Registries,Databases and Repositories for Developing Artificial Intelligence in Cancer Care

Modern artificial intelligence techniques have solved some previously intractable problems and produced impressive results in selected medical domains. One of their drawbacks is that they often need very large amounts of data. Pre-existing datasets in the form of national cancer registries, image/genetic depositories and clinical datasets already exist and have been used for research. In theory, the combination of healthcare Big Data with modern, data-hungry artificial intelligence techniques should offer significant opportunities for artificial intelligence development, but this has not yet happened. Here we discuss some of the structural reasons for this, barriers preventing artificial intelligence from making full use of existing datasets, and make suggestions as to enable progress. To do this, we use the framework of the 6Vs of Big Data and the FAIR criteria for data sharing and availability (Findability, Accessibility, Interoperability, and Reuse). We share our experience in navigating these barriers through The Brain Tumour Data Accelerator, a Brain Tumour Charity-supported initiative to integrate fragmented patient data into an enriched dataset. We conclude with some comments as to the limits of such approaches.     

A national stroke quality register: 12 years experience from a participating hospital

Registration of all hospitalized stroke patients is practiced in Sweden in order to assess care quality. Data in this register, Riks-Stroke (RS), may be biased due to incomplete registration. The purpose of this paper was to report changes in stroke outcome in relation to fluctuations in registration. Patients registered in RS at a hospital during the period 1994–2005 were analyzed. Case fatality at 28 days, living conditions, and activities of daily living (ADL) performance at 3 months were correlated to the number of patients registered and follow-up frequency. A total of 4994 stroke cases were registered during the period. A high annual registration rate was significantly correlated to a high case fatality ratio. A low annual follow-up rate was associated with a low proportion of patients living in their own home without any need of help. Quality parameters are sensible for selection bias, which make them difficult to compare over time and between hospitals. We suggest that by weighing outcome data against stroke severity, safer conclusions may be drawn. Additionally, hospitals considering setting up quality registers should make every effort to attain complete case ascertainment at all times, including patients managed outside the hospital, in order to avoid selection bias.     

Development and results of the Spanish registry of patients with alpha-1-antitrypsin deficiency

The Spanish registry of alpha-1 antitrypsin deficiency was founded in 1993 and became a member of the International Registry (AIR) in 1999. We describe the updating process following its incorporation into AIR and compare the data collected in the first period (1993–1999) and the second period (1999–2005), during which time patients were included exclusively by internet.The registry included 301 patients during period 1, 69% males and 46% had a history of smoking. Their mean age was 46 years (SD = 13) and 284 (94%) had the ZZ phenotype, 49% received augmentation therapy. During period 2, 161 new cases were included, 63% of whom were males with a mean age of 44 years (SD = 16). A total of 126 (78%) had the ZZ phenotype. Only 12% received augmentation therapy. A total of 462 different patients were included in both periods. Significant differences were observed in the number of cases with the SZ phenotype and the severity of FEV₁ impairment between the two periods.Implementation of an internet-based collection of data did not result in a lower rate of reporting to the registry. However, data from a significant number of patient included in period 1 could not be actualized in the new data base.     

Methodology used for "software for automated linkage in Italy" (SALI)

Linkage of epidemiological registries can provide cost-effective information on the associations between different diseases or exposures in the population under study and on completeness of surveillance system databases. We describe the program SALI (software for automated linkage in Italy) aimed at matching individual records from medium-sized registries (in the order of 100,000 records), where the desired outcome is to miss as few links as possible and, because of low link-likelihood (< 1%), a manual revision of matched pairs is feasible. SALI, developed in CA-Clipper language, uses registry files in dBase format. It requires only name, surname, and date of birth as key fields, and it allows for spelling errors in Italian or other Latin languages through a specific algorithm. Furthermore, a double-blind procedure ensures data confidentiality. The main linkage procedure is based on four stages, two automatic ones, and two where the operator can decide through specific windows whether to accept stage-selected matches. SALI takes into account possible errors in key fields thus reducing false negatives. It was used to solve the problem of linkage between AIDS and cancer registries in Italy. It can be used with every IBM-compatible computer system, assuring uniquely high portability.     

