Rapid Cancer Registration Dataset: data at 05 September (CAS2609)

The National Disease Registration Service (NDRS) has developed an algorithmically generated Rapid Cancer Registration Dataset (RCRD) using the standard administrative datasets which flow rapidly into NHS England (NHSE) and are incorporated into the Cancer Analysis System (CAS) of NDRS. The data takes the form of a series of significant events that occur to each patient as they proceed through the diagnostic and then therapeutic parts of the cancer pathway, and is available at approximately 4-5 months behind real time. The RCRD is shallower and narrower than the National Cancer Registration Dataset; it should be used and interpreted with reference to the caveats outlined within this document.

Coverage

An overview of the coverage of the Rapid Cancer Registration Dataset (RCRD) is:

• the dataset includes diagnoses from January 2018 to the most recently available data (at the date specified in the title to this document), plus additional event data for the same period

• all results in this document are based on invasive diagnoses only (ICD-10 C codes), although some D-codes are included within the some of the broad cancer groups used to derive the RCRD dataset (specifically for the Breast, Gynaecological and Brain and CNS groups)

• non-melanoma skin cancer (ICD-10 C44) tumours are excluded from the majority of data shown (Figure 3 onwards)

• many results in this document use a comparison period of April to September 2018, which is for consistency of results over time and was initially selected as a period with complete data in both RCRD and the National Cancer Registration Dataset (NCRD), while allowing for diagnosis differences of up to 90 days (3 months) in either direction

Main findings

Some of the main features of the data to be aware of when interpreting the Rapid Cancer Registration Dataset (RCRD) are:

• across all cancer groups included, approximately 11.5% of cases are missing and 5.6% of cases are included erroneously, with an incorrect cancer group or with an incorrect diagnosis date (when compared to the National Cancer Registration Dataset [NCRD] for data from April to September 2018)

• these figures vary strongly with cancer group; broadly, more common cancers (particularly breast and prostate cancer) perform best and less common cancers (particularly bone and soft tissue, and cancers of unknown primary) perform worst

• there are more missing tumours in those aged over 70 compared to younger age groups

• other factors that reduce data completeness include the patient’s Route to Diagnosis, mortality within 30 days of diagnosis, and the presence of multiple cancers

• usable data is available approximately 4 to 5 months after diagnosis or other clinical activity occurs

• data on cancer stage group at diagnosis is available for a number of common cancer groups, although completeness is lower than that for NCRD; where data is available it generally agrees with the NCRD stage group in 80 to 90% of tumours

Contents

Summary

Methodology

   Proxy registration events (Rapid Cancer Registration Dataset tumours)
   Data structures

Data Quality

   How do the number of tumours compare between the Rapid Cancer Registration Dataset and the National Cancer Registration Dataset?
   Comparing the matching quality of the Rapid Cancer Registration Dataset
   Counts of events over time
   Estimated completeness of the Rapid Cancer Registration Dataset and secondary datasets

Staging data in the Rapid Cancer Registration Dataset

   TNM stage group 1-4
   “Early” vs “Late” stage
   Stage trends over time
   Stage completeness by snapshot

Data completeness, source, and mortality comparisons

   Counts of missing data
   Ethnicity completeness
   Tumour source
   Mortality proportion by month

Appendices

   Appendix 1 - List of pathway events
   Appendix 2 - List of fields available in Rapid Cancer Registration Data
   Appendix 3 - Cancer groups used for matching
   Appendix 4 - Alternative defining events
   Appendix 5 - Counts and error tabulations
   Appendix 6 - False negative errors and basis of diagnosis
   Appendix 7 - False positive and false negative proportion by month
   Appendix 8 - Sensitivity testing of matching criteria
   Appendix 9 - Code changes to the RCRD build process

Summary

A need was identified in 2020 to rapidly make available information on ‘proxy cancer registrations’ (and associated clinical activity) for the COVID-19 period, to support the public health response by NHS England (NHSE) and other agencies, and service reorganisation by the NHS. These proxy registrations were called ‘rapid registrations’, in contrast to the more formal detailed registration process that are used in non-clinical cancer research and the Accredited Official Statistics for cancer incidence.

The National Disease Registration Service (NDRS) developed a Rapid Cancer Registration Dataset (RCRD) using all standard administrative datasets which flow rapidly into NHSE and are incorporated into the Cancer Analysis System (CAS) of NDRS. This data has been shown to helpful beyond the initial COVID-19 focus and so the data has continued to be produced and developed.

This document describes the dataset structure, creation methodology, and data quality caveats (due to the rapid automated creation process without additional data curation) behind this dataset.

These data structures and methodologies have evolved over the course of the public health response to COVID-19 and beyond, and are expected to continue evolving to improve the data completeness and quality. The data is updated monthly and is referred to by the monthly CAS snapshot upon which it is based, e.g. CAS2009 refers to the CAS snapshot from September 2020. This document is considered a ‘living document’ and strictly applies only to the snapshot of CAS identified in the title.

Methodology

Proxy registration events (Rapid Cancer Registration Dataset tumours)

Datasets available to NHSE were surveyed for how many months in arrears that they arrive within NDRS systems and are loaded in a usable format for analysis. From these datasets a selection of event types were defined similarly to those typically used for cancer pathway analysis pursued by NDRS.

The data takes the form of a series of significant events that occur to each patient as they proceed through the diagnostic and then therapeutic parts of the cancer pathway. These events include chemotherapy cycles, radiotherapy episodes and major cancer surgery as well as events based on the Cancer Waiting Times (CWT) and Cancer Outcomes and Services Dataset (COSD) datasets. These event types are numbered in the range 1-25 in the dataset.

Some events hypothesised to be indicative of a cancer diagnosis were defined, including ‘Diagnosis reported in COSD’ (event 51) and ‘CWT estimated diagnosis date’ (event 52). These are numbered in the range 50-57 in the dataset - see Appendix 1 for a full list.

The indicative events for diagnosis were explored as candidate rapid cancer registration events. These candidate rapid cancer registration events were judged as matching against an NCRD tumour diagnosis if it met the following two conditions:

• diagnosis dates for each event was 90 days or less
• both registrations fell into the same broad cancer group (as defined in Appendix 3)

Using these matching criteria False Positive errors and False Negative errors are defined as:

• False Positive Error (FPE): An RCRD tumour has been created which does not match against an NCRD tumour in the comparison period
• False Negative Error (FNE): There exists an NCRD tumour for which no RCRD tumour can be matched

Additional filtering was applied to the candidate events and eventually event 101 was defined to minimise both false positive and false negative errors and is recommended for use by researchers as the best candidate for a rapid cancer registration. Appendix 4 briefly examines some of the alternatives examined in the development of this event definition.

Data structures

The Rapid Cancer Registration Dataset (RCRD) consists of two tables:

AT_RAPID_PATHWAY: This is an event-based dataset with a number of types of event of interest defined based on the rapidly available datasets, see Appendix 1 for event definitions and properties. These are numbered in the range 1-25 for general purpose events, 50-57 for events that are candidates for combining into a rapid cancer registration, and 101 for the final rapid cancer registration event.

AT_RAPID_TUMOUR: This is a tumour level dataset that holds tumour and patient level data for each of the tumours defined by a rapid cancer registration.

The structure and contents of both tables are presented in Appendix 2.

The RCRD pathway and tumour tables can be linked together as shown in Figure 1, and also to other datasets that are timely enough via NHSnumber.

Figure 1: Linkage diagram for the Rapid Cancer Registration Dataset

Data Quality

How do the number of tumours compare between the Rapid Cancer Registration Dataset and the National Cancer Registration Dataset?

To illustrate the strengths and weaknesses of the Rapid Cancer Registration Dataset (RCRD) compared to the National Cancer Registration Dataset (NCRD), registrations for tumours diagnosed during 2018 are compared in Figure 2.

For most cancer groups the RCRD counts are significantly lower than those of NCRD. The COSD system does not attempt to record basal cell carcinoma non-melanoma skin cancers (but they are recorded by hospital pathology systems, and thereby registered), explaining the discrepancy there. There is only one group where this situation is reversed - bone and soft tissue - for which a precise morphology is required to properly record the diagnosis. These cancers are being preferentially coded to bone and soft tissue in COSD (as the COSD standard necessitates simpler site-based coding, and this is the best choice under the circumstances) and re-coded during the NCRD registration process where more sophisticated combination of site and morphological coding is possible.

Figure 2: The number of cancer registrations by registration and cancer group, England, 2018

Figure 3 shows the age dependence of the ratio between NCRD and RCRD. Non-Melanoma skin cancer is excluded. The proportion of diagnoses is consistently high for both males and females until the age of 70 is reached, where it declines. This is explored further in Figure 5 below.

Figure 3: The proportion of cancer registrations by gender, age and registration type, England, 2018 (all cancer groups combined)

Comparing the matching quality of the Rapid Cancer Registration Dataset

The quality of the Rapid Cancer Registration Dataset (RCRD) was judged by comparing it against the National Cancer Registration Dataset (NCRD) in the period April 2018 to September 2018. This period was chosen during initial development as available NCRD data was only finalised to December 2018 at the time and a matching period of 90 days was allowed (restricting comparison to the middle six months of the twelve-month period).

Figure 4 shows the proportions of false positive and false negative events, by broad cancer group (excluding non-melanoma skin cancer), measured in the cas2609 snapshot (the cancer groups are defined in Appendix 3). A more detailed tabulation is available by cancer group and ICD-10 code in Appendix 5.

In most cancer groups, there are more tumours missed by RCRD process (false negatives) than there are falsely identified as tumours (false positives).

For breast and prostate, very few incorrect proxy registrations are made. Breast, colorectal, lung, oesophago-gastric (OG) and prostate cancers are also least likely to be missing from the proxy dataset, whereas for cancers of unknown primary, and bone and soft tissue tumours more than 25% of cancers are missed. Bone and soft tissue tumours are not frequently diagnosed. These tumours often require multiple pathology reports to correctly diagnose a patient and RCRD has not attempted to reconcile differences in the reported diagnoses.

Figure 4: Types of error by cancer group, April to September 2018

The proportion of false positive errors is fairly stable across all ages (Figure 5); the proportion of false negative errors slowly declines until age 70 when it increases significantly. The age dependence was investigated and the age-dependence of the basis of diagnosis was found to be at least partially responsible for this - see Appendix 6 for details.

The proportion of false positive cases is less sensitive to the age of the patient.

Figure 5: False negative and false positive errors by age band at diagnosis, April to September 2018

The charts in Figure 6 (below) examine these patterns by cancer group. Please note that age groups for each cancer group must have a denominator of 25 patients or more or they are suppressed for reasons of statistical power.

The patterns of false negative and false positive vary significantly by cancer group. Most groups have a higher proportion of false negatives than false positives at each age.

