2

Israel has one of the fastest Covid vaccine rollouts in the world right now.

To get it so many doses they signed a sweetheart deal with Pfizer to share anonymized, but highly detailed data with Pfizer to let Pfizer figure out how well the vaccine is working. Israel is well-suited as it has both a small population and an advanced health care system.

Some concerns have been raised:

"Your insurance company will know all your medical history. Your employer will know it. The political campaigner who would like to convince you to vote for someone would know everything about your medical history, not to say about people who would like to marry your children," warns Shwartz Altshuler, describing what she calls a small concern.

The context is that:

Nearly every Israeli citizen and resident belongs to one of four public HMOs, a health care system rooted in the national trade union of Israel's early years. Every Israeli's full medical history – from physician visits to hospitalizations – is accessible to any health provider at the click of a mouse, a repository of digital records going back 30 years.

However, in countries like Canada, the UK, or France that have universal health care, what are the actual risks, as long the data is kept safe * and there are laws against employment discrimination for medical disabilities?

By the same token, Iceland has long managed a database of DNA profiles that is very useful to researchers. And, yes, Iceland has universal healthcare.

Not many countries seem to even contemplate such data sharing, even when they are on universal health care.

In the case of Covid, it can only be done once or twice, so Pfizer wouldn't have done such deals with everyone: there could only be 1 or 2 Israels. But in general terms we see very little systematic medical data aggregation for research even when there are no risks to individuals concerning health insurance within a given country.

What are the risks, other than so-and-so has XYZ disease potentially becoming public? Which is a huge risk outside of universal health care systems, admittedly.

* things that come to mind are for example forbidding aggregation with other population-identifying data sources and a bonded deposit for very penalties in case of a data breach, provision for indemnification of individuals whose privacy is compromised. Severe penalties for people disseminating, citing or accessing breached data sets.

2

Deidentification is not just about redacting names and ID numbers and getting everything else. When a company receives de-identified data, they would get a very limited data set meant to analyse something very specific.

One scenario you mention is that a very small number of people would have a particular disease: in that case, a proper deidentification technique will have that disease simply replaced with "Other" before being passed to the company (assuming the disease is actually important at all for this analysis, in which case it wouldn't be sent at all).

In addition to limiting the amount of data being shared, and ensuring that the data that is shared is deidentified to an appropriate level, there will be agreements in place that would strongly disincentivise a company from making it public.

The risk, of course, cannot be reduced to zero. But with sufficient education and proper tooling, it can be reduced to a level where the kind of re-identification you refer to is very small indeed.

Disclosure: I work for a company that offers deidentification solutions.

1
  • +1 I like this answer, but I was less asking about how to protect medical privacy, which you are answering, as, assuming that privacy gets breached, something that as an IT professional myself see as sadly likely... who cares? AIDS, for example, would have been cause for a public shaming before, but nowadays it would likely boomerang on whoever tried to out it. Alcoholism markers, like cirrhosis? Possibly. Point, aggressive and paranoid data cleansing can limit the risk greatly. But IF that fails, what are the really negative outcomes, once you rule out insurance refusals/higher premiums? Feb 28 at 7:31

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .