Skip to main content

Publication, Part of

Health and Care of People with Learning Disabilities Standardised Mortality Ratio Indicator

Experimental statistics, Other reports and statistics

Current Chapter

Supporting Information


Supporting Information

About this release

This release comprises a three-year rolling indicator measuring the national and local standardised mortality ratio of the learning disabilities population.

It is understood that people with learning disabilities have a shorter life expectancy than that of the general population. A number of the causes of mortality amongst this population are thought to be premature and preventable. This indicator is a useful contributor to monitoring the rates of mortality in the learning disabilities population and is intended to help in improving health outcomes and life expectancy in this group.

The methodology behind the indicator has been assessed by NHS Digital’s Methodology Review Group and the Indicator Governance Board. As such it has been assured and entered into the National Library of Quality Assured Indicators.

Interpreting this indicator

This indicator compares the number of deaths occurring in the learning disabilities population aged 0-74 years to those that would be expected for people of the same age and sex characteristics in the general population. The comparison is made at a national level (England) and is also broken down into different areas of the country.

A score of 100 would indicate that the number of deaths for people with learning disabilities would be the same as that for people with the same age and sex characteristics in the general population. A score of 200 would mean it was twice as high, then 400 would be four times as high and that trend would continue. For areas with smaller population or lower coverage, chance will have a greater influence on the score of an area so confidence intervals are also presented.

What the data can tell you

The primary use for this indicator is to compare a locality's number of deaths in the learning disabilities population with the number of deaths that would be expected, based on figures taken from the general population.

The 'indicator score' is a standardised mortality ratio, calculated as observed deaths divided by expected deaths multiplied by 100 (detailed construction methodology can be found in the Methodology section further down this page).

Health bodies and commissioning organisations can use this indicator as a smoke alarm to investigate mortality outcomes in specific localities.

What the data cannot tell you

This indicator should not be used to directly compare mortality outcomes between localities and it is inappropriate to rank localities according to their indicator score. Year-on-year comparison of a specific area is not possible due to the method of construction and calculation.

It is important to note that the difference between the number of observed deaths and the number of expected deaths cannot be interpreted as the number of avoidable deaths. Whether or not a death could have been prevented can only be investigated by a detailed case note review. It is not a direct measure of quality of care in a locality.

Further information on the review of deaths amongst the learning disabilities population is undertaken by the English Learning Disabilities Mortality Review (LeDeR) programme, the most recent annual report presents information about the deaths of people with learning disabilities aged 4 years and over notified to the programme between 1 July 2016 - 31 December 2018, it can be found here.

The expected number of deaths for each locality is not an actual count of deaths or patients, but it is a statistical construct which estimates the number of deaths that may be expected in the locality on the basis of average England figures in age and sex specific bands.

Time breakdown

This indicator uses a three-year pooled mortality figure. This means that the population total is taken at the beginning of the first year of the indicator period (the earliest collection being 2014), while the deaths total is taken for an entire three-year period (in this case the earliest is 2014 to 2017).

The indicator is published on an annual basis, as the Health and Care of People with Learning Disabilities data is extracted annually.

Geographical breakdowns

National (England)

Region

Regional Local Office (RLO)

Sustainability and Transformation Partnership (STP)

Clinical Commissioning Group (CCG)


Data Quality

Accuracy and reliability

There are two major gaps in the data. System suppliers play an important role, translating the specifications into code for their systems and managing the submission of queries to practice systems and collecting responses. There are gaps in this extract as data were not gathered from all GP practices. Not all practices using EMIS, INPS and Microtest have had their data collected and no practices using TPP have had their data collected. The second gap arose because practices had to agree to participate in the data collection.

This leaves the total national coverage for this indicator at approximately 50%.

Data quality is generally considered to be very good for all GPES extracts. The data that is collected is 100 per cent complete, as it is an automated extract GPES extracts data for all patients who have specified codes on their record.

