Estimates of opiate and crack use in England: main points and methods
Published 24 October 2023
Applies to England
© Crown copyright 2023
This publication is licensed under the terms of the Open Government Licence v3.0 except where otherwise stated. To view this licence, visit nationalarchives.gov.uk/doc/open-government-licence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: psi@nationalarchives.gov.uk.
Where we have identified any third party copyright information you will need to obtain permission from the copyright holders concerned.
This publication is available at https://www.gov.uk/government/publications/opiate-and-crack-cocaine-use-prevalence-estimates/estimates-of-opiate-and-crack-use-in-england-main-points-and-methods
Main points
National prevalence trends
These estimates of opiate and/or crack users (OCU) were produced by the Office for Health Improvement and Disparities and the UK Health Security Agency, using a revised methodology from the one used to produce the previous Estimates of opiate and crack cocaine use prevalence: 2016 to 2017 published by Public Health England.
There is an increase of over 13,000 OCU between the previous estimate for 2016 to 2017 (313,971) and the updated estimate for 2016 to 2017 (324,840). This difference is a result of the revised methodology, rather than actual changes in the prevalence of opiate and crack use. You can read more about the revised estimation methodology in the section below.
Based on the revised method, the national trend figures between 2016 to 2017 and 2019 to 2020 indicate a continuous upward trend of 5%. Figure 1 shows the increase of a little over 16,000 OCU, from 324,840 in 2016 to 2017 to 341,032 in 2019 to 2020.
There are confidence intervals for each of these estimates, which you can see in in the data tables for each year that accompany this publication. Since all the 95% confidence intervals around each year’s prevalence estimate overlap, we cannot say that the upward trend is definitive.
Figure 1: OCU prevalence estimates in England for 2016 to 2017, 2018 to 2019 and 2019 to 2020
| Year | Count |
|---|---|
| 2016 to 2017 | 324,840 |
| 2017 to 2018 | 0 |
| 2018 to 2019 | 336,531 |
| 2019 to 2020 | 341,032 |
Note: there is no source data for estimation available for 2017 to 2018.
National and regional prevalence rates
If the prevalence estimates are set in relation to the size of the population of England in the same period, the rate of OCU per 1,000 general population increased from 9.2 in 2016 to 2017, to 9.4 in 2018 to 2019, and to 9.5 in 2019 to 2020.
The North East, Yorkshire and the Humber and the North West had the highest OCU rates in 2019 to 2020 among all English regions. The rates in all 3 regions are notably higher than the England national average of 9.5. The lowest rates were in East of England and the South East. Table 1 shows the rates for all English regions.
Table 1: OCU prevalence rates per 1,000 general population by region in England for 2016 to 2017, 2018 to 2019 and 2019 to 2020
| Region | 2016 to 2017 | 2018 to 2019 | 2019 to 2020 |
|---|---|---|---|
| East Midlands | 8.5 | 8.0 | 8.0 |
| East of England | 6.6 | 6.8 | 6.7 |
| London | 9.9 | 10.6 | 10.9 |
| North East | 12.6 | 13.0 | 13.4 |
| North West | 11.1 | 11.8 | 11.9 |
| South East | 6.9 | 6.4 | 6.6 |
| South West | 8.4 | 8.9 | 8.8 |
| West Midlands | 9.4 | 9.6 | 9.6 |
| Yorkshire and the Humber | 11.2 | 12.2 | 12.0 |
| England | 9.2 | 9.4 | 9.5 |
Prevalence by substances used
The number of people who used both opiates and crack has increased by 24% from 2016 to 2017 to 2019 to 2020. In the same period, the number of people who only used crack (and not opiates) rose by 19% and the number of people who used opiates but not crack decreased by 9%. Table 2 shows a breakdown of rates by substances.
Table 2: breakdown of OCU prevalence numbers and rates by substance for 2016 to 2017, 2018 to 2019 and 2019 to 2020
| Substance group | Estimate (and rate) 2016 to 2017 | Estimate (and rate) 2018 to 2019 | Estimate (and rate) 2019 to 2020 |
|---|---|---|---|
| Opiates and crack | 104,864 (3.0) | 119,731 (3.4) | 129,584 (3.6) |
| Opiates only | 180,281 (5.1) | 171,168 (4.8) | 164,279 (4.6) |
| Crack only | 39,694 (1.1) | 45,632 (1.3) | 47,168 (1.3) |
| OCU total | 324,840 (9.2) | 336,531 (9.4) | 341,032 (9.5) |
Note: the total of all substances may not be the same as the OCU total due to rounding.
