Records analysis, 81,400 admissions

Public health poster

Eight years of emergency admissions in older adults set against night time temperature, laid out as an A0 portrait board that opens with four headline numbers. It is an analysis of records a city already holds, and it ends with a change a public health unit could make on Monday: watch the night minimum, not the daytime maximum.

Create a poster with OneCraftA0 Portrait, printed at 841 by 1189 millimetres

The whole board

The poster at full size, exactly as it prints. Every number, citation and caption on it was written for this example, so the layout is being judged on real content.

BU
Night time temperature and emergency admissions in older adults: eight years of city records
N. Haddad ¹, E. Barlow ¹,², C. Iwu ²
1 School of Population Health, Brightwater University · 2 Coastal City Public Health Unit
14.2%
more admissions on hot nights
95% CI 9.8 to 18.7
22.1°C
night threshold for the rise
change point, 95% CI 21.4 to 22.9
3 days
lag to peak effect
distributed lag model, 0 to 10 days
81,400
admissions analysed
2017 to 2024, age 65 and over
Abstract
Heat warnings in this city are issued on the forecast daytime maximum. We asked whether the night time minimum is the better trigger for older adults, using 81,400 emergency admissions in people aged 65 and over between 2017 and 2024, linked to hourly temperature from the airport station. A distributed lag model found a change point at a night minimum of 22.1 degrees (95% CI 21.4 to 22.9), above which admissions rose 14.2% (95% CI 9.8 to 18.7) with the effect peaking three days later. Daytime maximum added nothing once night minimum was in the model. Applying a night based trigger to the study period would have issued 31 warnings against the 19 actually issued, and would have covered 84% of the excess admissions rather than 52%. The finding held when humidity was added, when a second weather station was used, and when the 2020 lockdown months were dropped.
Introduction
A hot day is survivable if the night that follows it is cool, because the body sheds the heat load while asleep. When the night stays warm the load carries over, and it is that carry over rather than the peak itself that the physiology literature associates with cardiovascular strain in older adults. Warning systems have mostly not followed. This city, like most in the region, triggers on forecast daytime maximum, a figure that is easy to communicate and easy to forecast. Whether it is the right figure is an empirical question that can be answered from records this city already holds, and answering it costs nothing but the analysis. Eight years of admissions and hourly temperature from a station six kilometres from the centre is enough to test it. The question is not whether heat kills, which is settled, but which number a public health unit should watch on a Thursday afternoon when it decides whether to open cooling centres for the weekend.
METHODS
Data. All emergency admissions age 65 and over, 2017 to 2024, from the city hospital admissions register.
Exposure. Hourly dry bulb temperature from the airport station, 6 km from the city centre, no gaps over 3 hours.
Model. Quasi-Poisson distributed lag non-linear model, lags 0 to 10 days, adjusted for season and day of week.
Change point. Segmented regression on night minimum, bootstrap confidence interval over 2,000 resamples.
Sensitivity. Repeated with humidity, with a second weather station, and excluding the 2020 lockdown months.
REFERENCES
1.Gasparrini A, et al. Lancet 2015;386(9991):369-375.
2.Royé D, et al. Environ Res 2021;196:110412.
3.Murage P, Hajat S, Kovats RS. Lancet Planet Health 2017;1(3):e97-e106.
Relative risk
Under 18
18 to 20
20 to 22
22 to 24
Over 24
00.350.71.051.4
Figure 1. Relative risk of admission by night minimum temperature, lag 0 to 10 days pooled.
Results
Above a night minimum of 22.1 degrees admissions rose 14.2% (95% CI 9.8 to 18.7), and above 24 degrees the rise was 28%. The change point at 22.1 degrees (95% CI 21.4 to 22.9) was stable across every sensitivity analysis we ran. The effect peaked at lag three and was gone by lag eight, which is consistent with a cumulative load rather than an immediate collapse. Cardiovascular and respiratory admissions accounted for 71% of the excess, and falls for a further 9%. Adding the daytime maximum to the model changed the night coefficient by less than 2% and did not improve fit, so on these records the night figure carries the signal on its own.
Discussion
A trigger built on the night minimum would have issued 31 warnings across the study period rather than 19, and would have covered 84% of the excess admissions rather than 52%. The extra warnings are not free, since a warning issued too often is ignored, but twelve extra nights across eight years is a modest cost for that coverage. This is one city with one weather station and a temperate coastal climate, so the threshold itself should not be copied. The method can be: any city with an admissions register and an hourly station can run this analysis on data it already holds, and should before it sets a threshold.
Conclusion
Night time minimum predicts emergency admissions in older adults better than daytime maximum in this city, and a warning threshold set at 22 degrees would have covered most of the excess. The change requires no new data collection, only a different figure read from a forecast the city already receives, and no change to how the warning itself is issued or communicated.
Admissions up 14.2% above a 22.1 degree night
Effect peaks at three days, gone by eight
Daytime maximum adds nothing to the model
Night trigger covers 84% of excess, not 52%

Block by block

What each block on the board is for, in the order a reader walks it.