Bloom's syndrome. XIX. Cytogenetic and population evidence for genetic heterogeneity

Cells with abnormally high rates of sister-chromatid exchange (SCE) are uniquely characteristic of Bloom's syndrome (BS). However, in one in five persons a minor population of cells with a low-SCE phenotype circulates in the blood. The origin and significance of the low-SCE cells in BS have never been understood, although they are assumed to arise by somatic mutation. In the present investigation, the enigmatic high-SCE/low-SCE mosaicism was investigated by comparing the incidence in several subpopulations of persons in the Bloom's Syndrome Registry who exhibit the two types of cells, and a striking negative correlation emerged: in persons with BS whose parents share a common ancestor, the case in approximately half of registered persons, low-SCE cells are found only rarely; conversely, the mosaicism occurs almost exclusively in persons with BS whose parents are not known to share a common ancestor. Because those who share a common ancestor are predominantly homozygous-by-descent at the mutated BS locus, the negative correlation is interpreted to mean that the emergence of low-SCE cells in BS in some way depends on the pre-existence of compound heterozygosity. A corollary to this is that BS is genetically heterogeneous.     

Protocolo del Estudio Poblacional Multinivel de las Desigualdades Socioeconómicas en la Distribución Geográfica de la Incidencia,la Mortalidad y la Supervivencia Neta del Cáncer en España: Estudio DESOCANES

Incidence and mortality provide information on the burden of cancer morbidity and the potential years of life lost due to cancer. The Spanish Deprivation Index (SDI) has been developed as a standardized measure to study socioeconomic deprivation in Spain at the census tract level. In addition, SDI information can be combined with ecological variables at the population level and data from the High-Resolution European Studies in Cancer. The aim of this study is to characterize socioeconomic inequalities in incidence, excess mortality, premature mortality and net survival for three of the most incident cancers (lung, colon-rectum and breast) in Spain using the SDI. This national population-based study will assess the impact of socioeconomic inequalities using a multilevel modelling approach. Spatial analysis, multilevel modeling, net survival and economic impact assessment will be used. The results will be useful for supporting decision-making, planning, and management of public health interventions aimed at reducing the impact of socioeconomic inequalities in the diagnosis and prognosis of cancer patients in Spain.     

Cancer prevalence in Central Europe: the EUROPREVAL Study.

BACKGROUND: Information on cancer prevalence is either absent or largely unavailable for central European countries. MATERIALS AND METHODS: Austria, Germany, The Netherlands, Poland, Slovakia, Slovenia and Switzerland cover a population of 13 million inhabitants. Cancer registries in these countries supplied incidence and survival data for 465 000 cases of cancer. The prevalence of stomach, colon, rectum, lung, breast, cervix uteri, corpus uteri and prostate cancer, as well as skin melanoma, Hodgkin's disease, leukaemia and all malignant neoplasms combined was estimated for the end of 1992. RESULTS: A large heterogeneity was observed within central European countries. For all cancers combined, estimates ranged from 730 per 100 000 in Poland (men) to 3350 per 100 000 in Germany (women). Overall cancer prevalence was the highest in Germany and Switzerland, and the lowest in Poland and Slovenia. In Slovakia, prevalence was higher than average for men and lower than average for women. This was observed for almost all ages. As shown by incidence data, breast cancer was the most frequent malignancy among women in all countries. Among men, prostate cancer was the leading malignancy in Germany, Austria and Switzerland, and lung cancer was the major cancer in Slovenia, Slovakia and Poland. The Netherlands had a high prevalence of both prostate and lung cancer. Time-related magnitude of prevalence within each country and the variability of such proportions across the countries has been estimated and cancer prevalence is given by time since diagnosis (1 year, 1-5 years, 5-10 years, >10 years) for each site. The weight of 1-year prevalence (248 per 100 000 among men and 253 per 100 000 among women) was <15% of total prevalence. Prevalent cases between 1 and 5 years since diagnosis represented between 22% and 34% of the total prevalence. Prevalent cases diagnosed from 5 to 10 years before (335 per 100 000 for men and 505 per 100 000 for women) represented between 17% and 23% of prevalent cancers. Finally, long-term cancer prevalence (diagnosed >10 years before), reflecting long-term survival, and number of people considered as cured from cancer were 490 per 100 000 for men and 1028 per 100 000 for women, with a range between 26% (The Netherlands, men) and 50% (Slovakia, women). CONCLUSION: It is clear from observing countries in Central Europe, that high cancer prevalence is associated with well-developed economies. This burden of cancer could be interpreted as a paradoxical effect of better treatments and thereby survival. It could also be taken as a sign for not being satisfied with the advances in treating patients diagnosed with cancer, and for supporting more primary prevention.     