The proportion of false positives does not exhibit a trend by age for most cancer groups; the proportion rises with increasing age in the bone and soft tissue, head and neck groups and melanoma group and conversely falls with increasing age in the colorectal and unknown groups.

The proportion of false negatives rises with increasing age for all cancer groups except bone and soft tissue and endocrine. The most pronounced increases occur in the brain and central nervous system, colorectal, gynaecological, haematological, prostate, upper gastro-intestinal and unknown primary cancer groups.

The levels of both types of error are highest in cancer groups which are less likely to have solid-tissue pathology (haematological) or where survival rates are typically low. Conversely, the levels of error are lowest for cancer groups for which survival rates are typically higher.

Figure 6: False negative and false positive errors by age band at diagnosis and cancer group, April to September 2018

The variation of the false positive and false negative errors with deprivation quintile is shown in Figure 7. While there is an overall trend visible this is likely to be due to confounding due to the variation with cancer group shown above and the known association of the incidence of many cancer groups with deprivation.

Figure 7: False negative and false positive errors by deprivation quintile, April to September 2018

Figure 8 shows the variation of false negative and false positive errors with route to diagnosis. For false positives there is moderate variation with the lowest error rate being those cases identified through cancer screening or a two week wait referral. (These tumours are those that are likely to be captured in both the COSD dataset and the screening/Cancer Waiting Times datasets so the lower error rate is understandable.)

Most routes to diagnosis have a substantially higher false negative rate than the overall average. Urgent Suspected Cancer (USC) referral and screening routes have a substantially lower false negative rate (and make up between them 45% of the total cohort).

Figure 8: False negative and false positive errors by route to diagnosis, April to September 2018

Figure 9 below shows the variation of false negative and false positive errors with whether or not the patient died within 30 days of diagnosis. The false negative error rate varies substantially between patients who die in the 30 days post-diagnosis compared to those who did, meaning that patients who die within 30 days are more likely to be missing from the dataset.

Figure 9: False negative and false positive errors by 30-day mortality, April to September 2018

Figure 10 below shows the variation of false negative and false positive errors with the multiple tumour status of the patient, i.e. whether or not the patient had been diagnosed with more than one type of tumour in the period January 2018 onward. The false positive error rate varies substantially between patients with multiple cancer groups and those that don’t, meaning that these patients with multiple tumours are more likely to have incorrect cancer groups or diagnosis dates recorded.

Figure 10: False negative and false positive errors by multiple tumour status, April to September 2018

Figure 11 below shows the variation of false negative and false positive errors with the stage at diagnosis.

Figure 11: False negative and false positive errors by stage, April to September 2018

Figure 12 below shows the variation of false negative and false positive errors with the cancer alliance of residence of the patient at the time of diagnosis. The false negative error rate varies more in absolute terms than the false positive rate and may be driven by trust level variation (see figures 13 and 14 below).

Figure 12: False negative and false positive errors by cancer alliance, April to September 2018

Figures 13 and 14 below show the variation of false negative and false positive errors with the trust that diagnosed the tumour. Figure 13 shows the error proportion and figure 14 the numerator (count) of the errors. Trusts shown are limited to NHS secondary care trusts with a denominator of at least 50 patients over the assessment period. Both figures are ordered in descending order of the false negative statistic - but note that the order is not the same in each figure.

There is substantial variation in both false positive and false negative rates and counts. Some large trusts have several hundred or up to 1000 cases (over the six-month period under assessment).

Figure 13: False negative and false positive errors (proportion) by hospital trust, April to September 2018

Figure 14: False negative and false positive errors (count) by hospital trust, April to September 2018

Counts of events over time

This section examines the population of events by chronological time and when they appear in successive analytical snapshots in the CAS. Figure 14 shows that most data items in RCRD are stable with respect to the snapshot month.

Specific comments about the events shown below are:

• Cancer Waiting Times data (events 1–4) are received based on the treatment start date; this explains why for event 2 all lines lie exactly on top of each other. Other CWT events accumulate over successive snapshots where these events occur before the first treatment start event

• an issue with HES data that caused lower than expected completeness from 2020-04-01 was resolved in cas2102, leading to increased event counts in events 5, 6, 11, 12, 13 and 23

• the definition of event 17 only includes tumour diagnoses prior to 2018, so lack of data in the chart below is expected

• definitions of staging events may change between snapshots, which might explain higher or lower counts in one snapshot compared to others

• the vital status shown in event 19 is typically only assessed each January or at the completion of registering each diagnosis year, explaining the large peaks in the graph

• the raw data used to populate events 21, 54 and 56 is subject to ongoing deduplication, which explains lower counts in earlier time periods for later snapshots

• between snapshots, event 101–103 (inferred diagnoses) counts generally increase, particularly for recent months as additional COSD data is submitted. For some earlier months, there is a small decrease in these counts because the algorithm excludes potential diagnoses where the patient already has a confirmed diagnosis in the same cancer group more than 90 days before. These exclusions can change between snapshots as the NCRD data is processed, leading to more confirmed previous diagnoses. The effect has been measured as less than 1% of all cases in any given month

Figure 15: Population of data items to CAS snapshot

Estimated completeness of the Rapid Cancer Registration Dataset and secondary datasets

Detailed linked Rapid Cancer Registration data, CWT, SACT and RTDS data is available at approximately a four-month lag from real time. Linked HES and raw COSD data is available at approximately 4-5 months behind real time.

Table 1 below shows data usability and completeness for RCRD and the constituent datasets. The “latest usable” column shows the ‘hard limit’ on data that is considered fit for analytical purposes (90% completeness), even in months prior to this though data is not necessarily considered complete and the completeness is displayed below. This should be taken into account in any use of the rapid registration data and the secondary datasets.

For the Rapid Tumour data completeness is expressed as the proportion of CCG of residence which show a cancer incidence within the normally expected range (see Table 2 below). For other datasets except CWT completeness is computed as a percentage of the number of data providers who have supplied data over those who are expected to do so.

Data completeness within the Cancer Waiting Times dataset varies at patient level with event type. Figures for the Treatment Start Date and Treatment Period Start Date are given below. Completeness of other CWT events can be estimated by inspecting Figure 15 (events 1-4).

Table 1: RCRD and dataset usability/completeness in cas2609

Rapid registration and dataset usability/completeness

Data source	Latest useable	July 2025	August 2025	September 2025	October 2025	November 2025	December 2025	January 2026	February 2026	March 2026	April 2026	May 2026	June 2026
RCRD Tumours (COSD)	June 2026	Complete	98%	99%	98%	94%	96%	97%	95%	97%	93%	93%	96%
HES	April 2026	Complete	Complete	Complete	Complete	Complete	Complete	Complete	Complete	Complete	95%
SACT	October 2025	93%	93%	93%	91%
RTDS	February 2026	Complete	96%	98%	98%	98%	96%	92%	92%
CWT (TSD)	June 2026	Complete	Complete	Complete	Complete	Complete	Complete	Complete	Complete	Complete	Complete	Complete	97%
CWT (TPSD)	May 2026	Complete	Complete	Complete	Complete	Complete	Complete	Complete	Complete	Complete	Complete	96%
Note:
COSD = Cancer Outcomes and Services Dataset TSD = Treatment Start Date TPSD = Treatment Period Start Date

Table 2: Number of outlier CCGs in COSD dataset in cas2609

The table below shows the number of CCGs (using the April 2020 boundaries) which have 3-sigma outlier counts per month (either high or low) compared to the expectation of the fraction of the total number of new rapid cancer registrations in England. This can be used to judge to what extent there is large scale missing data in COSD (and therefore in RCRD in any particular month.)

Number of outlier CCGs in COSD dataset

Year and month	Outlier: High	Outlier: Low	In expected range	Total received	Prop.
2025-01	1	2	132	135	0.9777778
2025-02	2	2	131	135	0.9703704
2025-03	1	3	131	135	0.9703704
2025-04	2	2	131	135	0.9703704
2025-05	2	3	130	135	0.9629630
2025-06	0	1	134	135	0.9925926
2025-07	0	0	135	135	1.0000000
2025-08	2	1	132	135	0.9777778
2025-09	0	1	134	135	0.9925926
2025-10	0	3	132	135	0.9777778
2025-11	2	6	127	135	0.9407407
2025-12	0	5	130	135	0.9629630
2026-01	1	3	131	135	0.9703704
2026-02	3	4	128	135	0.9481481
2026-03	1	3	131	135	0.9703704
2026-04	4	5	126	135	0.9333333
2026-05	3	7	125	135	0.9259259
2026-06	0	6	129	135	0.9555556
2026-07	33	44	57	134	0.4253731

Staging data in the Rapid Cancer Registration Dataset

TNM stage group 1-4

The size and extent of a cancer is commonly described using the ‘TNM’ system for “Tumour”, “Node”, and “Metastases”. This is often abbreviated to a number between 1 (typically a localised tumour with limited spread) to 4 (typically a tumour that has invaded or spread to distant organs). The stage at diagnosis is very strongly associated with patient outcomes.

In the current version of the Rapid Cancer Registration Dataset (RCRD) partial staging data is provided for a number of different cancer groups (ICD-10 codes can be found in the labels for tables 4a-o). This has been benchmarked against the National Cancer Registration Dataset (NCRD) for cas2609. Table 3 shows the count and proportion of cases by TNM stage group for both RCRD and NCRD, for calendar year 2018. For example, 31% of breast cancers are TNM stage group 1 in RCRD, but 39% in NCRD. Compared to NCRD in 2018, RCRD under report breast cancers diagnosed at stages 1 or 2; colon and rectal cancers diagnosed at stage 4 are under reported; and prostate cancers have under reported stages 1 and 4. In all four cancer groups, there are more tumours allocated to the unknown or unstageable category. Lung cancers in the RCRD most accurately match NCRD and exhibits a broadly similar stage profile from both measures.

