Temporal distribution of Scottish COVID-19 deaths by location
In Figure 4, the time series of deaths due to COVID-19 suggests that the peak of COVID-19 deaths in care homes occurred after the peak was observed in Scottish hospitals. The three highest counts for hospitals were observed between 30th March and 13th April 2020. For care homes, the three highest counts were observed between 6th April and 20th April 2020. The CHCD were observed to be lower than that observed for hospitals except during the period of April 13th to April 27th 2020. From April 27th (up to June 28th), the observed CHCD has been consistently lower than that observed for hospitals.
The data used was updated up to June 28th 2020, week 26.
Figure 4: Time series of care home deaths in care homes compared to other locations in Scotland
COVID-19 deaths by age band and separately by sex (care homes compared with other locations)
Data for weeks 16-19 used (peak of CHCD)
The following graphical illustrations depict the distribution of COVID-19 deaths for care homes (with comparison to other locations such as hospitals) by age band and age band by sex. In Figure 5, the distribution by sex is shown for the data combined. Overall, it is observed that there are more slightly more male cases.
Figure 5: Boxplot of the distribution of the counts of all deaths by sex.
In Figure 6, the distribution of the number of COVID-19 deaths (all locations combined)is shown for each ageband separately for males and females. As the age is increased, more CHCD are observed. There is also an indication of an age-sex interaction for non-COVID-19 deaths; as the age increases the proportion of Females increases.
Figure 6: Boxplot of the distribution of the counts of COVID-19 (and separately, non-COVID-19) deaths by age and sex.
When the data is filtered so that only care home records are considered, strikingly higher counts for females are observed from ages 75+ for both COVID-19 and non-COVID-19 deaths. An age-sex interaction effect is evident for both COVID-19 and non-COVID-19 deaths. A recent study by Burton et.al (2019) indicate that 68% of residents in Scottish care homes are female median age in care homes is 84. This study was conducted using data from the Scottish Care Home Census from 1st April 2012 to 31st March 2016.
Figure 7: Boxplot of the distribution of the counts of COVID-19 deaths (and separately non-COVID-19) by age and sex for care homes.
When considering only hospital deaths, the age-sex interaction for COVID-19 cases is not as distinct as in that for care homes. There is a predominance however of male cases for most age groups. For non-COVID-19 deaths, there is a marked age-sex interaction with more cases occurring as age increases and at the highest age category, there are more females. This is the only category where more females are observed.
Figure 8: Boxplot of the distribution of the counts of COVID-19 deaths (and separately for non-COVID-19) by age and sex for hospitals.
T-tests and modelling
Data for weeks 16-19 used (peak of CHCD)
T-tests conducted (Table 1) for each age range (by sex) reveal that there were is a statistically significant difference between the observed CHCD between Females and Males at age 85+. In Table 2, we note that there were no statistically signifcant differences observed in COVID-19 deaths by ageband (between Females and Males) and overall (Females-Males) for hospital COVID-19 deaths.
When the CHCD were modelled using a Poisson glm, statistically significant effects were observed at age band levels of 75-84 and 85+, and a significant interaction effect (age x sex) was revealed. The results are in Table 3.
| age | mean.diff | F | M | t | p | df | CI low | CI high |
|---|---|---|---|---|---|---|---|---|
| 45-64 | 0.5 | 3 | 2.5 | 1 | 0.39 | 3 | -1.09 | 2.09 |
| 65-74 | -1.75 | 10.25 | 12 | -0.53 | 0.61 | 5.87 | -9.81 | 6.31 |
| 75-84 | 0.75 | 49.25 | 48.5 | 0.11 | 0.92 | 4.78 | -17.71 | 19.21 |
| 85+ | 65.75 | 120.5 | 54.75 | 6.88 | 0 | 5.31 | 41.61 | 89.89 |
| age | mean.diff | F | M | t | p | df | CI low | CI high |
|---|---|---|---|---|---|---|---|---|
| 45-64 | -10.75 | 11.5 | 22.25 | -2.78 | 0.05 | 4.03 | -21.47 | -0.03 |
| 65-74 | -14.25 | 17.25 | 31.5 | -2.59 | 0.05 | 4.54 | -28.84 | 0.34 |
| 75-84 | -18.75 | 29.75 | 48.5 | -1.48 | 0.22 | 3.47 | -56.08 | 18.58 |
| 85+ | -5 | 32.75 | 37.75 | -0.55 | 0.6 | 5.92 | -27.14 | 17.14 |
| Estimate | p | 2.5% | 97.5% | |
|---|---|---|---|---|
| (Intercept) | 2.33 | 0.00 | 2.00 | 2.62 |
| sexM | 0.16 | 0.46 | -0.26 | 0.58 |
| age45-64 | -1.23 | 0.00 | -2.19 | -0.45 |
| age75-84 | 1.57 | 0.00 | 1.24 | 1.92 |
| age85+ | 2.46 | 0.00 | 2.16 | 2.80 |
| sexM:age45-64 | -0.34 | 0.54 | -1.42 | 0.81 |
| sexM:age75-84 | -0.17 | 0.46 | -0.64 | 0.29 |
| sexM:age85+ | -0.95 | 0.00 | -1.40 | -0.50 |
| Note: | ||||
| R squared: | ||||
| 0.958699004531425 |
Discussion
The graphical descriptive summaries indicate differences in CHCD by age and sex; these are confirmed by Student’s t-tests and model esimates from a Poisson generalised linear model. It is worth noting, that when considering CHCD’s the human and epidemiological significance of a single CHCD outweighs that of statistical signifcance.
The main message which emerges from the analyses is that, when care homes are considered, there is potentially increased susceptibility amongst elderly (85+) female residents at care homes. In particular, our model estimate for the interaction component indicates that the difference in the expected number of CHCD’s as age is increased from 65-74 to 85+ is lowered when considering male residents (vs. female residents).
When considering only age, our model estimates indicate that increasing age to 85+ (from 65-74) results in an expected increase in CHCD by 12.
There is an indication that spatial neighbourhood transmission is a factor which affects the COVID-19 deaths in care homes.
Finally, more fine grained analyses will be generated as soon as more in-depth data on the epidemiology of COVID-19 in care homes is obtained. The availability of relevant linked data (Hanratty et.al 2020) would be valuable in similar studies.
It is important to note that the recorded deaths in this paper refers to that recorded by the NRS and includes deaths where Covid-19 was suspected as the cause of death. Note the records provided by other centers may record deaths only for cases where there was a positive Covid-19 test.
References:
- Hanratty, Barbara and Burton, Jennifer Kirsty and Goodman, Claire and Gordon, Adam L and Spilsbury, Karen, Covid-19 and lack of linked datasets for care homes, vol 369,2020 BMJ Publishing Group Ltd,https://www.bmj.com/content/369/bmj.m2463.full.pdf,BMJ.
- Burton Jennifer K , Lynch Ellen, Love Sharon, Rintou Julie, Starr John M , Shenkin Susan D ,Who lives in Scotland’s care homes? Descriptive analysis using routinely collected social care data 2012-16,2019,J R Coll Physicians Edin, doi: 10.4997/JRCPE.2019.103