Some examples of projects I have developed

Dispatchers often receive imprecise addresses. Comparing the outbound trip (destination uncertain) with the return trip (destination known), on Liguria emergency data, isolates the time lost looking for the patient.

Extra driving time to reach a patient at home, by trip length (minutes)

Read the first row: whether the trip is under 5 km or over 20 km, finding a patient at home costs roughly 4 to 5 extra minutes. It is a fixed cost of locating the patient, not a matter of distance; for urgent missions it is about 5 minutes, +30% on driving time.

So whatBetter caller location (handset location, address validation, dispatcher protocols) is one of the cheapest ways to save minutes.

Lombardy has a rule for where ambulances should take stroke patients: the nearest hospital of the right level within 45 minutes' drive. The region is now reclassifying hospitals and considering a new rule. Before changing the rule, I checked whether the current one is followed. For every patient I estimated the drive time to each of 37 hospitals and computed where the rule would have sent them.

Patients per hospital · actual (blue) vs. current rule (red)

So whatOne patient in three does not go where the rule says. Is it worth writing new rules, or changing them, if we do not follow the ones we have? Before redesigning the network, it pays to understand why practice departs from the rule.

Patients arrive at the emergency department following a smooth daily pattern. Doctors see them following the staff rota. Comparing the two, minute by minute across the whole region, shows where capacity is lost.

a · Patients arriving, per minute

b · Patients seen by a doctor, per minute

Each dot is one minute of the day, averaged over 2019 across all Lombardy emergency departments. Axis labels in Italian: pazienti in ingresso = arrivals, pazienti visitati = patients seen.

Demand peaks at about 14 arrivals per minute around 10:00, while doctors never see more than about 12. Instead of staying high to clear the queue, capacity collapses at the 14:00 and 20:00 shift changes. Overlapping shifts are not enough to keep the flow going.

So whatStaggering shift handovers and protecting capacity just after the morning peak would cut queues without hiring more doctors.

Hospitals increasingly fill emergency department shifts with contract doctors. Within each shift and treatment area, patients are assigned to doctors quasi-randomly, which allows a causal comparison of the two groups on the same kind of patients.

Balance test · predicted vs. actual length of stay

So whatThe gap comes from employment status, not from who is hired: employees put in extra effort for fellow employees, but not for contractors. Staffing contracts shape team behaviour and productivity, not only cost.

In Lombardy, ambulances are run by non-profit organisations, and employees and volunteers do exactly the same job on board: driver, crew leader, rescuer. Following every worker month by month shows how each group responded to the pandemic shock.

Average missions per worker per month · active workers

So whatVolunteers are a small but resilient share of capacity: when the system was under the greatest stress, their presence did not shrink. Emergency planning should count on them as a stabiliser, not only as extra hands.

Rain slows ambulances for reasons unrelated to the patient's condition. Using rainfall as a natural experiment, I estimate the causal effect of response time on mortality from cardiac events.

So whatCutting even one minute through vehicle positioning, dispatch speed or routing has a large, measurable return.

Ambulance services are judged on response time. But once the ambulance reaches the emergency department, the crew must wait until the patient is taken over by hospital staff. Until then the vehicle is blocked and cannot answer the next call. In Lombardy this handover takes 21 minutes on average, and it varies hugely from one hospital to another.

Average ambulance handover time by hospital (minutes)

So whatEvery minute an ambulance waits at the hospital is a minute it is not available. Where handovers are slow, more vehicles and crews are needed to guarantee the same response times. Fleet sizing and hospital emergency department capacity must be planned together.

In 11% of missions the patient refuses transport to hospital. Was an ambulance the right response? The refusal rate depends heavily on the reason for the call: it is close to zero for time-critical emergencies, but very high for calls that are not really medical emergencies.

Share of missions ending in refused transport, by reason for the call

So whatA large share of calls signal social and care needs rather than emergencies. Mapping who calls, for what and where shows which local services are missing, such as home assistance, lifting teams or community nursing, and frees ambulances for the patients who need them.

Lombardy's emergency calls are handled by four regional dispatch centres. In 2023 part of the Milan area was reassigned to the Laghi and Pianura centres, while the Alpina centre was unaffected. This sudden shift in workload works as a natural experiment: with Alpina as a control group, it is possible to measure how call pressure affects operators and the service, from calls per operator to call handling time, errors and the time for the ambulance to reach the patient.

Share of calls received by each dispatch centre, by week

So whatHow calls are split across dispatch centres is a management decision with measurable effects: lighter load means faster response, heavier load slows it. Redrawing territories should be planned together with staffing, and evaluated on the KPIs that matter to patients.

Which emergency workers are most at risk of burnout and secondary traumatic stress? I analysed the Professional Quality of Life (ProQOL) survey of the regional emergency system, covering physicians, nurses, rescuers and drivers, and dispatch operators, both employees and volunteers.

So whatPrevention should target clinical staff and experienced workers first, with dedicated support for secondary trauma.