Table 3: Summary proportions of stage at diagnosis for RCRD and NCRD, 2018

Summary proportions of stage at diagnosis for RCRD and NCRD

Cancer Group	Stage Group	Count (RCRD)	Percentage (RCRD)	Count (NCRD)	Percentage (NCRD)
Bladder	1	2044	20.1%	3569	39.5%
Bladder	2	1595	15.7%	2107	23.3%
Bladder	3	503	4.9%	1037	11.5%
Bladder	4	366	3.6%	799	8.8%
Bladder	Unknown	5669	55.7%	1521	16.8%
Breast	1	14627	31.1%	19060	38.7%
Breast	2	13555	28.8%	18002	36.5%
Breast	3	3320	7.0%	3994	8.1%
Breast	4	1488	3.2%	2372	4.8%
Breast	Unknown	14115	30.0%	5885	11.9%
Cervix	1	1223	44.6%	861	30.6%
Cervix	2	421	15.3%	429	15.3%
Cervix	3	172	6.3%	177	6.3%
Cervix	4	249	9.1%	230	8.2%
Cervix	Unknown	679	24.7%	1116	39.7%
Colon	1	2772	11.6%	3519	14.1%
Colon	2	5144	21.6%	6294	25.2%
Colon	3	5527	23.2%	6067	24.3%
Colon	4	3697	15.5%	5925	23.7%
Colon	Unknown	6658	28.0%	3211	12.8%
Kidney	1	2453	27.3%	4159	41.2%
Kidney	2	461	5.1%	629	6.2%
Kidney	3	1384	15.4%	1804	17.9%
Kidney	4	710	7.9%	1738	17.2%
Kidney	Unknown	3970	44.2%	1769	17.5%
Lung	1	6390	16.4%	7938	19.5%
Lung	2	2610	6.7%	2964	7.3%
Lung	3	7336	18.9%	8015	19.7%
Lung	4	15177	39.1%	18296	45.0%
Lung	Unknown	7339	18.9%	3411	8.4%
Lymphoma	1	938	7.3%	2017	13.7%
Lymphoma	2	962	7.5%	1772	12.1%
Lymphoma	3	1232	9.6%	2295	15.6%
Lymphoma	4	2716	21.1%	5677	38.7%
Lymphoma	Unknown	7010	54.5%	2910	19.8%
Melanoma	1	6490	44.6%	9635	61.9%
Melanoma	2	2488	17.1%	2888	18.6%
Melanoma	3	465	3.2%	1113	7.1%
Melanoma	4	233	1.6%	423	2.7%
Melanoma	Unknown	4873	33.5%	1509	9.7%
Oesophagus	1	470	5.3%	454	5.6%
Oesophagus	2	867	9.8%	915	11.4%
Oesophagus	3	2069	23.3%	1961	24.4%
Oesophagus	4	2688	30.3%	2951	36.7%
Oesophagus	Unknown	2768	31.2%	1762	21.9%
Ovary	1	1243	21.5%	1913	29.4%
Ovary	2	266	4.6%	339	5.2%
Ovary	3	1344	23.3%	1877	28.8%
Ovary	4	785	13.6%	1237	19.0%
Ovary	Unknown	2130	36.9%	1148	17.6%
Pancreas	1	372	4.3%	768	8.3%
Pancreas	2	634	7.3%	887	9.6%
Pancreas	3	762	8.7%	1114	12.1%
Pancreas	4	2083	23.9%	4544	49.3%
Pancreas	Unknown	4867	55.8%	1913	20.7%
Prostate	1	11900	25.0%	17933	34.6%
Prostate	2	5636	11.9%	6933	13.4%
Prostate	3	10591	22.3%	12071	23.3%
Prostate	4	5775	12.1%	8587	16.5%
Prostate	Unknown	13649	28.7%	6369	12.3%
Rectal	1	2169	18.9%	2644	21.8%
Rectal	2	1783	15.5%	2011	16.6%
Rectal	3	3082	26.8%	3865	31.9%
Rectal	4	1567	13.6%	2323	19.2%
Rectal	Unknown	2880	25.1%	1276	10.5%
Stomach	1	362	7.8%	471	8.8%
Stomach	2	434	9.4%	578	10.8%
Stomach	3	595	12.9%	990	18.6%
Stomach	4	1147	24.9%	2126	39.8%
Stomach	Unknown	2076	45.0%	1171	21.9%
Uterus	1	4707	56.5%	5626	66.3%
Uterus	2	502	6.0%	544	6.4%
Uterus	3	727	8.7%	848	10.0%
Uterus	4	506	6.1%	609	7.2%
Uterus	Unknown	1895	22.7%	863	10.2%

In Tables 4a-o below, the distribution of the stage allocations between the Rapid Cancer Registration Dataset (RCRD) and the National Cancer Registration Dataset (NCRD) are examined.

The figures indicate, by RCRD stage, the proportion of cases which match to an NCRD case at each 1-digit TNM stage group, along with the proportion without a matched NCRD tumour. Stages 1-4 in the RCRD agree with the NCRD stage variable for a high proportion.

For example, when examining the subset of RCRD breast tumours that are identified as TNM stage 1 (31%, Table 3), approximately 87% of these are found to be TNM stage group 1 in NCRD, with the other 13% distributed across TNM stages 2-4, the unknown groups, or being unmatched to an NCRD tumour.

For many but not all (e.g., late stage breast cancer), roughly 85% or more of staged cases in RCRD table have the same stage grouping as the equivalent tumour in the NCRD data - this can be seen in the table below by inspecting the figures where the stage metrics for RCRD and NCRD are the same.

Where the stage is labelled as unknown in the RCRD dataset it is known for at least 70% of those cases in the NCRD data.

Tables 4a-o: Stage comparison between RCRD and NCRD by cancer group, 2018

Stage comparison between RCRD and NCRD

1. bladder (ICD-10 C67)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	85.2%	4.3%	7.9%	6.0%	18.5%
2	3.8%	73.4%	18.7%	6.9%	9.5%
3	2.6%	10.4%	61.5%	21.7%	5.1%
4	1.1%	4.3%	4.7%	58.8%	5.9%
Unknown	6.7%	7.0%	5.7%	5.2%	56.1%
Unmatched	0.6%	0.6%	1.4%	1.4%	4.9%

Stage comparison between RCRD and NCRD

2. breast (ICD-10 C50)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	86.7%	5.0%	1.7%	4.3%	22.4%
2	6.6%	86.5%	11.4%	14.7%	25.0%
3	0.6%	2.7%	77.8%	5.2%	4.3%
4	0.2%	0.9%	2.9%	58.7%	6.2%
Unknown	3.6%	3.1%	4.2%	5.2%	29.0%
Unmatched	2.3%	1.9%	1.8%	11.9%	13.1%

Stage comparison between RCRD and NCRD

3. cervix (ICD-10 C53)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	57.2%	3.2%	3.5%	2.6%	16.9%
2	1.7%	76.4%	2.9%	2.6%	7.2%
3	1.1%	3.2%	68.8%	1.7%	6.7%
4	0.4%	1.5%	2.9%	69.4%	8.8%
Unknown	38.8%	15.6%	21.2%	22.6%	49.3%
Unmatched	0.7%	0.2%	0.6%	1.3%	11.2%

Stage comparison between RCRD and NCRD

4. colon (ICD-10 C18)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	82.6%	1.7%	1.9%	0.8%	9.9%
2	6.0%	88.0%	6.8%	1.3%	12.5%
3	3.7%	4.4%	82.4%	3.5%	11.5%
4	0.6%	2.3%	5.8%	91.7%	24.2%
Unknown	1.7%	1.5%	1.7%	1.0%	26.9%
Unmatched	5.4%	2.1%	1.4%	1.7%	14.9%

Stage comparison between RCRD and NCRD

5. kidney (ICD-10 C64)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	88.3%	6.6%	3.3%	2.1%	28.8%
2	0.4%	74.5%	0.9%	1.3%	4.8%
3	1.6%	6.3%	83.9%	4.1%	10.2%
4	0.5%	3.1%	6.3%	90.0%	22.0%
Unknown	5.6%	5.5%	3.6%	1.4%	23.5%
Unmatched	3.6%	4.1%	2.0%	1.0%	10.7%

Stage comparison between RCRD and NCRD

6. lung (ICD-10 C33-C34)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	90.9%	6.4%	1.4%	0.6%	8.3%
2	2.7%	82.7%	1.8%	0.4%	2.9%
3	1.9%	5.1%	89.7%	1.4%	10.4%
4	1.3%	3.3%	5.4%	96.3%	35.5%
Unknown	0.8%	1.0%	0.7%	0.4%	30.3%
Unmatched	2.4%	1.5%	0.8%	1.0%	12.7%

Stage comparison between RCRD and NCRD

7. lymphoma (ICD-10 C81-C86, C88)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	88.1%	1.4%	0.7%	0.6%	13.3%
2	0.9%	92.4%	1.6%	0.6%	10.2%
3	1.2%	1.4%	88.5%	1.5%	12.6%
4	6.5%	2.5%	7.1%	91.6%	33.7%
Unknown	2.5%	1.6%	1.2%	4.7%	25.5%
Unmatched	0.9%	0.7%	0.9%	1.0%	4.7%

Stage comparison between RCRD and NCRD

8. melanoma (ICD-10 C43)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	91.6%	3.4%	7.2%	12.9%	45.3%
2	2.1%	75.6%	9.2%	16.8%	11.5%
3	1.9%	12.0%	71.2%	12.9%	5.3%
4	0.2%	1.6%	2.9%	37.9%	4.1%
Unknown	1.5%	5.0%	5.2%	11.2%	12.6%
Unmatched	2.7%	2.4%	4.3%	8.2%	21.2%

Stage comparison between RCRD and NCRD

1. oesophagus (ICD-10 C15)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	59.9%	3.6%	0.6%	0.1%	4.9%
2	20.5%	62.1%	8.2%	1.0%	5.2%
3	8.5%	22.5%	69.4%	7.0%	11.0%
4	2.6%	4.2%	16.0%	82.4%	25.2%
Unknown	6.0%	6.1%	5.2%	8.8%	45.5%
Unmatched	2.6%	1.5%	0.6%	0.7%	8.3%

Stage comparison between RCRD and NCRD

10. ovary (ICD-10 C56)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	91.7%	5.3%	0.7%	0.3%	12.9%
2	0.5%	86.7%	0.5%	0.1%	1.9%
3	0.9%	2.3%	89.7%	9.7%	16.7%
4	0.2%	1.1%	4.7%	83.8%	17.6%
Unknown	1.7%	2.3%	2.8%	4.5%	28.7%
Unmatched	5.0%	2.3%	1.5%	1.7%	22.0%

Stage comparison between RCRD and NCRD

11. pancreas (ICD-10 C25)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	70.0%	3.4%	0.9%	0.3%	8.3%
2	15.0%	73.0%	2.4%	0.5%	5.8%
3	4.6%	12.3%	87.1%	0.6%	6.1%
4	3.3%	5.7%	6.3%	96.5%	44.9%
Unknown	3.3%	4.0%	1.9%	1.1%	29.3%
Unmatched	3.8%	1.6%	1.3%	1.0%	5.6%

Stage comparison between RCRD and NCRD

12. prostate (ICD-10 C61)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	85.2%	10.2%	4.6%	1.6%	37.6%
2	6.7%	81.7%	2.5%	0.8%	6.3%
3	4.2%	4.4%	85.6%	2.7%	13.1%
4	0.8%	0.8%	3.9%	91.9%	17.0%
Unknown	2.2%	2.3%	2.8%	2.3%	22.7%
Unmatched	0.8%	0.6%	0.6%	0.8%	3.4%

