Skip to main content

Table 1 Summary of findings and grading of quality of evidence

From: The effectiveness of non-exposure to incarceration in preventing COVID-19 and mitigating associated events: a systematic review and meta-analysis

Non-incarceration compared to incarceration for populations aged over 18

Patient or population: population aged over 18

Context: prevention of COVID-19 and mitigation of associated events

Intervention: non-incarceration

Comparison incarceration

Outcomes

No of participants (studies)

Certainty of the evidence (GRADE)

Relative effect (95% CI)

Potential absolute effects

Risk with incarceration

Difference in risk with non-incarceration

COVID-19 incidence: number of COVID-19 cases/population

2,613,319,533

(13 observational studies)

Lowa, b,c

RR 0.39

(0.34 to 0.45)

12,268 per 100,000

7,483 less per 100,000

(8,097 less to 6,747 less)

Sex- and age-adjusted COVID-19 mortality: relative risk

(5 observational studies)

Very lowa, c,d

RR 0.36

(0.27 to 0.49)

-

-

Intensive care admission rate: number of admissions to ICU/total number of hospital admissions

627,432

(3 observational studies)

Very lowa, d,e

RR 0.91

(0.74 to 1.13)

31,620 per 100,000

2,846 less per 100,000

(8,221 less to 4,111 more)

Use of Invasive mechanical ventilation: number of patients using IMV/total hospital admissions

627,432

(3 observational studies)

Very lowa, c,d

RR 0.77

(0.70 to 0.84)

19,093 per 100,000

4,391 less per 100,000

(5,728 less to 3,055 less)

COVID-19 hospital mortality: number of deaths/number of people admitted to hospital

627,432

(3 observational studies)

Very lowa, c,d

RR 0.81

(0.54 to 1.23)

15,119 per 100,000

2,873 less per 100,000

(6,955 less to 3,477 more)

  1. CI: Confidence interval; MD: Mean difference; RR: Risk ratio
  2. Explanations
  3. a. To assess risk of bias we used the Robins-E, a suitable tool for evaluating observational studies of exposure. None of the included studies controlled for length of incarceration, indicating high risk of bias
  4. b. The funnel plot (S2, Fig. 2.5). Egger’s test (p= 0.654)
  5. c. We downgraded data inconsistency by two levels
  6. d. As less than 10 studies described the outcome, it was not possible to test for presence of publication bias; bias was therefore considered highly likely
  7. e. We downgraded data inconsistency by one level