Actuarial studies · 2026
Radiology Workforce Model
Turn imaging activity into reviewable workforce estimates.
Reliable staffing figures are difficult to assemble across a clinic network. The model approaches the question through workload: allocate imaging activity to sites, estimate the reporting time involved and compare the implied workforce with published benchmarks. Scenario outputs show how much the answer depends on the assumptions.
- Status
- Production
- Domain
- Workforce modelling · Health economics
- Role
- Model design, pipeline, QA framework
- Year
- 2026
Explore the workflow with illustrative content.
Start with the place
The base layer establishes the streets and area around a sample location. Select a location below to move the focus.
Add the surrounding context
A separate catchment layer shows how the surrounding area relates to the selected place. Spread the layers to see how they fit together.
Compare the nearby activity
Sample sites and a route sit above the catchment. Switch between locations to see the area of interest move across the map.
Illustrative content · no live data
View activity and capacity in the same place
Stack
- Python
- pandas
- Playwright
- pydantic
- rapidfuzz
- YAML scenarios
- openpyxl
A workforce model that combines imaging volumes, clinic characteristics and published benchmarks to estimate reporting capacity by site.
Build a consistent clinic base
Discovery, entity matching and geocoding bring clinic records into a consistent register.
Rurality, service mix and site characteristics provide context for allocation. Cached geocodes and smaller geographic test runs help keep data preparation repeatable and manageable.
Connect volume to workload
The model allocates activity by geography, clinic and modality before translating it into reporting time.
Separate assumptions describe radiologist-owned work and productive capacity. This keeps the steps between a service count and a workforce estimate available for review.
Compare coherent scenarios
Low, base and high cases vary related assumptions together.
That lets an analyst explore combinations of activity allocation and capacity, instead of relying on a single estimate. The scenarios are assumption-based ranges rather than statistical confidence intervals.
Put the checks beside the result
Audit sheets compare aggregate totals, individual-site estimates and state distributions with reference information.
Flags distinguish cases that need review from more serious breaches. A geographic mismatch prompts investigation of both the model and the network’s actual footprint; it does not establish the explanation by itself.
Leave a trail someone else can follow
Each run records its assumptions, configuration and input hashes alongside the workbook.
Those records make it possible to revisit a result, identify what changed and discuss the source of a difference without reconstructing the whole run from memory.
Pipeline
Clinic data to workforce scenarios
Stage Can reject
-
01
Discover
Clinic records
-
02
Match and locate
Consistent site register
Cached geocodes reused
-
03
Allocate activity
Clinic and modality volumes
-
04
Model workload
Workforce scenarios
-
05
Reconcile
Checks and review flags
-
06
Export
Audit workbooks and run records
The mathematics
A closer look
at the method.
The ideas behind the analysis, with the main terms explained. These are method summaries; fitted parameters and project-specific calculations remain private.
From activity to reporting workload
A service count needs assumptions about time and responsibility before it can inform workforce estimates. This schematic highlights the share of work attributed to radiologists and an allowance for ancillary activity.
- w
- Radiologist reading workload
- m
- Reading minutes from allocated volume
- r
- Radiologist-owned share of reading time
- u
- Ancillary uplift — meetings, protocolling
The full implementation also handles allocation, annualisation, productive hours and clinic characteristics. Those details remain private. Workload and headcount-equivalent outputs are kept distinct in the model.
Allowing allocation to vary by modality
Clinic counts provide one input to activity allocation, alongside a baseline share assumption. A modality-specific exponent lets the model explore different relationships between site footprint and allocated activity.
- σb
- Baseline operator share in bucket b
- nb / Nb
- Operator clinics over all clinics in the bucket
- ηd
- Tilt exponent for modality d
Buckets group sites by state and rurality. Setting η to zero removes the site-count adjustment. Other values change its sensitivity; interpretation also depends on the model’s normalisation and calibration, which are not reproduced here.
Three checks on the estimates
The audit workbook checks the estimates at national, clinic and state level. These checks help identify results that need investigation; agreement with a benchmark does not establish that every site estimate is accurate.
- National total Modelled Australia-wide FTE against published workforce data scaled by operator share, within a stated band.
- Per-clinic ceiling Soft and hard thresholds flag unusually large estimates, with evidence such as hospital affiliation or credentialed headcount recorded for review.
- State distribution Modelled state shares compared with published workforce distributions. Differences prompt review of allocation assumptions and the clinic network’s footprint.
The checks can raise warnings or errors. A state-level difference may reflect a modelling problem or a real feature of the network. Additional evidence is needed to distinguish those explanations.
Behind the work
The skills
that came together.
Workforce modelling
- Activity allocation and workload conversion
- Related-parameter scenarios
- Benchmark reconciliation
Data engineering
- Clinic discovery and entity matching
- Geographic classification and cached geocoding
- Typed configuration and reproducibility records
Review
- Assumption registers
- Site and regional QA flags
- Audit-first workbook design
Project highlights
-
A traceable estimate
Activity, workload and capacity linked in stages
-
Scenario comparison
Related assumptions explored together
-
Checks included
Aggregate, site and regional review sheets