Why this matters
No two sites operate the same way. Without routine monitoring, small differences in how sites screen, consent, and enter data can balloon into missed enrollment targets or delayed database locks. By watching a small set of practical indicators—conversion from screened to enrolled, timeliness of data entry, backlog, and attrition—you can coach sites early, keep regulatory‑quality data flowing, and size your team to handle predictable surges.
Core concepts & definitions
Plain‑language definitions first, with the usual abbreviations introduced once.
- Electronic Data Capture (EDC): The study database where sites enter case report form (CRF) information.
- Screen→Enroll Conversion (Screening Conversion): The share of screened people who end up enrolled, counted the same way at every site: screened = consented and started screening; enrolled = entered the study (the protocol sets the point, often randomization). How to measure: count enrolled participants at the site and divide by the number screened for that same period, leaving out anyone still in screening. Track the complement, the screen‑failure rate, with the reasons (which criterion).
- Screening log and screen‑failure data: The site lists everyone it screens on a screening log. Consent is signed before any study procedure; checking existing records first is pre‑screening, not screening. For screen failures the sponsor collects a minimal data set (typically demographics, consent date, reason for failure, any serious adverse event) and no direct identifiers. Recording a serious adverse event there does not report it: the site still reports it to pharmacovigilance directly.
- Timeliness / Currency: How “current” the data are. How to measure: percent of expected forms that are already in the EDC (the expected‑form percent), plus the typical lag from visit date to data entry date, and the share of queries that are older than a set threshold (for example, 14 days).
- Attrition: Participants who withdraw from the study early or are lost to follow‑up. How to measure: number who discontinue or are lost out of those enrolled, tracked by site and time on study. Stopping study treatment is not attrition while follow‑up continues. Early warning: the share of a site’s participants with no data for two consecutive visit windows (a single late visit is only a timeliness issue).
- Backlog: Work that is due but not yet complete. How to measure: open queries per participant, and expected‑but‑missing forms by site and visit window.
- Key risk indicators (KRIs): The industry term for site metrics like these, with thresholds set in advance, reviewed in centralized monitoring (part of the risk‑based approach of ICH E6(R2) and E6(R3)). A quality tolerance limit (QTL) is the trial‑level counterpart: a preset limit whose breach triggers a documented evaluation (E6(R3) calls these acceptable ranges).
- Throughput / Full‑Time Equivalent (FTE): Converting routine work into staffing needs. How to measure: volume (items each time, e.g., files per transfer) × frequency (times per week) × time on task (minutes per item), then divide the weekly hours by a standard workweek (e.g., 40 hours) to estimate FTE, plus headroom: nobody is on task 100% of the time.
Tip: Set study‑specific thresholds early (for example, “query >14 days” or “entry lag >5 days”). Calibrate to the protocol and vendor schedules.
The process, step‑by‑step
- Define your key performance indicators (KPIs). Include: screening conversion, expected‑form percent, lag from visit to entry, query aging, and attrition.
- Segment by site. Compare each site to the network median (across all sites in the study) and to its own history (trends). Show counts beside percentages and require a minimum count before flagging a rate (1 of 2 screened is “50%”).
- Investigate outliers. Look for root causes: eligibility interpretation, staffing, training gaps, device/data transfer issues, or local workflows.
- Intervene. Offer targeted coaching, micro‑trainings, extra data management (DM) support, or vendor escalation where needed. DM supplies the metrics and raises queries; the monitor (clinical research associate, CRA) follows up with the site; the site enters and corrects its own data (DM never does); study leadership decides site‑level escalation.
- Model capacity. Turn operational load into hours and FTE so you can plan for surges (late site activations, batch vendor uploads, seasonal staffing).
- Close the loop. Re‑measure after interventions, adjust targets, and acknowledge improvements with sites.
What good looks like
A per‑site dashboard showing:
- Screening conversion
- Percent of expected forms present
- Lag from visit to EDC entry
- Query aging (share older than 14 days)
- Attrition (study withdrawals or loss to follow‑up)
Clear attention flags (red flags), for example:
- Screening conversion below 40% for two consecutive months
- Expected‑form percent below 90% for the current window
- More than 20% of queries older than 14 days
- Median entry lag above 5 days
A capacity view that projects workload vs. available FTE by month and highlights upcoming surge windows.
Good practice: Pair every red flag with a standard response (who will contact the site, what to review, and by when).
Worked examples
Self‑contained scenarios you can adapt.
Example 1 — Screening conversion and coaching
Context: Site A reports 40 screened participants in the last month: 10 enrolled and 30 failed screening.
What this shows: Conversion is 10 out of 40 (25%). The network median is 55%.
Likely causes to check: eligibility pre‑screening steps, inclusion/exclusion interpretation, timing of key labs, and coordinator turnover.
Action plan:
- The CRA walks through two recent screen failures with the site.
- Add a one‑page pre‑screen checklist to the site’s reference materials.
- Schedule a 30‑minute refresher on the top three exclusion criteria.