Headline number strip
Four stat blocks across the top: the 14.2% rise, the 22.1 degree threshold, the three day lag and the 81,400 admissions analysed, each with its interval or period underneath.
Abstract
The current trigger, the question, the model, the change point and what a night based trigger would have covered, in one block.
Introduction
The physiological reason night temperature matters, why warning systems use the daytime figure anyway, and why this is answerable from existing records.
Methods as a labelled protocol
Data, exposure, model, change point estimation and the three sensitivity analyses, each on its own labelled line.
Risk by temperature band
Relative risk across five night minimum bands, so the shape of the threshold is visible rather than asserted.
Results
The effect above the threshold, its stability across sensitivity analyses, the lag structure, and the finding that daytime maximum adds nothing.
Discussion
What a night based trigger would have issued and covered, the cost of over warning, and why the threshold itself should not be copied to another city.
Conclusion and references
The recommended change, four checked takeaway points, and three references in journal style. The three references are invented for this fictional study, because the reference block is required on this layout; on your board, replace them with your sources.
How to adapt this board
Lead with the four numbers, because a portrait board with a headline strip is read top down and those four are what a reader takes away from ten metres. Then the model, stated plainly enough that another analyst could rebuild it: the design, the lag window, and what was adjusted for. The threshold figure comes last and does one job, which is to show where the curve turns. Everything that is not one of those three things belongs behind the QR code, including the sensitivity analyses, the full model output and the data source terms.

What makes this board work

Four numbers before any prose

The effect, the threshold, the lag and the sample size run across the top as a strip. A passer by who reads nothing else leaves with the finding, which is what a poster is for.

The comparison is the point

Daytime maximum adds nothing once night minimum is in the model. Stating the rival explanation and showing it fails is more convincing than only showing your own variable works.

It ends in a decision, not a finding

Thirty one warnings instead of nineteen, covering 84% of excess admissions instead of 52%. Public health work is judged on what changes, so the poster converts the model into the trigger a unit would set.

Questions people ask

What goes on a public health poster?

The population and period, the data source, the analysis, the effect with its interval, and what a service should do differently. An analysis with no implication reads as unfinished to this audience.

How do I show a threshold effect?

A change point with its confidence interval, plus a figure showing risk across the range so a reader can see the shape rather than trust a single cut point.

Should I list sensitivity analyses?

Name them briefly in the methods and say in the results whether the finding held. This board ran humidity, a second weather station and a lockdown exclusion, and says the change point was stable across all three.

Can the poster be reused for a different city?

The method transfers, the threshold does not. The board says so explicitly, because a 22 degree trigger copied into a different climate would be wrong.

How do I choose the four headline numbers?

The effect, the threshold or exposure it applies to, the timing, and the sample. That is what this board shows: the rise in admissions, the temperature at which it starts, the lag, and the number of admissions analysed. Anything that is not one of those four is a result, not a headline.

Should the sensitivity analyses be on the board?

One line each in discussion, with the direction they moved the estimate. Readers want to know that the result survived a different lag window or a different threshold; they do not want the tables. Those go behind the QR code.

Can I reuse the board for another city?

Yes. Retype the four stats, the chart values and the counts in methods, and keep the structure. The layout, the block order and the headings are the same job for any city, which is why the numbers are the only thing you should have to change.

Build your own in about a minute

The button below opens the generator with this use case already described. Change the wording to match your own, generate, then edit anything you like.

Make my public health poster

Other poster examples

Undergraduate research poster

An undergraduate research poster does not need big numbers, it needs honest ones arranged in the standard scientific order. This A0 portrait example shows a four week audit of a 50 cent reusable cup discount at two campus cafes: 4,860 sales observed, use rising from 12.4 to 29.1 percent, and a conclusion that explicitly declines to claim causation.

Health economics poster

A health economics poster stands or falls on whether its table adds up and its uncertainty is on the board rather than in the appendix. This A0 portrait example carries a cost and QALY table whose totals are the sums of their rows, an ICER of 4,120 per QALY as the key number, and a tornado chart showing which input moves it.

Lab research poster

A lab research poster earns trust with its controls, its replicate counts and its willingness to print the doses where nothing happened. This A0 portrait example gives one dose response line chart most of the canvas, states the null results at 5 and 10 micrograms per millilitre as findings, and keeps the positive control on the same axes.

Want the steps in the builder? Read Add charts and diagrams, then choose the template, theme and size. For everything this generator can do, see the poster maker.

Sources

Written and checked by the OneCraft team. Last checked .