Skip to main content
SearchLogin or Signup

Review 1: "Disparities in COVID-19 Related Mortality in U.S. Prisons and the General Population"

This study draws much-needed attention to the higher COVID-19 mortality burden among US prison populations, however, reviewers raised several methodological concerns involved in the authors' calculations of the scale of the disparities.

Published onNov 06, 2020
Review 1: "Disparities in COVID-19 Related Mortality in U.S. Prisons and the General Population"
1 of 2
key-enterThis Pub is a Review of
Disparities in COVID-19 Related Mortality in U.S. Prisons and the General Population

We provide an analysis of COVID-19 mortality data to assess the potential magnitude of COVID-19 among prison residents. Data were pooled from Covid Prison Project and multiple publicly available national and state level sources. Data analyses consisted of standard epidemiologic and demographic estimates. A single case study was included to generate a more in-depth and multi-faceted understanding of COVID-19 mortality in prisons. The increase in crude COVID-19 mortality rates for the prison population has outpaced the rates for the general population. People in prison experienced a significantly higher mortality burden compared to the general population (standardized mortality ratio (SMR) = 2.75; 95% confidence interval = 2.54, 2.96). For a handful of states (n = 5), these disparities were more extreme, with SMRs ranging from 5.55 to 10.56. Four states reported COVID-19 related death counts that are more than 50% of expected deaths from all-causes in a calendar year. The case study suggested there was also variation in mortality among units within prison systems, with geriatric facilities potentially at highest risk. Understanding the dynamic trends in COVID-19 mortality in prisons as they move in and out of hotspot status is critical.

RR:C19 Evidence Scale rating by reviewer:

  • Potentially informative. The main claims made are not strongly justified by the methods and data, but may yield some insight. The results and conclusions of the study may resemble those from the hypothetical ideal study, but there is substantial room for doubt. Decision-makers should consider this evidence only with a thorough understanding of its weaknesses, alongside other evidence and theory. Decision-makers should not consider this actionable, unless the weaknesses are clearly understood and there is other theory and evidence to further support it.




STRENGTH OF EVIDENCE: potentially informative


This paper reports on rates of COVID-19 related death in US prisons in the first 6.5 months of 2020. Key findings are that (a) the overall rate of COVID-19 related death was 2.75 times higher than in the age- and sex-matched general population, and (b) there was marked variation in the rate of COVID-19 related death between states. Correctional settings are potentially important incubators for COVID-19, and the authors call for (a) COVID-19 mitigation in prisons, (b) better data to inform this, and (c) compassionate release for those at greatest risk. These are sensible (although not novel) recommendations.

Key limitations: The authors calculate COVID-19 related CMRs, but the appropriate denominator is person-years. The authors do not explain what denominator they used, but it was presumably a proxy. If the proxy was the number in prison on a given day, this is problematic since people cycle in and out of prisons. Since the entire paper rests of these numbers, clarification and justification of the method used to calculate CMRs is essential.


A structured Abstract would further improve clarity.


Why were prison staff not in scope for this work, given that the CPP publishes data on both prisoners and staff?

The manuscript cites 31 publications, all US-based. The paper is not well situated within the large and rapidly growing international literature on COVID-19 in prisons.


-        Did the states that did not report COVID-19 mortality (a) not report to the CPP, or (b) report that no deaths had occurred?

-        Specifically, what “additional national data” were obtained? Since the authors used “the latest year that data is publicly available”, these are evidently not COVID-19 mortality data.

-        Additional detail on data analysis is required. “Standard epidemiologic and demographic estimates” is inadequate. In particular, further details on the methods used to calculate CMRs and SMRs is essential.

The authors need to define “COVID-19 related death” (ICD codes, and any other means). Did the authors make a distinction between deaths where COVID-19 was the underlying cause, and those where it was a contributing cause?


Can the authors confirm that there were 97 COVID-19 related deaths on April 22, and not by April 22? Given that there was an average of 48 deaths per week after that, I assume they mean by April 22. If data on deaths up to April 22 were available, why did the authors not examine time trends from the beginning of the year?

It is not particularly meaningful to report a “roughly 600% increase” in COVID-19 related deaths from April 22 to July 15. The average of 48 deaths per week would be more informative if it could be contextualized, for example by indicating what proportion of all deaths in prison this represented (with reference to Figure 3).

Figure 1 is problematic for multiple reasons:

1.     Y axes are inadequately labelled.

2.     Denominator for prison CMR is unclear. Number of people in prison on a given day is probably unsuitable as a proxy. The appropriate denominator is person-years. Was this used? If not, what proxy was used, and on what basis is this considered adequate?

3.     Due to marked age, sex, and race differences between the prison and general populations, meaningful interpretation of a comparison between CMRs is complex. As such, simplistic interpretations such as that “Beginning in early May, the COVID-19 death rate in prisons began to outpace the general population rate” are arguably inappropriate. Perhaps these should be age-adjusted CMRs?

“…the SMR was not significant on April 25” – I think the authors mean that it was not significantly different from 1.

An SMR of 1.59 indicates that the rate is 59% higher, not 159% higher. It is probably clearer to report that the rate is “1.59 times higher.”

In text the CMRs on July 11 are stated as 50 and 40 respectively, but Figure 1 suggests slightly lower CMRs on this date.

Figure 2 — 95%CI should be included for all states.

Figure 3 — expressing COVID-19 deaths (presumably over a 6.5 month period in 2020?) as a proportion of all deaths in 2016 is flawed. First, rates of death in prison may change over time. Second, what does it mean to express COVID-19 deaths in 6.5 months of 2020 as a proportion of all deaths in 2016? Third, interpretation of the state-by-state comparison depends on the underlying rate of all-cause death in prisons in that state, as well as the rate of death due to COVID-19. Fourth, 95% CI should be added for all comparisons.

Given that 21/104 units in the TJDC had COVID-19 deaths, but 45% of deaths occurred in 3 units, some brief information on the outbreak response might be instructive. The authors may wish to refer to WHO interim guidance for COVID-19 response in prisons, summarized here:

Kinner et al (2020). Prisons and other custodial settings are part of a comprehensive response to COVID-19. Lancet Public Health, 5, 188-89.

The TX case study hints at higher case and fatality rates among staff, in units with high prisoner case and fatality rates. If staff death data are available for other states via the CPP, could the authors explore the association between prisoner and staff death rates? 


The finding that geriatric facilities were at highest risk is not at all surprising. The authors may wish to comment on the implications for the aging prison population in the US.

One limitation not acknowledged is potential differences between states in classifying deaths as COVID-19 related. Apparent differences in COVID-19 related deaths may in part be a function of differences in COVID-19 testing and/or ME/C practices.

Authors conclude that mitigation will require “higher quality data both within and across systems.” What sorts of data, and in what ways are currently available data limited? What are the key elements of a “focused approach to COVID-19 mitigation in prisons,” with reference to relevant literature?


The paper is written assuming a US-only audience (e.g., use of state abbreviations). Some minor edits would make it easier to follow, and more relevant, for readers outside of the US, particularly those who may not be familiar with US correctional systems.


No comments here