Research and analysis

Methods of the TB in inclusion health groups in England, 2021: supplementary analysis

Published 17 January 2024

Tuberculosis (TB) notifications

People who are diagnosed with TB in England, Wales and Northern Ireland must be notified through the National TB Surveillance (NTBS) system. In 2021, the NTBS was launched and replaced 2 historical surveillance systems, the systems were called the:

• Enhanced Tuberculosis Surveillance (ETS) system
• London TB Register (LTBR)

Data set production

This analysis includes data for individuals with TB who are resident in England or are treated in England (including individuals who are homeless or visiting from abroad). An exception is for the calculated rates of TB in people with a history of imprisonment, which includes people notified as resident or treated in England or Wales.

Data are from the cleaned and national data set for 2021 as used in the TB incidence and epidemiology in England, 2022 (data up to end of 2021) report and limited to notifications from 2018. Data sets from 2018 onwards were extracted from ETS and LTBR and merged with NTBS following a series of data migrations between July and December 2021. Data reported here were obtained from the merged data sets (NTBS, ETS, LTBR).

Data cleaning to improve data quality

Detailed data cleaning methods to improve data quality of denotifications, geography, site of disease, demographic and clinical characteristics can be found in the methods and definitions section of the TB incidence and epidemiology in England, 2022 (data up to end of 2021) report.

Social risk factors (SRFs) among TB cases

The presence or absence of SRFs (current or a history of drug misuse, alcohol misuse, homelessness, prison, mental health and asylum status; including if remanded in an immigration detention centre) were updated from missing or unknown if relevant information was found in the ‘comments’ variables within NTBS.

Homelessness was updated to ‘yes’ if mentioned in the ‘comments’ variables or if the address given was ‘no fixed abode’ or a shelter or hostel for homeless people was named.

Prison (current or in the past) was updated to ‘yes’ if mentioned in the ‘comments’ variable, if HMP (Her Majesty’s Prisons) or a prison name was recorded as the address or if the residential postcode corresponded with a prison. Up until 2020, data on incident TB cases reported to the Public Health in Prisons (PHiP) log were used to further identify people who had been imprisoned, but this was not conducted in 2021.

The immigration detainee variable (newly introduced in NTBS) was updated if the ‘comments’ variable or occupation field showed the person to be an immigration detainee. The asylum seeker variable (also introduced in NTBS) was updated as asylum seeker if it was recorded in the free text occupation sub-category of ‘no occupation’. For analysis, asylum seeker was then recorded as ‘yes’ if either asylum seeker variable or immigration detainee variable was ‘yes’. The asylum seeker variable was further updated so that all UK-born individuals with a missing value for this variable was updated to ‘no’.

Reporting methodology

Individuals with TB are reported by area of residence and by calendar year of notification.

Time periods

For this extended analysis of SRFs data are presented from 2018 or 2019 onwards, when more complete data collection started.

Detailed reporting methods for overall TB rates, geographical location, disclosure control methods and Index of Multiple Deprivation (IMD) can be found in the TB incidence and epidemiology in England, 2022 (data up to end of 2021) report.

Proportions of people with SRFs

People with TB are reported as having at least one SRF (‘yes’) if any of the 6 SRFs (current alcohol misuse, current or a history of homelessness, drug misuse, imprisonment, asylum seeker status and current mental health needs) had ‘yes’ recorded. When calculating the proportion of cases with at least one SRF, the denominator used is all notifications. This assumes that people for whom no data were recorded for individual SRFs were a ‘no’ and may result in underestimation. When calculating the proportion of people with 2 or more SRFs data are limited to notifications with a ‘yes’ or ‘no’ value for at least 2 SRFs. Thus, the denominator for this group is smaller than for the proportion with at least one SRF.

Data for individual SRFs reported are limited to those with recorded data, for example, a ‘yes’ or a ‘no’. As a result, the denominators for these are smaller than all notifications due to missing data. If there is significant under-reporting of SRFs in those with missing data, this should result in a better estimate of the true proportion of the people with each SRF. However, if data is more likely to be recorded if the response is a ‘yes’ this could result in overestimates. This may be the case for the asylum seeker SRF.

Mental health is recorded by TB case managers and is based on their judgement if mental health concerns are likely to affect the person’s ability to adhere to treatment. This was added to surveillance in the London UK Health Security Agency (UKHSA) region in 2018 and is a simple ‘yes’ or ‘no’ response. It was introduced nationally in 2021 with the introduction of NTBS. Here we report this as the person requires support for mental health and therefore has ‘mental health needs’.

Asylum seeker status and immigration removal centre were added to national surveillance as discrete variables in 2020. Prior to this, ‘asylum seeker’ status was extracted from ‘comments’ variable and the user entered values within occupation (LTBR). As a result, more complete data on this exposure is observed from 2020 to 2021 compared with previous years.

