Section 02

The national trend and where it broke

Turnaround remained near a tenth of a calendar day through 2021. It then rose to 0.123 in 2022 and 0.193 in 2023. The shape matters more than the level: this is not a drift, it is a step.

Verification ledger

Every headline claim in the synthesis, recomputed from the values as printed. Two entries do not reproduce exactly, which is expected when a ratio is quoted from unrounded data but the endpoints are printed rounded. Both are disclosed rather than smoothed over.

Section 03

CT and MRI carried the deterioration

The modalities most dependent on scarce subspecialty expertise, and on increasingly large and complex image sets, showed by far the largest increases. The overall figure of +113% sits closer to plain film than to advanced imaging, which is a clue about volume weighting rather than a reassurance.

Section 04

The distribution matters more than the mean

A mean of 0.193 calendar days sounds trivial. It is trivial, as a mean. The operational and clinical weight sits in the tail, and in the weekday-to-weekend split that a single national average dissolves entirely.

The weekend signal

The single clearest published evidence that the national mean hides structure is the day-of-week split. In 2023 the weekend mean was 0.325 calendar days against 0.189 on weekdays, a difference of 0.136 days and a ratio of 1.72. A department reporting one blended figure would see neither number.

Section 05

Capacity loss is not distributed neutrally

In 2023 the lowest-income communities recorded a mean of 0.246 calendar days against 0.112 in the highest-income communities. More consequentially, the relative disadvantage widened from 56% in the 2014 to 2020 period to 121% in the 2021 to 2023 period. Constrained capacity converted an existing access disadvantage into longer diagnostic uncertainty.

What this implies for measurement

• A single departmental turnaround figure cannot detect an equity gap. Stratification is not a refinement of the measure, it is the measure.
• The disadvantage widened during the same years the national mean broke from its plateau, which is consistent with the interpretation that scarce capacity is rationed along existing lines of advantage.
• Improvement targets set on a blended mean can be met by improving the already-fast cohort, which is the easiest cohort to improve.
• Section 07 lets you apply the published ratio to your own case mix to estimate the internal spread implied by your blended figure.

Section 06

Size did not confer resilience

Radiology-majority practices reached 0.455 calendar days in 2023, and practices with at least 100 radiologists reached 0.371, while radiology-only practices remained comparatively low. The organizations with the most scale performed worst on the measure scale was supposed to protect.

The reframe this forces

If scale were the binding constraint, the largest organizations would show the shortest turnaround, since they hold the deepest subspecialty benches and the most capacity to load-balance. They do not. The synthesis names the plausible contributors, and every one of them is an operating-model variable rather than a headcount variable.

• Rapid consolidation

  Growth by acquisition assembles capacity faster than it integrates it. Reading capacity that cannot see a worklist is not reading capacity.

• Fragmented PACS and worklists

  Several queues cannot be load-balanced against one another. Idle capacity in one silo coexists with a backlog in the next.

• Uneven credentialing

  A radiologist who is not credentialed at a site cannot read for that site, so nominal headcount overstates the pool available to any given study.

• Contract complexity

  Coverage obligations negotiated separately across entities constrain which cases can be routed where, independent of who is available.

• Post-merger workflow integration

  The interval during which two operating models coexist is precisely the interval in which neither runs at design capacity.

Section 07 · Interactive model 1

Later-day exposure calculator

National means are hard to act on. This model converts your own volume and turnaround into two figures a department can carry into a budget conversation: how many examinations cross into a later calendar day each year, and how many patient-days of diagnostic uncertainty your service generates.

Section 08 · Interactive model 2

Capacity inflection simulator

Projected radiologist supply growth of 40.3% against projected utilization growth of 26.9% reads as comfortable headroom. The synthesis warns that nominal headcount is not equivalent to effective capacity, and names six constraints on the conversion. This model prices them.

Section 09 · Interactive model 3

Artificial intelligence and the shape of the queue

The synthesis makes a precise claim: artificial intelligence will improve prioritization and reduce selected critical-case delays, but it will redistribute queues more readily than it will create radiologist capacity. That claim is arithmetically demonstrable, and the demonstration below is exact rather than illustrative.

Section 10 · Interactive model 4

Is your reporting set able to see this problem?

The synthesis names seven measures and three stratifications that together make a capacity inflection visible. A department reporting only a mean would have watched the national series move from 0.091 to 0.193 and seen a number still comfortably under a fifth of a day. Score your own reporting against the full set.

Section 11 · Interactive model 5

Leadership agenda readiness

The synthesis argues that radiology leaders should manage the queue as a clinical production system, and it names the components. Thirteen of them are scored here across five domains, with the operating-model items treated as load-bearing because the article closes on exactly that point.

Section 12

Consequences for patients, radiologists, and health systems

Delayed interpretation does not stay in radiology. It propagates through the care pathway, and it propagates back into the workforce that produced it.

The reinforcing loop

The most consequential dynamic in this section is not any single consequence but the circuit they form. Each stage is drawn from the synthesis, and the loop closes back on its own starting condition.

Section 13

A research-grounded outlook to 2055

Demand will probably remain ahead of usable capacity. National projections place imaging utilization in 2055 between 16.9% and 26.9% above 2023 levels by modality, with CT potentially 45.2% higher if recent trends continue through 2030. Radiologist supply is projected to grow 25.7% without further residency growth, or 40.3% with it.

Apparent headroom, recomputed as the supply index divided by the demand index. These figures assume every additional radiologist converts fully into timely reading coverage, which is the assumption the synthesis specifically warns against. Recomputed from published projections

Three developments the synthesis considers likely

• Turnaround becomes a patient-safety, equity, and contract-performance measure

  Rather than a narrow departmental statistic. The equity finding in Section 05 is what forces this reclassification: a measure that varies this systematically by community income is not an operational metric alone.

• Distributed reading networks and enterprise worklist orchestration expand

  Particularly during nights and weekends and in underserved markets, which is precisely where the published weekend and income gaps sit.

• Artificial intelligence improves prioritization and reduces selected critical-case delays

  But redistributes queues more readily than it creates radiologist capacity. Evidence shows active reprioritization can help, whereas passive flags may not, and maximum-wait guardrails remain essential to prevent routine or false-negative studies from aging invisibly. Section 09 models all three of those conditions.

Section 14

Critical appraisal and bottom line

Interpretation requires discipline. The limitations below are stated in the synthesis itself, and they bound what any of the models in this dashboard can legitimately claim.

What would strengthen the evidence

• Timestamp-level data rather than claim dates, which would convert the later-day share from a proxy into a direct measure of the delay distribution.
• Percentile reporting at national level, so the tail the article argues for could be observed rather than inferred.
• Linkage across payer types, to establish whether the outpatient Medicare fee-for-service cohort is genuinely the first visible buffer or simply the most observable one.
• Practice-level operating characteristics, so the consolidation and worklist-fragmentation hypotheses in Section 06 could be tested rather than named as plausible.

Section 15

Selected references

The six sources cited in the synthesis, with digital object identifiers.

Methodological note on this dashboard

Every published value used here is listed in the verification ledger in Section 02, where each headline claim is recomputed from the figures as printed. Two entries do not reproduce exactly and are disclosed as rounding artifacts rather than corrected silently. All figures are drawn as inline scalable vector graphics by this dashboard’s own code, with no charting library and no external requests, so nothing here depends on a content delivery network being reachable.