Research and analysis

Audit methods

Published 24 October 2019

1. Background

In brief, when a case of histologically-confirmed invasive cervical cancer has been identified, the clinician treating the individual must ensure that the hospital-based programme coordinator (HBPC) and the regional screening quality assurance service (SQAS) are informed. This will initiate a cascade of audit activities.

The role of the HBPC is to organise audit activities locally (within each trust).

The role of the regional SQAS is to ensure that local cytology, histology, and colposcopy review processes are coordinated according to the national audit protocol, and to work with cancer registries to make sure that the information captured includes a record of the diagnostic status of each cancer case. The regional SQAS also assembles all the data for a region ready for national SQAS collation. The collated national audit of invasive cervical cancer database is held at the Midlands and East regional SQAS.

Although there are minor differences in the procedure employed by different SQAS regions, the broad principles of the audit, including the allocation of main roles, are the same nationwide.

The Cervical Screening Programme Audit Management Group is the steering committee for the audit. Based on the audit data and findings, the group approves updates and makes recommendations regarding the audit process.

2. Audit protocol

2.1 Data aggregation

Names, addresses, and unique identifiers (such as NHS numbers) are deleted before data are transferred to the national audit database. In addition, date of birth, date of diagnosis and dates of screening for each case/control are shifted by a random number between -50 and +50. This ensures that the data received are de-identified, while allowing for meaningful analysis. The data are received under a service level agreement between the NHS Cervical Screening Programme (lead by Public Health England) and Queen Mary University London for the purpose of producing the report.

2.2 Selection of controls

To permit rigorous evaluation of the programme, cases of cervical cancer are compared to controls of the same age. Controls are identified using bespoke software within the national call and recall computer system (known as the Exeter system).

All controls were born within a year of the case and are registered with a GP in the same administrative district as the case. People who are known to have had a hysterectomy are excluded. Additionally, controls fall into the following groups, based on their similarity with cases:

• GP controls, from the same group practice as the case
• district controls, who share the same first half of a postcode with the case, but who are registered with a different GP
• screened controls, who underwent cytological tests over roughly the same period as the case (used where the case may have been diagnosed as a result of screening)
• abnormal controls, who received an abnormal cytology test report during roughly the same period as the case (but were not diagnosed with cervical cancer)

Each case is assigned 2 population controls (one GP control and one district control). In addition, some cases are assigned controls to match the person’s screening history (screened and abnormal controls).

Population controls are used to study the importance of coverage and the efficacy of the screening programme. Screened controls are used to explore the impact of the screening interval on the incidence of screen-detected cancers. Abnormal controls are used to compare the way in which cases and controls are managed by the screening programme after a cytological test is reported as abnormal.

3. Databases and other data sources

The audit is designed to collect data from a number of sources on a person’s age, cancer stage, and call and recall status, as well as on their cytology, colposcopy, and histology results. Information on a person’s screening invitations and results, and laboratory data on their cytology, are drawn from the National Health Application and Infrastructure Services (NHAIS) system ^{[footnote 1]}. Coordination between the HBPC and the regional SQAS is needed to obtain all other records, due to variability in the availability of data and level of access to the different databases. Colposcopy clinics are contacted for records of all appointments (such as information on patient attendance, details of the examiner, data on the colposcopic impression, account of any procedures performed). Histology results are also collated to produce a fuller picture, and to facilitate slide review.

An audit database exists for epidemiological analysis of all data collected.

3.1 Essential fields

To generate a minimum dataset, information about each case of cervical cancer is entered into a database via a number of essential fields (see Table A). A coding guide for the national dataset is available.

3.2 Cytology screening history

Before 2003, cytology samples took the form of conventional smears. Between 2004 and 2008, laboratories converted to liquid-based cytology (LBC). To reflect the use of both technologies during the audit period, this document refers to cytology ‘tests’ or ‘samples’, not to ‘smears’.

Details of every recorded cytology test for both cases and controls are downloaded from the NHAIS computer system. The dataset includes a large number of privately-taken samples, as well as information on all cervical screening programme tests. Since data for cytological tests are downloaded directly from NHAIS, completeness is assumed for all cases and controls. This is because cytological test results are recorded for all people who participate in the cervical screening programme, and for some of those who are tested privately. The audit does not attempt to capture screening events that take place outside the UK.

Information is obtained on the:

• date on which the test was taken
• result of the test
• action code resulting from the test
• reason for no cytology if no tests are recorded

The action code is the national code used to define the person’s recall type, the type of notifications required, and the period of time between recalls. It determines the management action for each person in the light of their latest test result and records any additional clinical input.

Additionally, since April 2012, the NHAIS download includes information on the:

• interval between tests
• date on which the next test is due
• high risk human papilloma virus (HR-HPV) test result (where performed)
• reason for the HR-HPV test being performed (it is possible to derive from a combination of codes whether the test was primary testing, self-sample, triage or test of cure)
• reason for screening being postponed (where appropriate)
• reason for cessation of screening (where appropriate)
• records of any HPV vaccination recorded on the NHAIS system

For cases, additional information is collected from NHAIS on:

• date of birth
• date of cancer diagnosis
• FIGO stage of the tumour
• histology of the tumour
• treatment received
• Index of Multiple Deprivation (IMD) decile

For controls, date of birth and IMD decile are collected.

3.3 Colposcopy

The quality and completeness of the colposcopic data are variable. This is principally because there is no central database to act as a repository for this information (indeed, most colposcopy records were not computerised until 2001). It is therefore difficult to determine where a person attended for colposcopy, particularly if they visited more than one clinic.