Prevalence by substance group and region
Opiates were the most misused substance in all regions of England in 2019 to 2020, either on their own or with crack. The 2 opiate groups (opiates only and opiates and crack) had a combined rate of more than 8.2 per 1,000 general population. The group of people who misused crack without opiates had a rate of 1.3 OCU per 1,000 general population.
In 2019 to 2020, the highest rate of opiate-only users was in the North East with 9.8 per 1,000 population. This was more than double the national average rate of 4.6. Yorkshire and the Humber and the North West followed with 6.4 and 6.1 respective opiate-only users per 1,000 population. London, the North West and Yorkshire and the Humber had the highest rates of crack use, either on its own or with opiates.
Table 3: rates per 1,000 general population by substance group and region, 2019 to 2020
| Region | OCU | Opiates and crack | Opiates only | Crack only |
|---|---|---|---|---|
| East Midlands | 8.0 | 2.9 | 4.1 | 1.0 |
| East of England | 6.7 | 3.0 | 2.7 | 1.0 |
| London | 10.9 | 4.8 | 4.2 | 1.9 |
| North East | 13.4 | 2.0 | 9.8 | 1.6 |
| North West | 11.9 | 4.3 | 6.1 | 1.5 |
| South East | 6.6 | 2.8 | 2.9 | 0.9 |
| South West | 8.8 | 3.2 | 4.5 | 1.2 |
| West Midlands | 9.6 | 4.1 | 4.4 | 1.1 |
| Yorkshire and the Humber | 12.0 | 4.2 | 6.4 | 1.5 |
| England | 9.5 | 3.6 | 4.6 | 1.3 |
Note: substance group rates may not add up to the same total as the OCU rate due to rounding.
Prevalence by age group
In 2019 to 2020, more than two thirds of the total estimated number of OCU were aged between 35 and 64. Less than 8% (25,094) of OCU are estimated to be between 15 and 24 years old.
Since 2016 to 2017, there has been a 13% increase in people aged 35 to 64 years old. This increase, combined with the declining numbers in the younger age groups, might reflect an ageing population of OCU.
Based on the 2019 to 2020 OCU estimates, the North East had the highest OCU rates for the 15 to 24 age group (6.8 per 1,000 population) and the 25 to 34 group (18.5 per 1,000 population). Both rates are much higher than the national average rate in both age groups. Only London comes close to the North East rate for the 15 to 24 years old, but is still notably lower (4.3 per 1,000 population). In the 35 to 64 age group, the highest OCU rates were in the North West, Yorkshire and the Humber, the North East and London.
Table 4: OCU prevalence estimates and rates per 1,000 general population by age group and region, 2019 to 2020
| Region | Estimate (and rate) 15 to 24 years | Estimate (and rate) 25 to 34 years | Estimate (and rate) 35 to 64 years | Estimate (and rate) all age groups |
|---|---|---|---|---|
| East Midlands | 1,903 (3.2) | 5,922 (9.7) | 16,575 (9.0) | 24,400 (8.0) |
| East of England | 1,994 (3.0) | 6,449 (8.4) | 17,502 (7.2) | 25,945 (6.7) |
| London | 4,488 (4.3) | 14,674 (9.1) | 47,791 (13.7) | 66,953 (10.9) |
| North East | 2,203 (6.8) | 6,467 (18.5) | 13,952 (13.7) | 22,622 (13.4) |
| North West | 3,427 (3.9) | 10,377 (10.5) | 41,620 (15.0) | 55,424 (11.9) |
| South East | 3,286 (3.2) | 9,756 (8.9) | 24,826 (6.90) | 37,868 (6.6) |
| South West | 2,460 (3.9) | 7,498 (11.3) | 20,322 (9.5) | 30,280 (8.8) |
| West Midlands | 2,640 (3.6) | 8,184 (10.2) | 24,799 (11.3) | 35,623 (9.6) |
| Yorkshire and the Humber | 2,692 (3.9) | 8,755 (12.0) | 30,468 (14.7) | 41,915 (12.0) |
| England | 25,094 (3.8) | 78,082 (10.3) | 237,856 (11.0) | 341,032 (9.5) |
Prevalence by sex
Over the period covered by the analysis (between 2016 and 2020), the prevalence estimates for male and female OCU show a very stable male (79%) to female (21%) ratio with only very little variation between regions.
