Healthcare data analyst role: hospitals, insurers and careful metrics

Three kinds of employer, three kinds of data

Diagnosis and procedure data are usually coded, for example with ICD-10 diagnosis codes, and billing uses package or procedure codes that vary by hospital and payer. Learning which code field is reliable in your organisation's system is an early and essential task.

Metrics that look simple and are not

Average length of stay (ALOS) depends on whether day-care cases are included, whether stays are counted by discharge date or admission date, and how a patient transferred between wards is handled. Readmission rate depends on the window (7 days, 30 days), whether planned readmissions such as chemotherapy cycles are excluded, and whether readmissions to another hospital in the chain are visible. Claim turnaround depends on whether the clock starts at document submission or at receipt of the last missing document.

None of these choices is universal. The analyst's duty is to write them down and hold them steady, so that a change in the number reflects a change in care or operations, not a change in counting.

The artefact: explaining a metric change to a stakeholder

Suppose the analyst at a hypothetical multi-speciality hospital discovers that ALOS for general medicine appears to have improved after a definition was corrected. The email that follows matters more than the query.

Subject: General medicine ALOS: definition correction from this month's pack

Dear Dr [Name],

This month's pack shows general medicine average length of stay at 4.1 days, down from
4.9 days last month. Most of this change is due to a correction in how we count, not a
change in patient care, and I want to flag it before the review meeting.

What changed
  Until last month, patients transferred from general medicine to ICU were counted
  as one continuous general medicine stay, including their ICU days. From this month,
  ICU days are attributed to ICU, as agreed with the medical administrator on 2 September.

Effect
  Recalculating last month on the new basis gives 4.3 days, not 4.9. So the like-for-like
  change is 4.3 to 4.1 days, a small improvement within normal month-to-month variation.

What I have done
  Restated the last 12 months on the new basis (attached), and added a footnote to the
  dashboard. The old series is kept in a separate tab for reference.

Please let me know if you would like the ward-level breakdown before the meeting.

Regards,
[Analyst name], Hospital Analytics

The email does three things: it warns before the number is misread, gives the like-for-like comparison, and shows that history has been restated so trends remain valid. All figures are illustrative.

Privacy is not optional

Health data is among the most sensitive personal data. Work with the minimum identifiable data needed, use patient IDs rather than names in analysis files, apply minimum cell sizes before sharing small breakdowns, follow your organisation's policies under the Digital Personal Data Protection Act, 2023 and any sector-specific rules, and never paste patient data into public AI tools.

A week in a hospital analytics team

Daily occupancy and admissions for the operations head; a request from the cardiology department for procedure volumes by consultant; an investigation into why pharmacy revenue does not match billing (a returned-medicines process books credits a day late); a monthly quality pack with infection and readmission indicators for the quality team; and time spent on the HIS vendor's export format, which changed after an upgrade.

Measures of the role

Accuracy and timeliness of the recurring packs, consistency of metric definitions, turnaround on department requests, and whether analysis informed operational decisions such as discharge planning or bed allocation.

Medical administrators, billing and the HIS vendor

Usually a head of MIS or analytics, the chief operating officer's office, or, at insurers, a claims analytics manager. Colleagues include medical administrators, department heads, the quality team, billing, IT and the HIS vendor, and at insurers, medical reviewers and fraud teams.

Confused with clinical data management

Clinical data managers work on clinical trial data under strict protocols, which is a different world from hospital operations; see the clinical data manager guide. Medical coders assign codes to records; healthcare analysts use those codes.

HIS, TPA and ICD-10: reading healthcare postings

• "HIS, MIS, bed occupancy, ALOS" — hospital operations analytics.
• "Claims analytics, TPA, denial management" — insurance or payer analytics.
• "ICD-10, CPT coding knowledge" — coding familiarity valued; sometimes a coder role.
• "Population health, outcomes research" — more statistical, often with clinicians.

Definitions sheet, reconciliation, one acted-on analysis

Month one: learn the HIS tables and write the definitions sheet for the five metrics management uses most. Month two: reconcile one metric to finance or billing. Month three: deliver one analysis that a department head acts on. Next roles include senior healthcare analyst, hospital MIS head, claims analytics lead or health-tech product analyst. Writing a definitions sheet and a restatement email like the one above is practised in the Data Analyst programme, alongside the SQL needed to recompute a year of history quickly.