Report 1: Care Processes and Treatment Targets 2020-21, Full Report

Information is collected from GP practice administrative data via pre-agreed extracts of their computer system. Data is extracted from General Practice (GP) clinical systems and specialist diabetes service units in secondary care hospitals. For the 2020-21 audit collection, a primary care specification was developed containing all the SNOMED codes for extraction and information about time periods the data needed to capture, which was used as the basis for each system supplier extract. The 2020-21 NDA primary care collection was delivered via the General Practice Extraction System (GPES).

GP practices were invited to participate in the NDA through their GP clinical system via the Calculating Quality Reporting Service (CQRS). By accepting the invitation, practices opted into GP system supplier-developed queries to automatically extract their NDA data, based on the NDA primary care specification. The NDA operates under an ‘opt in’ model to remain open and transparent with practices and services about what data is being collected. System suppliers went through a testing and certification process with GPES prior to the final data extraction to highlight any issues with the data and ensure the data that was extracted was acceptable and aligned to the primary care specification. Data were extracted and supplied to the Clinical Audit and Registries Management Service in XML format.     

For Wales, GP practices confirmed that they wanted to participate in the NDA and NHS Wales Informatics Service collected the data on their behalf and transferred it to NHS Digital.

For specialist services, data has been collected directly through the Clinical Audit Platform (CAP) since the 2018-19 audit.  This is a continuous collection platform where a final submission window was set, for data to be included in NDA 2020-21. Any submissions within that window may be used in the 2020-21 audit if observation dates fall in the appropriate period. Data was then matched to primary care records to generate a report for the specialist services.

Data from the NHS Diabetic Eye Screening Programme (NHS DESP) was collected for the 2020-21 audit. This covers the eye screening service for England only. This data was used to improve the eye check care process data for England meaning that all 9 care processes could be reported on (for England only).

The audit is collected and disseminated in England under S254 of the Health and Social Care Act 2012, known as a Direction. This means in England it is mandatory for GP practices and specialist services to participate. In Wales the Confidentiality Advisory Group has granted permission to collect the NDA under Regulation 5 of the Health Service (Control of Patient Information) Regulations 2002 (known as Section 251 support). It is voluntary for GP practices and specialist services in Wales to participate in the NDA.

The NDA participation stood at 99.3% of all GP practices in England and Wales for 2020-21. Further information on submitting data to the NDA is available. 

The information collected from GP practices for the audit is individual level data and so contains demographic information such as age, sex, ethnicity, diabetes duration and some geographic variables such as postcode. 

Technical specifications of the collection are available. 

Cleaning rules are applied to supplied data to ensure valid information is included in the audit. The following cleaning rules are applied.

Data outside acceptable limits is deemed to be invalid and set to unknown or null.

Certain records are excluded completely from the extracts submitted by GP practices and specialist services. Records are rejected for the following reasons:

• A specialist service care record submitted with only NHS number and no corresponding primary care record.
• A primary care record does not come from a GP practice which is eligible for the NDA.

A list of eligible GP practices is obtained from the NHS Digital corporate reference library, which is updated weekly using data from the Organisation Data Service (ODS) for England.

NHS Wales Informatics Service (NWIS) supplied the list of eligible GP practices and associated LHB in Wales.

Data were automatically extracted from GP systems in England via system supplier-developed queries and GPES. Internal validation checks were completed within the NDA team to ensure that data had been received from all eligible participating practices and there were no systematic issues with the data. Comparisons were made with GP practice and CCG level counts from previous years.

Any data quality issues with the data are included alongside analysis to ensure users can interpret the results appropriately. For example, eye screening information was removed from the 2011-12 publication as the data was not deemed reliable, and the following statement was added to the report:

“To improve alignment with NICE guidelines, a revise Read code set of terms describing digital eye screening was used. This identified that variation in the use of terminology and its impact on the consistency of data extraction from electronic clinical records rendered it unreliable as a measure of this care process. The NHS Diabetic Eye Screening Programme (NHS DESP) records every digital eye screening and we believe that its records should now be used as the preferred measure for this annual care process. Presently this is reported only nationally.”

Most individuals within the NDA will have 1 record only which can be used to determine demographic and diagnostic information. Most records will come from primary care (GP extracts). 