Stage comparison between RCRD and NCRD

13. rectal (ICD-10 C19-C20)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	82.3%	3.4%	2.4%	1.0%	17.1%
2	5.0%	76.4%	4.0%	1.7%	8.6%
3	8.9%	13.6%	86.9%	5.9%	15.7%
4	1.1%	3.4%	4.1%	89.3%	20.4%
Unknown	1.5%	2.5%	1.8%	1.4%	28.9%
Unmatched	1.2%	0.6%	0.7%	0.8%	9.3%

Stage comparison between RCRD and NCRD

14. stomach (ICD-10 C16)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	59.8%	3.7%	0.5%	0.1%	5.6%
2	16.9%	56.7%	8.7%	0.8%	5.3%
3	6.1%	24.9%	68.3%	3.0%	7.7%
4	1.9%	6.5%	17.9%	91.2%	25.6%
Unknown	5.0%	3.9%	2.4%	2.2%	36.1%
Unmatched	10.2%	4.4%	2.2%	2.7%	19.7%

Stage comparison between RCRD and NCRD

15. uterus (ICD-10 C54-C55)

Stage Group	RCRD
NCRD	1	2	3	4	Unknown
1	95.9%	10.6%	5.1%	2.3%	40.4%
2	0.7%	82.0%	1.1%	1.7%	4.0%
3	0.6%	1.8%	87.3%	5.5%	6.7%
4	0.2%	1.8%	2.1%	81.5%	7.6%
Unknown	1.5%	1.8%	3.3%	6.8%	29.8%
Unmatched	1.2%	2.0%	1.1%	2.1%	11.5%

“Early” vs “Late” stage

Below in table 5 we repeat the above tabulations but now grouping RCRD and NCRD tumour stages into “Early” (TNM stage group 1 & 2) or “Late” (TNM stage group 3 & 4) categories. We see that 60% of breast cancers are identified as “Early” stage in the RCRD compared to 75% in NCRD due to the higher proportion of “Unknown” stage tumours (30% vs 12% respectively).

As with the more detailed stage data, there is a high degree of concordance between the NCRD and RCRD stage fields if a known stage can be identified.

Table 5: Summary proportions of “Early” vs “Late” stage for RCRD and NCRD, 2018

Summary proportions of Early vs Late stage for RCRD and NCRD

Cancer Group	Stage Group	Count (RCRD)	Percentage (RCRD)	Count (NCRD)	Percentage (NCRD)
Bladder	Early	3639	35.8%	5676	62.8%
Bladder	Late	869	8.5%	1836	20.3%
Bladder	Unknown	5669	55.7%	1521	16.8%
Breast	Early	28182	59.8%	37062	75.2%
Breast	Late	4808	10.2%	6366	12.9%
Breast	Unknown	14115	30.0%	5885	11.9%
Cervix	Early	1644	59.9%	1290	45.9%
Cervix	Late	421	15.3%	407	14.5%
Cervix	Unknown	679	24.7%	1116	39.7%
Colon	Early	7916	33.3%	9813	39.2%
Colon	Late	9224	38.8%	11992	47.9%
Colon	Unknown	6658	28.0%	3211	12.8%
Kidney	Early	2914	32.5%	4788	47.4%
Kidney	Late	2094	23.3%	3542	35.1%
Kidney	Unknown	3970	44.2%	1769	17.5%
Lung	Early	9000	23.2%	10902	26.8%
Lung	Late	22513	57.9%	26311	64.8%
Lung	Unknown	7339	18.9%	3411	8.4%
Lymphoma	Early	1900	14.8%	3789	25.8%
Lymphoma	Late	3948	30.7%	7972	54.3%
Lymphoma	Unknown	7010	54.5%	2910	19.8%
Melanoma	Early	8978	61.7%	12523	80.4%
Melanoma	Late	698	4.8%	1536	9.9%
Melanoma	Unknown	4873	33.5%	1509	9.7%
Oesophagus	Early	1337	15.1%	1369	17.0%
Oesophagus	Late	4757	53.7%	4912	61.1%
Oesophagus	Unknown	2768	31.2%	1762	21.9%
Ovary	Early	1509	26.2%	2252	34.6%
Ovary	Late	2129	36.9%	3114	47.8%
Ovary	Unknown	2130	36.9%	1148	17.6%
Pancreas	Early	1006	11.5%	1655	17.9%
Pancreas	Late	2845	32.6%	5658	61.3%
Pancreas	Unknown	4867	55.8%	1913	20.7%
Prostate	Early	17536	36.9%	24866	47.9%
Prostate	Late	16366	34.4%	20658	39.8%
Prostate	Unknown	13649	28.7%	6369	12.3%
Rectal	Early	3952	34.4%	4655	38.4%
Rectal	Late	4649	40.5%	6188	51.1%
Rectal	Unknown	2880	25.1%	1276	10.5%
Stomach	Early	796	17.3%	1049	19.7%
Stomach	Late	1742	37.8%	3116	58.4%
Stomach	Unknown	2076	45.0%	1171	21.9%
Uterus	Early	5209	62.5%	6170	72.7%
Uterus	Late	1233	14.8%	1457	17.2%
Uterus	Unknown	1895	22.7%	863	10.2%

In Table 6a-o below the distribution of the stage allocation between RCRD and NCRD are examined, aggregated into Early and Late stage.

Tables 6a-o: “Early” vs “late” stage comparison between RCRD and NCRD, 2018

Early- vs late-stage comparison between RCRD and NCRD

1. bladder (ICD-10 C67)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	84.0%	20.8%	28.0%
Late	8.5%	72.3%	11.0%
Unknown	6.9%	5.5%	56.1%
Unmatched	0.6%	1.4%	4.9%

Early- vs late-stage comparison between RCRD and NCRD

2. breast (ICD-10 C50)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	92.4%	14.9%	47.4%
Late	2.1%	75.6%	10.5%
Unknown	3.4%	4.5%	29.0%
Unmatched	2.1%	5.0%	13.1%

Early- vs late-stage comparison between RCRD and NCRD

3. cervix (ICD-10 C53)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	64.2%	5.7%	24.1%
Late	2.3%	71.4%	15.5%
Unknown	32.9%	22.0%	49.3%
Unmatched	0.6%	1.0%	11.2%

Early- vs late-stage comparison between RCRD and NCRD

4. colon (ICD-10 C18)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	89.3%	6.0%	22.4%
Late	5.8%	91.0%	35.8%
Unknown	1.6%	1.4%	26.9%
Unmatched	3.3%	1.5%	14.9%

Early- vs late-stage comparison between RCRD and NCRD

5. kidney (ICD-10 C64)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	87.5%	4.0%	33.6%
Late	3.2%	91.5%	32.2%
Unknown	5.6%	2.8%	23.5%
Unmatched	3.7%	1.7%	10.7%

Early- vs late-stage comparison between RCRD and NCRD

6. lung (ICD-10 C33-C34)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	92.3%	1.7%	11.2%
Late	4.7%	96.8%	45.9%
Unknown	0.8%	0.5%	30.3%
Unmatched	2.1%	0.9%	12.7%

Early- vs late-stage comparison between RCRD and NCRD

7. lymphoma (ICD-10 C81-C86, C88)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	91.4%	1.5%	23.5%
Late	5.8%	93.9%	46.3%
Unknown	2.0%	3.7%	25.5%
Unmatched	0.8%	0.9%	4.7%

Early- vs late-stage comparison between RCRD and NCRD

8. melanoma (ICD-10 C43)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	89.6%	21.0%	56.7%
Late	5.3%	66.1%	9.5%
Unknown	2.5%	7.2%	12.6%
Unmatched	2.6%	5.6%	21.2%

Early- vs late-stage comparison between RCRD and NCRD

1. oesophagus (ICD-10 C15)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	70.9%	4.5%	10.1%
Late	21.2%	87.7%	36.2%
Unknown	6.1%	7.2%	45.5%
Unmatched	1.9%	0.7%	8.3%

Early- vs late-stage comparison between RCRD and NCRD

10. ovary (ICD-10 C56-C57)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	92.1%	0.9%	14.9%
Late	1.5%	94.1%	34.4%
Unknown	1.8%	3.4%	28.7%
Unmatched	4.5%	1.6%	22.0%

Early- vs late-stage comparison between RCRD and NCRD

11. pancreas (ICD-10 C25)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	79.6%	1.5%	14.1%
Late	14.2%	96.2%	51.0%
Unknown	3.8%	1.3%	29.3%
Unmatched	2.4%	1.1%	5.6%

Early- vs late-stage comparison between RCRD and NCRD

12. prostate (ICD-10 C61)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	91.9%	5.5%	43.8%
Late	5.1%	91.3%	30.1%
Unknown	2.3%	2.6%	22.7%
Unmatched	0.8%	0.6%	3.4%

Early- vs late-stage comparison between RCRD and NCRD

13. rectal (ICD-10 C19-C20)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	83.9%	5.2%	25.7%
Late	13.2%	92.4%	36.1%
Unknown	2.0%	1.6%	28.9%
Unmatched	0.9%	0.8%	9.3%

Early- vs late-stage comparison between RCRD and NCRD

14. stomach (ICD-10 C16)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	67.8%	3.7%	10.9%
Late	20.8%	91.5%	33.3%
Unknown	4.4%	2.3%	36.1%
Unmatched	7.0%	2.6%	19.7%

Early- vs late-stage comparison between RCRD and NCRD

15. uterus (ICD-10 C54-C55)

Stage Category	RCRD
NCRD	Early	Late	Unknown
Early	96.2%	5.4%	44.4%
Late	1.0%	88.4%	14.3%
Unknown	1.5%	4.7%	29.8%
Unmatched	1.2%	1.5%	11.5%

Stage trends over time

Figure 16 shows the monthly variation of the incidence count by stage at diagnosis for a number of common cancers. Allowing for variation in the number of working days in each month (which affects the overall number of tumours diagnosed per month) and for statistical fluctuation there is little evidence of any stage shift in the period displayed. The feature around May 2018 in the prostate cancer trends can be ascribed to the so called ‘Turnbull-Fry effect’.

Figure 16: Stage trends over time

Stage completeness by snapshot

Figure 17 shows the completeness of stage by cancer group for one snapshot per quarter. Stage completeness continues to increase and lags behind the incidence completeness due to staging activity happening up to several months after diagnosis.

Figure 17: Stage completeness by snapshot

Data completeness, source, and mortality comparisons

Counts of missing data

Figure 18 shows the count of tumours per month where the indicated data item is missing. The data items are: basis of diagnosis, birth date best, ethnic category, NHS number, postcode, deprivation quintile, gender and trust code. Larger counts in the most recent months are to be expected.