Timeliness and punctuality

For the Health and Care of People with Learning Disabilities publication, data from the previous year (1st April – 31st March) is extracted from GPES each year.  This extraction period usually lasts 4 days.  We then aim to publish by December of that year.  For example, 2017-18 data is extracted in October.  Analysis begins in October 2019 and we aim to publish in December 2019.

Integrity

The integrity of the data used was assessed thoroughly. A number of checks were carried out to ensure the data used were internally consistent, for example:

  1. No individual indicator counts exceeded the count of the relevant denominator
  2. The total number of patients registered at a general practice equalled the sum of the number of patients recorded on the general practice's learning disability register and the number of patients who are not recorded on the general practice's learning disability register e.g. LDOB001 = (LDOB003a+LDOB003b)
  3. The sum of the number of patients registered at a general practice broken down by age band was equal to the "all ages" aggregation within the data

Learning Disabilities Population Figure Choice

There are two learning disabilities population measures within the Health and Care of People with Learning Disabilities dataset. LDOB002 and LDOB003a, the preference would have been to use LDOB002 throughout this indicator, however this was not possible due to inconsistencies with the collection of this measure, as stated below.

LDOB002

Data for LDOB002 'The number of patients recorded on their general practice's learning disabilities register at the reference year start date' in fact only included patients who were on their general practice's learning disabilities register at both the reference year start date and reference year end date. This field therefore systematically under-represents the true value. As such these data were not used for the production of this indicator.

Unallocated data

Within each CSV, there will be some fields labelled 'UNALLOCATED'. 

‘UNALLOCATED’ refers to GP Practices that could not be historically mapped to current NHS geographies but are included in the England total.

In any current publication the aim is to provide the most up to date mapping. Where practices close or Clinical Commissioning Group borders move for example, it is sometimes not possible to map these historic practices to current localities.


Accessibility

Accessibility and clarity

All data produced are available on the website. The publication comprises a CSV containing all indicator data for the specified year.

The data within the CSV files are presented in columns, a description of the contents of each column is as follows:

YEAR - Indicator coverage in years, for example 2014-17

ORG_TYPE - Type of organisation, for example Region or Clinical Commissioning Group (CCG)

ONS_CODE - The Office for National Statistics geographical identifier, identifying the specified health geography

CODE - The NHS specified Organisation Data Service (ODS) code, identifying the specified health geography

NAME - The name of the specified health geography

MEASURE - This column identifies which data value lies within the corresponding VALUE column

VALUE - This column contains the data values. These are: the learning disabilities population figure for the specified health geography, observed number of deaths for the specified health geography, expected number of deaths for the specified health geography, the indirectly standardised mortality ratio (ISR), the lower confidence interval of the ratio and the upper confidence interval of the ratio

Assessment of user needs and perceptions

Comments and questions can be submitted to NHS Digital general enquiries by emailing [email protected] or by telephone on 0300 303 5678.

NHS Digital consult with NHS England and The Department of Health and Social Care on a regular basis and make adjustments to the publication based on their feedback. 

Confidentiality, transparency and security

The Health and Care of People with Learning Disabilities publication is subject to a standard NHS Digital risk assessment prior to issue. Disclosure control is implemented where judged necessary.

Detailed methodology specification documents and other supporting material are available on our website.

The UK Statistics authority code of practice for statistics is followed regarding security and release of information prior to publication.


Methodology

Data sources

Health and Care of People with Learning Disabilities Dataset

The reason for use of this dataset is that no other source identifies both the subject population and corresponding mortality data for this specific population with age and sex stratification on a national scale.  The ideal alternative would be linkage between GP learning disabilities register and mortality data.  This would provide additional data about causes nationally but this is not currently available.  

Coverage in the Learning Disabilities and Care Dataset is just over half of the population of England (around 50%), with CCG boundaries, with variable coverage between CCGs. Data is 100% complete from practices and systems that do contribute data, further information on this is found in the Accuracy and Reliability section. Further background information on this dataset and associated publication can be found here.  