The demographic fields include:

• Age
• Sex
• Ethnicity
• Lower Super Output Area (LSOA)
• GP practice code
• Diagnosis year
• Diabetes type
• Diagnosis of Learning Disability (LD)
• Diagnosis of Serious Mental Illness (SMI)

Diagnostic information includes dates and values for each of the care processes.

Insulin pump data, submitted by participating specialist diabetes services, is included in the NDA from the 2015-16 audit year.

Multiple readings data for people with diabetes is included in the NDA from the 2016-17 audit period. These data sets hold all readings for a person, within the audit period, for blood pressure, body mass index (BMI), cholesterol and HbA1c (extracted from GP systems and dependent on GP practice participation).

Drug data for people with diabetes is included in the NDA from the 2017-18 audit period. This data set holds all drug prescriptions for a person ,within the audit period, for diabetes-related medication: diabetes mellitus (including metformin), antihypertensive medication and statins (extracted from GP systems and dependent on GP practice participation).

Where data is received from more than 1 organisation a single record per person is created through the ‘golden record’ process.

Information for the same person can be submitted by several organisations, for instance if the person has moved GP practices during the period, and/or attended 1 or more hospital-based specialist services. For analysis purposes a single record is created for each person.

The GP recording the most recent activity is assigned as the person’s ‘current’ GP practice, as at the end of the audit period. Similarly, the specialist service recording the most recent activity is assigned as the person’s ‘current’ hospital. If no GP or hospital data was received for a given data field for a person then the data field is left blank.

Demographic information is taken from the current GP practice where this is valid. If this isn’t the case, any other records are considered, and if they are not contradictory then this information is used. For example, if the current GP practice listed the sex as female, we would use this in the first instance. If the current GP practice listed the sex as ‘unknown’, other records for that person would be considered. If these records were a mixture of ‘male’ and ‘female’, sex would be left as ‘unknown’. If the other records agreed that the only valid value for that person was ‘female’ then they would be assigned as ‘female’ in the data.

The exception to this in demographics is diabetes type. For this, the value from specialist services is used where this is available, and not contradictory. Where it isn’t available from hospital data, diabetes type is taken from the current GP practice. If this is ‘unknown’ then other primary care records are considered, and the value used from these if they are not contradictory.

For care processes the most recent date per person is used, with the lowest measurements taken that day assigned to the golden record.

The record can therefore be made up of information from a mixture of GP practices and specialist services.

The NDA provides a detailed picture of the clinical processes and care pathways for those diagnosed with diabetes. Some of the information collected as part of the audit is collected and published via the Quality and Outcomes Framework (QOF). QOF is an aggregated return which provides information for nearly all GP practices in England. The QOF information is used within the report when discussing coverage.

Clinical Commissioning Groups (CCGs), Local Health Boards (LHBs), GP practices and adult outpatient specialist service units submit data about the care that is being delivered for people with diabetes in their organisation. This will include children that have been treated in an adult care setting, however it does not cover paediatric units. For the full picture on paediatric care for children with diabetes, please refer to the National Paediatric Diabetes Audit (NPDA) which is conducted by the Royal College of Paediatrics and Child Health (RCPCH)¹.

The suite of NDA reports uses 3 separate cohorts of patients from England and Wales:

• Primary care patients – These are patients who are registered with a participating GP practice.
• Specialist service patients – These are patients that have a specialist service record. However, they may or may not be registered with a participating GP practice. Information is collected from hospitals and trusts via a bespoke data collection system.
• All patients – Any patient from participating GP practices and specialist service units.

The audit changed from an 'opt out' to an 'opt in' model from 2013-2014. A reduction in participation was observed in 2013-2014 and 2014-2015. Due to variation in participation caution should be taken when looking at local level analysis (CCGs and LHBs) for low participation years, and across time periods.

Prior to 2013-2014 the approach to the audit was opt out, this meant that GP practices had to tell NHS Digital that they did not wish to participate; this was normally a small number of practices.  The drop in participation in 2012-2013 was expected due to the restructuring of the NHS in April 2013.  This organisational transition disrupted many well-established organisations and systems.
 

A large amount of work was completed to improve participation in the audit, resulting in over 98% of GP practices in England and Wales participating in the last 4 years.
 

1. National Paediatric Diabetes Audit, Royal College of Paediatrics and Child Health https://www.rcpch.ac.uk/work-we-do/quality-improvement-patient-safety/national-paediatric-diabetes-audit

Data is collected covering a 15-month period between 1st January in the first year and 31st of March in the second year.