Follow‑up: Next month, Site A reports 44 screened participants with an outcome: 20 enrolled and 24 failed screening. Conversion improves to 45%. Keep the checklist; reduce monitoring intensity if sustained.
Example 2 — Timeliness and expected‑form percent
Context: Over the last 30 days, Site B should have 120 forms. Only 90 are present in the EDC; 30 are missing. Median entry lag is 7 days.
What this shows: Expected‑form percent is 90 out of 120 (75%). Risk of missing data at interim analysis.
Likely causes to check: missed visits (an attrition signal, not late entry), coordinator workload or turnover, new staff still awaiting training and their own EDC access.
Action plan:
- The CRA agrees a twice‑weekly data entry block on the site calendar.
- DM enables visit‑window reminders in the EDC (through change control if the study build changes).
- DM sends a short “how to find missing forms” guide.
Follow‑up: Two weeks later, expected‑form percent rises to 95% and median entry lag drops from 7 days to 3 days.
Example 3 — Query aging and backlog focus
Context: Site C has 220 open queries; 80 are older than 14 days.
What this shows: About 36% are past the threshold. This will delay interim and final lock.
Likely causes to check: site staffing; queries answered but not yet closed by DM; a few edit checks or unclear wording causing most queries (DM’s to fix).
Action plan:
- DM sorts the queries by impact (primary endpoints, safety, and critical protocol fields first).
- DM bundles similar queries (for example, missing vitals) so the CRA can cover them with the site in a single coaching call.
- The CRA agrees a daily target with the site (for instance, 25 query responses per day for one week).
Follow‑up: After five business days, the site has answered 125 queries, DM has reviewed and closed them, and 15 of the 95 still open are older than 14 days (about 16%, under the 20% flag).
Example 4 — Capacity modeling for vendor batch files
Context: A device vendor begins 12 weeks of weekly batch transfers: 80 files each week; average review and reconciliation time is 30 minutes per file.
Workload: 80 files × 1 transfer per week × 30 minutes = 2,400 minutes each week (40 hours).
Staffing implication: This is about 1.0 FTE of task time (one full‑time person, before headroom) for the 12 weeks of transfers.
Action plan:
- Assign a dedicated reviewer for the 12‑week surge window.
- Pre‑build a reconciliation checklist.
- Track actual minutes on a sample of 20 files to confirm the estimate in week 1.
Follow‑up: If the time per file falls to 20 minutes with the checklist, weekly effort drops to about 27 hours; adjust staffing accordingly.
Pitfalls & misconceptions
- Looking only at enrollment totals. Totals hide screen failures and growing backlogs that will hurt data quality later.
- Blaming sites for everything. Vendor schedules, slow lab feeds, and unclear protocol notes can create backlogs that look like “site problems.”
- Planning to the average. Peaks—not averages—drive real staffing needs. Plan for surge windows and set aside capacity.
- Ignoring payment triggers. If sites are paid on milestones, show data that map to those triggers (for example, visit completions and current forms) to keep payments smooth.
Practice in the wild (checklist)
- Build a per‑site KPI dashboard (conversion, expected‑form percent, entry lag, query aging, attrition).
- Define thresholds and a short playbook for coaching and escalation.
- Create a simple FTE calculator (volume × frequency × minutes per item) and validate it with real timing samples.
- Review surge scenarios quarterly (late site additions, device/lab batch windows).
- Align dashboard views to payment milestones so site finance teams can reconcile quickly.
Key points to remember
- Track the few leading indicators that predict delays—screen→enroll conversion, expected‑form percent, entry lag, query aging, and attrition—by site and over time.
- Conversion = enrolled ÷ screened (for the same period). When low, check eligibility interpretation, timing of key labs, and staffing; use a simple pre‑screen checklist and short refreshers.
- Set timeliness thresholds up front (e.g., ≥90% expected forms current, median entry lag ≤5 days, ≤20% queries older than 14 days). Review against them on a routine cadence.
- Use backlog views (expected‑but‑missing forms, open queries) to focus work; close high‑impact items first (safety, endpoints, critical protocol fields).
- Monitor attrition (early withdrawal or loss to follow‑up, LTFU) by site and time on study, and flag two consecutive visit windows without data; coach early to protect endpoint completeness.
- Convert workload to staffing: volume × frequency × minutes per item → hours/week → FTE. Plan to peaks, not averages (late activations, device/lab batch windows), and validate estimates with real timing samples.
- Segment by site and compare to the network median and each site’s trend; investigate outliers and escalate vendor/system causes when root issues are upstream of the site.
- Pair every red flag with a standard response (owner, action, due date); the site corrects its own data, never DM. Re‑measure after interventions and adjust targets; recognize improvements.
- Align dashboards to payment milestones so coordinators and finance see the same “current/counted” work, reducing friction.
- Document thresholds, decisions, and coaching outcomes to support interim and final lock readiness.
Practice prompt: Given last month’s site metrics (screened, enrolled, expected forms, forms present, open queries with age buckets), identify two red flags and write a 3‑step coaching plan.