Primary Care Mortality Database

This database is required for use within this indicator to calculate the rate of mortality in the general population. The Primary Care Mortality Database (PCMD) holds mortality data as provided at the time of registration of the death along with additional GP details, geographical indexing and coroner details where applicable. It is managed by NHS Digital and is accessed via a secure SQL database. 

The PCMD currently holds data on deaths starting from 1997 up until the most recent monthly data extract (monthly data are added in the second week of the following month). Further background information on the data can be found on the Office for National Statistics website and on the NHS Digital website.

Construction

Calculation of the general population reference mortality rate

Numerator

All deaths from any cause for ages 0 to 74 at time of death are included in the numerator for the general population reference mortality rate (taken from the PCMD extract) throughout the three year period (for example 1st April 2014 - 31st March 2017. The data are based on the registered date of death. 

Denominator

The general population used here is derived from the health and care of people with learning disabilities dataset (population measure LDOB001), using only people aged 0 to 74. This population count is taken at the beginning of the first year (for example in 2014 - 2017 the population count would be as at 1st April 2014) 

The simple reference mortality rates are then calculated as follows:

\(r=\frac{O_i}{n_i}\)

Where:

\(r_i\) is the age and sex specific rate in the population in the age and sex group \(i\).

\(O_i\) is the observed number of events (deaths) in the population in age and sex group \(i\).

\(n_i\) is the number of individuals in the subject population in age and sex group \(i\).

Calculation of the standardised mortality ratio (the indicator value)

The standardised mortality ratio (SMR) represents the indicator value. It is an indirectly standardised ratio.  

The SMR is calculated by using the general population age and sex specific reference mortality rates to get expected deaths in the learning disabilities population, then comparing the observed deaths with the expected deaths. The age groups used are 0-9, 10-17, 18-24, 25-29, 30-34, 35-39, 40-44, 45- 49, 50-54, 55-59, 60-64, 65-69 and 70-74.

The SMR is calculated as follows:

\(SMR=\frac{O}{E}\times100=\frac{\Sigma_iO_i}{\Sigma_iE_i}\times 100=\frac{\Sigma_iO_i}{\Sigma_i\eta_i\lambda_i}\)

Where:

\(O_i\) is the observed number of deaths in the learning disabilities population in age and sex group \(i\).

\(E_i\) is the expected number of deaths in the learning disabilities population in age and sex group \(i\).

\(\eta_i\) is the number of individuals in the learning disabilities population in age and sex group \(i\).

\(\lambda_i\) is the age and sex specific rate in the general population in age and sex group \(i\).

Calculation of the confidence intervals

95% confidence intervals are used to identify whether a difference in mortality rate is or is not statistically significant. When calculating 95% confidence intervals for an indirectly standardised ratio such as this, it is assumed that the standard or reference rates come from a population sufficiently large as to assume their sampling variance is negligible, and that the observed number of events (\(O\)) follows a Poisson distribution.

95% confidence intervals were calculated as follows:

\(ISR_{lower}=\frac{O_{lower}}{E}\)

\(ISR_{upper}=\frac{O_{upper}}{E}\)

Where:

\(O_{lower}\) and \(O_{upper}\) are the lower and upper confidence intervals for the observed number of events.

 

Using Byar's method, the confidence limits for the observed number of events are given by:

\(O_{lower}=O\times\bigg(1-\frac{1}{9O}-\frac{z}{3\sqrt{O}}\bigg)^3\)

\(O_{upper}=(O+1)\times\bigg(1-\frac{1}{9(O+1)}+\frac{z}{3\sqrt{(O+1)}}\bigg)^3\)

Where:

\(z\) is the 100(1-\(\alpha\)/2)th percentile value from the Standard Normal Distribution (for 95% confidence interval \(\alpha\) = 0.05 and \(z\) = 1.96)


Last edited: 18 December 2019 10:03 am


Pages in this publication