Deliverable D9.2 Statistical Analysis of the EPIRARE Survey on Registries Data Elements Michele Santoro, Michele Lipucci, Fabrizio Bianchi
CONTENTS Overview of the documents produced by EPIRARE... 3 Disclaimer... 3 I. Introduction... 4 II. Methods... 5 III. Results... 6 IV. Conclusions... 18 2
Overview of the documents produced by EPIRARE Disclaimer The contents of this document is in the sole responsibility of the Authors; The Executive Agency for Health and Consumers is responsible for any use that may be made of the information contained herein. 3
EPIRARE WP6 ANALYSIS OF THE EPIRARE SURVEY ON REGISTRIES DATA ELEMENTS Michele Santoro, Michele Lipucci, Fabrizio Bianchi I. Introduction The EPIRARE Survey on the data elements collected by the Rare Disease Registries represents a useful basis to assess the feasibility of developing an information structure common to all Registries. The EPIRARE WP6 carried out exploratory analyses on data collected by the first EPIRARE Survey, which identified the groups of Registries containing quite similar information (cfr. Epirare WP6 - Collection of released documents, 26 th April 2013). By using different statistical approaches and techniques it was possible to identify Registries with common characteristics, which allowed us to classify them in macro-groups defined according to the agglomerative tendencies and associations emerged. A first group of Registries tends to be concentrate on the area of Public Health. A second group of Registries deals mainly with the Clinical Research and it looks highly correlated with the activity of genetic research. A third group is characterized by the actions on Treatment monitoring and Treatment evaluation, but it shows a tendency to pursue also other goals. This third group appears similar to the second one and it is more complex to be defined, although the collection of data on the Treatment activities represents its specific trait. Based on the results obtained by the first Epirare Survey on 219 Registries, the WP6 analyzed data collected by a second Survey concerning Data Elements used by Rare Disease Registries. The questionnaire consists of 45 specific questions addressing to deepen the information collected by Registries. The analyses were aimed to evaluate the feasibility of collecting specific information and assess if the potential capability of collection is different among the different types of 4
Registry. This analysis will give information aimed to assess the feasibility of defining data elements common to all Registries and data elements specific for one or more types of Registry. II. Methods The analysis was performed on 147 Registries which responded to the Survey. In order to identify the types of Registry, responders Registries were classified using different sequential criteria. We started to identify the Registries, which operate in the area of Public Health as follows: Registries declaring exclusively the aims belonging to the group: Epidemiological-Public Health as defined in the Survey; Registries declaring to collect data on All rare diseases ; Registries declaring to collect data on many groups of diseases; Registries declaring to be part of network Eurocat. After making crossed and combined checks of the above listed criteria, these Registries have been labeled as "Public Health". Then we explored on a possible classification of the remaining Registries. The Registries that have declared aims for the group "Treatment", have been labeled as "Clinical"; they are Registries involved in clinical research, but without dealing with Treatment. The remaining Registries were classified as "Clinical-Treatment". Two Registries did provide any information on aims and on number of diseases. For these Registries information on the Web were searched in order to classify them. A Multiple Correspondence Analysis was performed using variables characterizing Registries: single aims and number of diseases (one disease, a group of disease, several/all diseases). The scopes of this analysis were: - to verify the associations between variables identified by the first Survey data analysis (cfr. Epirare WP6 - Collection of released documents, Statistical Analysis of the EPIRARE Survey data - Part 1 -, 26 th April 2013) performing again the same technique but on a smaller sample of Registries; - to evaluate the classification of Registries adopted. 5
