/
Adolescent Health, World Health Background The Child Health Epidemiolo Adolescent Health, World Health Background The Child Health Epidemiolo

Adolescent Health, World Health Background The Child Health Epidemiolo - PDF document

olivia-moreira
olivia-moreira . @olivia-moreira
Follow
458 views
Uploaded On 2016-09-30

Adolescent Health, World Health Background The Child Health Epidemiolo - PPT Presentation

s from LSHTM ID: 471206

from LSHTM

Share:

Link:

Embed:

Download Presentation from below link

Download Pdf The PPT/PDF document "Adolescent Health, World Health Backgrou..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.


Presentation Transcript

Adolescent Health, World Health Background The Child Health Epidemiology Refereneen reviewing the major causes of child mortality and morbidity in developing countries, with a view to improving available estimates and informing the modelling of burden of disease. It has become increasingly important to CHERG to ensure common methods and procedures across the various disease areas (e.g. ARI/pneumonia, diarrhoea, malaria, neonatal, measles). CHERG utilises a number of different approaches to developing estimates for the burden of disease in children aged under 5 years including abstracting data from research papers that report on childhood mortality and morbidity. This estimation process involves (1) systematic searches of the literature, (2) review of the research articles to abstract key information, (3) standardisation of specific data points (e.g., mortality rates, age ranges), and (4) the estimation process itself. In a working meeting of selected CHERG members in mid 2002, it was agreed that a new approach was needed to speed up the abstraction process and to improve the reliability of the data abstracted. It was agreed that the Public Health Interventions Research Unit (PHIRU) of the London School of Hygiene and Tropical Medicine (LSHTM) would carry out a data abstraction exercise using standardised methods with careful quality control for 3 major categories of under-5 mortality (diarrhoeal disease, malaria and neonatal deaths). Subsequently, abstraction of data on measles mortality and malaria morbiditTo review published research reports and abstract the key data and parameters needed to estimate child mortality due to malaria, measles and diarrhoeal disease, the proportional distribution of causes of death occurring period and child morbidity due to malaria. To prepare databases that include these key data and parameters related to under-5 mortality due to malaria, measles and diarrhoeal disease, mortality during the neonatal period and malaria morbidity. To develop, implement and report on quality control procedures to ensure the validity and reliability of the abstracted data points and the resulting databases. Research papers for review were proviPinto, Alex Rowe, Joy Lawn, William Moss, Joanna Schellenberg, Ilona Carneiro). s from LSHTM’s Masters degree programme and a current PhD student. Quality control procedures included: development and pre-testing of data collection tools for each of the categories of deaths (diarrhoeal disease, malaria, measles and neonatal) and for malaria morbidity; formal data abstraction training of the reviewers by an epidemiologist; double data abstraction (abstractors were grouped into pairs randomly and each individual in a pair abstracted the data from the same paper); close supervision of the abstraction process and random checks of 10% of the forms by a supervisor; inter-rater reliability checks between the 2 abstracters (using variables representing the most important content of the data collection forms); double data entry with verification and range and consistency checks; and subsequent identification and correction of all errors in the data base. Data were entered and validated using Epi Data Version 2.1b® (2002) Abstracted data included information on: the design of the study, any intervention included in the study, the study setting and population and its representativeness of the wider population, the methods used to ascertain deaths or morbidity, the methods used to identify cause of death and case definitions, all cause and cause-specific mortality rates, morbidity rates, case-fatality rates, protective efficacy (if an intervention study), co-morbidity, co-mortality estimates, and potential flaws in the study. Examples of the data collection forms and the level of agreement forms have already been provided to CAH. The data abstraction forms for diarrhoea, malaria, measles and neonatal mortality ran to 20, 21, 19 and 27 pages respectively. The form for malaria morbidity ran to 39 pages. The number of data fields in the different databases ranged from 179 to 856, though not all of these fields would be completed 5 new data abstractors were recruited and trained, and a total of 6 abstractors reviewed papers during rounds 2-4. Over the course of the whole exercise a total of 624 research reports were reviewed: 109 for diarrhoea mortality; 176 for malaria mortality; 100 for measles mortality; 89 for deaths in the neonatal period; 150 for malaria morbidity. Review and abstraction of papers eligible for inclusion in the process produced a total of 605 completed abstraction forms; 104 for diarrhoea mortality; 129 for malaria mortality; 102 for measles mortality; 90 for deaths in the neonatal period; 180 for malaria morbiditall forms were double Data quality control procedures were as to data entry, the abstractors resolved all differences with assistance, when necessary, from the project supervisors. Random checks of 10% the forms were performed by the project supervisors. (Melissa Thumm took over from Karen Edmond as project supervisor on 17 February) All errors identified were corrected. Once they were entered, data were tency checks were performed. Data were converted using Stat Transfer Version 6.0® to Excel Version 6.0®, SAS Version 6.0® and Stata Version 6.0®, depending on the preferences of the CHERG investigators. A coding manual for each database was prepared. The finalised databases for were sent together with coding manuals, copies of the final data collection instruments, and copies of