- S-201: J Morrison, 2024-08-30, Site A
- S-208: Jane Morrison, 2024-08-30, Site A
Resolve Duplicate Identities
- S-201: Jane Morrison
Megan Kern
I live in Chattanooga, TN, and contract with research labs and small companies. I'll work with you to understand what tools should exist for more of your time to be yours. Data integration steps are built to always trace each number back to its original source, making your workflows observable. I start with a no-obligation discovery call to understand your current data workflow and share solutions. If you'd like me to handle implementation, I'll follow up with a fixed-price contract and statement of work covering a turnkey solution documented for handoff.
Clean, repeatable, analysis-ready data verified at every step.
Have an automated cleaning process that counts every step, flags exceptions instead of silently dropping them, and traces every output back to its source. To build it, we start with examples of raw data and cleaned results from your current process. From those examples, I extract the rules for which records are duplicates, which fields get normalized, what happens when two sources disagree, and where manual review is still needed. Those rules run as an automated process with an audit trail. The trail gives you counts at each step and lets you test different cutoffs. If you don't have a before-and-after yet, we can start from the raw formats (exports, spreadsheets, scanned forms, API responses) and work from the target data structure. This process scales to datasets too large for Excel or SPSS, with privacy and access controls, harmonization across sites, and narrowly scoped uses of AI where they make sense. De-identification tools can also be built without seeing your originals, so you run them on your side before I receive any test samples.- field_notes: "applied 2.5kg urea per hectare, plots 3–7, Aug 12"
Extract From Free Text
- treatment: urea
- dose: 2.5 kg/ha
- plots: 3, 4, 5, 6, 7
- date: Aug 12
- site_db: subj 4401, 2024-06-12, hgb 11.2
- central_lab: #4401, 12-Jun-24, hgb 10.8, venous draw
Reconcile Across Sources
- 4401, 2024-06-12: hgb 10.8, method: venous draw
- site_A: 5 traps, 3 days, 12 specimens
- site_B: 250 traps, 1 day, 89 specimens
- site_C: 40 traps, 14 days, 280 specimens
Normalize By Collection Effort
- site_A: 0.80 / trap-day
- site_B: 0.36 / trap-day
- site_C: 0.50 / trap-day
- 03-15: ED admit, site A, dx: heart failure
- 03-16: inpatient, site A
- 03-29: outpatient, site B, follow-up
- 04-14: ED admit, site A, dx: heart failure
Construct Care Episodes
- ep_1: 03-15 to 03-29, visits: 3, dx: heart failure
- ep_2: 04-14, dx: heart failure, 30d readmission
Bring all your data into one dashboard built around the entities that matter to you.
Records from every source end up merged into one dataset that is ready for analysis, with every match decision logged so you can audit its lineage at any time. To build it, we first identify the real entity or event you are reconstructing and which pieces each source contributes. We then review the fields you currently match on, how often those matches succeed, whether a wrong match or a missed match is costlier, and how long manual assembly takes. Those rules are built and versioned. Sources with different formats, granularities, and update cadences are standardized as they come in. Any identifying information stays behind user access controls.
Survey data across all forms.
Lab and field data from different instruments.
Institutional records behind database queries.
Government portal downloads.
Collaborator files and shared drives.
One Unified Dataset
- Standardized dates, units, codes.
- Matched records across sources.
- Enriched data with AI tagging.
- Auto updates for every source.
Cross-source analysis.
Discrepancies and coverage gaps.
Export any data on demand.
Stay ahead of deadlines with a live dashboard built on your protocol and connected to your tools.
A live tracker shows you what needs attention today: overdue visits, incomplete assessments, and enrollment against target. To build it, we first map each stage a participant, specimen, or case passes through, including what happens at each step, what must come first, and how long it should take. The tracker then connects to tools you already use, such as REDCap, email, and institutional databases. It sends event and time based reminders. It routes exceptions to the right coordinator. Progress is measured against enrollment goals, visit windows, or whatever your protocol defines.Turn high volume decisions into fast, consistent, fully auditable calls, built around your own real examples.
A custom coding interface keeps your team's judgment consistent across high volumes of repetitive decisions. Each screen shows only the context needed for one decision, so reviewers move faster and onboard in less time. Every choice is logged and reference cases are one click away, which makes decisions easy to audit and stops drift over time. To scope it, we start with a handful of your real example decisions: what data you look at and how you reason about it before deciding.