Alcohol misuse is recorded by case managers and is based on their judgement if current alcohol misuse is likely to affect adherence to treatment.

History of drug misuse, homelessness and prison are self-reported by individuals and are first asked as a ‘yes’ or ‘no’ response and then with additional information on duration; as current, within last 5 years or more than 5 years ago. Unless indicated otherwise, analyses here present these SRFs as ‘yes’ if either history of, or a duration value, was recorded. Data dictionary definitions for SRFs within NTBS are shown in Table 1 below.

Table 1. Social risk factors data dictionary definitions

Field name	Definition	Values	Further information
Patient’s ability to self-administer treatment is affected by alcohol misuse or abuse	Is patient’s ability to self-administer treatment affected by alcohol in the absence of directly observed treatment (DOT)?	Yes/No/Unknown	Not applicable
Patient has a history of drug misuse	Is the patient misusing drugs, or has misused drugs in the past?	Yes/Current/In the past 5 years/More than 5 years ago/No/Unknown	Problem drug use is defined as illicit injecting drug use or long duration/ regular use of illicit opiates, cocaine and/or amphetamines.
Patient has a history of homelessness	Is patient currently homeless or has been in the past?	Yes/Current/In the past 5 years/More than 5 years ago/No/Unknown	Includes people sleeping rough, living in temporary accommodation such as bed and breakfast, hostels, hotels, and squats and people who are involuntarily dependent on friends.
Patient has a history of imprisonment	Is the patient a prisoner or been a prisoner in the past?	Yes/Current/In the past 5 years/More than 5 years ago/No/Unknown	Include any history of incarceration within the UK including incarceration in a prison or a remand centre.
Patient has mental health problem	Does the persons have a mental health problem likely to affect their ability to effectively take anti-TB treatment?	Yes/No/Unknown	Most likely with major mental health such as schizophrenia or depression/anxiety but could be more minor problems if their TB treatment may be affected in the opinion of the clinical team.
Patient is an asylum seeker	Is the patient seeking asylum in the UK?	Yes/No/Unknown	Not applicable
Patient is an immigration detainee	Is the patient currently being held in an immigration detention centre?	Yes/No/Unknown	Not applicable

Sources of population data for calculation of TB notification rates in at-risk population groups in 2021

TB notification rates in the general population in this report have been estimated using the recently released population estimates for 2021, updated using the 2021 population census and released in December 2022 from the Office for National Statistics (ONS). Further methods can be found in the TB incidence and epidemiology in England, 2022 (data up to end of 2021) report. All rates are presented as per 100,000 population.

Sources of population data for estimating TB notification rates in the ‘at-risk’ prison population in England and Wales 

The population ‘at risk’ can be separated into prison and probation populations (as separate or combined denominators) and are reported for England and Wales combined and not for England alone.

The 2021 TB notification rate in the ‘current’ prison population is estimated using the prison population mid-year estimate for 2021, obtained from UK government (Prison population statistics). The current prison population is formed of 4 main custody categories: remand, sentenced, recall, and non-criminal. Detailed prison population definitions can be found in the Guide to Offender Management Statistics.

The 2021 TB notification rate for people with current imprisonment or current and recent imprisonment (previous 5 years) is estimated using:

• prison populations using the mid-year population estimates (for each year of the 5-year period 2017 to 2021), obtained from UK government (Prison population statistics)
• the prison population in 2021 plus the probation population in 2021 (this includes supervised offenders given community sentence orders and suspended sentence orders; detailed probation population definitions can be found in the Guide to Offender Management Statistics)

Sources of population data for estimating TB notification rates in the homeless population in England

The annual TB notification rate for 2021 in the homeless population has been estimated using the 2021-point prevalence (one night) estimate for all homeless individuals in England, obtained from Shelter’s annual Homelessness in England Report 2021. The total homeless population is formed of 4 categories: people in temporary accommodation, street sleepers, people in hostels, and those in social services accommodation. When compared to alternative homelessness data sources, the data from Shelter most closely align with UKHSA’s definition of homelessness in NTBS (Table 1).

Detailed methods for how the total prevalence estimate is calculated can be found within the above Shelter report. Sources include:

• Department for Levelling Up, Housing and Communities (DLUHC)
• statutory homelessness statistics
• Ministry of Housing, Communities and Local Government (MHCLG) rough sleeping data
• Homeless Link’s Annual review of single homelessness support in England 2020, November 2021
• ONS, mid-year population estimates 2021

Sources of population data for estimating TB notification rates in those who misuse drugs in England

The annual TB notification rate in those who currently misuse drugs is calculated using 2020 ONS drug misuse statistics. These include a prevalence estimate of adults (age 16 to 59 years) in England who have taken any class A drug in 1) in the last year, and 2) in those with any history of misusing a class A drug. Any class A drug comprises powder cocaine, crack cocaine, ecstasy, lysergic acid diethylamide (LSD), magic mushrooms, heroin and methadone plus methamphetamine.

