Independent report

Appendix 1: UKHSA report estimating the number needed to vaccinate to prevent COVID-19 hospitalisation for booster vaccination in autumn 2023 in England

Updated 5 October 2023

In this report the UK Health Security Agency (UKHSA) estimates the number needed to vaccinate (NNV) to prevent a hospitalisation, severe hospitalisation and a death due to COVID-19. This is to inform decisions around use of booster vaccination in autumn 2023. The data relevant to these calculations is incidence of these outcomes by vaccination status as well as incremental vaccine effectiveness of additional doses and how this wanes. Stratification by risk group is shown using data in the National Immunisation Management System (NIMS). The analysis focuses on ages 15 and over and stratifies by age in 5-year bands to over 90 years.

Methods

Outcomes

The outcomes were:

• hospitalisations: secondary user services (SUS) hospital admissions from 15 November 2022 to 14 January 2023, with a respiratory code in the primary discharge field and with a positive PCR or lateral flow test in the period 14 days before to 2 days after admission. In addition, any individuals without evidence of a test but with an International Classification of Diseases (ICD) code for COVID (U071-COVID virus identified and U072-COVID epi diagnosis) in the primary diagnostic field
• severe hospitalisations: at least a 2 day stay and with codes to indicate use of oxygen, ventilation or ICU admission
• death: ONS coded deaths with COVID-19 listed as one of the causes

Vaccination status

The outcomes are linked to NIMS and vaccination status as of 14 December 2022, and categorised as:

• unvaccinated
• 1 or 2 doses
• 3 or more doses with no autumn 2022 bivalent booster
• 3 or more doses which included an autumn 2022 bivalent booster

Covariates

From NIMS, age (as of 31 December 2022) and risk group were obtained. Risk group was defined as having a flag for any of the risk group fields in NIMS as follows:

• no risk: no risk group flags
• risk with no immunosuppression: at least one risk group flag but not a flag for immunosuppression
• immunosuppressed: a flag for immunosuppression or severely immunosuppressed

The full list of flags followed the Cohorting as a Service (CaaS) (created autumn 2022 for the booster) cohorts:

• asplenia
• chronic heart or vascular disease
• chronic kidney, liver or digestive disease
• chronic neurological disease
• chronic respiratory disease
• diabetes and endocrine disease
• morbid obesity
• severe mental illness
• serious genetic abnormalities
• immunosuppression
• severely immunosuppressed

The latter 2 are used for immunosuppression.

Vaccine effectiveness of booster doses

Vaccine effectiveness (VE) of the autumn bivalent booster dose during the period when the rates were calculated (15 November 2022 to 14 January 2023) can be used to adjust the rates in those with a bivalent booster to the expected rate in the absence of a booster (for example, the rate after this additional protection has waned). This was calculated using the test-negative case-control (TNCC) design and gave a VE of 37% (95% CI: 32 to 41) for any time post booster when using an end point of at least a 2 day stay which is the end point used by UKHSA for most VE analyses.

To estimate the incremental effectiveness (iVE) of a further booster dose, the recent VE estimates of the bivalent booster were used along with data from the spring 2022 booster for which longer follow up is available (in those not boosted again) to assess waning. This allowed estimates of iVE by month since a further boost with a 6-month window.

The VE estimates based on the TNCC run on 16 March 2023 for the hospital end point with at least a 2 day stay were as follows:

• 2 to 4 weeks post bivalent vaccine: 53% (95% CI 48 to 57)
• 5 to 9 weeks: 47% (43 to 50)
• 10 to 14 weeks: 36% (32 to 40)
• greater than 15 weeks: 28% (21 to 35)

See recent published estimates

Based on this and data from the autumn 2022 booster the iVE assumed by month since booster was assumed to be as follows for all end points:

• month 1: 40%
• month 2: 50%
• month 3: 40%
• month 4: 30%
• month 5: 20%
• month 6: 20%

Method to estimate NNV

To estimate NNV the monthly incidence was assumed to be:

• over 50 years of age or in a risk group: 15 November 2022 to 14 January 2023 incidence observed in the bivalent boosted population increased by a factor of: 1 divided by (1 minus 0.37) equals 1.587 (that is 1/(1 - 0.37) = 1.587) to allow for the effectiveness of the bivalent booster in this period and assuming this would wane prior to further doses
• aged 15 to 49 and not in a risk group: 15 November 2022 to 14 January 2023 incidence in those with more than 3 doses and not boosted with a bivalent booster

Due to small numbers in some ages the incidence calculated above was smoothed using a 4-knot cubic spline and the fitted values used. When combining across no-risk and risk to get an overall rate (called ‘all’ in the figures and tables) the predicted rate comes from combining across each of the predicted counts within the no-risk and 2 risk groups rather than a separate model fitted to these data.

The NNV was then calculated by applying the iVE to this incidence for a 6-month period post a booster dose. For example, an incidence of 10 per 100,000 per month would be reduced with vaccination by:

• 10 times 0.4 equals 4 in month 1
• 10 times 0.5 equals 5 in month 2
• 10 times 0.4 equals 4 in month 3
• 10 times 0.3 equals 3 in month 4
• 10 times 0.2 equals 2 in months 5 and 6

This accumulates to observing 20 per 100,000 population vaccinated fewer cases which can be inverted to an NNV of 100,000 divided by 20 equals 5,000.

