An audit of the quality of cancer registration data

Integrating predictive coding and a user-centric interface for enhanced auditing and quality in cancer registry data

Data curation for a hospital-based cancer registry heavily relies on the labor-intensive manual abstraction process by cancer registrars to identify cancer-related information from free-text electronic health records. To streamline this process, a natural language processing system incorporating a hybrid of deep learning-based and rule-based approaches for identifying lung cancer registry-related concepts, along with a symbolic expert system that generates registry coding based on weighted rules, was developed. The system is integrated with the hospital information system at a medical center to provide cancer registrars with a patient journey visualization platform. The embedded system offers a comprehensive view of patient reports annotated with significant registry concepts to facilitate the manual coding process and elevate overall quality. Extensive evaluations, including comparisons with state-of-the-art methods, were conducted using a lung cancer dataset comprising 1428 patients from the medical center. The experimental results illustrate the effectiveness of the developed system, consistently achieving F1-scores of 0.85 and 1.00 across 30 coding items. Registrar feedback highlights the system's reliability as a tool for assisting and auditing the abstraction. By presenting key registry items along the timeline of a patient's reports with accurate code predictions, the system improves the quality of registrar outcomes and reduces the labor resources and time required for data abstraction. Our study highlights advancements in cancer registry coding practices, demonstrating that the proposed hybrid weighted neural-symbolic cancer registry system is reliable and efficient for assisting cancer registrars in the coding workflow and contributing to clinical outcomes.

Quality control on digital cancer registration

Population-based cancer registration methods are subject to internationally-established rules. To ensure efficient and effective case recording, population-based cancer registries widely adopt digital processing (DP) methods. At the Veneto Tumor Registry (RTV), about 50% of all digitally-identified (putative) cases of cancer are further profiled by means of registrars' assessments (RAs). Taking these RAs for reference, the present study examines how well the registry's DP performs. A series of 1,801 (putative) incident and prevalent cancers identified using DP methods were randomly assigned to two experienced registrars (blinded to the DP output), who independently re-assessed every case. This study focuses on the concordance between the DP output and the RAs as concerns cancer status (incident versus prevalent), topography, and morphology. The RAs confirmed the cancer status emerging from DP for 1,266/1,317 incident cancers (positive predictive value [PPV] = 96.1%) and 460/472 prevalent cancers (PPV = 97.5%). This level of concordance ranks as "optimal", with a Cohen's K value of 0.91. The overall prevalence of false-positive cancer cases identified by DP was 2.9%, and was affected by the number of digital variables available. DP and the RAs were consistent in identifying cancer topography in 88.7% of cases; differences concerned different sites within the same anatomo-functional district (according to the International Agency for Research on Cancer [IARC]) in 9.6% of cases. In short, using DP for cancer case registration suffers from only trivial inconsistencies. The efficiency and reliability of digital cancer registration is influenced by the availability of good-quality clinical information, and the regular interdisciplinary monitoring of a registry's DP performance.

Electronic medical record: research tool for pancreatic cancer?

BACKGROUND

A novel data warehouse based on automated retrieval from an institutional health care information system (HIS) was made available to be compared with a traditional prospectively maintained surgical database.

METHODS

A newly established institutional data warehouse at a single-institution academic medical center autopopulated by HIS was queried for International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis codes for pancreatic neoplasm. Patients with ICD-9-CM diagnosis codes for pancreatic neoplasm were captured. A parallel query was performed using a prospective database populated by manual entry. Duplicated patients and those unique to either data set were identified. All patients were manually reviewed to determine the accuracy of diagnosis.

