Cancer registries in epidemiologic research

Validation of primary and outcome data quality in a Swedish population-based breast cancer quality registry

BACKGROUND

Population-based cancer quality registries are of great importance for the improvement of cancer care. However, little is known about the quality of recurrence data in cancer quality registries. The aim of this study was to evaluate data quality in the regional Breast Cancer Quality Registry of Central Sweden, with emphasis on the validity of recorded information on recurrence.

METHODS

Validation by re-abstraction was performed on a random sample of 800 women with primary invasive breast cancer stage I-III diagnosed between 1993 and 2010, of which 400 had at least one registered recurrence and 400 had no registered recurrence. Registry data were compared with data from medical records. Exact agreement, correlation and kappa values, sensitivity and specificity were calculated.

RESULTS

Seven hundred forty-seven women (93%) were available for analysis. Exact agreement was high for diagnostics, tumor characteristics, surgery, and adjuvant oncological treatment (90% or more for most variables). The registry's sensitivity was low for regional recurrence (47%), but higher for local and distant recurrence (80% and 75%), whereas specificity was overall high (≥ 95%). Combining all recurrence categories irrespective of localization improved sensitivity to 90% with a specificity of 91%. In 87% of women, the date of first recurrence according to medical records fell within ± 90 days of the date recorded in the registry.

CONCLUSIONS

While the quality of data in the regional Breast Cancer Quality Registry was generally high, data accuracy on recurrences was lower. The overall precision of identifying any recurrence, irrespective of localization, was high. However, the accuracy of classification of recurrences (local, regional or distant) was lower, with evidence of underreporting for each of the recurrence categories. Given the importance of recurrence-related outcomes in the assessment of quality of care, efforts should be made to improve the reporting of recurrences.

Unsupervised anomaly detection of implausible electronic health records: a real-world evaluation in cancer registries

BACKGROUND

Cancer registries collect patient-specific information about cancer diseases. The collected information is verified and made available to clinical researchers, physicians, and patients. When processing information, cancer registries verify that the patient-specific records they collect are plausible. This means that the collected information about a particular patient makes medical sense.

METHODS

Unsupervised machine learning approaches can detect implausible electronic health records without human guidance. Therefore, this article investigates two unsupervised anomaly detection approaches, a pattern-based approach (FindFPOF) and a compression-based approach (autoencoder), to identify implausible electronic health records in cancer registries. Unlike most existing work that analyzes synthetic anomalies, we compare the performance of both approaches and a baseline (random selection of records) on a real-world dataset. The dataset contains 21,104 electronic health records of patients with breast, colorectal, and prostate tumors. Each record consists of 16 categorical variables describing the disease, the patient, and the diagnostic procedure. The samples identified by FindFPOF, the autoencoder, and a random selection-a total of 785 different records-are evaluated in a real-world scenario by medical domain experts.

RESULTS

Both anomaly detection methods are good at detecting implausible electronic health records. First, domain experts identified [Formula: see text] of 300 randomly selected records as implausible. With FindFPOF and the autoencoder, [Formula: see text] of the proposed 300 records in each sample were implausible. This corresponds to a precision of [Formula: see text] for FindFPOF and the autoencoder. Second, for 300 randomly selected records that were labeled by domain experts, the sensitivity of the autoencoder was [Formula: see text] and the sensitivity of FindFPOF was [Formula: see text]. Both anomaly detection methods had a specificity of [Formula: see text]. Third, FindFPOF and the autoencoder suggested samples with a different distribution of values than the overall dataset. For example, both anomaly detection methods suggested a higher proportion of colorectal records, the tumor localization with the highest percentage of implausible records in a randomly selected sample.

CONCLUSIONS

Unsupervised anomaly detection can significantly reduce the manual effort of domain experts to find implausible electronic health records in cancer registries. In our experiments, the manual effort was reduced by a factor of approximately 3.5 compared to evaluating a random sample.

Evaluation of Data Quality of Four New Population Based Cancer Registries (PBCRs) in Chandigarh and Punjab, North India- A Quality Control Study

BACKGROUND

Population based Cancer Registries(PBCRs) are hallmark of cancer surveillance and cancer control activity .The value of cancer registries rely heavily on underlying quality of their data. Current study assessed data quality of four new PBCRs of Chandigarh, SAS Nagar, Mansa and Sangrur covering a total population of 4.5 millions on three quality parameters i.e. comparability, validity and completeness as recommended by International Agency of Research on Cancer(IARC), Lyon, France.