After that, an analysis on the data elements collected by Registries was performed. The distribution of each data element was analyzed stratifying the Registries in the three categories according to the classification described above. The capacity of collection of the data elements (,,, Not Adaptable, ) was distinguished as defined in the questions. A specific focus was made on the Registries that operate in the field of genetic research. The use of the question "Is your Registry able to collect data on Genetic features of the rare disease?", was able to identify these Registries that operate across the field of clinical research and treatment. III. Results Classification of Registries The distribution of the three categories of Registries is reported in Table 1. The 26.5% of Registries falls in the group "Public Health" and the 25.9% in the group called "Clinical". The remaining 47.6% of the Registries is classified in the Group "Clinical-Treatment." Tab.1 Number and Percentage of Registry by Type Number Percentage CLINICAL 38 25.9 CLINICAL-TREATMENT 70 47.6 PUBLIC HEALTH 39 26.5 Total 147 100 In order to check the plausibility of the classification, the distribution of the responses to the specific question on the collection of clinical data was analyzed. The results (Tab. 2) clearly show a greater tendency to collect this type of data by "Clinical" and "Clinical-Treatment" Registries, whereas the percentage for the Registries "Public Health" is lower, as expected, even if a high percentage of collected information is worth of e. 6
Tab.2: Clinical data CLINICAL 78.4 10.8 0.0 0.0 10.8 CLINICAL-TREATMENT 85.7 12.9 0.0 0.0 1.4 PUBLIC HEALTH 33.3 10.3 2.6 0.0 53.9 Total 69.9 11.6 0.7 0.0 17.8 Interesting results about the distinction of the Registries emerged by the question about the collection of data on patients with a confirmed diagnosis only or also with suspected diagnosis (Tab.3). "Public Health" Registries tend mainly to collect data of patients with a confirmed diagnosis only. Tab.3: Confirmed and suspected diagnosis Confirmed only Both Confirmed and suspected CLINICAL 39.5 60.5 CLINICAL-TREATMENT 54.3 45.7 PUBLIC HEALTH 71.8 28.2 Total 55.1 44.9 In the Figure 1 the factorial plan obtained by Multiple Correspondence Analysis using the variable aims and number of diseases is represented. All the aims relating to the group Epidemiology and Public health (as defined in the Survey) are placed in the upper part of the plan; in the lower part all the aims relating to the groups Clinical and Treatment are collocated. The aims of these 2 groups are placed on the same direction, so it means they are associated, but all the 3 aims of treatment (efficacy, effectiveness and safety) have higher x values on the horizontal axis, that means they are distinguished although slightly. The Registries which deal with several or all rare diseases, are placed in the upper part, opposite to treatment and clinical aims, as expected. This representation confirms the tendency of the Registries to aggregate into three groups: Public Health, Clinical Research, Treatment. 7
Fig.1: Factorial plan defined by variables: aims and number of disease; projection of the Registries classification Key connections The need to have a unique key for the identification of the patients is a challenging issue for the interconnection of the Registries. This key, defined here as Key connection, it must be generated through algorithms derived by using personal and demographic information, capable of producing a unique and anonymous code for each patient. In order to ensure the uniqueness of the patient, the following data elements might be at least included: First name, Last name, Sex, Date of birth, City of birth. On the basis of the data collected by the Survey, we estimated the percentage of Registries that would be potentially able to produce the Key connection, in the perspective of identifying the common data elements. The condition for defining a Registry capable to produce the key 8