RESULTS

A total of 1107 patients were identified from the HIS-linked data set with pancreatic neoplasm from 1999-2009. Of these, 254 (22.9%) patients were also captured by the surgical database, whereas 853 (77.1%) patients were only in the HIS-linked data set. Manual review of the HIS-only group demonstrated that 45.0% of patients were without identifiable pancreatic pathology, suggesting erroneous capture, whereas 36.3% of patients were consistent with pancreatic neoplasm and 18.7% with other pancreatic pathology. Of the 394 patients identified by the surgical database, 254 (64.5%) patients were captured by HIS, whereas 140 (35.5%) patients were not. Manual review of patients only captured by the surgical database demonstrated 85.9% with pancreatic neoplasm and 14.1% with other pancreatic pathology. Finally, review of the 254 patient overlap demonstrated that 80.3% of patients had pancreatic neoplasm and 19.7% had other pancreatic pathology.

CONCLUSIONS

These results suggest that cautious interpretation of administrative data rely only on ICD-9-CM diagnosis codes and clinical correlation through previously validated mechanisms.

Histological confirmation of breast cancer registration and self-reporting in England and Wales: a cohort study within the UK Collaborative Trial of Ovarian Cancer Screening

BACKGROUND

In research studies, accurate information of cancer diagnosis is crucial. In women with breast cancer (BC), we compare cancer registration (CR) in England/Wales and self-reporting with independent confirmation.

METHODS

In the UK Collaborative Trial of Ovarian Cancer Screening, notification of BC diagnosed between randomisation and 31 December 2009 was obtained through (1) CR (17 October 2011) and (2) self-reporting using postal-questionnaire. Breast cancer was confirmed using a detailed questionnaire (BC questionnaire BCQ) completed by the treating clinician (gold standard). Apparent sensitivity and positive-predictive value of CR/self-reporting vs BCQ were calculated.

RESULTS

Of 1065 women with possible BC notification, diagnosis was confirmed in 932 (87.5%). A total of 3.1% (28 out of 918) of BC CR and 12.4% (128 out of 1032) of women with self-reported BC only had in-situ carcinoma on BCQ. Another 4.6% (43 out of 932) of BCQ-confirmed cancer did not have a BC registration, and 3.6% (34 out of 932) did not self-report BC. Apparent sensitivity of CR and self-reporting vs BCQ were 95.4 and 96.4%, respectively. Positive-predictive value of self-reporting (87.1%) was significantly lower than that of CR (96.8%). Women aged<65 were more likely to over report in-situ carcinoma as BC. Overall, 73 (6.8%) women would have been misclassified/missed if CR, and 167 (15.6%) if self-reporting data alone was used.

CONCLUSION

This study confirms the reliability of BC registration in England/Wales and highlights the fact that 1 in 10 women self-reporting BC might only have in-situ breast carcinoma.

Tumor registry versus physician medical record review: a direct comparison of patients with pancreatic neuroendocrine tumors

PURPOSE

Tumor registry (TR) data are becoming more prominently cited in research through increased use of the National Cancer Database. We aimed to establish the accuracy of TR data by comparing them with physician medical record review (MD review) using pancreatic neuroendocrine tumors (NETs) as an example.

METHODS

For MD review, the health information system of an academic medical center was queried for patients with pancreatic International Classification of Diseases, ninth revision (ICD-9), codes from January 2000 to August 2008. A single physician investigator analyzed those medical records and identified patients with pancreatic NETs. For TR data, patients with pancreatic NETs were identified by two separate strategies. For the period of January 2000 to December 2006, patients were identified through manual review of pathology reports, admission and discharge sheets, and clinic visit logs. For January 2007 to August 2008, patients were identified using an automated case-finding program.

RESULTS

In MD review, 1,192 patients with pancreatic ICD-9 codes were identified, 34 of whom were found to have pancreatic NETs. The TR indicated 15 patients with pancreatic NETs, four of whom were not identified during MD review. Of the total 38 patients identified by either strategy, pancreatic NET identification rate of the TR was 39.5% compared with 89.5% in MD review.

CONCLUSION

Academic TR analysis indicates a substantial proportion of patients with pancreatic NETs are not identified when compared with MD review. Most instances of patients going unidentified are the result of registry time lag and case-finding methodologies; specifically, physicians may define tumors with malignant potential differently. This may be applicable to other individual tumor registries as well as aggregate registry-based national studies.

