The PACS worklist is the first thing a radiologist sees at the start of a shift. It's also one of the least thoughtfully designed interfaces in all of clinical informatics, and the design problems are not primarily visual. They're cognitive.
When the worklist is a flat list of 60 studies sorted by arrival time, the radiologist faces a non-trivial decision problem before reading a single image: what do I open first? That decision requires scanning the list for indication clues, patient age, ordering location, and any flags that suggest urgency. Doing this for 60 studies takes time and mental energy that most radiologists don't consciously account for, because it feels like pre-work rather than work. But the decision quality matters, and the cognitive overhead accumulates across a shift.
The Prioritization Decision Is Made Under Load
The classic PACS worklist presents each study as a row: patient name, MRN, study type, ordering date/time, ordering location, and perhaps a note field with the indication in free text. The radiologist reads down the list, building a mental model of what's urgent, what's complex, and what can be batched as routine.
Cognitive load theory, developed by Sweller in the 1980s and applied extensively to medical decision-making since, distinguishes between intrinsic load (the inherent complexity of the task itself), extraneous load (the effort imposed by how information is presented), and germane load (the productive mental effort that leads to learning and schema formation). The worklist scan is almost entirely extraneous load: the information is there, but the format forces the radiologist to extract and reformat it mentally before they can act on it.
This extraneous load is particularly costly because it occurs at the start of the shift, before the radiologist has settled into a reading rhythm. Decision fatigue research consistently shows that the quality of complex decisions degrades after sustained cognitive effort. If the first 10-15 minutes of a shift are spent parsing an unstructured list to identify reading priorities, that's not a neutral cost.
What the Worklist Gets Wrong, Specifically
Indication fields as free text are a consistent cognitive friction point. An indication field that says "r/o PE, 52M, presented SOB" requires parsing multiple pieces of clinical information. An indication field that says "CT chest PE protocol" with no clinical context tells the radiologist almost nothing about urgency. And an indication field that's blank, which is more common than it should be, forces the radiologist to work backward from the study type and ordering location to guess at clinical context.
Time-of-order as the default sort is the wrong primary signal for most use cases. Studies that have been waiting longer are not necessarily more urgent; they may simply have been ordered first. The radiologist scanning a list sorted by arrival time needs to apply urgency filters mentally on every line, which is the very thing the worklist should be doing computationally.
Visual density compounds both problems. A typical PACS worklist presents 8-10 columns of text in a compact table. All rows look the same. There is no visual hierarchy: the stroke rule-out CT and the routine knee MRI occupy identical visual weight. Humans are very good at pattern recognition but only when patterns are visually distinct. A flat list of identical-weight rows provides no such distinction.
What Priority-Ordered Worklists Change
Priority-ordered worklists, whether assigned by a radiologist manually, by PACS configuration rules, or by an AI triage layer, address the extraneous load problem by moving the prioritization decision out of the individual radiologist's cognitive space and into the system.
When the first three studies on the worklist are flagged as urgent-review, the radiologist's first decision is not "which of these 60 things should I open first?" but "should I agree with this prioritization?" That's a verification decision rather than a generation decision, and it requires substantially less cognitive effort when the prioritization is accurate.
We're not saying that AI triage eliminates the need for radiologist judgment about reading order. For genuinely complex cases or unusual presentations, a radiologist who understands the clinical context should override a priority assignment. But for the 70-80% of worklist items where the urgency level is unambiguous once the indication is known, computational prioritization frees cognitive bandwidth for the actual reading task.
The Column Display Problem
Even with priority ordering in place, the column design of most PACS worklists undercuts the value of that ordering. If urgent and routine studies look visually identical, the prioritization signal has to be carried entirely by row position. Radiologists who scan lists rather than reading them top-to-bottom will miss the urgency hierarchy.
Effective worklist design for cognitive load reduction uses at least three visual distinctions: color or icon coding for urgency level, visual weight variation (bold or size differences) for the primary identification fields (patient name, study type), and explicit suppression of columns that are not useful at the scan-and-decide stage. Most radiologists don't need to see the referring physician's NPI number during the worklist scan. Showing it adds visual noise without adding decision-relevant information.
The dictation platform integration compounds this: if the worklist and the dictation window are in separate application panels that don't share context, the radiologist navigates between them multiple times per study. Every navigation is a context switch, and context switches are a documented source of cognitive load increase and error risk in complex cognitive work. Worklist design that reduces navigation steps, even by one or two per study across a 60-study shift, has a compounding effect.
What Worklist Redesign Cannot Fix
Better worklist design reduces extraneous cognitive load. It does not reduce intrinsic load, which is set by the complexity of the studies being read. A worklist containing 30 high-complexity neuroimaging studies will still require the cognitive effort those studies demand regardless of how elegantly they're sorted and displayed.
Worklist redesign also doesn't address the fundamental capacity constraint: if more studies arrive than a radiologist panel can read in a shift, a better-designed worklist tells them more clearly what to read first, but it doesn't reduce the volume. The cognitive benefit of a well-designed worklist is most pronounced at moderate load levels. At extreme overload, the primary problem is volume, and worklist design is a second-order concern.
The goal we're working toward is a worklist that surfaces clinical context, provides prioritization support, and minimizes navigation overhead so that the radiologist's mental effort is directed almost entirely at the image reading task. That's a better design problem than "how should we arrange the columns." It requires thinking about the radiologist's cognitive state at different points in the shift, and designing the information architecture to match that state rather than treating the worklist as a neutral display of everything the system knows.