The best indicator of whether a person is likely to have had colposcopy is the presence of a ‘suspend’ code in their cytology record. Similarly, a record from the histology laboratory would suggest that a sample was taken at colposcopy. However, neither the cytology nor the histology record provides conclusive information regarding colposcopic examination.

For cases, colposcopy data are obtained for:

• total number of colposcopy appointments
• date of appointment
• attendance at appointment
• whether the examination was satisfactory
• information on any surgical procedure(s) performed

Non-essential additional fields include:

• colposcopic impression
• pathological diagnosis
• whether the person was pregnant
• time to next follow-up appointment

Colposcopy review guidelines and data collection forms were introduced in April 2013. People who have colposcopy appointments within 5 years of the referral to colposcopy resulting in diagnosis have their histories reviewed to establish whether:

• management at colposcopy was appropriate and if not whether it could have failed to prevent (or delay) the diagnosis of cancer
• a delay in the treatment of high grade disease led to a diagnosis of cancer and if so whether the colposcopy clinic is responsible for the delay

3.4 Histology

The quality and completeness of the data on histology in this audit are also variable, as there is no national link between histology laboratories. The proportion of histological samples reviewed in the audit is based on the total number of samples recorded in the database, rather than the total number of histological samples taken within the cervical screening programme.

For cases histology data are obtained for:

• date of specimen
• type of specimen
• pathological diagnosis
• excisional margins

3.5 Cytology and histology reviews

Audit guidelines mandate that when a case of cervical cancer is confirmed, all cytology samples and histology specimens obtained over the 10 years preceding diagnosis (including those that led to diagnosis) must be reviewed.

Essential data obtained for the audit from the review process are the:

• date of the original sample or specimen
• date of the review
• type of reviewer (screener, checker, consultant biomedical scientist, consultant pathologist)
• original sample or specimen result
• review/consensus result

Following the implementation of revised guidelines for cytology and histology review in April 2012, fewer slides need to be reviewed and, in the case of cytology, fewer reviews per slide are required. The introduction of the new audit guidelines was followed by a 3-month period (April to June 2012) during which, reviews of cytology and histology samples as part of the audit were suspended. Since July 2012, the review workloads for cytology and histology have been reduced by 68% and 78%, respectively. However, it takes close to a year after a diagnosis of cervical cancer before full slide review data are entered into the database.

3.6 GP notes

Information derived from GP notes is no longer required as part of the invasive cervical cancer audit.

However, where a person’s screening history is unclear the HBPC is encouraged to collect information from GP notes. This may yield additional information on the patient’s symptoms (if the cancer is symptomatic) and may also explain any non-attendance at appointments (for example where there is evidence of pregnancy, travel, ill health, or private treatment). There are currently several other projects exploring the possibility of obtaining information from GP notes. The Audit Management Group will evaluate the results from these projects and assess whether there is a feasible way to obtain data from this source before revising the audit protocol.

3.7 Index of Multiple Deprivation

The Department of Communities and Local Government produces the English indices of deprivation, from which the IMD (used by this audit) is derived ^{[footnote 2]}. For the audit, the IMD has been divided into deciles, from the most deprived (0) to the least (9). IMD is derived from the postcode of the home address. To enable the collection of this field, home postcode is now captured as part of the NHAIS download and the IMD decile is automatically assigned by the audit database. Only the deprivation deciles, not the postcode, are collected nationally.

3.8 Treatment

Data on treatment are obtained by the HBPC from patient notes or from the meeting notes of the multidisciplinary team (MDT). These fields tend to be provided as data become available, which may mean that the information is missing for a few months after cases are first recorded in the audit. Obtaining treatment data can be especially challenging when people are diagnosed in one centre and treated in another.

There has been some confusion over the use of the category ‘none’ to report treatment. The intended meaning is that the treating hospital has given only palliative care, but at least one SQAS area interpreted the category as ‘no treatment was reported’. Additionally, some HBPCs used ‘none’ when micro-invasive cancers were treated solely with the diagnostic ‘loop’ excision/cone biopsy. While efforts have been made to correct this miscommunication for the future, some cases with treatment classified as ‘none’ in the audit may, in fact, have received treatment. From 2011 onwards, we are able to distinguish between ‘palliative care’ and ‘no treatment’.

3.9 Table A: Essential fields

Section	Field	Detail
SECTION A & A1	Personal details	NHS number (to be held locally) Date of birth Index of Multiple Deprivation For cases only: Date of diagnosis Stage of tumour (FIGO) Histology Treatment
SECTION B	Cytology	Reason for no cytology Date test was taken Result of cytology test HR-HPV result
SECTION C	Colposcopy	For cases only: Number of colposcopic appointments Date of colposcopy Attendance type Colposcopist Surgical procedure
SECTION C2	Colposcopy review	All fields should be completed
SECTION D1	Histology cancer diagnosis	For cases only: Date of specimen Type of specimen Pathological diagnosis FIGO stage
SECTION D2	Specimen history	Date of specimen Type of specimen Pathological diagnosis Excision status
SECTION E Cytology Review of cases	E1. Original slide	Slide ID Cytology type Date of original test Original test result
	E2. Review results	Reviewed location Review result Original result no further review
SECTION F Histology Review of cases	F1. Original specimen	Specimen ID Date of original specimen Pathological diagnosis Evidence of TZ sampling
	F2. Review results	Reviewed at Review pathological diagnosis Excision status

1. Open Exeter is a portal that allows bodies such as NHS trusts, GP practices, and laboratories to access the NHAIS system. ↩

2. For more information see the 2010 technical report. ↩