The expenses of the Portuguese Public Health Service with doctors working as independent workers won’t stop increasing. In 2018, Portugal spent 104 million euros in these type of contracts — the highest amount ever registered. By collecting and categorizing all the contracts made by public hospitals in Portugal, Rádio Renascença was able to understand how this business works, who is profiting from it and what are the consequences for the Public Health Service. Among 4.987 contracts, our data team was able to find a doctor who worked 100 hours weekly for a year.
Through data analysis and interviewing doctors representatives and hospital administrators, we showed that there are doctors doing excessive amounts of work in some areas in which lack specialists. We found a case of an anesthetist who was working 100 hours weekly during one year, sleeping between surgeries and not resting enough. This doctor earned 50 euros per hour, amounting to 326.000 euros in one year, almost five times more than a senior doctor at the highest level of the medical career.
We also found there are no rules to enter this independent provider system, you just have to be a doctor. As such, there are unprepared doctors, right after college, assuming the urgency service without any supervision.
Another big conclusion we came to was that many hospitals don’t publish their contracts. Portuguese law makes it mandatory that all public contracts should be available online, but there’s a loophole in the law targeting specifically these cases. This opens up questions about transparency because many of these contracts are in the millions, with big recruitment agencies, making it impossible to get a proper overview of the issue and how to report it as a whole.
Our investigation contributed to the debate about the quality of the Portuguese Public Health Service and the work conditions of medical people, bringing an important debate to the public sphere and to other publications. The Health Minister was questioned in the parliament about this investigation and assumed there were “evident problems” in the Public Health Service.
This was the first journalistic investigation in Portugal using data analysis with a 100% transparent methodology: besides writing an article explaining how the data was collected and analyzed, Renascença published the code and all the data in a reproducible and open-source regime.
Since the public authorities refused to deliver the list of contracts that hospitals did with private companies to hire freelance doctors, we had to be creative.
Using BASE, the Portuguese portal where public institutions are obligated to publish all contracts they do with private companies, we discovered there was an non-public API behind the website that allowed us to request all the contracts by their fiscal number. Using R and this non-documented API, we found a way to request all the contracts the 67 public hospitals did between 2011 e 2019 – about 263.000 of them.
We elaborated a big list of words that could immediately disqualify a contract as pertinent to freelancer doctors – words like “blankets” ou “screw” usually mean those contracts were not for doctors. Because we were afraid this methodology could have false negatives, we decided to request all contracts made between these flagged companies and public institutions. We came to a new database of 28.007 contracts that we needed to classify. Since we were afraid of misclassified categories, we used keyword evidence to classify some cases, but then we went by all the 28 thousand contracts to make sure they were not false positives/negatives. For that, some cases required to read the PDF file of the contract.
This way, we were able to get a database of 4.987 contracts that we were sure were about freelance doctors.
Having all the data requested and cleaned, we used R to do some data analysis and find outliers that were the basis of stories published.
What was the hardest part of this project?
The hardest part of the project was the lack of access to public data – in spite of having a portal for transparency, the Portuguese authorities don’t analyze it properly nor share it with journalists. Therefore, we had to collect them and then clean the database, which took long hours to do.
After all, this long task turned out to be useful, because if we had a clean data set to work with, we wouldn’t be able to find the story of the doctor who worked 100 hours per week, for example.
What can others learn from this project?
Even though automatized some parts of the process, this project showed us why the data cleaning process can be really relevant to an investigation. It was during the manual classification process that our journalist was able to find the contract of the doctor who worked 100 hours weekly for one year, bare sleeping.
Because these contracts could refer to one single doctor or 100 doctors – the BASE website requires contracts to tell how much they are going to spend, for how long, and what they are buying, but not the number of people – the case would not be classified automatically as an outlier. But while doing the data cleaning our team had to check the PDF of the contract. That was when we realized it was regarding only one person.
This project also taught us that sometimes public authorities not releasing data makes journalists find the real stories. The investigation started because we didn’t knew who were the biggest companies profiting from the lack of professionals in the National Public Health Service.