However, in 2019 to 2020, the North East, Yorkshire and the Humber, and the North West had considerably higher OCU rates than the English national average for both male and female OCU. See figure 2 for more information of the prevalence rates by sex for each region.
Figure 2: OCU prevalence rates per 1,000 general population by sex and region, 2019 to 2020
| Region | Men | Women |
|---|---|---|
| East Midlands | 12.63 | 3.41 |
| East of England | 10.65 | 2.84 |
| London | 17.21 | 4.5 |
| North East | 21.22 | 5.69 |
| North West | 18.97 | 4.96 |
| South East | 10.4 | 2.84 |
| South West | 13.9 | 3.74 |
| West Midlands | 14.96 | 4.11 |
| Yorkshire and the Humber | 19.08 | 5.01 |
| England | 15.07 | 4.01 |
Methodology
Data sources for the OCU estimate modelling
The modelling uses 3 data sources:
- National Drug Treatment Monitoring System (NDTMS) information on people in community drug treatment.
- Criminal justice system (CJS) information, combining data from the Home Office’s Police National Computer, the Ministry of Justice’s Offender Assessment System, and NDTMS data on people in drug treatment in prisons.
- Drug-related mortality information from the Office for National Statistics’ (ONS) data on deaths registered in England.
Estimation method
Capture-recapture technique
To produce these estimates, we used a technique known as ‘capture-recapture’. The capture-recapture technique partly evolved from estimating the levels of hidden wildlife populations and is one of the most widely used in estimating hidden populations.
The basic premise of the capture-recapture technique is that individuals may be in multiple linked data sources. The technique uses information about how individuals are observed in different data sources to estimate the number of people who are not observed in any of them. To produce OCU prevalence estimates, we identified people across the 3 data sources using probabilistic matching (a statistical approach to measure the probability that 2 records represent the same person). We applied this approach across the used data sources by matching a person’s:
- first initial of their first name and surname
- date of birth
- sex
- local authority or region of residence
If there is a large amount of overlap between data sources, this implies that most of the total population has been captured and the hidden population is relatively smaller. However, if only a small proportion overlap, this implies there is a larger hidden population that is only rarely observed in each source.
The basic model structure
The possible 8 different combinations (A to H) of being observed or not observed in the 3 data sources are listed in table 5. The combinations A to G in represent the observed number of people in each local authority and combination H represents the number of OCU not observed in any of the 3 data sources.
Table 5: combinations of observed and unobserved records across used data sources
| Combination | Observed in NDTMS data | Observed in CJS data | Observed in ONS mortality data | Observed population |
|---|---|---|---|---|
| A | Yes | No | No | Yes |
| B | Yes | Yes | No | Yes |
| C | Yes | Yes | Yes | Yes |
| D | No | Yes | No | Yes |
| E | No | Yes | Yes | Yes |
| F | No | No | Yes | Yes |
| G | Yes | No | Yes | Yes |
| H | No | No | No | No |
To model the dependence structure in the observed cells (combinations A to G), we fitted a single statistical model (a Poisson model) to all local authorities using:
- the observed counts in combinations A to G
- information on a person’s age, sex, substance use and injecting status
- the number of the general population in each local authority by age and sex
We used the resulting model to estimate the number of people not observed (combination H). The total of all calculated combination counts (A to H) is the estimated prevalence for each local authority.
Additions to the basic model structure
Our model includes interaction terms between data sources. This allows for dependencies between sources, which helps us to estimate whether a person appearing in one source is more or less likely to appear in another source (for example, people in contact with the criminal justice system may be more likely to enter treatment).
We used the same basic model structure across all local authorities but where indicated by the data, we included random effects to allow for local differences. These random effects model if there are local differences for the chance of being in treatment, or if there are local variations in the composition of the age groups.
The random effects we have included in the model account for variation by:
- overall OCU prevalence in the population (estimated baseline prevalence)
- relative proportions of crack or opiate only use
- current and past injecting
- age group between local authority areas
Also, we included variation in the proportions observed in community drug treatment, the criminal justice system and drug-related mortality sources in the modelling. The random effect terms allow for the effect of shrinkage. This is where data shows a difference to the national average. So, estimates can pull away from the estimated baseline prevalence, but where data is sparse (and potentially subject to random variation) estimates shrink towards the estimated baseline prevalence.
Since the model estimates all OCU sub-populations (by age group, sex, substance use and injecting behaviour) jointly for each local authority, we can aggregate the OCU numbers to higher level OCU numbers (for example at a regional or national level).