This relates to the number of people registered with a GP or with a specialist service and covers the “all patients” cohort. It is expected that most patients should be registered with a GP, however there are instances where this is not the case. This may be because their registered GP practice has failed to participate. 

It is vital that people with diabetes receive all 9 of the NICE recommended key health tests and measures. These tests help to monitor and manage diabetes, as well as to reduce the risk of complications such as stroke, heart disease and amputations.

For each of the care processes the date fields are used to determine if the care process has been performed.  The most recent date is used in the audit year and these are matched back onto the relevant cohort of patients.  The denominator for the care processes is the total number of people aged 12 or over within the relevant cohort except for HbA1c (where all patients are used).

A care process can take place in any setting for example the blood test or the smoking review can be done in hospital or in the registered GP practice and it will still count towards the GP practice, CCG, LHB and the specialist service unit.

The 9 annual health checks for people with diabetes are:

Blood pressure is a measurement of the force of blood flow inside the arteries. It is stated as 2 figures, e.g.130/80. The first figure is known as the systolic pressure and relates to the pressure as the heart contracts and pushes blood through the arteries. The other figure is the diastolic pressure measured when the heart relaxes to refill with blood.

A blood pressure measurement should be taken for all registered patients with diabetes at least once a year. This information is collected as part of the audit.  Records are only used within the care process completion indicator if there is a diastolic and systolic date and these are the same. Where a patient aged 12 or over has a diastolic and systolic date in the audit year at any organisation participating in the audit the care process is considered complete.

Body Mass Index this should be recorded for all people with diabetes. This should be measured and recorded annually by the GP or specialist service unit.

The date of this record is used to determine if the care process has been completed in the audit period. Where a patient aged 12 or over has a BMI date in the audit year at any organisation the care process is considered complete.

From 2017-18, height and weight data were collected for people with diabetes, alongside BMI, as 1 GP system supplier (Vision) do not have the facility to submit a calculated BMI.

This should be measured and recorded annually by the GP or specialist service unit. The date of this record is used to determine it the care process has been completed in the audit period. Where a patient aged 12 or over has a creatinine date in the audit year at any organisation the care process is considered complete.

This should be measured and recorded annually by the GP or specialist service unit. The date of this record is used to determine if the care process has been completed in the audit period. Where a patient aged 12 or over has an albumin date in the audit year at any organisation the care process is considered complete.

If you have raised cholesterol and raised triglycerides (triglycerides are another type of fat in the blood) you have an increased risk of cardiovascular disease.  This should be measured and recorded annually by the GP or specialist service unit. The date of this record is used to determine it the care process has been completed in the audit period. Where a patient aged 12 or over has a cholesterol date in the audit year at any organisation the care process is considered complete.

Diabetic retinopathy is a common complication of diabetes. It occurs when high blood sugar levels damage the cells at the back of the eye, known as the retina. Patients with diabetes should be offered screening on an annual basis.

Unfortunately, the variation in the use of terminology and its impact on the consistency of data extraction from electronic clinical records rendered this element of the data unreliable as a measure of this care process. The NHS Diabetic Eye Screening Programme (NHS DESP) records every digital eye screening and we were able to use these records in 2019-20 as the preferred measure for this annual care process for England only. For Wales, we have not reported this care process.

Patients with diabetes should have an annual foot check, as diabetes can limit the blood supply to feet and cause a loss of feeling. This can mean foot injuries do not heal well. A patient is more likely to have a limb amputated when they have diabetes.  Where a patient aged 12 or over has a foot exam date in the audit year at any organisation the care process is considered complete.

There was a change to the definition of the smoking care process in 2017-18. This was to bring the care process into line with updated QOF rules. The care process is considered complete where:

1. a patient aged 12 or over has a smoking status date in the audit year, or
2. a patient has a smoking status of “Never smoked” dated prior to the audit period, but after the patient was 25 years old.

This indicates your blood glucose levels for the previous 2 to 3 months. The HbA1c (glycated haemoglobin) test measures the amount of glucose that is being carried by the red blood cells in the body.  This care process is appropriate for all ages. Where a patient of any age has an HbA1c date in the audit period at any organisation the care process is considered complete.