METHODS

For assessing comparability, data of the registries were reviewed in terms of system of classification and coding, definition of incidence date and rule for multiple primaries. For assessing validity (Accuracy) four different methods i.e. re-abstraction and re-coding, percentage morphologically verified cases (MV%), percentage of death certificate only (DCO%) cases and percentage of cases with other and unspecified sites (O and U%) were used. For assessing completeness of coverage, different semi-quantitative methods were used.

RESULTS

Re-abstraction done for 10% of the total incident cases yielded overall percentage agreement of 97.4%, 97.2%, 95.4% and 94.9% for PBCR Chandigarh, SAS Nagar, Mansa and Sangrur respectively. MV% was found to be 96.3% for PBCR Chandigarh, 92.8% for PBCR SAS Nagar , 89.3% for PBCR Mansa and 82.9% for PBCR Sangrur. Percentage of DCO cases and O and U cases were 1.4% and 2.8% for PBCR Chandigarh, 3.9% and 5.3% for SAS Nagar, 6.4% and 16.4% for Mansa and 6.3% and 8.3% for Sangrur. Completeness assessed through the various methods showed good level of completeness at PBCR Chandigarh and SAS Nagar and somewhat lower but acceptable level of completeness at PBCR Mansa and Sangrur.

CONCLUSIONS

All the four PBCRs are comparable internationally. PBCR Chandigarh and SAS Nagar, predominantly urban registries, have higher accuracy of their data and good completeness levels as compared to predominantly rural registries of Mansa and Sangrur. Cancer estimates given by all the four registries are reliable and data from these registries can be utilized for planning cancer prevention and control activities in the region.

Cancer staging at diagnosis data comparisons in South Australia

Cancer stage at diagnosis is an important gap for Australian population based cancer registries. The study aims to understand the quality and completeness of three different collections of cancer staging data. The South Australian Cancer Registry data collection for breast and colorectal cancer (CRC) cases diagnosed in 2011, was linked to Registry Derived Stage (RDS) data, pathology plus hospital metastasis codes (pathology stage), and the South Australian Clinical Cancer Registry Stage (SACCR stage). The agreement between staging systems was examined using kappa statistics. Kaplan-Meier curves and Cox regression were used to examine the difference in survival by staging methods. Among 2,530 breast and CRC cases 98.8% were stageable (n = 2,500) according to histology. Among stageable cases, 84.6% had RDS, 51.2% had pathology stage and 29.5% had SACCR stage. The kappa statistic for RDS and pathology stage was 0.930 for breast cancer and 0.973 for CRC, and 0.574 for RDS and SACCR stage for breast cancer and 0.632 for CRC. The agreement between pathology stage and SACCR stage was 0.430 for breast cancer and 0.528 for CRC. The distribution of stage was similar across staging methods, although more stage four cancers by pathology stage, and survival patterns were similar but not the same. The agreement was high between different staging systems. Pathology stage had a higher than expected stage 4 proportion. This study highlights an opportunity to collect stage information in a cost-effective manner, while collecting data that usefully represent stage at diagnosis across the population, for population based epidemiological analyses.

Increased incidence of thyroid cancer in Latina, Italy: a possible role of detection of subclinical disease

OBJECTIVE

To describe the thyroid cancer incidence trends and geographical patterns in the Latina Province of Lazio, Italy using the population-based cancer registry.

METHODS

We extracted from the Latina cancer registry all cases of thyroid cancer from 1997 to 2006. Cases were classified according to morphological type and diameter. Data for diagnostic procedures for Latina Province residents from 2001 to 2006 were extracted from the regional outpatient procedures information system.

RESULTS

A total of 982 cases were diagnosed, for a standardized incidence of 8.3 and of 27.9 per 100,000 in males (n, 220) and in females (n, 762), respectively. The annual percent change (APC) was +16.7% (95% CI +7.2, +27.2) and +10.5% (95% CI +6.5, +14.6) in males and females, respectively. The increase was mostly due to papillary (n, 759) and small (≤20 mm) cancers (n, 617), with no difference by age (<45 years; n, 431). The APC of neck ultrasound performed was +8.7% (95% CI +0.1, +18.1) and +9.0% (95% CI +1.1, +17.4) and that of biopsy/cytology was +17.0% (95% CI +13.0, +21.3) and +16.6% (95% CI +6.2, +28.1) in men and women, respectively. The geographic pattern of biopsy/cytology was similar to that of cancer incidence but not to that of neck ultrasound.

CONCLUSIONS

In Latina, the increase in thyroid cancer incidence was more rapid than in the rest of Italy, particularly for types with a good prognosis. While tumor size and histotype suggest an increase in detection instead of an increase in disease occurrence, data on diagnostic procedure reimbursements cannot provide an explanation.