Socioeconomic status and cancers of the female breast and reproductive organs: a comparison across racial/ethnic populations in Los Angeles County, California (United States)

Objectives: Despite the fact that socioeconomic status (SES) has been shown to have important implications in health related issues, population-based cancer registries in the United States do not routinely collect SES information. This study presents a model to estimate the SES of cancer patients in the registry database.Methods: At the Los Angeles Cancer Surveillance Program (CSP), we developed a model to estimate each cancer patient's SES from aggregate measurements of the census tract of residence (n = 1,640) at time of diagnosis. We then applied the SES estimates to observe the relationship between SES and risk of cancers of the female breast and reproductive organs including cancers of the ovary, cervix uteri, and corpus uteri. The analyses were performed on the cumulative records (n = 127,819) of cancer patients diagnosed between 1972 and 1992 in Los Angeles County, California, for the mutually exclusive racial/ethnic groups of non-Hispanic Whites, Hispanic Whites, Blacks, Asians, and persons of other ethnic groups.Results: We found SES is positively associated with female breast cancer, ovarian cancer, and cancer of the corpus uteri, but inversely associated with cervical cancer. These SES trends were quite consistent across age groups among non-Hispanic White women. Variations by race/ethnicity in the SES patterns were also found, with Asians exhibiting little association.Conclusions: Our model of measuring SES is sufficiently sensitive to capture the trends. Adopting the aggregate approach to measure SES in population-based registry data appears to be useful.     

Survival rates for primary malignant brain tumours in Europe

In the framework of EUROCARE, a concerted action between 45 population-based cancer registries, in 17 European countries, survival of patients with primary malignant brain tumours was investigated. Survival analysis was carried out on 16 268 patients diagnosed between 1985 and 1989 and followed-up for at least 5 years. The mean European age-standardised 5-year relative survival was 17% in men and 20% in women, with minimal intercountry variations, except for markedly lower rates in Scotland, Estonia and Poland. The age-specific analysis showed a relatively uniform survival in patients aged more than 65 years at diagnosis, but there were more marked intercountry differences in younger patients. In the 15–44 year age group (25% of the total study population) 5-year relative survival ranged between 55% (Finland and Sweden) and 27% (Poland). Generally, survival decreased with increasing age at diagnosis. The analysis of a temporal trend in survival was carried out on a subset of registries with available data from 1978–1989. Overall, there was an increase in survival over the considered study period, mostly confined to 1-year survival, suggesting that it was mostly related to improved diagnostic techniques. The most important survival increase occurred in the younger patients, both for 1- and 5-year survival, suggesting that younger patients have less biologically aggressive tumours, benefiting from the combined effect of diagnostic accuracy and effective therapies. The most marked survival increase was seen in England and Denmark, countries with low survival rates at the beginning of the study period, whereas in Finland and Germany, where survival was relatively high to begin with, no important temporal trend was seen.