Monitoring indicators of health care quality by means of a hospital register of tumours

RATIONALE

Hospital registers of tumours provide, on a continuous basis, information on differences in patterns of neoplasias and the results of the treatment strategies employed.

OBJECTIVE

In view of the scant publications on measures of health care quality in hospital tumour registers, the aim of our paper is to present the outcome of a study to monitor the results related to health care quality in oncology.

METHODS

Data are presented for cases recorded at the Hospital Costa del Sol over a period of 8 years. The sources of information are fundamentally the patient's medical record and the database of the Pathology Department.

RESULTS

A high proportion of patients (mean 50%, range 45-68%) were admitted to the hospital by the Emergency Department; there was a notably long delay between the appearance of the first symptoms and the occasion of the first hospital visit (median 65 days; range 60-75 days). Particularly striking was the corresponding delay for breast cancer patients, in most cases superior to 3 months. As was the case for the percentage of admissions by the Emergency Department, most of the indicators evaluated in this study present a significant improvement compared with the initial years of the Hospital Register of Tumours. Thus, non anatomic-pathological diagnoses represented around 7% (range 3-13%), while 43% of patients (range 28-57%) were given adjuvant treatment in the form of radiation therapy or chemotherapy. In 40% of cases (range 20-50%), the tumour stage was included in the clinical record by the doctor who was treating the patient (in the remaining cases, these data were recorded by the Tumour Registry); the date of appearance of the first symptoms was included in the medical record in 65% of cases (range 54-80%). According to the stage classification, the following 5-year survival rates were recorded: (I) 98%, (II) 94%, (III) 69% and (IV) 39% for breast cancer; (I) 93%, (II) 83%, (III) 68% and (IV) 12% for cancer of the colon; and (I) 100%, (II) 94%, (III) 79% and (IV) 53% for prostate cancer.

CONCLUSION

The high percentage of patients admitted by the Emergency Department and the long delay between the appearance of the first symptoms reflect the deficient attention paid to this problem by patients and by primary health care services. Our results suggest that the Hospital Register of Tumours could constitute an excellent tool for monitoring the quality of health care systems for oncological patients.

Probabilistic classifiers and automated cancer registration: an exploratory application

A test of the performance of two probabilistic classifiers (random forests and multinomial logit models) in automatically defining cancer cases has been carried out on 5608 subjects, registered by the Venetian Tumour Registry (RTV) during the years 1987-1996 and manually checked for possible second cancers that occurred during the 1997-1999 period. An eightfold cross-validation was performed to estimate the classification error; 63 predictive variables were entered into the model fitting. The random forest allows to automatically classify 45% of subjects with a classification error lower than 5%, while the corresponding error is 31% for the multilogit model. The performance of the former classifier is appealing, indicating a potential drop of manually checked cases from 1750 to 960 per incidence year with a moderate error rate. This result suggests to refine the approach and extend it to other categories of manually treated cases.

Consistency and accuracy of diagnostic cancer codes generated by automated registration: comparison with manual registration

Background

Automated procedures are increasingly used in cancer registration, and it is important that the data produced are systematically checked for consistency and accuracy. We evaluated an automated procedure for cancer registration adopted by the Lombardy Cancer Registry in 1997, comparing automatically-generated diagnostic codes with those produced manually over one year (1997).

Methods

The automatically generated cancer cases were produced by Open Registry algorithms. For manual registration, trained staff consulted clinical records, pathology reports and death certificates. The social security code, present and checked in both databases in all cases, was used to match the files in the automatic and manual databases. The cancer cases generated by the two methods were compared by manual revision.

Results

The automated procedure generated 5027 cases: 2959 (59%) were accepted automatically and 2068 (41%) were flagged for manual checking. Among the cases accepted automatically, discrepancies in data items (surname, first name, sex and date of birth) constituted 8.5% of cases, and discrepancies in the first three digits of the ICD-9 code constituted 1.6%. Among flagged cases, cancers of female genital tract, hematopoietic system, metastatic and ill-defined sites, and oropharynx predominated. The usual reasons were use of specific vs. generic codes, presence of multiple primaries, and use of extranodal vs. nodal codes for lymphomas. The percentage of automatically accepted cases ranged from 83% for breast and thyroid cancers to 13% for metastatic and ill-defined cancer sites.

