Covid injections didn’t reduce deaths in care homes but there’s evidence they slightly increased covid deaths, a new study finds

A recently published paper in the European Economic Review estimated the impact of covid “vaccines” on care home mortality using double-debiased machine learning for the first time. 

It found that a high covid “vaccine” booster take-up in elderly care homes did not reduce the covid deaths of residents.  In fact, the study found there was some evidence that covid deaths increased after the rollout of the booster dose.

Sourafel Girma is a professor at the School of Economics at the University of Nottingham and David Paton is a professor at the Nottingham University Business School.

In June 2023 the two researchers published a paper examining the impact of the vaccine mandate for elderly care homes in England on vaccine take-up, staffing levels and mortality.   “Our results suggest strong evidence of harms (lower staffing) from the vaccine mandate and no evidence of mortality benefit,” Paton tweeted when the paper had been published.

The findings of their latest study were similar. Titled ‘Using double-debiased machine learning to estimate the impact of covid-19 vaccination on mortality and staff absences in elderly care homes’, their study was published in the European Economic Review in November 2024. For this study, the two researchers used double-debiased machine learning (“DDML”) to examine the impact of differential covid vaccination rates on care home mortality and other outcomes.

DDML is a statistical technique used to estimate the effect of a treatment or intervention (e.g., a medicine, a policy change) on an outcome (e.g., health, economic growth). It’s called “double-debiased” because it addresses two types of biases that can occur when using machine learning algorithms to estimate the effect of a treatment: overfitting and regularisation bias.

In their paper, Girma and Paton noted that “machine learning is still relatively novel in the context of healthcare in general and vaccination in particular … To our knowledge, our paper is the first to use DDML to estimate the causal impact of vaccination itself on healthcare outcomes in elderly care homes.”

The researchers focused on care homes because in response to the (publicised) significant numbers of covid related deaths in elderly care homes throughout the pandemic, many governments, including in the UK, focused vaccination efforts on care homes including priority allocation of resources, vaccination promotion campaigns and, in many jurisdictions, compulsory vaccination for care workers.

When tweeting about the paper, Paton said: “We examine the impact of vaccination take-up on mortality in elderly care homes across 150 local authorities in England.  We use machine learning to isolate the causal effect of vaccination from other factors like prior immunity, demographics etc.”

He continued: “Care homes are an important context to study as elderly residents were the most vulnerable and, hence, most likely to be able to pick up any benefit.”

The paper stated the study’s objectives: “The key research questions of this paper are whether vaccination efforts in elderly care homes led to reductions in resident mortality and staff absences and, if so, what was the magnitude of such effects.”

One of the analyses the researchers performed was to split the data into two time periods, the initial rollout period for doses 1 and 2, and the booster period for dose 3, the cut-off between the two periods being September 2021 (week 39 of 2021):

• the initial rollout period began from the start of the study’s sample until the end of the primary course rollout (2020 week 23 of 2020 to week 39 of 2021); and,
• the booster period began from the start of the rollout of the booster doses to the end of the study’s sample (week 39 of 2021 to week 26 of 2022).

As Paton noted on Twitter, the key results of their study were:

• Higher staff vaccination did not reduce deaths or staff absences from covid.
• Higher resident vaccination did lead to fewer deaths.  However, the effect was very small and restricted to the initial rollout period, and higher vaccination of residents in the booster period did not reduce deaths.

Not only did the booster injections (the third dose) not reduce deaths but the paper noted that there was evidence that higher vaccination rates are associated with more covid deaths.  The paper said: 

 … we are unable to identify strong evidence that vaccination rates amongst care home staff reduced mortality or that resident vaccination reduced mortality during [the] booster rollout period (from September 2021). Indeed, in the later period, we find some evidence that higher vaccination rates are associated with higher covid mortality.

The increase in covid deaths associated with booster injections is larger than the “very small” reduction in deaths seen after the initial rollout period, but it is small.

The paper concluded:

Using standard panel data regression analysis and doubly debiased machine learning (DDML) techniques suggests both that vaccination had only a limited impact on care home mortality and that any impact was restricted to the initial rollout period of the vaccine.

Our analysis casts doubt on the hypothesis that high rates of vaccination were a particularly important factor in reducing covid mortality after the initial waves. In turn, this has implications for public policy relating to covid vaccination. In particular, it may be appropriate to look again at the case for continuing to expend resources on offering regular booster vaccination doses to vulnerable populations such as care home residents.

Using double-debiased machine learning to estimate the impact of Covid-19 vaccination on mortality and staff absences in elderly care homes., European Economic Review, Volume 170, 2024, 104882, ISSN 0014-2921, https://doi.org/10.1016/j.euroecorev.2024.104882