Community-led educational interventions to reduce non-prescription antibiotic sales in retail pharmacies: a systematic review.
Skills / interests: Data analysis and organisation, Data extraction
Methodological skills / interests: Artificial intelligence, Comparing multiple interventions (network meta-analysis and overviews), Individual participant data meta-analysis, GRADEing
The detailed description of the task are as follows:
Phase 1: The Protocol (Months 1–6)
Before we look at a single trial's results, we must publish our "rules of engagement." This prevents us from changing our mind later to make the data look better.
-
Defining PICO: We must specify exactly who counts (e.g., only adults? only children?), what the intervention is (e.g., specifically a 10-minute blood test), and what success looks like (e.g., total prescriptions written vs. actual clinical recovery).
-
Search Strategy: We will work with a Cochrane Information Specialist to write a "Search String"—a massive, complex list of keywords used to scan databases like PubMed, Embase, and CENTRAL.
-
Peer Review: The Protocol itself will be peer-reviewed and published.
Phase 2: The Search & Screen (Months 6–9)
Once the protocol is live, you hit "Go" on the search.
-
The "Haystack": Our search might return 8,000 to 12,000 abstracts.
-
Sifting: Two authors must independently read every single title and abstract to decide if it fits your PICO. If anyone disagree, a third author breaks the tie.
-
Full-Text Review: We’ll likely whittle those thousands down to 50–100 papers that you must read in full to ensure they meet your strict quality criteria.
Phase 3: Data Extraction & "Risk of Bias" (Months 9–15)
This is the most labor-intensive part of the task.
-
Extraction: We will create a massive spreadsheet (or use RevMan) to pull data from every study: How many patients? What dose? How many got antibiotics? What were the side effects?
-
Risk of Bias (RoB 2): We must judge the quality of every study. Did the doctors know which patient got the test? Was the randomization truly random? If a study is "low quality," its data will carry less weight in your final result.
Phase 4: The Meta-Analysis (Months 15–18)
This is where the math happens.
-
Pooling Data: We have to use a statistical method to combine the results of all the small studies into one giant "Forest Plot."
-
Heterogeneity: we have to calculate I^2 to see if the studies are too different to be compared. If I^2 > 75%, you have a problem—the studies are telling very different stories.
-
GRADE Assessment: To provide a "Certainty of Evidence" (High, Moderate, Low, or Very Low) to each outcome.
Phase 5: Writing & Final Peer Review (Months 18–24)
-
Plain Language Summary (PLS): We will write a version of our findings that a regular patient or a journalist can understand.
-
The "Editorial Gauntlet": The Cochrane Group will tear the draft apart, checking every calculation and every citation. It may take 2 or 3 rounds of heavy revisions.
The Workload Reality
-
Team Size: Usually 8 to 10 people.
-
Software: RevMan Web, COVIDENCE, GRADE pro
-
Evidence-Based Medicine: We will not just say that "antibiotic misuse is bad." We will find out the process "How to reduces antibiotic initiation by 22% (95% CI 18% to 26%) with high-certainty evidence.”
Cochrane
I would be eager to know the status of my inclusion in the team