Data Visualisation & Health Metrics

Blood pressure, glucose and lab results now travel from devices and records into patient portals and monitoring dashboards. One reading can reach two very different readers: a clinician scanning a long list of patients between appointments, and a patient looking at a single number about their own body, possibly with worry attached. A chart that is misread can cause needless alarm, false reassurance or a missed change, so how a number is drawn is a safety question as well as a visual one.

Two readers, two questions

A clinician's questions are usually comparative: who has changed, who needs attention first, and how does this reading sit against the earlier ones. That calls for density: many patients in one list, trends over a chosen time window, and flags that say why they were raised. In a 2025 study of a remote blood pressure dashboard, clinicians asked for an adjustable time window and easy switching between patient details and readings, and the chart showed readings alongside a 14-day moving average (a rolling average that smooths day-to-day swings).

A patient's questions are different: what does this number mean, is it typical for me, and what happens next. That calls for fewer marks, plain words and context next to the number. A screen can describe where to find the care team without giving medical direction; the clinical wording itself belongs to clinicians to write and review.

A number needs a range, and the right one

A reference range is the span of values a laboratory treats as typical for a population. A systematic review of 18 studies on lab results in patient portals found that most tested displays were either numbers with ranges or horizontal bars with coloured sections. Its authors reported that bars combined with ranges tended to support understanding better than numbers alone, but 89% of the studies used hypothetical data and only 3 involved actual patients. A separate review of 39 articles on patient-facing charts found number lines and bar graphs were generally understood better than line graphs, while borderline results stayed hard to interpret.

The standard range is not always the one that applies to a person. In an experiment with 6,776 participants imagining a diabetes test result, a result that fell between the standard range and a treatment goal range was understood better, and felt less discouraging, when the goal range replaced the standard range than when both were shown. In another experiment with 1,618 adults, adding a marker for the point where clinicians typically become concerned lowered the perceived urgency of near-normal results without dulling the response to extreme ones. Both are online experiments with hypothetical results, so they show what to test, not what to assume.

Colour, scale and missing data

Red, amber and green are common in health displays; in the patient-facing review, about three quarters of the charts used colour and traffic-light schemes were prominent. Under WCAG 1.4.1 (Level A), colour cannot be the only way information is conveyed, and under 1.4.11 (Level AA) chart lines and other meaningful graphics need at least a 3:1 contrast ratio against their surroundings. UK government chart guidance adds that colour-vision differences affect roughly 8% of men, and recommends labelling lines directly and providing a table or text alternative. That 8% figure comes from European populations; a 2012 review of prevalence studies put it at 4% to 6.5% in men of Chinese and Japanese ethnicity, so the share varies by population.

The same guidance says bar chart axes should not be cut short because it distorts proportions, and that points should not be joined across a gap in the data. A gap in home readings is information in itself: a missed week is not a normal week. Its uncertainty guidance, written for official statistics rather than health screens, favours showing ranges and describing changes within the margin of error as broadly stable, a principle that has not been tested on clinical dashboards.

Alerts and risk numbers

In monitoring, a review of remote patient monitoring alarm research describes sensitive detection rules that raise frequent incorrect alerts, leading to alarm fatigue: staff who stop responding because too much is flagged. In the blood pressure dashboard study, alerts for out-of-range readings were generated each morning and were triggered for 69% of the 190 participants in the intervention group. The study did not judge whether those alerts were false, but it shows how quickly a flag list can fill, which makes thresholds and their presentation design decisions.

Risk figures have their own traps. A published primer for decision-aid developers advises stating which group a figure applies to and over what time period, keeping the denominator (the bottom number of a fraction) the same across comparisons, and avoiding the 1-in-X format. An icon array, a grid of small figures with some shaded to show how many in a group are affected, can help people with lower number skills, though not everyone benefits equally from graphics. The CDC Clear Communication Index, a checklist for public health materials, includes separate items on how numbers and risk are presented.

Two illustrative examples

Take one week of home blood pressure readings, invented for illustration. The clinician view is a row in a caseload list: latest reading, a moving average, the number of days with no reading, and a short reason if the row is flagged, with the time window adjustable. The patient view shows the same week as seven plain points against the range the care team set, states in words how many fell outside it, marks any missing day as missing, and gives a contact route. Both views rely on a text label or symbol as well as colour.

For risk, a screen comparing two options can write 1 in 100 against 2 in 100 with the same denominator and a stated time period, rather than 1 in 100 against 1 in 50, and can add an icon array showing the whole group.

Related Showcase cards

• Humanising the number - a lab result shown with its reference range instead of as a bare number.
• Status is never colour alone - a status reading shown with a shape and a word, not colour alone.