The person must have received all 8 care processes (excluding eye screening) where appropriate. The requirement is that everyone has their HbA1_C measured annually, and everyone aged 12 or over should have all 8 care processes annually. So, an 11-year-old will only be required to have their HbA1c value tested and so if this has been done they will have achieved the all care process element, However, an adult aged 25 will have to have received all 8 different care processes to have achieved the all 8 care processes.

For England only: the person must have received all 9 care processes (now including eye screening) where appropriate. The requirement is that everyone has their HbA1_C measured annually, and everyone aged 12 or over should have all 9 care processes annually. So, an 11-year-old will only be required to have their HbA1c value tested and so if this has been done they will have achieved the all care process element, However, an adult aged 25 will have to have received all 9 different care processes to have achieved the all 9 care processes.

Treatment targets in the NDA relate to 2 care processes (HbA1c and blood pressure measurements) and the prescription of statins for prevention of cardiovascular disease.

The most recent date is used in each audit year and the best result for the care process is taken if 2 tests were completed on the same day. These are matched back on to the relevant cohort of patients. A care process, and therefore the target result, are taken from any setting and it will count towards the GP practice/CCG and specialist service unit targets.

This test is vital to monitoring the long-term glucose levels within the patients’ blood; the aim should be to keep these at normal levels. The most recent result is taken for use within the calculation and if there were 2 tests completed on the same day the lowest value is used.

There are 3 cumulative level targets:

1. The HbA1c value is less than or equal to 48 mmol/mol (≤48)
2. The HbA1c value is less than or equal to 58 mmol/mol (≤58)
3. The HbA1c value is less than or equal to 86 mmol/mol (≤86)

In addition, the percentages of people achieving the following levels are also reported in the treatment target section of the report:

1. The HbA1c value is less than or equal to 53 mmol/mol (≤53)
2. The HbA1c value is less than or equal to 75 mmol/mol (≤75)

This is a flat target that the patient’s blood pressure (BP) should be less than or equal to 140/80.  Both systolic and diastolic pressures must be within their respective limits for this treatment target to be met. The most recent result is taken for use within the calculation and if there were 2 tests completed on the same day the lowest value is used.

This shows people without cardiovascular disease (CVD) or ischaemic heart disease (IHD) between the age of 40 and 80 years or with CVD/IHD at any age that are recorded as being prescribed statin medication at any point in the audit period.

This measure shows people achieving an HbA1c value ≤ 58mmol/mol, blood pressure (BP) ≤140/80 and, if their use is indicated, being prescribed statins. For children under the age of 12 years, meeting 'all 3 treatment targets' is defined as meeting the HbA1c requirement only, as other treatment targets are not recommended in the NICE guidelines for this age group. For people whom prescription of statins is not indicated, meeting 'all 3 treatment targets' is defined as meeting the HbA1c and blood pressure requirement only.

Indirect standardisation was used to compare the actual number of people completing care processes with the expected number of people completing care processes.

Logistic regression models were built to estimate expected numbers of people completing care processes for each care provider and commission. The models were constructed for each audit year for the whole of England and Wales.

The observed numbers of people completing care processes or achieving treatment targets in each CCG/LHB were compared to these expected numbers. CCGs/LHBs having the greatest positive difference between their observed and expected care process completion and treatment target achievement were highlighted as having been particularly successful in the audit period and were contacted to provide tips on how they achieved this.

 A ‘banding score’ was assigned to each care provider or commission according to whether the observation was above, in-line-with or below expectation.  Noting that when the expected number of people with diabetes completing a care process is small a banding of ‘Insufficient data’ is shown.

However, banding is not a measure of quality of care and a higher or lower than expected number of people completing care processes should not immediately be interpreted as indicating poor or good performance. The banding should not be treated as an absolute assessment of performance and cannot be used to directly compare performance between GPs and CCGs/LHBs, but rather as a tool to aid local investigation.[SJ1] 

Demographic variables, such as ethnicity, were included in the models to determine, as an example: if a younger male of Asian ethnicity is more likely to have had a particular outcome compared to an older female of white ethnicity. 

The models allow each variable to be considered independently by controlling for the effects of other, sometimes related factors.