connection, was the collection or adaptability of all the five key variables. Less than half of the responder s Registries, would be potentially capable to produce a Key connection as previously defined. A different collection ability exists among the three types of Registry, with a minor capability for the "Clinical-Treatment" Registries. Considering that the capacity to produce a key connection represents a very critical issue to assess the potential degree of interconnection between Registries and having observed a variability by nation of the Register, a geographical focus has been realized. Checking Registries by country of location, Italian Registries show a higher capability to produce Key connection, whereas only a low percentage of German and British Registries collect all the data needed to calculate the Key (Tab. 5). Tab.4: KEY CONNECTION = First Name + Last name + Sex + Date of birth + City of birth Yes No CLINICAL 55.3 44.7 CLINICAL-TREATMENT 38.6 61.4 PUBLIC HEALTH 48.7 51.3 Total 45.6 54.4 Chi-square test: p value = 0.22 Tab.5: KEY CONNECTION by nation of Registries Yes No Country France 41.7 58.3 Germany 23.8 76.2 Italy 87.5 12.5 Spain 56.5 43.5 UK 36.4 63.6 Not European Countries 0.0 100.0 Other European Countries 45.0 55.0 International Coverage 50.0 50.0 Total 45.6 54.4 In the following tables the distributions of each variable forming the Key connection are reported. Almost all the Registries collect data on gender and date of birth, whereas the data on 9
names and place of birth seem to be very critical for the collection. In particular, the Public Health" Registries show difficulties in collecting data on name of patients, whereas the "Clinical" and "Clinical-Treatment" Registries present more deficiencies in collecting data on the city of birth. Tab.6: First name of patient CLINICAL 68.4 10.5 0.0 2.6 18.4 CLINICAL-TREATMENT 50.7 14.5 2.9 8.7 23.2 PUBLIC HEALTH 56.4 2.6 0.0 2.6 38.5 Total 56.9 10.3 1.4 5.5 26.0 Tab.7: Last name of patient CLINICAL 68.4 10.5 0.0 2.6 18.4 CLINICAL-TREATMENT 52.9 11.4 2.9 10.0 22.9 PUBLIC HEALTH 59.0 2.6 0.0 2.6 35.9 Total 58.5 8.8 1.4 6.1 25.2 Tab.8: Gender of patient CLINICAL 76.3 15.8 2.6 2.6 2.6 CLINICAL-TREATMENT 88.6 8.6 0.0 1.4 1.4 PUBLIC HEALTH 94.9 5.1 0.0 0.0 0.0 Total 87.1 9.5 0.7 1.4 1.4 10
Tab.9: Date of birth of patient CLINICAL 86.8 5.3 0.0 2.6 5.3 CLINICAL-TREATMENT 84.3 5.7 1.4 2.9 5.7 PUBLIC HEALTH 97.4 2.6 0.0 0.0 0.0 Total 88.4 4.8 0.7 2.0 4.1 Tab.10: City of birth of patient CLINICAL 31.6 21.1 2.6 0.0 44.7 CLINICAL-TREATMENT 31.4 12.9 10.0 2.9 42.9 PUBLIC HEALTH 66.7 12.8 0.0 0.0 20.5 Total 40.8 15.0 5.4 1.4 37.4 Other personal data The Patient's National security code could represent an alternative way to identify the patient ensuring uniqueness and anonymity, but the limited ability to collect this data highlights the criticality of their use (Tab.11). Other personal data were analyzed in function of the definition of Key connection, but mainly for exploring their information content. The city of residence is a variable collected in a consistent way by the Public Health Registries whose action is based on the epidemiological research (Tab.12). This result confirms the trait of this type of Registry and the need to identify specific data elements based on different scopes of the Registries. The responses to the question on the Patient s contact, shows that RD Registries are inclined to collect this information (tab.13). Regarding to the question on the collection of the date of death, about 25% of "Public Health" and "Clinical" Registries declared that they do collect this data (tab.14). It should be taken into consideration that this data is more related with a complex activity of follow-up, or a systematic 11
linkage with mortality Registries that could be active in all areas. A relevant coverage of this data element is present on the "Clinical-Treatment" Registries. Tab.11: Patient s national Security Code Collected Adaptable Not collected CLINICAL 13.2 23.7 63.2 CLINICAL-TREATMENT 22.1 44.1 33.8 PUBLIC HEALTH 38.5 10.3 51.3 Total 24.1 29.7 46.2 Tab.12: City of residence of patient CLINICAL 68.4 5.3 0.0 0.0 26.3 CLINICAL-TREATMENT 57.1 12.9 5.7 2.9 21.4 PUBLIC HEALTH 84.6 5.1 0.0 2.6 7.7 Total 67.4 8.8 2.7 2.0 19.1 Tab.13: Patient s contact CLINICAL 73.7 7.9 0.0 0.0 18.4 CLINICAL-TREATMENT 53.6 14.5 4.4 2.9 24.6 PUBLIC HEALTH 35.9 2.6 0.0 7.7 53.9 Total 54.1 9.6 2.1 3.4 30.8 Tab.14: Date of death CLINICAL 44.7 13.2 5.3 7.9 29.0 CLINICAL-TREATMENT 70.0 15.7 1.4 2.9 10.0 PUBLIC HEALTH 66.7 7.7 0.0 0.0 25.6 Total 62.6 12.9 2.0 3.4 19.1 Diagnosis 12