Conclusion

Since 59% of cases were accepted automatically and contained relatively few, mostly trivial discrepancies, the automatic procedure is efficient for routine case generation effectively cutting the workload required for routine case checking by this amount. Among cases not accepted automatically, discrepancies were mainly due to variations in coding practice.

Quality control of automatically defined cancer cases by the automated registration system of the Venetian Tumour Registry. Quality control of cancer cases automatically registered

BACKGROUND

In the Venetian Tumour Registry a substantial quota of cases (55%) is accepted using an algorithm that automatically evaluates diagnostic evidence: this study aims at assessing the reliability of the information produced in this way.

METHODS

A reabstraction study was conducted, which put a stratified sample of 1539 automatically accepted cases through a double-blind manual revision.

RESULTS

A significantly higher proportion of prevalent cases were found among breast, prostate and larynx cancer cases without microscopic confirmation, while there is a clear strong inverse relationship between the number of concordant diagnostic sources and the proportions of discordant diagnoses: cases based only on a single cytology record are particularly unreliable. A small number of multiple cancers are not detected because of one of the rules applied.

CONCLUSION

The overall proportion of incorrect decisions is not high and similar to those reported by other registries, but errors are correlated to the diagnostic evidence pattern. As a further check, we decided to revise clinical cases for the three sites mentioned manually, in order to reduce the numbers proportion of both prevalent cases, and all cytology-based diagnoses, so as to reduce the number of 'false positives'. Coverage of hospital discharge source has been extended in order to decrease the proportion of cases based only on pathology records.

Study of completeness of registration at the Estonian cancer registry

Every cancer registry should be able to quantify the level of completeness of registration. The current study describes a routine quality control procedure in the Estonian Cancer Registry (ECR) for assessing the completeness of registration. The registry's database was compared with the databases of the Tartu University Lung Clinic and the Maarjamõisa Hospital of the Tartu University Clinics, and active retrieval to obtain missing cancer cases diagnosed in 1998 was carried out. The overall completeness of case ascertainment based on this study was 90.8%. As a result of this procedure, 67 cases of malignant neoplasms (1.1% of the total number of incident cancer cases for 1998) and 11 cases of other reportable neoplasms were detected and recorded at the ECR. Cancers of the lung, thyroid gland and prostate were most frequently under-notified. For these sites, the number of cancer cases for 1998 for Estonia as a whole increased 2.6%, 11.8% and 2.2%, respectively. To conclude, the existence of electronic databases is a positive development, but cancer registrars still need to employ labour-intensive methods to validate diagnostic codes and to decide whether to include in the ECR cases found by active retrieval. Based on the findings of our study, which is the first one of its kind in Estonia, the completeness of cancer reporting varied by cancer site, and it appeared to be a substantial concern for several sites.

Determining the quality of breast cancer care: do tumor registries measure up?

BACKGROUND

Hospital tumor registries, which provide data that inform health services research and cancer control policies, may be a source of information about quality of cancer care. However, the accuracy of data from such registries is unknown.

OBJECTIVE

To determine the accuracy of tumor registry data by comparing it with data collected from numerous sources for a breast cancer quality improvement project.

DESIGN

Retrospective cohort study.

SETTING

Three teaching hospitals with tumor registries in the New York metropolitan area that had participated in the quality improvement project.

PATIENTS

All women with newly diagnosed primary breast cancer (stage I or stage II) who were surgically treated at the study hospitals between 1 November 1994 and 31 August 1996.

MEASUREMENTS

Sensitivity and specificity were calculated, and data from the quality improvement project were used as the gold standard.