Figure 18: Counts of missing data

Ethnicity completeness

Figure 19 shows the count of tumours per month where the indicated data item is missing. Larger counts in the most recent months are to be expected.

Figure 19: Ethnicity completeness

Tumour source

Figure 20 shows the number of tumours created by the source of the diagnosis - i.e., which dataset was used to create them, by month

Figure 20: Tumour source dataset

Mortality proportion by month

Figure 21 shows the mortality proportions by month mortality within 30, 90 and 182 days in the RCRD compared to the NCRD, for all cancers included in RCRD (C00-C97 excl C44).

Figure 21: Monthly mortality proportions at 30, 90 and 182 days

Figure 22 shows the proportions of Route to Diagnosis by month, for all cancers included in RCRD (C00-C97 excl C44).

Figure 22: Monthly Route to Diagnosis

Appendices

Appendix 1 - List of pathway events

Table A1: AT_RAPID_PATHWAY: event list

AT_RAPID_PATHWAY: event list

EVENT_TYPE	EVENT_DESC	EVENT_PROPERTY_1	EVENT_PROPERTY_2	EVENT_PROPERTY_3	EVENT_DATE	LINKAGE
1	CWT treatment period start	First treatment flag derived from cancer treatment event type	CWT recorded primary diagnosis (ICD-10)	CWT cancer treatment event type	Treat period start date	NHSNUMBER
2	CWT treatment start	CWT treatment modality	CWT cancer treatment event type		Treatment start date	NHSNUMBER
3	CWT MDT begin	CWT MDT cancer care plan discussed indicator			MDT date	NHSNUMBER
4	CWT faster diagnosis pathway end		Faster diagnosis pathway site		Faster diagnosis pathway end date	NHSNUMBER
5	HES Admitted Patient Care episode	Treatment speciality: Admission method: Ethnicity	All ICD-10 codes (for episode)	All OPCS-4 codes (for episode)	Episode start date - Episode end date	NHSNUMBER
6	HES Admitted Patient Care operation	All OPCS codes (for date) in POS order	All ICD-10 codes (for episode)		Operation date	NHSNUMBER
7	SACT cycle	Benchmark group	Cycle number	Treatment intent	Cycle start date	PATIENTID
8	RTDS episode	Radiotherapy intent	ICD-10 diagnosis code		Episode treatment start date	PATIENTID
9	Tumour diagnosis (Provisional)	Statusofregistration	ICD-10 diagnosis code	Stage_best	Diagnosis date best	PATIENTID
11	HES major surgery	OPCS-4 code	ICD-10 diagnosis code		Operation date	NHSNUMBER
12	HES major surgery (further constraints)	OPCS-4 code	ICD-10 diagnosis code	Further notes/constraints	Operation date	NHSNUMBER
14	RAWDATA major surgery	OPCS-4 code	ICD-10 diagnosis code		Operation date	PATIENTID
15	RAWDATA major surgery (further constraints)	OPCS-4 code	ICD-10 diagnosis code	Further notes/constraints	Operation date	PATIENTID
17	Prior tumour diagnosis	Status of registration	ICD-10 diagnosis code	Stage_best	Diagnosis date best	PATIENTID
18	Tumour diagnosis (Final)	Status of registration	ICD-10 diagnosis code	Stage_best	Diagnosis date best	PATIENTID
19	Patient vital status	Vital status code A = Alive; D = Dead; D2 = Death recorded but with uncertain date of death; D3 = Death recorded but there is a later diagnosis recorded; D4 = Death recorded but there is a later treatment recorded; I = Patient record inactive, as identified as a duplicate of another active patient record; X = Embarked	ICD-10 Underlying cause of death	Place of death code 1 = Hospital; 2 = Hospice not otherwise specified; 4 = Private home; 5 = Nursing home; 6 = Other; X = Unknown	Vital status date	PATIENTID
20	RAWDATA holistic needs assessment record	COSD version : HNA point of pathway : HNA offered : HNA staff role : (for V10 only) HNA and PCSP status	Primary diagnosis	Laterality	Date of HNA	PATIENTID
21	RAWDATA staging	Inferred best stage	ICD-10 diagnosis code	T/N/M components and full stage for each of pre-treatment, pathological and integrated stage and type of data used for the inferred best (‘combined’) stage	Collected stage date	PATIENTID
22	CWT first seen	Source of referral	Categorisation of USC referral, screening and consultant upgrade cases, where relevant	Suspected cancer referral type	Date first seen	NHSNUMBER
23	HES diagnostic event	OPCS-4 code	Description of OPCS-4 code	BX/LD	Operation date	NHSNUMBER
24	RAWDATA personal care and support plan	COSD version : PCSP point of pathway : PCSP offered : PCSP staff role : (for V10 only) HNA and PCSP status	Primary diagnosis	Laterality	PCSP date	PATIENTID
25	RAWDATA end of treatment summary		Primary diagnosis	Laterality	End of treatment summary date	PATIENTID
50	Skeleton tumour creation	E_base_record type (COSD = England, CANISC = Wales)	ICD-10 diagnosis code		Diagnosisdate	PATIENTID
51	Diagnosis reported in COSD	Number of times reported	ICD-10 diagnosis code	E_base_record type	Diagnosisdate	NHSNUMBER
52	CWT estimated diagnosis date	Broad modality group, used for median offset	CWT recorded primary diagnosis (ICD)	CWT cancer treatment event type	Adjusted treat period start	NHSNUMBER
53	HES inferred tumour	HES cancer group	ICD-10 diagnosis code		Episode start date	NHSNUMBER
54	COSD diagnosis submission	Diagnosis trust	ICD-10 diagnosis code (submission)	Gender	Diagnosis date (submission)	NHSNUMBER
55	RAWDATA biopsy record	Laterality	ICD-10 diagnosis code		Collected date/authorised date	PATIENTID
56	RAWDATA imaging record	Laterality	ICD-10 diagnosis code	Procedure_date - diagdate	Diagnosis date (Imaging record)	PATIENTID
57	RAWDATA HNA diagnosis	Laterality	Primary diagnosis (ICD-10)		Diagnosis date (HNA record)	PATIENTID
101	Inferred diagnosis	Event_property_1	ICD-10 diagnosis code	Cancer group	Diagnosis date (derived)	PATIENTID
102	Inferred diagnosis, with adapted diagnosis dates (from 101, adapted)	source_id from second record used to adapt diagnosis date	ICD-10 diagnosis code	Cancer group	Diagnosis date (derived)	PATIENTID
103	Alternative inferred diagnosis (work in progress)	source_id from second record used to adapt diagnosis date	ICD-10 diagnosis code	Cancer group	Diagnosis date (derived)	PATIENTID

*: Data dictionary: Primary cancer site for cancer faster diagnosis pathway

**: Data dictionary: Holistic needs assessment and personalised care and support plan point of pathway for cancer

Appendix 2 - List of fields available in Rapid Cancer Registration Data

Table A2: AT_RAPID_TUMOUR: field list

AT_RAPID_TUMOUR: field list

COLUMN_NAME	DATA_TYPE	Notes
INDIVIDUALID	NUMBER(19,0)	Matches AT_RAPID_PATHWAY for each event with event_type = 101
PATIENTID	NUMBER(19,0)	Matches AT_RAPID_PATHWAY for each event with event_type = 101
NHSNUMBER	VARCHAR2(12 CHAR)	Matches AT_RAPID_PATHWAY for each event with event_type = 101
TUMOUR_AVPID	NUMBER	Matches AT_RAPID_PATHWAY for each event with event_type = 101
SURNAME	VARCHAR2(64 CHAR)	Taken from Encore
FORENAME	VARCHAR2(64 CHAR)	Taken from Encore
BIRTHDATEBEST	DATE	Taken from Encore
VITALSTATUS	VARCHAR2(260 CHAR)	Records whether the patient is currently alive or deceased at the time of the snapshot
VITALSTATUSDATE	DATE	The date of the last known vital status for the patient
GENDER	VARCHAR2(260 CHAR)	Taken from Encore
ETHNICCATEGORY	VARCHAR2(1 CHAR)	Taken from Encore or the HESAPC dataset
DIAGNOSISDATE	DATE	Matches AT_RAPID_PATHWAY for each event with event_type = 101
TUMOUR_SITE	VARCHAR2(260 CHAR)	Matches AT_RAPID_PATHWAY for each event with event_type = 101 (event_property_2)
CANCER_GROUP	VARCHAR2(40 CHAR)	Broad cancer group derived from TUMOUR_SITE, according to groupings used for RCRD derivation and RCRD dashboard
CANCER_GROUP_DETAILED	VARCHAR2(40 CHAR)	Detailed cancer group derived from TUMOUR_SITE, according to groupings used for RCRD dashboard
STAGED_SITE_FLAG	NUMBER (0 or 1)	Indicator flag (0 = No; 1 = Yes) for whether the TUMOUR_SITE (ICD-10 site code) can be staged (according to the RCRD algorithm)
STAGE	VARCHAR2(260 CHAR)	Defined for selected cancer sites
STAGE_EXPERIMENTAL	VARCHAR2	Used for data under development (RCRD staging algorithm in testing)
TUMOUR_MORPHOLOGY	VARCHAR2(5 CHAR)	Tumour morphology as recorded in the COSD system
BASISOFDIAGNOSIS	VARCHAR2(260 CHAR)	The basis of diagnosis (eg clinical; pathological; etc)
TUMOUR_PERFORMANCESTATUS	VARCHAR2(1 CHAR)	Patient performance status at time of diagnosis
FINAL_ROUTE	VARCHAR2(22 CHAR)	Final Route to Diagosis using an adapted version of the standard NDRS methodology
CHRL_TOT_27_03	NUMBER(10,0)	Charlson score defined using the standard NDRS methodology
SURGERY_FLAG	NUMBER(1,0)	Indicator flag (0 = No; 1 = Yes) for whether the patient has an associated surgical tumour resection record
SURGERY_DATE	DATE	Where the SURGERY_FLAG indicates one or more associated surgical tumour resection record, this is the earliest such date
RADIOTHERAPY_FLAG	NUMBER(1,0)	Indicator flag (0 = No; 1 = Yes) for whether the patient has an associated radiotherapy record (from RTDS)
RADIOTHERAPY_DATE	DATE	Where the RADIOTHERAPY_FLAG indicates one or more associated radiotherapy record, this is the earliest such date
SACT_FLAG	NUMBER(1,0)	Indicator flag (0 = No; 1 = Yes) for whether the patient has an associated systemical anti-cancer therapy (SACT) record (from SACT dataset)
SACT_DATE	DATE	Where the SACT_FLAG indicates one or more associated SACT record, this is the earliest such date
AGE	NUMBER(3,0)	Age at diagnosis (from DIAGNOSISDATE AND BIRTHDATEBEST)
DIAGNOSIS_TRUST	VARCHAR2(260 CHAR)	Trust of diagnosis (derived)
POSTCODE	VARCHAR2(255 CHAR)	Postcode at diagnosis, taken from Encore
DEP_QUINTILE	VARCHAR2(120 CHAR)	Index of Multiple Deprivation quintile (2019 or 2025, defined using the standard NDRS methodology)
LSOA21	VARCHAR2(27 CHAR)	2021 census LSOA of residence at time of diagnosis
PRACTICE_CODE	VARCHAR2(7 CHAR)	General practice at diagnosis (derived)
SOURCE	VARCHAR2(11 CHAR)	The dataset used as the primary source for the RCRD registration
SOURCE_ID	VARCHAR2(69 CHAR)	The unique ID of the record used as the primary source for the RCRD registration