When modelling data, the aim is to produce a model which both satisfies certain statistical criteria and maintains a connection with the real-world understanding of the behaviours we are trying to explain. The ability of the model to explain the variation seen in the data and to correctly predict outcomes is assessed using the c statistic. The values for this measure range from 0.5 to 1.0. A value of 0.5 indicates that the model is no better than chance at predicting care process completion and a value of 1.0 indicates that the model perfectly identifies those who will and those who will not complete a care process. Models are typically considered reasonable when the c statistic is higher than 0.7 and strong when the c statistic exceeds 0.8.

Explanatory variables were selected by consultation with our clinical lead and advisory group. However, not all variables made significant improvements to the model and, where this was the case, these were removed. 

 Variables included in the care process models were:

• Age band
• Ethnic group
• Indices of multiple deprivation quintile group (a measure of deprivation based on a patient’s postcode)
• Sex
• Smoking status 
• BMI group 
• Duration of diagnosed diabetes
• Whether or not the person was prescribed statins

The outputs from the logistic regression models are presented in the supporting tables to allow users to review the quality of the models derived and these tables include the c statistic associated with each model.

Banding has been investigated for treatment target achievement results but was judged not to be appropriate. This is because the statistical models do not predict with enough certainty whether an individual was likely to achieve the treatment target. Accordingly, it is likely that achievement of treatment targets is largely driven by factors other than the patient characteristics captured in the NDA.

The electronic Frailty Index (eFI) has been autogenerated by GP Electronic Records for several years. It is included in the 2020-21 NDA dataset for the first time. It has 3 grades: mild, moderate and severe. In the NDA 2019-20 publication only severe frailty is used: the cohort of people with severe frailty is compared to the cohort without severe frailty (i.e., with mild, moderate, or no frailty on record).

Frailty describes a dynamic state of increased vulnerability to adverse health outcomes resulting from loss of physiological reserve. The prevalence of frailty increases with increasing age. However, frailty is not universal among older people (aged 65+ years) and can also be identified in younger people (aged <65 years), particularly in the context of long-term conditions, including diabetes.

Care process completion and treatment target achievement is calculated for each audit year using data from that audit year collection. Structured education information is calculated on a different basis, as it looks at people diagnosed within a calendar year and whether education was offered/attended in the 12 months following diagnosis. The diagnosis dates for each person are taken from the most recent audit year’s data collection (i.e. NDA 2019-20) and reported by calendar year (i.e 2017, 2018). As new NDA data is received, the structured education figures are re-calculated to consider audit participation increases and attempts at a local level to retrospectively update offered/attended data in GP systems.  

The latest structured education data reported in the NDA 2019-20 publication are for people diagnosed in calendar year 2018. This is because if a person is diagnosed in 2019, they would need a full 12 month opportunity to be offered/attend structured education, however the NDA, at that point in time, would only hold data up to March 2020. As an example, a person diagnosed in November 2019 may attend structured education in July 2020 (within 12 months of diagnosis) but the attendance would not yet be recorded in NDA data.

Learning disability (LD) data were first collected during the 2015-16 audit where it was found that the LD population had a substantially different age and sex profile compared to the NDA population as a whole.

Therefore, to permit a fair comparison, results for the LD population were directly standardised by age and sex to the whole NDA population. Comparison to the whole NDA population, of which only a very small proportion have LD, is consistent with other sections of the report. Individuals with a learning disability diagnosis were identified using the codes listed in the NDA Primary Care Extract Specification.

Disclosure control has been applied to mitigate the risk of patient identification.

In years prior to 2016-17, the data for practices was removed from the publication where the practice list size was less than 1,000. Data for additional practices was removed where differencing would reveal suppressed data.

This was compliant with the Anonymisation Standard for Publishing Health and Social Care Data, but data for all practices that had submitted data was not available. In addition, providing data for CCG, LHBs, specialist services or GP practices broken down by demographics such as age required additional ad-hoc suppression that was time consuming and meant much of this information was not practical to produce.

From 2016-17, a new suppression method was implemented. Zeros are reported, and all numbers are rounded to the nearest 5, unless the number is 1 to 7, in which case it is rounded to ‘5’. This allows for more granular data to be made available, and for data for all GP practices to be made available.

Rounded numbers are used to calculate rates such as care process completion and treatment target achievement. At CCG/LHB level and above this makes virtually no difference to the resultant rates. At GP level, where the numbers are small, this rounding can have a relatively large impact on the resultant rates, but where numbers are small, rates are volatile and should already be treated with caution.