Provider Analytics career guide

• Academic medical center (AMC)

  Mayo, Hopkins, Cleveland Clinic, UCSF, MSK, NYU, Penn, MGB. Heavy research support, IRB-sensitive data, more R / Python, deeper statistics. Physicians in the interview loop are common. Best pay ceiling in the space.

• Community hospital

  Smaller teams of 2–10 people, broader responsibilities, more operations and finance, less statistics. Less gated than AMCs — a good entry door if you come from a clinical background with basic SQL.

• Multi-hospital IDN

  HCA, CommonSpirit, Ascension, Trinity. Standardization work, facility-comparison dashboards, timezone/ID-merge pain, governance-heavy. The highest-leverage work here is getting 15+ hospitals to report the same metric the same way.

• Physician group / MSO

  Ambulatory-centric, wRVU productivity, panel size, no-show rates. Less inpatient complexity, more compensation-model analysis.

• ACO (MSSP, ACO REACH)

  Population health, total-cost-of-care benchmarks vs CMS, shared-savings calculations. More claims-data work than a hospital analyst typically sees — a bridge role toward payer-side careers.

1. 1

   Recruiter screen (30 min)

   Lightweight fit check. Be ready to explain why healthcare and why provider-side (vs payer). AMCs will ask about IRB / research interest; community hospitals will ask about operations interest.

2. 2

   Hiring manager (45–60 min)

   Behavioral + domain depth. Expect a case-style walkthrough: "a physician says our CHF readmission rate is higher than what Epic shows — troubleshoot." They’re testing whether you decompose numerator/denominator, observation handling, planned-exclusion lists, and ETL refresh timestamps on the fly.

3. 3

   Technical SQL screen (60 min, live or take-home)

   Clarity-style schemas and window-function emphasis. Common asks: 30-day readmission with LAG, recursive department hierarchy, flowsheet EAV pivot. Edge cases matter — "what if a patient has two admissions on the same day?"

4. 4

   Epic-specific round

   Chronicles vs Clarity vs Caboodle, PAT_ENC vs PAT_ENC_HSP, flowsheet structure, ETL lag, SlicerDicer vs Reporting Workbench vs Radar. If you haven’t worked in Epic before, say so — faking it is always caught.

5. 5

   Clinician panel (AMCs especially)

   A physician or nurse manager in the room. They want to see you can translate clinical-speak into data-speak and back without losing the intent. Ask them the first question, don’t wait.

6. 6

   Behavioral + director/executive

   Communication, conflict resolution, governance temperament. At AMCs, expect questions about working with faculty PIs on research requests.

“Walk me through the difference between PAT_ENC, PAT_ENC_HSP, and HSP_ACCOUNT.”

PAT_ENC is every patient contact including outpatient, phone, tele, and inpatient daily rows. PAT_ENC_HSP is the inpatient overlay with HOSP_ADMSN_TIME and ADT_PAT_CLASS_C. HSP_ACCOUNT is the financial grouping — one hospital account can aggregate many PAT_ENC_HSP daily rows. A "true inpatient admission" is often reconstructed from CLARITY_ADT rather than HOSP_ADMSN_TIME alone because of observation-to-inpatient status conversions mid-stay.

“How would you report on flowsheet vital signs?”

Join PAT_ENC.INPATIENT_DATA_ID → IP_FLWSHT_REC.INPATIENT_DATA_ID → IP_FLWSHT_MEAS.FSD_ID. Filter by FLO_MEAS_ID for the vital you want. CAST MEAS_VALUE (always a string). Distinguish RECORDED_TIME (entry timestamp) from TAKEN_TIME (observation timestamp) — using RECORDED_TIME for a sepsis bundle clock gives wrong answers because nurses chart retroactively.

“Clarity refresh cadence — what does it mean for operational reporting?”

Clarity refreshes nightly (~24hr lag). For real-time operational reporting (e.g., current ED census, open OR cases) use Reporting Workbench against Chronicles. Caboodle adds another 12–24hr behind Clarity so it’s never the right place for operational decisions.

