patient expectations and patient adherence to medication.4 For this intervention the investigators focused on two behaviour change theories previously demonstrated to be associated with prescribing behaviour; Theory of Planned Behaviour (TBP) and Operant Learning Theory (OLT).5 The letter will provide some social normative comparison by indicating to these physicians that they prescribe more antibiotics than their peers (TPB). By providing change ideas and information on the consequences of unnecessary antibiotic use, the investigators are attempting to change the physicians' risk perception (TPB) and anticipated consequences (OLT) of prescribing antibiotics for acute respiratory tract infections. This letter represents a form of persuasive communication behaviour change technique (BCT). This form of BCT has shown to be effective in modifying OLT driven behaviours.6, 7
To test different change ideas, the investigators will simultaneously evaluate two types of letters to be randomly distributed to the physicians in the top 25% of prescribers allocated in 3:3:1 ratios; a) Providing change ideas from Choosing Wisely Canada on appropriate initiation of antibiotics for acute respiratory conditions; b) Providing change ideas on appropriate duration of antibiotics.
Designs of the letters went through user testing and were iteratively refined by a sample of Ontario primary care physicians. The letters will be co-branded with the Ontario Medical Association (OMA) Section of General and Family Practice and Choosing Wisely Ontario. The letters were also reviewed by the Ontario College of Family Physicians, HQO and members of the study team to provide input and edits to the language and content. Based on the trial results the investigators will send a letter to all 3500 physicians the following year (12 months later).
Sample Size: Sample size calculations are based on Antibiotic outcome 1. A total of 3500 primary care physicians will be included in the randomized controlled trial: 3000 will receive a letter and 500 will not receive a letter. In an analysis of covariance (ANCOVA), sample sizes of 500, 1500, and 1500 physicians achieve >80% power to detect a 1% absolute difference in the primary antibiotic outcome between the average of the three intervention arms versus control using an F-Test at the 5% level of significance. The calculations assumed a standard deviation of 12 calculated using routinely collected data available within March 1 2017 and February 28 2018, and accounted for the baseline measure of the primary outcome assuming a correlation of 0.8.
The primary outcome will be an intention to treat analysis of the RCT at 12 months for both outcomes 1 and 2. The unit of analysis will be the physician. The primary analysis will use the modified robust Poisson regression approach of Zou and Donner.(Zou and Donner 2013) The dependent variable will be the numerator for each outcome, with the log of the denominator specified as an offset term. The distribution will be Poisson and the link function will be the log-link. The model will include a fixed term for time (pre/post), and the interaction between time and group (intervention). Note that, in order to constrain the differences between the arms at baseline to 0, the model will omit the main effect for the intervention. Pairwise least square mean differences will be obtained from the model (each intervention arm versus control) and expressed as Relative Risk (RR) and 97.5% confidence intervals. Statistical significance will be assessed at the Bonferroni-corrected level of 0.05/3. Robust standard errors will be used, with the correlation structure specified as Exchangeable. Secondary pairwise comparisons will also be obtained from the model (e.g., the average of the two letter versus control, and direct comparison between the two letters).
The investigators will use the same approach for the quarterly data, except that the least square mean pairwise differences will be obtained at the first, second, third and final quarters after randomization, to examine short and long-term effects of the intervention. The investigators will evaluate subgroups including baseline physician prescribing (>90th %ile vs 75-90th %ile), years in medical practice (<10y, 11-24y, vs 25y), physician gender (M vs F), rural vs urban practice, and patient age/sex groupings (<18yM, <18yF, 18-64yM, 18-64yF, 65+M, and 65+F) using IQVIA data.
A secondary outcome will be new physician enrollment into HQO primary care practice reports. Working with HQO, the investigators will track physicians in our study that sign up in response to receiving the letter from Public Health Ontario (PHO) by February 28, 2019. Physicians will be invited to sign up via a special weblink created for this study. The primary analysis will calculate the absolute difference between the proportions of new sign ups in the two intervention arms versus the control arm together with 95% confidence interval. Multiplicity-adjusted pairwise comparisons will also be made between each of the intervention arms versus the control and between the different versions of the letters. The investigators will evaluate predictor variables of new enrollment including baseline physician prescribing (>90th %ile vs 75-90th %ile), years in medical practice (<10y, 11-24y, vs 25y), physician gender (M vs F), rural vs urban practice, and patient age/sex groupings (<18yM, <18yF, 18-64yM, 18-64yF, 65+M, and 65+F) using logistic regression analysis.
To evaluate the overall effect of receiving any letter, compared to no letter, on prescribing outcomes in Ontario, The investigators will use a regression discontinuity design (RDD) by including antibiotic prescribing data from the rest of Ontario's primary care physicians. In the traditional RDD, subjects are assigned to treatment or control based on a cut-off value of a continuous assignment variable. The investigators will use the 3000 primary care physicians who received letters (excluding the 500 physicians in the control arm) as the treatment group and compare them to the remainder of Ontario physicians as controls. The assignment variable in our design is the antibiotic prescription rate (defined as antibiotics per 100 total medications over the year before the intervention).The investigators will present effects graphically by plotting the continuous assignment variable on the x-axis and the continuous outcome variable (antibiotic prescription rate over the year after the intervention) on the y-axis. Different regression lines are then fit to the data on each site of the cut-off. If the intervention is effective, there will be a discontinuity in the regression lines at the value of the cut-off. The intervention effect is expressed as the vertical displacement at the point of discontinuity.8 The investigators will also conduct this analysis for antibiotic duration.
