Groups & Experiences

SciK-Health: what does research produce in university hospitals?Case study · Science mappingA dataset on the 49 Italian public university hospital centres from 2000 to 2024, indicators in five dimensions and two tools for citizens and decision-makers.

PRIN 2022 project (code 2022825Y5E) · University of Naples Federico II and University of Campania “Luigi Vanvitelli” · Principal investigator: Massimo Aria

The problem

In Academic Health Science Centres (AHSCs), patient care, teaching and research coexist: in Italy they are university hospitals and research hospitals (IRCCS). Their scientific output should translate into better care, but how much research do they produce, on which topics and with what impact? Answering requires making different sources comparable, and results readable by citizens, professionals and decision-makers.

The method

The project builds a longitudinal dataset on the 49 public centres, from 2000 to 2024: about 229,000 scientific publications and 15,800 clinical trials, together with competitive funding, patents, datasets and online attention (sources: OpenAlex, Dimensions, Altmetric). The centres follow three organisational models: IRCCS research hospitals (23), university hospitals (17) and hospitals integrated with the national health service (9). On this basis:

• science mapping of the conceptual and social structure of research: topics, collaborations, change over time;
• indicators in five dimensions: bibliometric, social and digital, industrial impact and innovation, clinical research, funding;
• a composite score summarising the five dimensions;
• validation with stakeholders: focus groups with three university hospitals, one research hospital and a regional health agency.

The result

A web app makes more than 25 years of the centres’ scientific output searchable in a single query, with three perspectives and profile-based access: citizens find centres, experts and studies by disease; managers follow researchers’ productivity, impact and talent; funders assess grants, patents, clinical trials and online attention. For managers and funders, the composite R-daisy indicator summarises the five dimensions in a “daisy” profile. The case shows how methods developed to study science become tools for transparency and for supporting public decisions in health care.

Further information: SciK-Health project website

Can dogs smell our emotions?Case study · Experimental design and multivariate analysisAn experiment with 40 dogs shows that the smell of human fear and happiness changes their behaviour and heart rate.

D’Aniello, Semin, Alterisio, Aria, Scandurra, Animal Cognition, 2018 · University of Naples Federico II (Departments of Biology and of Economics and Statistics) and ISPA – Instituto Universitário, Lisbon

The problem

“If you are afraid, the dog will notice”: people often say so, but is it true? When we feel fear or happiness, the body releases chemical signals through sweat. The question was whether these signals are enough to pass an emotion from one species to another, from humans to dogs, and how to show it with data rather than anecdotes.

The method

A controlled experiment, with the choices typical of a sound statistical design:

• odours collected from the sweat of donors watching videos that induced fear or happiness, pooling the samples of several people to neutralise individual differences;
• 40 pet dogs (31 Labradors and 9 Golden retrievers) assigned at random to three conditions: fear odour (15), happiness odour (15) and an odourless control (10);
• two-minute trials in a room with the owner and a stranger, both unaware of the condition (“blind”), with video recording and a heart rate monitor;
• behaviours towards the owner, towards the stranger and stress signals coded every 5 seconds, together with heart rate;
• linear discriminant analysis to see whether the odour can be recognised from the data, smoothing curves for the trend over time, ANOVA and Kruskal–Wallis tests for the comparisons.

The result

Yes, dogs “sense” our emotions. With the fear odour, dogs showed more stress signals and a higher heart rate; with the happiness odour, they sought their owner less and approached the stranger more. The discriminant analysis correctly assigns 87.5% of the observations to the odour condition they come from. Statistics was needed three times: to design the experiment properly, to collect comparable data and to measure how far the results support the hypothesis.

Further information: the article in Animal Cognition

How do you slow cars down at intersections?Case study · Driving behaviour and experimental designHow do drivers behave near an intersection? In a simulator, 23 people and ten versions of the same road: driving styles compared to choose the safest treatment for a rural municipality.

Montella, Aria, D’Ambrosio, Galante, Mauriello, Pernetti, Accident Analysis and Prevention, 2011 · University of Naples Federico II (Departments of Transportation Engineering and of Mathematics and Statistics) and Second University of Naples

The problem

In a rural municipality the main road is a long straight and cars reach the intersections at more than 100 km/h, with a high risk of crashes. Road safety is first of all a matter of behaviour: speed is chosen by the driver, depending on how they perceive the road and the hazard, and not everyone reacts in the same way. Low-cost “perceptual” treatments (road markings, colours, narrowings) work precisely on this: change perception to change behaviour. The administration has to choose how to intervene on a limited budget. Which solution really changes the way people drive? And how can we know before opening the building site?

The method

An experiment on driver behaviour, run in the VERA driving simulator at the TEST laboratory in Naples (the cabin of a real car on a motion platform, in front of screens covering 180 degrees):

• 30 drivers drawn at random from those meeting the selection criteria, balanced by gender and age; 23 completed the test (the others suffered from simulator sickness);
• repeated measures: each driver went through all ten versions of the same stretch (a straight road without intersection, two untreated intersections and seven with treatments), so each person is compared with themselves; the order was counterbalanced across drivers to neutralise the habit effect;
• speed and lateral position recorded over the 500 metres before the intersection: 207 individual driving profiles in all;
• driving styles: each profile was classified, by similarity to theoretical profiles, into one of three ways of approaching the intersection: speeding up, braking late at a constant rate, or braking early and gradually;
• cluster analysis to find homogeneous groups of behaviour; analysis of variance for categorical data (CATANOVA, based on Gini’s heterogeneity index) to compare styles across versions; ANOVA and t tests for speed and lateral position.

The result

Without treatments, one driver in three speeds up when approaching the intersection (35%): the riskiest style. And the closer they get, the more behaviours diverge: the spread of speeds across drivers grows from 15 to 21 km/h. With perceptual treatments the driving style changes significantly: drivers who speed up fall to 9–22% and the cautious style, braking early and gradually, becomes prevalent (up to 83% with the raised island).

Dragon teeth markings (white triangles at the edges that optically narrow the lane), the colored intersection area and the raised median island also cut speed by 13–23 km/h from 250 metres before. The islands, however, by narrowing the lane, shift cars 40–60 cm towards the road edge. For the municipality the choice therefore falls on the colored intersection area with rumble strips: braking starts 330 metres earlier, speed at the centre of the intersection falls from 97 to 76 km/h, and cars stay in the middle of the lane. A low-cost treatment, chosen by observing how people drive before laying a single metre of asphalt.

Further information: the article in Accident Analysis and Prevention