“A physician says our CHF readmission rate is higher than the Epic dashboard shows. How do you troubleshoot?”

Check five things in order: numerator / denominator definitions (index admission vs discharge-based cohort), observation status handling (is the 3-day stay in the denominator?), planned-readmission exclusions (CMS CCS list), cohort dx code set (ICD-10 principal dx list for CHF), and ETL refresh timestamps on both sources.

“How would you identify a CAUTI?”

Urinary catheter in place >2 calendar days via LDA flowsheet + positive urine culture via LOINC on the index or subsequent day + symptoms. The trap: people forget the >2-calendar-day window and end up counting catheters placed that same day.

“ORDER_MED vs MAR_ADMIN_INFO — when do you use each?”

ORDER_MED tells you what was ordered. MAR_ADMIN_INFO tells you what was actually administered. Reconciled-at-discharge meds are a third list. Use the one that matches the question: quality measure compliance usually wants administered, not ordered.

• Epic Cogito Fundamentals (COG170)

  Gold standard

  The entry-point Epic cert. Gated — only available through employer sponsorship at an Epic customer, consulting partner, or Epic itself in Verona.

• Epic Clarity Data Model Fundamentals (CLR110) + one of 220/230/240

  Gold standard

  Clinical (CLR230) is the broadest — covers encounters, orders, results. Access (CLR220) is revenue-cycle flavored. Revenue (CLR240) is deeper RCM. Pick to match the role.

• Epic Caboodle Data Model Fundamentals (CDW110)

  Gold standard

  Increasingly a separate hiring filter — Caboodle is replacing hand-rolled Clarity data marts at most Epic shops.

• CPHIMS (HIMSS Certified Professional in Healthcare Information & Management Systems)

  Nice to have

  Signal for management track. Useful once you’re targeting director-and-above.

• Tableau Desktop Specialist / Power BI PL-300

  Nice to have

  A portfolio dashboard demonstrates the same thing more credibly. Cheap to do — only worth it if the target employer lists it as a requirement.

• Microsoft Certified: Azure Data Fundamentals (DP-900)

  Nice to have

  Relevant if the target is a health system using Azure Synapse (many are). Low cost, short study.

• SAS Base

  Skip

  Provider-side rarely uses SAS. It matters in pharma clinical and some public-health departments, not here.

• Thinking "Epic" is one tool. Epic has 50+ modules — Ambulatory ≠ Inpatient ≠ HB ≠ PB ≠ OpTime ≠ Willow ≠ Beaker ≠ Cogito — and they don’t share all their tables.
• Confusing encounter / visit / admission. Every provider interview tests this.
• Not understanding Clarity’s ~24-hour ETL lag. You cannot run an operational ED-census report off Clarity; you need Reporting Workbench against Chronicles.
• Trusting HOSP_ADMSN_TIME blindly. Observation-to-inpatient conversions mid-stay move the start time.
• Ignoring POA (Present-On-Admission) flags. A postoperative infection coded without POA is a hospital-acquired condition; with POA it’s a comorbidity. The two get paid differently.
• Averaging flowsheet values without dedup. Multiple entries per vital sign per hour is normal, and a naive AVG over-counts nurse-chart-artifact entries.
• Using ORDER_MED when the question wanted MAR_ADMIN_INFO — ordered ≠ administered.
• Over-indexing on Python / ML early. Provider-side wants rock-solid SQL + stakeholder skills + clinical fluency first; Python is the Year-3 lever.
• Applying only to the top four health systems. Community hospitals and regional IDNs are less gated and pay comparably after Epic sponsorship.

The caseSQL Provider Analytics path can’t teach you Epic Clarity specifically — the schema is a hospital-flavored star schema, not a Chronicles/Clarity replica. What it does teach is the reasoning: encounter grain, readmission cohort construction, department hierarchy rollup, LOS tie-outs, ED throughput, quality-measure numerator/denominator discipline. Those are the transferable skills Epic sponsorship then lets you apply against the real Clarity tables. If you’re preparing for a provider-side interview loop, the Hard and Expert missions are the closest approximation of the questions you’ll be asked on the whiteboard.