Limitations: The investigators can only monitor antibiotics prescribed per 100 total medications with this dataset, instead of per patient volume. This measure appears to be a reasonable antibiotic use measure within a physician as it is stable over time in high prescribing physicians in our datasets.
Ethical considerations: Antimicrobial stewardship and combating antimicrobial resistance is a provincial and national priority. The Xponent database with prescribing data at an individual prescriber level is already held by PHO and is kept in a secure location. The agreement with IQVIA allows PHO to provide direct feedback to physicians. The letters will be addressed to the physician and marked confidential. The only data within them will be that they have been identified as a high antibiotic prescriber. No specific prescription counts or any patient-level data are being provided.
The investigators have designed this study to maximize HQO sign up and potential impact through a 6:1 allocation. Furthermore, the investigators will send a debrief and follow-up letter to all physicians the following year so that all high prescribing physicians will have the opportunity to receive this information and sign up for HQO reports. Physicians will be given an email address to contact PHO with any questions or concerns regarding the letter. All results will be aggregated and reports will not divulge any prescriber details.
A waiver of consent for this study was provided. This intervention involves minimal risk and burden to participants (letters are sent to physicians from government agencies routinely; these letters do not require an immediate response). There is no anticipated impact on the welfare of physicians by receiving this letter. The intervention would be impractical if consent were required. A consent process even if feasible, would likely create selection bias and make it impossible to answer the research question posed. In addition, the consent process would include a greater burden to physicians than this single letter intervention itself. This quality improvement initiate has the potential to benefit physicians by providing helpful recommendations on optimal antibiotic prescribing practices. A PHO contact email will be provided on the letter to provide physicians the opportunity to contact PHO and opt-out of future letters.
|Study Type :||Interventional (Clinical Trial)|
|Actual Enrollment :||3500 participants|
|Intervention Model:||Parallel Assignment|
|Intervention Model Description:||3:3:1 for intervention 1: intervention 2: control|
|Masking:||Single (Outcomes Assessor)|
|Masking Description:||The data for the outcome are being provided from a third party with no involvement in the study|
|Primary Purpose:||Health Services Research|
|Official Title:||Ontario Healthcare Implementation Laboratory: A Learning Health System Approach That Leverages Data to Improve Quality in Primary Care: Stage 1|
|Actual Study Start Date :||December 1, 2018|
|Estimated Primary Completion Date :||November 30, 2019|
|Estimated Study Completion Date :||November 30, 2020|
Active Comparator: Antibiotic use feedback letter 1
Antibiotic use feedback letter 1 provides physicians with information on their antibiotic use plus change ideas on appropriate antibiotic prescribing for acute respiratory conditions
Other: Antibiotic use feedback letter
Mailed letters indicating that the physician prescribes more antibiotics than 75% of their peers
Active Comparator: Antibiotic use feedback letter 2
Antibiotic use feedback letter 2 provides physicians with information on their antibiotic use plus change ideas on appropriate antibiotic durations for common infections
Other: Antibiotic use feedback letter
Mailed letters indicating that the physician prescribes more antibiotics than 75% of their peers
No Intervention: Control
Controls will not receive a letter
- Antibiotic prescribing rate [ Time Frame: 12 months ]Number of antibiotics prescribed per 100 total prescriptions
- Prolonged duration prescribing rate [ Time Frame: 12 months ]Proportion of antibiotic prescriptions that are >7 days per 100 total prescriptions
- New enrollment for Health Quality Ontario primary care practice reports [ Time Frame: 24 months ]Proportion of new sign ups for Health Quality Ontario primary care practice reports
- New enrollment for Health Quality Ontario primary care practice reports [ Time Frame: 6 months ]Proportion of new sign ups for Health Quality Ontario primary care practice reports
- New enrollment for Health Quality Ontario primary care practice reports [ Time Frame: 12 months ]Proportion of new sign ups for Health Quality Ontario primary care practice reports
- Antibiotic prescribing rate [ Time Frame: 3 months ]Number of antibiotics prescribed per 100 total prescriptions
- Antibiotic prescribing rate [ Time Frame: 6 months ]Number of antibiotics prescribed per 100 total prescriptions
- Antibiotic prescribing rate [ Time Frame: 9 months ]Number of antibiotics prescribed per 100 total prescriptions
- Antibiotic prescribing rate [ Time Frame: 24 months ]Number of antibiotics prescribed per 100 total prescriptions
- Prolonged duration prescribing [ Time Frame: 3 months ]Proportion of antibiotic prescriptions that are >7 days per 100 total prescriptions
- Prolonged duration prescribing [ Time Frame: 6 months ]Proportion of antibiotic prescriptions that are >7 days per 100 total prescriptions
- Prolonged duration prescribing [ Time Frame: 9 months ]Proportion of antibiotic prescriptions that are >7 days per 100 total prescriptions
- Prolonged duration prescribing [ Time Frame: 24 months ]Proportion of antibiotic prescriptions that are >7 days per 100 total prescriptions
To learn more about this study, you or your doctor may contact the study research staff using the contact information provided by the sponsor.
Please refer to this study by its ClinicalTrials.gov identifier (NCT number): NCT03776383
|Public Health Ontario|
|Toronto, Ontario, Canada, M5G 1V2|
|Principal Investigator:||Kevin Schwartz, MD MSc||Ontario Agency for Health Protection and Promotion|