The date of diagnosis is collected by the 64.6% of the Registries; more than 16% collects the information in a potentially adaptable way. About 10% of the Registries does collect this type of data, in particular the "Clinical" Registries. "Clinical-Treatment" Registries show a high capacity to collect this data element. About the date on the onset of the symptoms, the capability of collection is quite low for all Registries, especially for "Public Health" Registries. It should be underlined that such information is difficult to collect because the patient could remind the exact date of onset of the first symptoms. Tab.15: Date of diagnosis CLINICAL 39.5 26.3 10.5 0.0 23.7 CLINICAL-TREATMENT 78.6 14.3 2.9 0.0 4.3 PUBLIC HEALTH 64.1 10.3 7.7 7.7 10.3 Total 64.6 16.3 6.1 2.0 10.9 Tab.16: Date of disease onset CLINICAL 39.5 21.1 10.5 5.3 23.7 CLINICAL-TREATMENT 74.3 5.7 5.7 1.4 12.9 PUBLIC HEALTH 30.8 5.1 2.6 7.7 53.9 Total 53.7 9.5 6.1 4.1 26.5 Health services The data on co-morbidity are recorded in a fairly satisfactory way by the Clinical and Clinical- Treatment Registries, whereas they are inadequately collected by Public health Registries (tab.17). All Registries show a low capability to collect data on hospitalization and on patient s disability (tab.18, tab.19). Tab.17: Comorbidity 13
CLINICAL 60.5 10.5 2.6 0.0 26.3 CLINICAL-TREATMENT 68.6 11.4 0.0 1.4 18.6 PUBLIC HEALTH 18.0 10.3 7.7 2.6 61.5 Total 53.1 10.9 2.7 1.4 32.0 Tab.18: Hospitalization CLINICAL 5.3 26.3 2.6 7.9 57.9 CLINICAL-TREATMENT 36.8 19.1 4.4 1.5 38.2 PUBLIC HEALTH 13.2 5.3 7.9 2.6 71.1 Total 22.2 17.4 4.9 3.5 52.1 Tab.19: Patient's disability Collected Adaptable Not collected CLINICAL 26.3 26.3 47.4 CLINICAL-TREATMENT 22.9 41.4 35.7 PUBLIC HEALTH 10.3 20.5 69.2 Total 20.4 32.0 47.6 Treatment The Survey-question on the treatment with orphan drugs shows a lack of collection, with exception for the Registries classified as Clinical-Treatment, which reaches a satisfactory capability of coverage (tab.20). Clinical-Treatment Registries shows also a higher capacity to collect data on the treatment with other drugs (Tab. 21). 14
Tab.20: Current orphan drugs Treatment CLINICAL 7.9 18.4 7.9 0.0 65.8 CLINICAL-TREATMENT 51.5 25.0 5.9 1.5 16.2 PUBLIC HEALTH 33.3 2.6 5.1 2.6 56.4 Total 35.2 17.2 6.2 1.4 40.0 Tab.21: Current drugs Treatment CLINICAL 23.7 18.4 13.2 2.6 42.1 CLINICAL-TREATMENT 60.0 21.4 4.3 1.4 12.9 PUBLIC HEALTH 28.2 10.3 7.7 2.6 51.3 Total 42.2 17.7 7.5 2.0 30.6 Recruitment The question concerning the willingness to be recruited for clinical trials shows a different response among the types of Registries. About 85% of "Clinical" and "Clinical-Treatment" Registries collects this data, whereas the "Public Health" Registries tend to collect this information in a limited way. The data on donation of biological samples has a real or potential capability of collection in about 70% of "Clinical" Registries and in 85% of "Clinical-Treatment" Registries; "Public Health" Registries show a poor capacity to collect this data. Tab.22: Patient recruitment for Clinical trial Collected Adaptable Not collected CLINICAL 56.8 27.0 16.2 CLINICAL-TREATMENT 46.4 40.6 13.0 PUBLIC HEALTH 7.7 23.1 69.2 Total 38.6 32.4 29.0 15
Tab.23: Patient recruitment for donating biological samples Collected Adaptable Not collected CLINICAL 16.2 51.4 32.4 CLINICAL-TREATMENT 50.0 36.8 13.2 PUBLIC HEALTH 7.7 23.1 69.2 Total 29.9 36.8 33.3 Genetic data Genetic research is carried out in a clinical setting and across the Registries that operate in the Treatment too. About 60% of "Clinical" Registries declares also to pursue objectives on genetic research; the percentage is higher (about 75%) for "Clinical-Treatment" Registries. As expected, a low percentage of the Registries classified as "Public Health" declares to perform genetic research. Tab.24: Registries with aim of Geype-pheype correlation/mutation database Yes No CLINICAL 59.5 40.5 CLINICAL-TREATMENT 76.5 23.5 PUBLIC HEALTH 12.8 87.2 Total 54.9 45.1 The distribution of responses to the question on the collection of genetic features clearly shows that this need of information is a characteristic of "Clinical" and "Clinical-Treatment" Registries and of "Public Health" Registries. This distribution represents a further confirmation of the reliability on the classification criteria identified. The Survey requested also to specify the types of genetic data collected in case that the Registries declared previously to collect genetic data. The data on Gene (HGNC Gene Symbol) is collected by 71% of the Registries that collected genetic data; 34% collects HGVS Variant description format, 20% Chromosome number. The other genetic data are collected in a limited way. 16
