Was Alpha deadlier than wild-type COVID? Analysis in rural England

The dataset described COVID + patients with medical needs or occupational exposure (Pillar 1 status, described in more detail below) treated within the English county of Norfolk and a single district (Waveney) in the adjacent county of Suffolk. Only Pillar 1 cases are described in the analysis in this article, not cases found under any other testing framework. Provision of health care in this combined area was concurrently under the commissioning powers held by the Norfolk and Waveney clinical commissioning group (NWCCG). NWCCG was only permitted to receive and share records of patients registered with NWCCG general practice surgeries. Norfolk and Waveney is a coastal and predominantly rural area in Eastern England, UK that extends approximately 40 × 55 miles. The population is approximately 1 million. The median age of Norfolk residents is 46 years which compares to a median age ~ 40 years for all UK residents in mid-2018 [9]. Comparisons elsewhere showed that Norfolk and Waveney is fairly representative of rural areas of England (UK) with regard to population deprivation indicators, air quality, road network access to employment centres and population density [10]. Permission to analyse these records for epidemiological purposes was granted by our Faculty of Medicine and Health Sciences Ethics Research Ethics Committee, their reference 2019/20–127.

The supplied dataset comprised 8784 unique records of patients who had COVID-19 Pillar 1 positive test results and who received a + COVID Pillar 1 test as reported by National Health Service (NHS) Trusts in the Norfolk and Waveney administrative region through 31 March 2021. Supplementary Table 1 lists the NHS Trusts who provided Pillar 1 records to NWCCG. Case counts and 28-day mortality outcome data were available complete for all patients in all trusts through 31 March 2021 except for the Queen Elizabeth Hospital (QEH); the QEH data feed was unavailable after 17 Feb 2021. Therefore, with regard to mortality outcomes among QEH patients, we ignore cases identified after 17 January 2021. This truncation point allows for 28-day mortality outcomes plus 3 extra days in case of late recording. Partial omission of the QEH data does not bias our analysis of relative risk factors for mortality outcome, because the QEH data contribution is relatively small and we focus on outcomes after diagnosis, rather than outcomes with full population as denominator. Cases detected under the Pillar 1 framework were tested for possible COVID-19 because of medical need for urgent treatment or occupational exposure [11]. The dataset did not distinguish those tested for medical treatment needs from people with occupational risk. We believe most of the records relate to persons with medical need, because 56% of the records were for persons age 65 or older, beyond the recent average age of retirement in England [12], while 74% of the records were for persons age 50 +. The data were collected, cleaned and provided to us by NWCCG. The dataset generally reported which NHS Trust requested the test, residence area resolved to lower super output area (LSOA) geography, age, sex, date that COVID-positive swab was taken and date of death when applicable.

All patients had recorded the date of their + swab test. Information about home residence area for each COVID + patient was available for most records (85%), resolved to LSOA. LSOAs are standard census units in England for which socioeconomic and other indicators are often calculated. LSOAs are designed to be fairly consistent in population but not geographic size. LSOAs typically each contain about 650 households [13]. For each LSOA we accessed the Index of Multiple Deprivation 2019 score (IMD2019) [14]. The IMD2019 is a nationally standardised ranking of relative deprivation, which were categorised here as quintile ranks. Each quintile contains 20% of all LSOAs in England, with most deprived rank = 1 to least deprived rank = 5.

We used only the data records that were complete for all of these attributes: age, sex and residential area. Supplementary Table 1 lists the full number of records received and the number of records that contributed to the descriptive summaries reported in this article.

The data were analysed in two separate time periods. The first period with high case counts (“wave”) was considered to comprise cases swabbed in the period from 1 March to 31 May 2020 inclusive. The second wave was considered to include cases in the period (inclusive) 1 October 2020 to 28 Feb 2021. The dataset was extracted on 8 April 2021 to allow for delayed reporting of outcomes. There were 1997 unique Pillar 1 cases in the first wave, and 6388 unique Pillar 1 cases in the second wave.

We characterise the N and W Pillar 1 populations with regard to demographic traits (age and sex) and deprivation profile over time and between waves. Given the emergence of variants of concern in the UK in the autumn of 2020 [15], and the initially unclear implications of new COVID-19 variants for patient outcomes [8], we were interested in whether the age, sex or deprivation profiles in cases or deaths substantially differed between the two wave periods. Norfolk was identified using genomic analysis by end December 2020 as having a relatively large proportion of cases that were coronavirus novel variant of concern B1.1.7 by the middle of the second wave period [see Supplementary Fig. 1; B.1.1.7 variant comprised 45% of all community samples sequenced in Norfolk in December 2020; 16].