RESULTS

The tumor registries and the quality improvement project had similar information on tumor stage and surgery type. Sensitivity ranged from 0.91 to 0.96, and specificity ranged from 0.93 to 0.97. When both sources were used to calculate quality measures, the overall rate of radiation therapy after breast-conserving surgery was 80% in the quality improvement project and 48% in the tumor registries (sensitivity, 0.58; specificity, 0.94). For receipt of adjuvant systemic treatment, the rate was 78% in the quality improvement project and 22% in the tumor registries (sensitivity, 0.27; specificity, 0.97).

CONCLUSIONS

Data from tumor registries provide accurate measures for hospital-based surgical treatments but not for outpatient treatments. Unverified tumor registry data should not be used to measure quality of care.

Benefits and limitations of pathology databases to cancer registries

In order to assess the benefits and limitations of pathology databases to cancer registries, computerised pathology records of malignant neoplasms diagnosed during 1992 were obtained for a defined area of Scotland for which pathology data were not routinely being used for cancer registration. Apparently 'missed' cancer registrations were identified by computerised probability matching with cancer registration records and their eligibility for registration was determined by reference to medical records, or when these were unavailable, by reference to the text of the original pathology report in conjunction with the local Community Health Index (to establish residency at the time of diagnosis). Misclassifications of site or incidence year were not regarded as 'missed' cases. Of 218 apparently 'missed' cancer registrations identified from computerised pathology records, 133 (5.7% of the revised total number of registrations for the study area in 1992) should have been registered. A further 14 cases were already registered but with misclassified site, morphology and/or behaviour codes. Ascertainment of cases by the Scottish Cancer Registration Scheme seems to be high for most sites. Pathology databases represent a useful additional source of cases but the fact that 71 apparently 'missed' cases were found to be ineligible for registration as independent primary malignant neoplasms suggests that unverified computerised pathology data should not be used uncritically nor independently for cancer registration purposes.

Quality of prostate cancer data in the cancer registry of Norway

Completeness of reporting and internal validity of the coding of prostate cancer in the Cancer Registry (CR) in Norway were examined. Data were matched and evaluated against diagnostic indices at eight selected hospitals in the country and against death certificates from Statistics Norway. Validity control was based on detailed re-analysis of an approximately 1% sample of the registered data during the period 1957-1986. The deficiency in reporting of prostate cancer was less than 1%. The grave deficiencies in hospital patient registers were considered to be of non-systematic nature and should, therefore, not impair the reliability of our investigation of incompleteness. The validity control revealed errors in 0.5% of the data elements, or, illustrated differently, 6% of the patient files had an error, of importance or not, in one of the data elements. One false positive registration was found among 298 controlled patient files (0.3%).

Registration of non-melanoma skin cancers in Scotland--how accurate are site and morphology codes?

Although under-reporting of non-melanoma skin cancers to cancer registries is widely acknowledged, less is known about the accuracy of information held about registered cases. In 1993, the accuracy of a random sample of cancer registrations in Scotland attributed to the year 1990 was assessed by reference to relevant medical records. The sample contained 290 registrations of non-melanoma skin cancers, 251 (90%) of which had records available for scrutiny. Here we report the results of a detailed analysis of the accuracy of site and morphology coding of non-melanoma skin cancers. Following reabstraction of details from available medical records, only three cases (1.2%) did not retain the same first three digit ICD-9 site code (173.--), although a further three cases were judged to have been registered in error. There were 56 (21.5%) discrepancies in morphology coding, but 21 of these arose through inferences about morphology in the absence of microscopic confirmation, and most of the remainder were of a relatively minor nature. In summary, it does seem possible to collect data about non-melanoma skin cancers to a reasonably high standard of accuracy. This provides some justification to those who advocate an increased effort directed towards improving levels of completeness of case ascertainment.

Reliability of data of the Thames cancer registry on 673 cases of colorectal cancer: effect of the registration process

OBJECTIVE

To measure the reliability of data collected by the Thames cancer registry and to identify factors in the registration process affecting reliability.