Cancer suppression at old age

Increased age is regularly linked with heightened cancer risk, but recent research suggests a flattening around age 80. We report that, independent of cancer site or time period, most incidence rates decrease in the more elderly and drop to or toward zero near the ceiling of human life span. For all major organ sites, male and female, we use 1979 to 2003 Surveillance, Epidemiology, and End Results registry records (8-26% of the U.S. population) to construct three sequential cross-sections at 10-year intervals, totaling 129 sets of age-specific cancer data. To compute incidence rates, we estimate older populations at risk with census counts and NIH life tables. This article provides both a minimal and a more comprehensive extension of Surveillance, Epidemiology, and End Results cancer rates to those above 85. Almost all cancers peak at age approximately 80. Generally, it seems that centenarians are asymptomatic or untargeted by cancers. We suggest that the best available justification for this pattern of incidence is a link between increased senescence and decreased proliferative potential among cancers. Then, thus far, as senescence may be a carcinogen, it might also be considered an anticarcinogen in the elderly. We model rising and falling incidence rates with a beta curve obtained by appending a linearly decreasing factor to the well-known Armitage-Doll multistage model of cancer. Taken at face value, the beta model implies that medical, diet, or lifestyle interventions restricting carcinogenesis ought to be examined for possible effects on longevity.

Comparison with manual registration reveals satisfactory completeness and efficiency of a computerized cancer registration system

Automated software for cancer registration, called Open Registry and developed by ourselves was adopted by the Varese (population-based) Cancer Registry starting from 1997. Since the use of automated cancer registration is increasing, it is important to assess the quality and completeness of the automated data being produced. In this study, we assessed the completeness of the automatically generated data by comparison with a gold standard of all cases identified by manual and automatic systems for the year 1997 when the automated system was introduced, and the manual system was still in operation. We also evaluated the efficiency of the automated system. 5027 cases were generated automatically; 2959 (59%) were accepted automatically and 2068 (41%) were flagged for manual checking. Sixty-nine cases (1.3%) were not recorded automatically, the most common reason (0.8%) being that the incidence record was dated 1998, even though the case was incident in 1997. A total of 98.7% of all cases found were picked up by the automated system. A completeness figure of 98.7% indicates that the automatic procedure is a valid alternative to manual methods for routine case generation. The fact that 59% of cases were registered automatically indicates that the system can speed up data production and enhance registry efficiency.

Do incident and new subsequent cases of non-melanoma skin cancer registered in a Danish prospective cohort study have different 10-year mortality?

BACKGROUND

The Danish Gerda Frentz Cohort (GFC) was created for registering all incident and new subsequent cases of non-melanoma skin cancer (NMSC) among patients seen by Danish dermatologists in 1995. We have recently found, in this cohort, a lower 10-year mortality than in the general population in patients with basal cell carcinoma (BCC). Differences in mortality between incident and new subsequent cases, incomplete registration or selection bias may be responsible for this finding.

METHODS

We aimed to quantify differences in mortality between incident and new subsequent cases of NMSC in the GFC and to compare mortality among incident cases recorded in the GFC and those recorded in the Danish Cancer Registry (DCR). We followed 10,830 skin cancer patients and 106,696 age-, gender- and residence-matched population controls through 2006 and computed their cumulative mortality and mortality rate ratio (MRR).

RESULTS

One-, 5-, and 10-year cumulative mortality of incident and new subsequent cases of BCC and SCC in the GFC were similar. Likewise, MRR for incident BCC (MRR=0.91; 95% CI 0.84-0.98) and incident SCC (MRR=1.29; 95% CI 1.05-1.56) among patients registered in the GFC were similar to their counterparts in the DCR (MRR=0.96; 95% CI 0.91-1.00 and MRR=1.36; 95% CI 1.22-1.52).

CONCLUSION

Mortality of incident and new subsequent cases of NMSC was similar and thus did not explain the reduced mortality of BCC patients.

Risk of breast cancer after night- and shift work: current evidence and ongoing studies in Denmark

Breast cancer is the most frequent cancer among women, and the number is increasing worldwide. This tumour is strongly associated with Western lifestyle, but the specific risk factors behind this observation are not well known. Exposure to light-at-night, including disturbance of the circadian rhythm, possibly mediated via the melatonin synthesis and clock genes, has been suggested as a contributing cause of breast cancer. Since shift- and night-time work is prevalent and increasing in modern societies, this exposure may be of public health concern, and contribute to the continuing elevation in breast cancer risk. Until now only few epidemiological studies have evaluated breast cancer risk after shift and night work. Although these studies are all suffering from methodological problems, especially concerning assessment of light exposure, results have consistently shown an increase in risk associated with night and shift work. Good opportunities for epidemiological cancer research exist in Denmark, and several studies on different aspects of breast cancer, work schedules, light exposure and melatonin levels are ongoing in order to further examine different aspects of this issue.