the data entry interfaces to the investigators by pre-arranged deadlines. The CHERG investigators reviewed the databases from the current and previous rounds and reported missing or unclear data to the project supervisor. In the malaria database five records contained a misspelling of the authors’ last name and for a few records the “Country where the study was done,” had not been entered despite the availability of this information in the paper. A more generalised inconsistency in the method of reporting deaths associated with a disease and deaths attributed to a disease was also identified in the databases for both diarrhoeal disease and malaria. The project supervisor reviewed all the papers and abstraction forms involved to resolve inconsistencies and omissions, made the necessary corrections in the databases, and then sent the revised files to the investigators. Inter-abstractor agreement Levels of agreement between abstractors on different sections of the mortality abstraction forms information All Diarrhoea Malaria Neonatal Measles Study design 79% 84% 72% 80% 41% Study population 61% 57% 55% 75% 58% 50% 62% 55% 60% 21% All-cause 52% 59% 48% 56% 43% 67% 62% 67% 81% 59% Table 1 presents the level of agreement between the two data abstractors prior to reconciling their findings. Under each heading the abstractors could record a number of different options or combinations of options. For example, for agreement to be recorded under “all-cause mortality” the information recorded in 5 out of 5 fields (person-time at risk, number of children in sample, number of deaths, reported death rate, reported death risk) had to be in agreement. Agreement was highest for neonatal mortality, followed by diarrhoea, then malaria and then measles. Levels of agreement were across the different areas were broadly similar for diarrhoea and malaria. Agreement on the neonatal mortality papers appeared substantially better than that for measles papers. Why was there relatively poor agreement for the measles data abstraction? Some ng. The measles form was not developed, tested and modified in the same way as the other disease forms (which were developed with significant input from data Inter-abstractor agreement over timeMeasles mortality Early papers (N=43) Later papers (N=52) Study design 44% 38% Population 67% 50% Determining deaths 16% 25% All-cause mortality 49% 38% Measles specific mortality 70% 50% No clear pattern emerges, with agreement on some areas appearing to improve while comparison is that the early papers tended to be measles-specific papers while the later papers included papers with multiple causes, measles among them. Table 4 presents a comparison of the two pairs of abstractors who performed the measles data abstraction. E Comparison of inter-abstractor agreement between two pairs of abstractors Measles mortality Pair A (N=50) Pair B (N=45) Study design 40% 42% Population 56% 60% Determining deaths 26% 16% All-cause mortality 42% 44% Measles specific mortality 54% 64% No major differences are evident between the two pairs. Agreement between data abstractors and supervisor l of agreement between the abstractors reconciled forms and those of the supervisor (performed for 10% of forms. There was a much higher level of agreement than between the initial abstractions performed by the two data abstractors (Table 1). This is perhaps not surprising because the supervisors abstraction was not always entiabstractors (if for example the data abstractached the supervisor agreement than between abstractors? In verbal autopsy studies ( a source of much of l of agreement between two physorder of 75-80%. Generally the agreement bereement between abstractors The measured level of agreement is likely to be dependent on a number of factors. The most important of these, in our experience, is the quality of the paper. Well abstractors record their level of agreement - it is clear that there was some variability in this – how often they meet together informally to discuss papers and identify discrepancies/problems, and how often theysome cases, identified minor errors, discrepancies and omissions in the data. To minimize such problems subsequently, the project supervisors increased the emphasis on relevant aspects during the data abstractors’ training. In resolving any discrepancies or omissions, it was helpful when the CHERG investigators reviewed the databases within a short time of receiving them. The forms used in the abstraction process were long. Long forms allow for more variation in the form in which informaminimize the subjectivity that to enter into the fields. They also, again hopefully, make e papers to abstract further information. Whenever it was available abstractors were asked to record raw data (numerators and denominators). The forms also allowed data abstractors to information that will help investigators interpret the data. There was inevitably some variation between forms for the different disease areas. However, greater standardisation of some sections could have achieved and might have helped to improve comparability and consistency of reporting and minimize Time is required to develop the wording, pilot the form, to have it mirror the structure investigators, the project supervisor and the data abstractors. Spending more time on the forms initially will save time in the long-run and improve the accuracy of the data. On some occasions this process was rushed. It was noticeable that different selection criteria appeared to have been used (or the same criteria had been interpreted differently) by different CHERidentify papers for abstraction. For example, a number of hospital-based studies were included in the papers submitted for abstraction. A misunderstanding on the part of the abstraction team over selection criteria also led to the initial exclusion of several malaria papers, which the investigators had intentionally included for abstraction. These papers had to be reviewed a second time in order to be abstracted. In any future such exercises, criteria for inclusion of papers should be documented and disseminated to all CHERG investigators, the people who are conducting the Also, it should be clear WHO