DESIGN

A retrospective study of data from the registry, including death certificate only registrations, and hospital case notes on cases of colorectal cancer diagnosed in 1983 or 1988.

SETTING

Four districts in South Thames region.

SUBJECTS

673 cases of colorectal cancer in resident patients.

MAIN MEASURES

Dates of birth, diagnosis of cancer, and death; sex; tumour site; whether treatment was given; type of treatment; and district of residence.

RESULTS

Among the 416 (62%) case notes retrieved, including 66 death certificate only registrations, full or high agreement between registry data and hospital notes was recorded for sex, district of residence, and dates of birth and death. Only 12% of cases had the same date of diagnosis, which may be due to failure of registry policy. Lower agreement rates occurred for tumour site (87%), whether treatment occurred (84%), and treatments administered (80%, 1983; 72%, 1988). 20% of surgical treatments and 37% of adjuvant therapy, radiotherapy, and chemotherapy were not recorded by the registry. Disagreements were common among death certificate only registrations. Such registrations accounted for 16(32%) disagreements over tumour site, 33(41%) major disagreements over date of diagnosis (difference > 30 days), and 47(44%) disagreements over treatment. In 65 cases the registry failed to capture all treatments carried out within the six month follow up period, 38(58%) of which were for death certificate only registrations. In 36% of death certificate only registrations the patients survived more than one year from diagnosis, indicating a failure of registry policy over retrospective follow up.

CONCLUSIONS

Registry data on district of residence; sex; dates of birth, diagnosis, and death are highly reliable, but treatment and tumour site data are less so. Lack of follow up in death certificate only registrations and failure to monitor treatments during follow up period seemed to be associated with disagreements.

Registration of lung cancer in Scotland: an assessment of data accuracy based on review of medical records

Lung cancer represents a major public health problem in Scotland. Cancer registration data permit the approximate incidence of this disease to be measured directly and the projected incidence to be modelled. Thus, in addition to epidemiologic studies and survival analyses, cancer registration data may be used for planning and monitoring relevant health services. Since the value of the data depends on their quality, we undertook a large-scale study of the accuracy of cancer registration data in Scotland. The medical records of a random sample of cancer registrations attributed to the year 1990 were sought. The sample contained 340 registrations of lung cancer, 309 (91 percent) of which had relevant medical records available for scrutiny. Registration details were reabstracted from available records and compared with data in the registry. Results revealed 19 discrepancies in identifying items of data (surname, forename, gender, and date of birth) involving 16 (5.2 percent) patients. Most were trivial and would not disturb record linkage. Discrepancy rates were found to be: 7.8 percent in postcode of residence at the time of diagnosis, 10 percent in 'anniversary date' (excluding differences of six weeks or less), 12.5 percent in histologic verification status; 4.2 percent in ICD-9 site code (the first three digits), and 15.5 percent in four digit ICD-O morphology code (excluding 'inferred' morphology codes). This relatively high level of accuracy gives weight to routinely published incidence figures and supports the use of these data for exploratory epidemiologic studies, assessment of health care needs, and calculation of survival.

Registration of colorectal cancer in Scotland: an assessment of data accuracy based on review of medical records

Colorectal cancer accounts for a substantial burden of morbidity and mortality in the population. While the need for reliable incidence data may be self-evident, the quality of cancer registration data has rarely been assessed. In Scotland during 1993, the medical records of a random sample of cancer registrations attributed to the year 1990 were sought. The sample contained 238 registrations of colorectal cancer, 217 (91%) of which had relevant medical records available for review. Registration details were reabstracted from available records and compared with data in the registry. Discrepancies in identifying items of data (surname, forename, sex and date of birth) were recorded in eight cases (3.7%, 95% confidence intervals 1.2-6.2%). None would have disturbed record linkage. Discrepancy rates of 3.7% (1.2-6.3%) in postcode of residence at the time of diagnosis (excluding differences arising through boundary changes), 8.3% (4.6-12.0%) in 'anniversary date' (excluding differences of 30 days or less), 2.8% (0.6-5.0%) in histological verification status and 13.5% (8.9-18.1%) in morphology code (excluding 'inferred' morphology codes) were recorded. Twelve cases (5.5%, 2.5-8.6%) were deemed not to warrant site codes for cancer of the colon (ICD-9 153) or rectum (ICD-9 154). In many respects, therefore, the data held about registrations of colorectal cancer in Scotland appear to show a high level of accuracy. Completeness of case ascertainment has still to be formally assessed.