Cancer Surveillance Research: a Vital Subdiscipline of Cancer Epidemiology

Public health surveillance systems relevant to cancer, centered around population-based cancer registration, have produced extensive, high-quality data for evaluating the cancer burden. However, these resources are underutilized by the epidemiology community due, we postulate, to under-appreciation of their scope and of the methods and software for using them. To remedy these misperceptions, this paper defines cancer surveillance research, reviews selected prior contributions, describes current resources, and presents challenges to and recommendations for advancing the field. Cancer surveillance research, in which systematically collected patient and population data are analyzed to examine and test hypotheses about cancer predictors, incidence, and outcomes in geographically defined populations over time, has produced not only cancer statistics and etiologic hypotheses but also information for public health education and for cancer prevention and control. Data on cancer patients are now available for all US states and, within SEER, since 1973, and have been enhanced by linkage to other population-based resources. Appropriate statistical methods and sophisticated interactive analytic software are readily available. Yet, publication of papers, funding opportunities, and professional training for cancer surveillance research remain inadequate. Improvement is necessary in these realms to permit cancer surveillance research to realize its potential in resolving the growing cancer burden.

Cancer incidence in postwar Lebanon: findings from the first national population-based registry, 1998

PURPOSE

Cancer incidence rates in Lebanon have been lacking for over three decades. National data based on a total of 4388 cases diagnosed during the year 1998 were reviewed and analyzed.

METHODS

Crude and age-standardized rates (ASRs) per 100,000 population were calculated and results were contrasted with estimates from developed and selected developing countries in the region.

RESULTS

Among males, bladder (18.5%), prostate (14.2%), and lung cancer (14.1%) were the most frequently reported malignancies. Among females, breast cancer alone accounted for over a third of all cancers, followed by colon cancer (5.8%), and cancer of the corpus uteri (4.8%). Sex-differentials in incidence rates were highest for tobacco-related cancers (lung, larynx, and bladder). Compared with current estimates worldwide, ASRs for bladder cancer in Lebanon showed strikingly high rates. Whereas ASRs for breast and prostate cancer remained lower than those observed in developed countries, they were greater than those estimated from neighboring countries with a similar epidemiological transition as Lebanon.

CONCLUSIONS

Findings of the comparative assessments most likely reflect differentials in prevalence of risk factors and lifestyle variables (e.g., lung and breast cancers) and can be partly explained by improvement in cancer detection rate in recent years (for prostate cancer). The implications of the results in light of primary prevention activities, screening practices, and research initiatives in Lebanon are discussed.

Global registries for measuring pharmacoeconomic and quality-of-life outcomes: focus on design and data collection, analysis and interpretation

Disease registries have traditionally been vehicles for the collection of clinical data, in most instances following a large number of patients for a long time period in an observational manner, and enhancing our understanding of disease aetiology and epidemiology. However, over recent decades, the potential for additional data collection and analyses to be conducted within the framework of a registry has been recognised and utilised. This is evident by the sheer number of registries that are now referenced in the medical literature, covering a vast array of therapeutic areas and topics much more varied than incidence, prevalence and survival. The opportunity to collect QOL and pharmacoeconomic data has been utilised within the registry framework as more and more countries have increased their demands for such information for regulatory procedures, including pricing and reimbursement decisions. This increased need for information has led to a marked increase in the number of registries undertaken that are primarily sponsored by the pharmaceutical industry. Disease registries offer tremendous opportunities to realise improvements in care. The length of data collection and the large number of patients involved offer some unusual advantages for QOL and health economic analyses; however, these advantages are not yet fully exploited.

The potential and limitations of data from population-based state cancer registries

Cancer incidence varies markedly among states because of population heterogeneity regarding risk, genetic, and demographic factors. Population-based cancer registries are essential to monitoring cancer trends and control. The Centers for Disease Control and Prevention and the North American Association of Central Cancer Registries, through the National Program of Cancer Registries, are helping state registries generate more and better data nationwide. The National Program of Cancer Registries has supported the enhancement of 36 registries and the creation of 13 new registries in 45 states, 3 territories, and the District of Columbia, providing national standards for completeness, timeliness, and quality; financial support; and technical assistance. Users must be aware of diverse issues that influence collection and interpretation of cancer registry data, such as multiple cancer diagnoses, duplicate reports, reporting delays, misclassification of race/ethnicity, and pitfalls in estimations of cancer incidence rates. Attention to these issues and intense use of the available data for cancer surveillance will enable maximum societal benefit from the emerging network of population-based state cancer registries.