Drug misuse statistics (produced by ONS) examine the extent and trends in illicit drug use among a sample of residents aged 16 to 59 years from households in England and Wales. Detailed methodology and source material of these estimates can be found on GOV.UK.

Sources of population data for estimating TB notification rates in asylum seekers in England

The annual TB notification rate in asylum seekers is estimated using the total number of people seeking asylum in each year from 2018 to 2021. Asylum summary statistics are produced by the home office and obtained from UK government statistics. Detailed methodology and source material of these estimates can be found in Asylum applications, initial decisions and resettlement.

Statistical methods

Estimating relative risk of SRFs in people notified with TB and of clinical outcomes in people with and without SRFS notified with TB

Risk ratios and associated confidence intervals are model-derived using the binomial distribution for proportions. Methods for calculating risk ratios can be found in TB incidence and epidemiology in England, 2022 (data up to end of 2021).

Estimating TB rates in England, 2021

All 95% confidence intervals were model-derived and calculated using assumptions of the Poisson distribution for rates. For more details please see the methods section in TB incidence and epidemiology in England, 2022 (data up to end of 2021).

Rates and 95% confidence intervals for asylum seekers were generated using 1,000 bootstrap resamples and the 25th and 975th values and the model allowed for increasing uncertainty in and increasing undercount with increasing year of entry in asylum seeker denominator numbers (up to 15% undercount in 2021).

Estimating TB incidence rates in those with SFRs

Estimating TB rates in prison population in England and Wales

An annual TB notification rate is calculated for those recorded as ‘currently in prison’ by dividing the numerator (the number of TB notifications reporting ‘current prison history’, in that year) by the denominator (the mid-year prison population estimate for the same year) and multiplying by 100,000.  This method probably underestimates the numerator by under reporting cases and may not include those on remand.  The rate is probably an underestimate resulting in underestimation of TB cases.

An annual rate per 100,000 for people who were in prison in the 5-year period from 2017 to 2021 can be calculated for those recorded as ‘currently in prison’ and ‘in prison less than 5 years ago’ by dividing the numerator (the number of TB notifications in the 5-year period) by the denominator (the sum of the mid-year prison population estimates for the same 5-year period) and multiplying by 100,000. This method may also underestimate the numerator and likely overestimates the denominator population as it does not account for people imprisoned for more than one year. The rate is likely to be an underestimate.

As a simple alternative, the rate for this group was also calculated using themed-year prison population in 2021 plus the probation population in 2021, resulting in a smaller denominator than in the first method (and avoiding the problem of trying to estimate the effect of long term prisoners and those with multiple short sentences), but which should capture all those that have recently been in prison. However, the probation population will only capture those that were sentenced and then released on probation. This rate may be an under or overestimate.

Estimating TB incidence rates in asylum seekers adults in England

TB notification rate of asylum seekers (all ages) notified in 2021 who entered the UK within the previous 4 years (2018 to 2021) was calculated per 100,000 by dividing the numerator of total number asylum seekers recorded in NTBS by the denominator (the sum of asylum applications made in 2018, 2019, 2020 and 2021). Those with missing data for time since entry were included in the TB notified within 3 years of entry group (2018).  The number of asylum applications includes both main applicants and dependents.

Estimating TB incidence rates in the homeless (all ages) in England, 2021

The 2021 TB notification rate in the homeless was calculated by dividing the numerator (the number of TB notifications reporting ‘current homelessness in 2021) by the denominator (the annual point prevalence of homelessness for 2021) and multiplying by 100,000.

Estimating TB incidence rates in adults with current drug misuse in England, 2020

The 2021 TB notification rate in adults (aged 16 to 59 years) who currently misuse drugs was calculated by dividing the numerator (the number of TB notifications in 2021 in those aged 16 to 59 years reported as currently misusing drugs) by the 2020 prevalence estimate of adults (aged 16 to 59 years) in England who have taken any class A drug in the last year and multiplying by 100,000. This rate may be an underestimate due to the denominator population capturing those that have used any class A drug in the previous year, whilst the numerator population is specific to people with misuse, indicating dependence on just 3 types of class A drugs.

Estimating TB incidence rates in adults with any lifetime drug misuse in England, 2020

The 2021 TB notification rate in adults (aged 16 to 59 years) who have ever misused drugs is calculated by dividing the numerator (the number of TB notifications in 2020 reported as having ever misused drugs) by the 2020 prevalence estimate of adults (age 16 to 59 years) in England who have ever taken any class A drug and multiplying by 100,000. This is likely an underestimate due to the denominator capturing people who have ever used any class A drug, whilst the numerator population is specific to people with misuse indicating dependence, rather than occasional use.

Software packages

All statistical analysis was carried out using Stata 17.0.