Results

Monthly incidence

The observed and modelled monthly incidence is shown for each end point in figures 1a to 1c. Note that for observed rates those with zero cases are shown as 0.5 cases for hospitalisations and 0.01 for severe and death. Also, for severe hospitalisation where modelled rates drop below 0.01 per 100,000 a value of 0.01 is regarded as the lowest possible value.

Figure 1a: hospitalisation rates by age and risk group status

Figure 1b: severe hospitalisation

Figure 1c: death

Notes:

• ‘all’ means combining across no risk, risk and immunosuppressed (IS) risk
• squares are the observed data and the line is the fitted spline
• ‘IS risk’ means immunosuppressed risk group

The figures show that hospitalisation rates (figure 1a), severe hospitalisation rates (figure 1b) and death rates (figure 1c) increased with increasing age in the whole population (‘all’ group). The increasing age trend was also seen in each when people were separated into 3 groups:

• those with no clinical risk
• those with a clinical risk but no immunosuppression
• those who are immunosuppressed

NNV estimates

NNV estimates are shown in tables 1a to 1c. As expected NNV decreases with risk and age. It is notable NNV is lower for death than for severe hospitalisation. This likely relates to deaths occurring outside hospital and possibly older ages in hospital dying before oxygen, ventilation or ICU use. It may also be that some ONS coded deaths were not due to COVID-19.

Notes on tables 1a to 1c:

• ‘IS’ means immunosuppressed
• ‘all’ means combining no risk and risk groups

Table 1a: NNV estimates for a hospitalisation by age and risk status

Age (years)	No risk	Risk (not including IS)	IS	All
15 to 19	48,300	4,100	480	27,000
20 to 24	43,700	4,300	550	29,700
25 to 29	39,900	4,500	640	26,400
30 to 34	36,900	4,600	710	23,800
35 to 39	34,900	4,500	780	21,600
40 to 44	34,000	4,400	820	19,100
45 to 49	34,100	4,100	810	15,800
50 to 54	33,900	3,600	770	6,600
55 to 59	31,900	3,100	700	5,200
60 to 64	27,100	2,400	600	3,800
65 to 69	19,800	1,800	490	2,600
70 to 74	12,100	1,200	390	1,600
75 to 79	6,300	800	290	940
80 to 84	3,000	490	210	540
85 to 89	1,300	290	150	310
Over 90	530	170	110	180

Table 1b: NNV estimates for a severe hospitalisation by age and risk status

Age (years)	No risk	Risk (not including IS)	IS	All
15 to 19	5,000,000	177,200	41,500	2,960,500
20 to 24	5,000,000	134,400	38,100	2,862,700
25 to 29	5,000,000	102,200	34,600	1,689,000
30 to 34	2,847,500	78,100	30,800	1,032,100
35 to 39	1,612,400	60,300	26,600	644,900
40 to 44	1,022,200	47,000	22,000	404,700
45 to 49	735,300	37,100	17,300	246,100
50 to 54	572,600	29,500	13,200	76,500
55 to 59	453,900	23,400	9,800	53,000
60 to 64	344,100	18,400	7,200	35,400
65 to 69	234,600	14,300	5,300	23,200
70 to 74	137,400	10,800	4,000	15,000
75 to 79	70,300	8,100	3,000	9,600
80 to 84	32,600	5,900	2,300	6,300
85 to 89	14,100	4,300	1,800	4,200
Over 90	5,900	3,100	1,400	3,000

Table 1c: NNV estimates for death by age and risk status

Age (years)	No risk	Risk (not including IS)	IS	All
15 to 19	2,065,100	1,630,900	22,900	1,486,200
20 to 24	1,872,800	770,500	26,100	1,476,200
25 to 29	1,687,400	373,700	29,000	1,238,300
30 to 34	1,500,800	191,000	30,700	984,700
35 to 39	1,309,100	105,600	30,200	730,800
40 to 44	1,112,600	64,900	27,000	496,500
45 to 49	911,700	44,800	21,400	301,000
50 to 54	701,100	33,300	15,400	87,900
55 to 59	490,500	25,200	10,200	56,500
60 to 64	302,600	18,400	6,300	33,600
65 to 69	159,600	12,200	3,800	18,400
70 to 74	70,600	7,000	2,200	9,100
75 to 79	26,900	3,600	1,300	4,200
80 to 84	9,100	1,700	740	1,800
85 to 89	2,900	740	420	790
Over 90	870	310	240	330

Comments

Assumptions in this calculation include:

• bivalent iVE wanes back to the pre-booster protection level by the time of future doses. In practice some incremental immunity may remain for those revaccinated at about 6 months. This will lead to slight under estimation of NNV
• the 15 November 2022 to 14 January 2023 rates are a best approximation of future rates as they should reflect the recent immune state of the population (past infections and so on). Nevertheless, future rates are likely to differ. A reasonable uncertainly would be doubling to halving the rates (and hence NNV)
• new variants will not change the iVE appreciably