Predictions from standard epidemiological models of consequences of segregating and isolating vulnerable people into care facilities

Abstract

Objectives: Since the declaration of the COVID-19 pandemic, many governments have imposed policies to reduce contacts between people who are presumed to be particularly vulnerable to dying from respiratory illnesses and the rest of the population. These policies typically address vulnerable individuals concentrated in centralized care facilities and entail limiting social contacts with visitors, staff members, and other care home residents. We use a standard epidemiological model to investigate the impact of such circumstances on the predicted infectious disease attack rates, for interacting robust and vulnerable populations.

Methods: We implement a general susceptible-infectious-recovered (SIR) compartmental model with two populations: robust and vulnerable. The key model parameters are the per-individual frequencies of within-group (robust-robust and vulnerable-vulnerable) and between-group (robust-vulnerable and vulnerable-robust) infectious-susceptible contacts and the recovery times of individuals in the two groups, which can be significantly longer for vulnerable people.

Results: Across a large range of possible model parameters including degrees of segregation versus intermingling of vulnerable and robust individuals, we find that concentrating the most vulnerable into centralized care facilities virtually always increases the infectious disease attack rate in the vulnerable group, without significant benefit to the robust group.

Conclusions: Isolated care homes of vulnerable residents are predicted to be the worst possible mixing circumstances for reducing harm in epidemic or pandemic conditions.

Fig 1. Example epidemic curves for different values of the degree of segregation, x.

Epidemic curves showing the number of new cases per day in population v (vulnerable, minority group, solid lines, left y-axis) and population r (robust, majority group, dashed lines, right y-axis), for different values of x, and for the fixed model parameters indicated above the figure. Inset: attack-rates A_r and A_v as functions of x (coloured circles indicate the x values listed in the main figure legend).

Fig 2. Attack rate contour maps.

Contour maps of A_r (blue lines, see scale at the upper right) and A_v (red lines, see scale at the upper right) for a range of contact frequencies c_r and c_v. Each column of panels corresponds to a different γ_v and each row to a different x, as indicated.

Fig 3. Variation of attack-rates with x, for γ_v = γ_r/4.

Attack-rates A_r and A_v as functions of x, for a range of contact frequencies c_r and c_v, for γ_v = γ_r/4.

Fig 4. Variation of attack rates with x, for γ_v = γ_r/2.

Attack-rates A_r and A_v as functions of x, for a range of contact frequencies c_r and c_v, for γ_v = γ_r/2.

Fig 5. Variation of attack rates with x, for γ_v = γ_r.

Attack-rates A_r and A_v as functions of x, for a range of contact frequencies c_r and c_v, for γ_v = γ_r.