Appendix 3 - Cancer groups used for matching

Table A3: Rapid Cancer Registration Data ICD-10 tumour inclusion list

Rapid Cancer Registration Data ICD-10 tumour inclusion list

ICD	CANCER_GROUP	CANCER_GROUP_DETAILED	SCOPE
C00	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C01	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C02	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C03	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C04	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C05	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C06	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C07	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C08	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C09	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C10	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C11	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C12	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C13	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C14	Head and neck	Lip, oral cavity and pharynx	DQ & CD
C15	Oesophago-gastric	Oesophagus	DQ & CD
C16	Oesophago-gastric	Stomach	DQ & CD
C17	Upper GI excl. OG		DQ & CD
C18	Bowel	Colon	DQ & CD
C19	Bowel	Rectal	DQ & CD
C20	Bowel	Rectal	DQ & CD
C21	Bowel		DQ & CD
C22	Upper GI excl. OG	Liver	DQ & CD
C23	Upper GI excl. OG		DQ & CD
C24	Upper GI excl. OG		DQ & CD
C25	Upper GI excl. OG	Pancreas	DQ & CD
C26	Upper GI excl. OG		DQ & CD
C27	Other		DQ & CD
C28	Other		DQ & CD
C29	Other		DQ & CD
C30	Head and neck	Nasal cavity, sinuses and larynx	DQ & CD
C31	Head and neck	Nasal cavity, sinuses and larynx	DQ & CD
C32	Head and neck	Nasal cavity, sinuses and larynx	DQ & CD
C33	Lung/ Respiratory and intrathoracic	Lung	DQ & CD
C34	Lung/ Respiratory and intrathoracic	Lung	DQ & CD
C35	Other		DQ & CD
C36	Other		DQ & CD
C37	Lung/ Respiratory and intrathoracic		DQ & CD
C38	Lung/ Respiratory and intrathoracic		DQ & CD
C39	Lung/ Respiratory and intrathoracic		DQ & CD
C40	Bone and soft-tissue		DQ & CD
C41	Bone and soft-tissue		DQ & CD
C42	Other		DQ & CD
C43	Melanoma		DQ & CD
C44	Non-melanoma skin cancer
C45	Lung/ Respiratory and intrathoracic		DQ & CD
C46	Bone and soft-tissue		DQ & CD
C47	Bone and soft-tissue		DQ & CD
C48	Gynaecological	Ovary	DQ & CD
C49	Bone and soft-tissue		DQ & CD
C50	Breast		DQ & CD
C51	Gynaecological		DQ & CD
C52	Gynaecological		DQ & CD
C53	Gynaecological	Cervix	DQ & CD
C54	Gynaecological	Uterus	DQ & CD
C55	Gynaecological	Uterus	DQ & CD
C56	Gynaecological	Ovary	DQ & CD
C57	Gynaecological	Ovary	DQ & CD
C58	Gynaecological		DQ & CD
C59	Other		DQ & CD
C60	Urological excl. prostate		DQ & CD
C61	Prostate		DQ & CD
C62	Urological excl. prostate	Testes	DQ & CD
C63	Urological excl. prostate		DQ & CD
C64	Urological excl. prostate	Kidney	DQ & CD
C65	Urological excl. prostate		DQ & CD
C66	Urological excl. prostate		DQ & CD
C67	Urological excl. prostate	Bladder	DQ & CD
C68	Urological excl. prostate		DQ & CD
C69	Brain and CNS		DQ & CD
C70	Brain and CNS		DQ & CD
C71	Brain and CNS		DQ & CD
C72	Brain and CNS		DQ & CD
C73	Endocrine		DQ & CD
C74	Endocrine		DQ & CD
C75	Endocrine		DQ & CD
C76	Unknown		DQ & CD
C77	Unknown		DQ & CD
C78	Unknown		DQ & CD
C79	Unknown		DQ & CD
C80	Unknown		DQ & CD
C81	Blood cancer (haematological neoplasms)	Hodgkin lymphoma	DQ & CD
C82	Blood cancer (haematological neoplasms)	Non-Hodgkin lymphoma	DQ & CD
C83	Blood cancer (haematological neoplasms)	Non-Hodgkin lymphoma	DQ & CD
C84	Blood cancer (haematological neoplasms)	Non-Hodgkin lymphoma	DQ & CD
C85	Blood cancer (haematological neoplasms)	Non-Hodgkin lymphoma	DQ & CD
C86	Blood cancer (haematological neoplasms)	Non-Hodgkin lymphoma	DQ & CD
C87	Blood cancer (haematological neoplasms)		DQ & CD
C88	Blood cancer (haematological neoplasms)		DQ & CD
C89	Blood cancer (haematological neoplasms)		DQ & CD
C90	Blood cancer (haematological neoplasms)	Myeloma	DQ & CD
C91	Blood cancer (haematological neoplasms)	Leukaemia	DQ & CD
C92	Blood cancer (haematological neoplasms)	Leukaemia	DQ & CD
C93	Blood cancer (haematological neoplasms)	Leukaemia	DQ & CD
C94	Blood cancer (haematological neoplasms)	Leukaemia	DQ & CD
C95	Blood cancer (haematological neoplasms)	Leukaemia	DQ & CD
C96	Blood cancer (haematological neoplasms)		DQ & CD
C97	Unknown		DQ & CD
D00	Insitu oral and OG
D01	Insitu digestive
D02	Insitu respiratory
D03	Insitu skin
D04	Insitu melanoma
D05	Breast
D06	Gynaecological
D07	Insitu genital
D09	Urological excl. prostate
D32	Brain and CNS
D33	Brain and CNS
D35	Brain and CNS
D37	Unknown behaviour oral and OG
D38	Unknown behaviour respiratory
D39	Unknown behaviour female genital
D40	Unknown behaviour male genital
D41	Urological excl. prostate
D42	Brain and CNS
D43	Brain and CNS
D44	Brain and CNS
D45	Unknown behaviour blood cancer
D46	Unknown behaviour blood cancer
D47	Unknown behaviour blood cancer
D48	Unknown behaviour other
D72	Unknown behaviour blood cancer
E85	Unknown behaviour blood cancer
M72	Unknown behaviour other
Scope: DQ = ‘Included in this data quality document’; CD = ‘Included in NDRS Cancer Data Hub https://digital.nhs.uk/ndrs/data/data-outputs/cancer-data-hub/rapid-cancer-registration-data-dashboards’

Appendix 4 - Alternative defining events

Several options were considered as to the defining events for RCRD. Both standalone datasets, subsets of standalone datasets, and combined datasets were explored and their FNE and FPE figures quantified. A subset of these alternatives are presented below (Table A4a) as a demonstration of the process but the majority of this exploratory work is out of scope for this document.

Candidates for diagnosis events from the three main datasets that are rapidly available and have nominally full coverage of cancer patients are shown below (SACT and RTDS were also examined but data is not presented). Of the three, the CWT data has the best FPE but the FNE is substantially higher than the COSD dataset. HES produced the worst results in both measures. A filtering process was applied to the standalone COSD data to remove apparently new diagnoses that were actually recurrences of prior tumours. This improved the FPE at a cost of increasing the FNE.

Further changes have been made to combine data from COSD, CWT, HES and death certificates, and to refine the filtering applied. We continue to test whether this process can be further refined to improve the combined FPE and FNE figures, and monitor changes in the underlying datasets that might also give new opportunities to do so.

Table A4a presents a history of FPE and FNE proportions as the algorithm was refined, up to CAS2608. All these calculations were were based on a cohort of tumours coded to ICD-10 C00-C97 exc C44, along with some D-codes (D05, D09, D32-D35, D41-D44). From CAS2609 onwards, the process for this calculation was refined and it was updated to be based on only a cohort of tumours coded to ICD-10 C00-C97 exc C44, for more consistency with other routine RCRD outputs. Table A4b presents a record of the FPE and FNE proportions following these changes.

Table A4a: Rapid Cancer Registrations: alternative defining events, C00-C97 exc C44, along with some D-codes (D05, D09, D32-D35, D41-D44), April to September 2018

Rapid Cancer Registrations: alternative defining events

Event	FPE	FNE
Event 52 - standalone CWT	7.6%	28.3%
Event 53 - standalone HES	13.2%	38.9%
Event 54 - standalone COSD	8.1%	15.8%
Event 101 (up to cas2106) - filtered COSD	5.2%	17.7%
Event 101 (cas2107) - filtered combined COSD/CWT	5.6%	16.4%
Event 101 (cas2108) - filtered combined COSD/CWT	5.1%	16.5%
Event 101 (cas2109) - filtered combined COSD/CWT	5.1%	16.6%
Event 101 (cas2110) - filtered combined COSD/CWT/HES	5.1%	14.7%
Event 101 (cas2111) - filtered combined COSD/CWT/HES	6.2%	13.4%
Event 101 (cas2112 to cas2202) - filtered combined COSD/CWT/HES and Death Certificates Only	5.3%	13.4%
Event 101 (cas2203 to cas2204) - filtered combined COSD/CWT/HES and Death Certificates Only	6.3%	12.2%
Event 101 (cas2205) - filtered combined COSD/CWT/HES and Death Certificates Only	6.1%	12.3%
Event 101 (cas2206) - filtered combined COSD/CWT/HES and Death Certificates Only	5.6%	12.5%
Event 101 (cas2207) - filtered combined COSD/CWT/HES and Death Certificates Only	6.0%	11.8%
Event 101 (cas2208 to cas2210) - filtered combined COSD/CWT/HES and Death Certificates Only	6.0%	11.6%
Event 101 (cas2211 to cas2304) - filtered combined COSD/CWT/HES and Death Certificates Only	6.1%	11.5%
Event 101 (cas2305) - filtered combined COSD/CWT/HES and Death Certificates Only	6.1%	11.3%
Event 101 (cas2306) - filtered combined COSD/CWT/HES and Death Certificates Only	6.1%	11.4%
Event 101 (cas2307 to cas2308) - filtered combined COSD/CWT/HES and Death Certificates Only	6.1%	11.3%
Event 101 (cas2309 to cas2311) - filtered combined COSD/CWT/HES and Death Certificates Only	6.1%	11.4%
Event 101 (cas2312 to cas2409) - filtered combined COSD/CWT/HES and Death Certificates Only	6.0%	11.5%
Event 101 (cas2410) - filtered combined COSD/CWT/HES and Death Certificates Only	5.9%	11.7%
Event 101 (cas2411) - filtered combined COSD/CWT/HES and Death Certificates Only	5.8%	12.3%
Event 101 (cas2412 to cas2501) - filtered combined COSD/CWT/HES and Death Certificates Only	6.0%	12.0%
Event 101 (cas2504 to cas2505) - filtered combined COSD/CWT/HES and Death Certificates Only	6.0%	12.1%
Event 101 (cas2505 to cas2506) - filtered combined COSD/CWT/HES and Death Certificates Only	6.0%	12.0%
Event 101 (cas2507 to cas2608) - filtered combined COSD/CWT/HES and Death Certificates Only	6.0%	12.1%