We were interested in assessing risk factors linked to mortality outcomes within this patient group, and whether the mortality risk factors changed over time. Some concurrent national data suggested declining mortality rates among hospitalised COVID-19 patients in the second wave period, and that these improvements were greatest for persons age 70 + (see Supplementary Fig. 2). We restrict the mortality analysis to deaths that occurred no later than 28 days after + swab date, which is appropriate and in line with concurrent national practice about identifying COVID-attributable deaths in the absence of specific causes of death that may be stated on death certificates [17]. Concurrent causes of death on death certificates for N&W in this period were not available to us. We expected that the cases and deaths would both be strongly skewed towards males and older adults. We therefore consider the proportions of the cases and deaths that were male, adults in specific age bands (< 40 years old, 40–49, 50–59, 60–69, 70–79, and 80 +) and in specific deprivation quintiles in the NWCCG area. Possible differences in the age distributions of cases or deceased were formally tested using the Mann-–Whitney U test.

We expected the UK vaccination programme to potentially change the age-related risks for mortality following COVID-19 diagnosis. However, determining when an age-related change in mortality statistics might be observable is complex. Community vaccination against SARS-CoV-2 infection began in the UK on 8 December 2020. By 15 February 2021, all health or social care workers and persons age 70 + had been offered a first dose of one of the licensed SARS-CoV-2 vaccines in England [23]. Exactly when vaccination might reduce COVID-19 mortality depended on the speed of the vaccination rollout, which groups received the vaccine first, time elapsed for immune system response, likely delay from exposure to any mortality and social distancing behaviours following vaccination.

Other analyses have found little evidence of immune system protection before 10–12 days post-vaccination, but definite reduced risk of severe illness by 14 days after a single vaccination dose, depending on the specific vaccine product. The estimated risk reductions for severe disease following single doses of the vaccines used in the UK through February 2021 was high, at 66.7% [Vaxzevria; 24] and 90% [Pfizer/BioNTech COVID-19 BNT162b2 vaccine: 25].

The median latent period for SARS-CoV-2 is about 5 days [26]. Of those in our NWCCG cohort who died following Pillar 1 COVID + swabs, 37% died by day 7 afterwards, 50% by day 10 and 80% by day 24. Hence, we observe that most within-28 day mortality happened by day 14 following + swab. Vaccination in the highest risk groups (age 70 +) in NWCCG did not exceed 50% before early February 2021 (see Supplementary Fig. 5). Immediate (within 3 weeks after first vaccine dose) reductions in social distancing precautions were reported by about 40% of persons age 80 + [27]. Consequently, given the vaccine programme start date, likely delay times to mortality outcomes and common behavioural changes post-vaccination, we expect any reduction in mortality from COVID-19 in the NWCCG dataset to not be ascertainable within or before January 2021 data. Fewer cases and subsequent deaths in persons age 70 + might be possible from the start of February 2021, but would be much more confidently expected in March 2021 and later (after our wave periods ended).

Our dataset did not contain information about the ethnicity of patients. Ethnic diversity is quite low in N and W, especially among older adults who are most at risk of severe illness or death from COVID-19 (age 65 +). 96.5% of all-age Norfolk residents self-identified as ‘White’ in the 2011 national census [28]. Assessing ethnic diversity as part of characterising the N and W waves was unlikely to be informative.

Comorbid diagnoses such as diabetes or dementia were available for some but not all patients in our dataset. It is possible that there were changes in case identification or mortality over time or between waves that we did not find, but could have been detected with more complete and detailed individual patient data. Our comparisons between waves are descriptive rather than quantitative precisely because of the lack of prognostic indicators (such as baseline morbidity). There is debate about under-ascertainment of COVID + deaths due to lack of testing and misattribution, especially in the first wave period; we do not have data to assess if our sample was biased with this problem. Similarly, it has been reported [29] that hospitalisations of relatively younger persons were more common in the second UK COVID-19 wave than in March–May 2020. The exact dates of hospitalisation for individual patients were not available in our dataset so we could not consider risk of hospitalisation.