Tab. 25: Genetic features Collected Adaptable Not collected CLINICAL 84.2 5.3 10.5 CLINICAL-TREATMENT 75.7 15.7 8.6 PUBLIC HEALTH 38.5 18.0 43.6 Total 68.0 13.6 18.4 Tab. 26: Genetic data collected by Registries which collect Genetic features Number % Data collected Gene (HGNC Gene Symbol) 71 71.0 Chromosome number 21 21.0 Chromosomal reference sequence accession and version number RefSeqGene accession and version number 16 16.0 13 13.0 Locus Reference Genomic (LRG) 9 9.0 Variant description in HGVS format 34 34.0 Variant description in other format 16 16.0 Other 32 32.0 17
IV. Conclusions The results obtained from the analysis of Survey on data elements provide a useful information framework for assessing the feasibility concerning the definition of a Common Data Element (CDE). It appears quite evident, even from this Survey, a different profile of collecting data among the Rare Disease Registries that operate mainly in the field of Public Health and of those which work within the general scope of Clinical research. With few exceptions, the Registries defined as "Clinical" and those operating in the area of treatment - called "Clinical-Treatment" - show common profiles. A first important result, common to the different types of Registries, is the difficulty to build a unique and useful key of interconnection. The critical situation is highlighted in different ways for the types of Registry: the Public Health Registries show more difficulties on collection of the names of the patients, whereas the Clinical Research Registries on the collection of the place of birth. The Public Health Registries tend to collect more data on patients with only confirmed diagnosis. These Registries, being based on an epidemiological function, tend to collect more the data on place of residence. Instead they collect in a limited way the contact information on the patient and almost 30% of this Registries does collect the information on the date of death. This last data element is of particular relevance for the production of accurate estimates on prevalence and survival. This data should be validated by a follow-up of the patient and/or by a linkage with other information systems (mortality, hospitalizations, other registries). For this reason it should be strengthened, for these Registries, the capability of linkage with other health information systems in order to obtain information also on hospitalization, comorbidity and disability. As expected, these Registries do collect clinical and genetic data, data on the treatment, patient recruitment for clinical trials and biospecimen donation. On the other side the Registries conducting clinical research, collect more data on the identification of patient s contact, on co-morbidity, on recruitment for clinical trials. The Clinical research Registries which deal with Treatment, show some expected peculiarities, such as an increased ability to collect data on treatments; they also show a greater attitude to record the date of death and the date of diagnosis. 18
The collection of genetic data, as expected, is a trait of the Registries conducting clinical research. The genetic data mostly collected among the Registries that are involved in genetic research, is the data on Gene (HGNC Gene Symbol); 34% also collects the data on the Variant description in HGVS format, only 20% of the Chromosome number. The collection of other genetic data are very limited. The framework shows that the collection of some information are common to all Registries, whereas other information are collected only by some types of Registry. The specific collection of some data elements should be taken into account for the definition of the Common Data Elements. It is desirable that the elaboration of a CDE for Rare Diseases Registries includes data elements with the different information needs. 19