Improving the quality of cancer registration data

Cancer registration is an essential element of any cancer control strategy. Data quality is, however, of paramount importance. This paper sets out some of the ways in which the quality of cancer registration data might be improved. In particular, the potential contribution of clinicians and pathologists is highlighted.

How accurate are Scottish cancer registration data?

In order to assess the accuracy of Scottish cancer registration data, a random sample of 2,200 registrations, attributed to the year 1990, was generated. Relevant medical records were available for review in 2,021 (92%) cases. Registration details were reabstracted from available records and compared with data in the registry. Discrepancies in identifying items of data (surname, forename, sex and date of birth) were found in 3.5% of cases. Most were trivial and would not disturb record linkage. Discrepancy rates of 7.1% in post code of residence at the time of diagnosis (excluding differences arising through boundary changes), 11.0% in anniversary date (excluding differences of 6 weeks or less), 7.7% in histological verification status, 5.4% in ICD-9 site codes (the first three digits) and 14.5% in ICD-O morphology codes (excluding 'inferred' morphology codes) were recorded. Overall, serious discrepancies were judged to have occurred in 2.8% of cases. In many respects, therefore, Scottish cancer registration data show a high level of accuracy that compares favourably to the reported accuracy of the few other cancer registries undertaking such analyses.

Registration quality and descriptive epidemiology of childhood brain tumours in Scotland 1975-90

Children (0-14 years) with malignant brain and central nervous system (CNS) tumours (ICD9 191 and 192) were listed from the Scottish Cancer Registration Scheme for the years 1975-90. These cases formed the basis for validation and verification procedures aimed at providing a complete and accurate data set for epidemiological analyses. A variety of data sources were cross-checked to optimise ascertainment, and resulting from this 5.7% of validated cases were found on the cancer registry with diagnostic codes outside the ICD-9 range 191-192. A further 8.4% were newly registered cases. Analyses were conducted on the validated data set showing a significant temporal increase in incidence rates over the 16 year study period with an average annual percentage change of +2.6%. Large-scale geographical heterogeneity was also found, with a particularly high incidence in the Fife and Lothian areas and a low incidence in Grampian. Examination of associations with socioeconomic status, using the Carstairs deprivation index, revealed a rising trend in incidence strongly linked to areas with increasing levels of affluence. Our results suggest that for studies of childhood CNS tumours validation of cancer registry data is necessary and large-scale geographical variation and socioeconomic factors should be taken into account in any investigation of distribution in small geographical areas.

Quality of cancer registry data: a comparison of data provided by clinicians with those of registration personnel

The quality of cancer registry data is of great importance to the usefulness of a cancer registry. To investigate the quality of its data the IKL cancer registry (Integraal Kankercentrum Limburg) performed a study with the aim of comparing data supplied by clinicians with data collected by registration personnel. Twenty clinicians reabstracted the information of a random sample of about ten of their patients, who were diagnosed with cancer in 1989 or 1990. After coding, the information was compared with the contents of the cancer registry records. For comparison of agreement the information of 190 cases was available. The relative frequency of major disagreements was 0% for date of birth, 0% for gender, 5% for date of incidence, 6% for primary site, 2% for laterality, 2% for histologic type and 2% for behaviour code. In general, the disagreements could be attributed to the handling of different coding rules (incidence date), or to a lower level of precision by the clinician in comparison to registration personnel (primary site, laterality). This study has shown that registration personnel are able to collect data with a high degree of accuracy.