New variant clustering in households analysis

Summary of request

This analysis uses data from the COVID-19 Infection Survey up to 21 December 2020.

The analysis looks to establish whether there are patterns in the associations of different types of positives within household pairs.

This analysis was undertaken by Thomas House, University of Manchester

Date of publication: 29th December 2020

Contents:

Data:

    1. Survey to date (26 April – 21 December)
    1a. Household pair counts
    1b. Pearsons Residuals
    2. 6 weeks to 21 December 2020
    2a. Household pair counts
    2b. Pearsons Residuals
    3. 4 weeks to 21 December 2020
    3a. Household pair counts
    3b. Pearsons Residuals

Methodology and findings

Data of New variant clustering in households analysis

1. Survey to date (26 April – 21 December) Table 1a.
Household pair counts: counts of pairs of individuals in household with at least one member ever positive by each member of the pairs strain

New Variant Compatible 1924 481 626
Other Positives 9307 4812 481
Negative 14844 9307 1924
Negative Other Positives New Variant Compatible

Table 1b.
Pearsons Residuals: Pearsons residuals compares to the null hypothesis of independence. Reults larger than 2 are indicative of a significant positive association (and less than -2 of a significant negative association)

New Variant Compatible 40 -5 44
Other Positives 85 29 -5
Negative 141 85 40
Negative Other Positives New Variant Compatible

2. 6 weeks to 21 December 2020 Table 2a.
Household pair counts: counts of pairs of individuals in household with at least one member ever positive by each member of the pairs strain

New Variant Compatible 1646 239 802
Other Positives 5300 2276 239
Negative 9756 5300 1646
Negative Other Positives New Variant Compatible

Table 2b.
Pearsons Residuals: Pearsons residuals compares to the null hypothesis of independence. Reults larger than 2 are indicative of a significant positive association (and less than -2 of a significant negative association)

New Variant Compatible 34 -11 49
Other Positives 66 19 -11
Negative 117 66 34
Negative Other Positives New Variant Compatible

3. 4 weeks to 21 December 2020 Table 3a.
Household pair counts: counts of pairs of individuals in household with at least one member ever positive by each member of the pairs strain

New Variant Compatible 1264 97 722
Other Positives 3315 1066 97
Negative 6588 3315 1264
Negative Other Positives New Variant Compatible

Table 3b.
Pearsons Residuals: Pearsons residuals compares to the null hypothesis of independence. Reults larger than 2 are indicative of a significant positive association (and less than -2 of a significant negative association)

New Variant Compatible 28 -13 46
Other Positives 54 8 -13
Negative 94 54 28
Negative Other Positives New Variant Compatible

Methodology and findings

This analysis is based on data from the COVID-19 Infection survey up to 21 December 2020.

The household pair residual analyses, run on the type of positive, was performed for the positivity pattern compatible with the new variant (ORF1ab+N) and all other types of positives. It was run for the entirety of the survey to date (from 26 April), and then separately from 4 weeks ago and from 6 weeks ago.

The new variant of COVID-19 has genetic changes in the S gene. This means the S-gene is no longer detected in the current test, and cases that would have previously been positive on all three genes are now positive only on the ORF1ab and the N gene (not the S gene). There are also other reasons why a swab may be positive for only these two genes, including lower viral load in the sample, which is why we have always seen a small percentage of this type of positive result. Our data suggests the dropping of the S-gene became a relatively reliable indicator of the new variation in COVID-19 from mid-November. Prior to that, the data should not be read as being an indicator of the variant.

The analysis looks to establish whether there are patterns in the associations of these different types of positives household pairs The households with at least one positive test were taken, and pairs of individuals were tabulated by the type of positive for each member of the pair.

The Household pair counts table counts how many times we see each possible pairing, across all pairs of people at household visits (all or in the last 4 or 6 weeks). The Pearsons Residuals tables compare this to a random distribution (taking into account the different sizes of households) – the residual is like observed minus expected, so positive means more likely than chance and negative less likely than chance.

The analysis shows:

– More within-household clustering of positive cases that are compatible with the new variant, measured as deviation from the null of independence, compared to other types of positives. As such it is consistent with more within-household transmission for the new variant but cannot be interpreted causally.

– This is shown in the tables as we observe the “both new variant compatible positives” residual value (top right corner value in Pearsons Residuals table) as being larger than the middle value (both other positives). The larger the residuals the greater the deviation from chance.

– The negative to negative residuals are even larger in each analysis. This is what we expect because negatives do not transmit, and so should cluster together very frequently. More importantly, the new variant compatible – other positive residual is negative, meaning less likely than chance, which is what we would expect if these really are different viruses and so cannot “transmit within household” because they are different.

– The findings are consistent regardless of whether the analysis is done over the last 4 or 6 weeks or over the entire survey, but the associations are strongest over the last 4 weeks. This is consistent with the majority of these ORF1ab+N positive strains now being the new variant.