Table A4b: Rapid Cancer Registrations: alternative defining events, C00-C97 exc C44, April to September 2018

Rapid Cancer Registrations: alternative defining events

Event	FPE	FNE
Event 101 (cas2609) - filtered combined COSD/CWT/HES and Death Certificates Only	5.6%	11.5%

Appendix 5 - Counts and error tabulations

Figure A1 shows an example for a very small dataset of how counts and error proportions are derived. This dataset has 10 tumours from the National Cancer Registration Dataset and 7 tumours from the Rapid Cancer Registration Dataset overall (both indicated by the dots in the figure, with time running vertically over the course of 2018 and NCRD versus RCRD divided horizontally). Successful linkages between NCRD and RCRD are indicated by blue lines. False negatives and false positives are indicated. Only tumours in the 6-month assessment period are included in the tabulations below, although these can link to tumours outside the period as shown, and many-to-one linkages are also allowed. The false negative rate is therefore 3 in 7 and the false positive rate 1 in 6 below.

Figure A1: Illustration of counts and errors tabulation

Tables A5a and A5b below tabulate counts of NCRD and RCRD together with the numbers of false positive and false negative errors. When considering comparisons between figures the nature of the linkage and relationships displayed in the diagram above should be kept in mind.

Table A5a: Counts and errors tabulation by cancer group, April to September 2018

Counts and errors tabulation by cancer group

Cancer group	NCRD tumours	RCRD tumours	Difference	Percentage RCRD/ NCRD	FP count	FPE	FN count	FNE
Bone and soft-tissue	1234	1129	105	91.5%	360	31.9%	445	36.1%
Bowel	18985	17848	1137	94.0%	907	5.1%	1696	8.9%
Brain and CNS	2754	2599	155	94.4%	222	8.5%	289	10.5%
Breast	25139	24356	783	96.9%	1336	5.5%	1330	5.3%
Endocrine	1911	1486	425	77.8%	138	9.3%	509	26.6%
Gynaecological	9727	9276	451	95.4%	626	6.7%	936	9.6%
Haematological	14230	12449	1781	87.5%	683	5.5%	2484	17.5%
Head and neck	5289	4931	358	93.2%	399	8.1%	694	13.1%
Lung	21728	20109	1619	92.5%	616	3.1%	2105	9.7%
Melanoma	8259	7688	571	93.1%	693	9.0%	1068	12.9%
Oesophago-gastric	6622	6471	151	97.7%	363	5.6%	470	7.1%
Prostate	27241	25170	2071	92.4%	299	1.2%	2462	9.0%
Unknown	3426	2569	857	75.0%	665	25.9%	1530	44.7%
Upper GI excl. OG	9276	8675	601	93.5%	793	9.1%	1441	15.5%
Urological excl. prostate	11650	11472	178	98.5%	607	5.3%	1853	15.9%
All cancers (C00-C97 exc C44)	167471	156228	11243	93.3%	8707	5.6%	19312	11.5%

Table A5b: Counts and errors tabulation by ICD-10 code, April to September 2018

Counts and errors tabulation by cancer group

ICD-10 code	NCRD tumours	RCRD tumours	Difference	Percentage RCRD/ NCRD	FP count	FPE	FN count	FNE
C00	110	151	-41	137.3%	66	43.7%	25	22.7%
C01	647	471	176	72.8%	13	2.8%	60	9.3%
C02	602	604	-2	100.3%	18	3.0%	91	15.1%
C03	234	108	126	46.2%	5	4.6%	65	27.8%
C04	256	239	17	93.4%	10	4.2%	34	13.3%
C05	211	185	26	87.7%	8	4.3%	32	15.2%
C06	271	287	-16	105.9%	21	7.3%	51	18.8%
C07	239	288	-49	120.5%	101	35.1%	54	22.6%
C08	82	89	-7	108.5%	15	16.9%	14	17.1%
C09	912	765	147	83.9%	16	2.1%	59	6.5%
C10	152	243	-91	159.9%	12	4.9%	30	19.7%
C11	113	111	2	98.2%	6	5.4%	13	11.5%
C12	157	99	58	63.1%	1	1.0%	11	7.0%
C13	142	130	12	91.5%	11	8.5%	22	15.5%
C14	25	66	-41	264.0%	15	22.7%	13	52.0%
C15	3996	4265	-269	106.7%	122	2.9%	218	5.5%
C16	2626	2206	420	84.0%	241	10.9%	252	9.6%
C17	822	669	153	81.4%	129	19.3%	269	32.7%
C18	12441	11762	679	94.5%	664	5.6%	1237	9.9%
C19	996	935	61	93.9%	41	4.4%	90	9.0%
C20	4901	4506	395	91.9%	112	2.5%	327	6.7%
C21	647	645	2	99.7%	90	14.0%	42	6.5%
C22	2648	2525	123	95.4%	261	10.3%	452	17.1%
C23	477	467	10	97.9%	27	5.8%	62	13.0%
C24	645	529	116	82.0%	32	6.0%	84	13.0%
C25	4533	4189	344	92.4%	136	3.2%	498	11.0%
C26	151	296	-145	196.0%	208	70.3%	76	50.3%
C30	162	158	4	97.5%	27	17.1%	24	14.8%
C31	93	65	28	69.9%	5	7.7%	26	28.0%
C32	881	872	9	99.0%	49	5.6%	70	7.9%
C33	13	11	2	84.6%	1	9.1%	3	23.1%
C34	20263	18743	1520	92.5%	541	2.9%	1919	9.5%
C37	170	95	75	55.9%	13	13.7%	60	35.3%
C38	73	352	-279	482.2%	45	12.8%	21	28.8%
C39	0	12	-12	Inf%	4	33.3%	0	NaN%
C40	119	107	12	89.9%	12	11.2%	25	21.0%
C41	117	149	-32	127.4%	78	52.3%	43	36.8%
C43	8259	7688	571	93.1%	693	9.0%	1068	12.9%
C45	1209	896	313	74.1%	12	1.3%	102	8.4%
C46	69	43	26	62.3%	3	7.0%	24	34.8%
C47	28	12	16	42.9%	5	41.7%	10	35.7%
C48	287	401	-114	139.7%	106	26.4%	77	26.8%
C49	836	782	54	93.5%	250	32.0%	305	36.5%
C50	25139	24356	783	96.9%	1336	5.5%	1330	5.3%
C51	645	594	51	92.1%	55	9.3%	78	12.1%
C52	95	108	-13	113.7%	15	13.9%	12	12.6%
C53	1321	1313	8	99.4%	42	3.2%	71	5.4%
C54	4086	3690	396	90.3%	103	2.8%	196	4.8%
C55	72	333	-261	462.5%	23	6.9%	17	23.6%
C56	3002	2529	473	84.2%	242	9.6%	485	16.2%
C57	274	320	-46	116.8%	34	10.6%	37	13.5%
C58	10	24	-14	240.0%	18	75.0%	1	10.0%
C60	304	314	-10	103.3%	50	15.9%	37	12.2%
C61	27241	25170	2071	92.4%	299	1.2%	2462	9.0%
C62	1056	1064	-8	100.8%	83	7.8%	70	6.6%
C63	33	31	2	93.9%	13	41.9%	19	57.6%
C64	4905	4371	534	89.1%	273	6.2%	794	16.2%
C65	419	321	98	76.6%	25	7.8%	92	22.0%
C66	362	260	102	71.8%	13	5.0%	123	34.0%
C67	4475	5055	-580	113.0%	144	2.8%	679	15.2%
C68	96	56	40	58.3%	6	10.7%	39	40.6%
C69	384	352	32	91.7%	48	13.6%	64	16.7%
C70	19	42	-23	221.1%	5	11.9%	1	5.3%
C71	2268	2115	153	93.3%	134	6.3%	205	9.0%
C72	83	90	-7	108.4%	35	38.9%	19	22.9%
C73	1735	1356	379	78.2%	78	5.8%	413	23.8%
C74	117	84	33	71.8%	27	32.1%	57	48.7%
C75	59	46	13	78.0%	33	71.7%	39	66.1%
C76	94	210	-116	223.4%	109	51.9%	54	57.4%
C77	271	126	145	46.5%	62	49.2%	129	47.6%
C78	594	55	539	9.3%	24	43.6%	332	55.9%
C79	2444	133	2311	5.4%	54	40.6%	1006	41.2%
C80	23	2045	-2022	8891.3%	416	20.3%	9	39.1%
C81	895	863	32	96.4%	15	1.7%	73	8.2%
C82	1246	1047	199	84.0%	15	1.4%	137	11.0%
C83	3032	2704	328	89.2%	39	1.4%	307	10.1%
C84	290	229	61	79.0%	14	6.1%	99	34.1%
C85	985	1026	-41	104.2%	63	6.1%	233	23.7%
C86	156	101	55	64.7%	2	2.0%	28	17.9%
C88	626	349	277	55.8%	12	3.4%	143	22.8%
C90	2564	2218	346	86.5%	62	2.8%	435	17.0%
C91	2373	1911	462	80.5%	79	4.1%	567	23.9%
C92	1504	1570	-66	104.4%	237	15.1%	203	13.5%
C93	278	187	91	67.3%	21	11.2%	82	29.5%
C94	142	78	64	54.9%	53	67.9%	112	78.9%
C95	51	66	-15	129.4%	10	15.2%	12	23.5%
C96	88	100	-12	113.6%	61	61.0%	53	60.2%
C00-C97 exc C44	167471	156228	11243	93.3%	8707	5.6%	19312	11.5%

Appendix 6 - False negative errors and basis of diagnosis

This appendix explores the reason for the overall age-dependence of the false negative error rate.

