Abstract
The Italian social security administration (INPS) in 2017 was given the task to perform home visits (HV) on public sector employees claiming sickness benefits. Inps decided to replicate a setup similar to that already operating in the private sector. Thus, it was decided to begin by randomly selecting the certificates potentially subject to home visits in order to maximize the informational content of the data collected in the home visits.
The experiment took place in the 45 days between November 22, 2017 and January 5, 2018. It assigned the HV treatment through a “typical” stratified randomized experiment, following a five-step procedure. Visits were performed over 29 working days out of the 45 of the experiment period. The main steps, were as follows:
1. On each working day from November 22, 2017 until January 5, 2018, INPS drew daily, for each local office, a random sample of certificates from the universe of absence certificates (around 400,000 certificates).
2. Exempted certificates, i.e. those involving chronic or very serious disease, recent surgery, etc., were excluded.
3. Among the selected sample, a second random sample was drawn. Its size was determined in such a way that for each worker subject to HV there were an additional eleven workers in the control group. This sample, made up of about 60,000 certificates, was then used for the experiment.
4. Selected certificates were randomly ranked, and the resulting order defined the HV priority.
5. Visits were assigned to doctors optimizing travel time. During the period of the experiment INPS performed about 4,200 HVs, that is, about 145 visits per day.