The most common methods of confirming a diagnosis (histology and cytology) account for the lowest proportion of false negatives (Figure A2). Where diagnosis comes from specific tumour markers, RCRD are much more likely to “miss” the significant event or events. Patients diagnosed clinically (from imaging, consultation by a doctor but without a pathological sample being taken) are also more likely to be “missed” in RCRD.

Those patients for whom a diagnosis method cannot be determined (unknown) or died before they could be offered cancer treatment (death certificate), are most likely to be “missed” in RCRD. As Figure A3 indicates though, these account for a small proportion of those falsely omitted from RCRD.

The marked reduction in the proportion of patients having their diagnosis confirmed from a pathological specimen (histology or cytology) explains the increase often observed at older ages in Figure A3, from the age of around 70, reflecting fewer patients having an invasive procedure performed on them as age increases. This is likely to be the reason behind the increasing false negative proportions by age observed overall and in most cancer groups (Figures 5 and 6).

Figure A2: The proportion of false negative diagnoses by cancer group and basis of diagnosis, England, April to September 2018

Note: Some cancer group and basis of diagnosis combinations excluded from graph, where the total number of cases in the group is 10 or fewer.

Figure A3: The proportion of false negative diagnoses within each basis of diagnosis, England, April to September 2018 (all cancer groups combined)

Appendix 7 - False positive and false negative proportion by month

Figure A4 shows the False Negative and False Positive error proportions by month for the broader matching criteria and a matching period of 90 and 30 days.

Figure A4: Monthly False Positive and False Negative proportions

Appendix 8 - Sensitivity testing of matching criteria

In this section, the sensitivity of RCRD is illustrated for different matching criteria.

As expected, the stricter the criteria about the timing of events, more errors (both false negative and false positive) are observed. Not including a match specification on cancer group (the bottom line of Table A6) improves both matching criteria and demonstrates that approximately 45% of false positive tumours have a cancer diagnosis of some sort when the necessity of matching by cancer group is removed.

Table A6: Proportions of false positive and negative errors under alternative matching criteria

Proportions of false positive and negative errors under alternative matching criteria

Tumour matching	Match within N days	False Negative %	False Positive %
Broader	90	11.5%	5.6%
Broader	60	13.1%	7.2%
Broader	30	18.7%	12.8%
Broader	14	29.6%	24.5%
Broader	7	46.0%	42.1%
Broader	0	81.3%	79.8%
Narrow	90	17.7%	12.6%
None	90	9.6%	3.0%

Appendix 9 - Code changes to the RCRD build process

In this section, code changes introduced in each monthly snapshot are described.

Table A7: RCRD change log

AT_RAPID_PATHWAY: event list

snapshot	change_id	code_change
CAS2609	CAS2609-1	Updates to various HNA related events (specifically 20, 24 and 57) to improve handling of COSDv10 data following the code updates applied in CAS2608.
CAS2609	CAS2609-2	Removal of records with null diagnosis code information from event 55 (biopsy records).
CAS2609	CAS2609-3	Updates to this QA and data quality document, to focus more clearly on ICD-10 C00-C97 exc C44, and refresh various aspects of the code base and formatting.
CAS2608	CAS2608-1	Code base updated for extraction of data from rawtext fields due to a change of underlying data format, but also including some wider improvements to code.
CAS2608	CAS2608-2	Geographies updated to reflect 2026 boundaries.
CAS2608	CAS2608-3	Deprivation updated to use information from the most relevant version of IMD (2019 or 2025) for the diagnosis year.
CAS2606	CAS2606-1	Cleaning of pathway data to avoid inclusion of rare condition records
CAS2605		None
CAS2604	CAS2604-1	Updated stage code to use a lookup table method rather than hard-coding values
CAS2604	CAS2604-2	Error corrected to avoid truncating the last character extracted from a COSD rawtext field.
CAS2603		None
CAS2602		None
CAS2601		None
CAS2512	CAS2512-1	Internal code changes to optimise build with regard to HES data
CAS2511	CAS2511-1	Update to HNA related events to include COSDv10 data and add de-duplication
CAS2510	CAS2510-1	Further updates to support assignment of C48 tumours to a cancer group depending on patient gender
CAS2510	CAS2510-2	Cleaning of odd values from stage field
CAS2510	CAS2510-3	Cleaning of dates of death prior to diagnosis
CAS2509	CAS2509-1	Further updates to support assignment of C48 tumours to a cancer group depending on patient gender
CAS2509	CAS2509-2	Update to include persons with death-only information in group of proxy tumours.
CAS2508	CAS2508-1	Further updates to support assignment of C48 tumours to a cancer group depending on patient gender
CAS2508	CAS2508-2	Minor changes to surgery lookup table to align with standard treatment reporting
CAS2508	CAS2508-3	Adding D48, D72, E85, M72 ICD-10 overall lookup table to align with current cancer registration practice
CAS2507	CAS2507-1	C53 and C57 staging values moved into STAGE field from EXPERIMENTAL_STAGE
CAS2507	CAS2507-2	C48 tumours now assigned to a cancer group depending on patient gender
CAS2507	CAS2507-3	Resective surgery lookup table better aligned with 2025 Cancer Flags output
CAS2506	CAS2506-1	Internal changes to deal with multiple NHSnumbers per personid
CAS2506	CAS2506-2	Internal changes to prepare for improvements to assigning C48 tumours
CAS2506	CAS2506-3	Further development of event 102 and 103 experiemental events
CAS2505		None
CAS2504	CAS2504-1	Further development of event 102 and 103 experiemental events
CAS2504	CAS2504-2	Update to basis of diagnosis code for 2023 cases onward to make consistent with updated registration practice
CAS2501	CAS2501-1	Permanent fix to enact deuplication of experimental event 102
CAS2412	CAS2412-1	Include staging of C53 (cervical cancer) in experimental stage field
CAS2412	CAS2412-2	Correcting issue that excluded rapidly fatal cancers being included from the HES data
CAS2412	CAS2412-3	Deduplication of experimental event 102 (hotfix)
CAS2412	CAS2412-4	Excluded lung screening Routes to Diagnosis prior to January 2019
CAS2411	CAS2411-1	Update to surgery code to use a combined table of all 3-digit ICD-10 codes, for all-stage and stage-specific procedures.
CAS2411	CAS2411-2	Filter OPCS4 procedure codes saved in initial HES tables, to include only those relevant to later lookups.
CAS2411	CAS2411-3	Added filtering to exclude Welsh only patients within the rapid_fatality section of event 101.
CAS2411	CAS2411-4	Two proposed new events, 102 and 103.
CAS2410	CAS2410-1	Refactored surgical lookup table code to be consistent with those used in treatment flag output
CAS2410	CAS2410-2	Added GP Practice code to tumour table
CAS2409	CAS2409-1	Added C33 to allowed list for lung screening
CAS2409	CAS2409-2	Updated NSPL postcode lookup to NSPL published May 2024
CAS2409	CAS2409-3	Internal refactoring of surgical lookup table to prepare for a simpler update process
CAS2409	CAS2409-4	Created internal experimental table showing patient GP practice at time of diagnosis
CAS2408	CAS2408-1	Changed criteria for including Event 54 in rapid pathway table such that there is a known nhsnumber instead of a known patient id (motivated by changes to COSD v10 data submissions)
CAS2407	CAS2407-1	Added STAGE_EXPERIMENTAL field
CAS2407	CAS2407-2	Added staging for C57 ovarian tumours (into STAGE_EXPERIMENTAL field)
CAS2407	CAS2407-3	Opened selection for screening cases to include C34 lung cancers
CAS2406		None
CAS2405	CAS2405-1	Updated assignment of trusts (reversing effect of cas2305-2 change), reducing numbers of patients diagnosed at tertiary trusts and increasing numbers diagnosed in near-by trusts.
CAS2405	CAS2405-2	Refactored order of properties in event 5 for consistency throughtout code while maintaining fix for ethnicity made in cas2404.
CAS2404	CAS2404-1	Fixed issue with ethnicity ‘top up’ from HES data which was incorrectly assigning ethinicity where it was present in HES but missing in COSD.
CAS2404	CAS2404-2	Update to allow creation of HES identified endocrine tumours based on event 11, restoring diagnoses previously identified from event 13.
CAS2403	CAS2403-1	Added place of death to event 19, property 3.
CAS2403	CAS2403-2	Merging event 13 into event 11 and event 16 into 14. This has the effect of no longer distinguishing surgery codes consistent with the CASSOP 4.5 with those specific to the RCRD build.
CAS2403	CAS2403-3	Add LSOA21 and age at diagnosis to AT_RAPID_TUMOUR table.
CAS2402		None
CAS2401		None
CAS2312	CAS2312-1	Update ICD-10 site lookup table to include more D-coded tumour groups.
CAS2311	CAS2311-1	Filter ethnicity to 1 digit only.
CAS2311	CAS2311-2	Updated postcode lookup table to nspl_202305.
CAS2311	CAS2311-3	Added filter to morphology codes to only allow those beginning with ‘8’ or ‘9’.
CAS2311	CAS2311-4	After review of fields removed ‘received_date’ from pathway table.
CAS2311	CAS2311-5	After review of fields removed event type 10 as an effective duplicate of event type 19.
CAS2310		None
CAS2309	CAS2309-1	Allow HES and CWT records to create event-type 52 and 53 events even if there is no patientid. Screen these out so that they don’t go on to create event-type 101 events, but are now available for testing.
CAS2308		None
CAS2307	CAS2307-1	Expose path and integrated TNM stage components in event 21.
CAS2307	CAS2307-2	Change offset for CWT diagnosis events to a fixed lookup table rather than re-calculating each time.
CAS2307	CAS2307-3	Update CWT surgery codes to reflect changes to CWT data dictionary.
CAS2307	CAS2307-4	Updated surgery lookup table to reflect changes implemented in cancer treatment flags output.
CAS2306	CAS2306-1	Move comparison of diagnosis date to date of death to earlier in the processing (and using vital status date for date of death if appropriate).
CAS2305	CAS2305-1	Remove duplicate patients with multiple patientid and same nhsnumbers.
CAS2305	CAS2305-2	Revert to prior order to prioritise creation of event 101s without prioritising those with a known trust.
CAS2305	CAS2305-3	Bring diagnosis trust through to AT_RAPID_TUMOUR table.
CAS2305	CAS2305-4	Added new basis of diagnosis codes to reflect changes to ENCR definitions for diagnoses from 2023 onwards.
CAS2305	CAS2305-5	Replace diagnosisdate with date of death for cases where date of death would otherwise have been within the 3 months before diagnosisdate.
CAS2304		None
CAS2303		None
CAS2302		None
CAS2301		None