The end of 2009 marks completion of a full decade of Household Surveillance. We started 4th January 2000. Every six months during the last 10 years we have visited about 13,500 homesteads, filled out questionnaires, and entered them into the ACDIS database, and thereby built up a detailed picture of the population of this area – their births, migrations, marriages, deaths, living arrangements. For our Individual Surveillance 2010 is the 7th annual round of collecting HIV samples and other personal health-related information from about 40,000 adults.
The last 2-3 years have generally been a period of stability in most departments – the Operations managers, the Data Centre management, data capture staff, IT management and software developers have all broadly remained the same. With this stability comes knowledge and experience and this has allowed us to make many significant improvements at all steps of the process from printing the forms, through fieldwork, back to quality control, data capture, document archiving, and issuing of data to scientists. The result is that capture of the 2009 data was completed on time by 10th December; the “Round RollOver” process went smoothly, such that the first two weeks of forms for the 2010 Round were printed and ready before Christmas, and took a day less than planned. Finally, datasets containing all data up to end 2009 were made available to scientists from 7th January, less than a month after the end of data collection, and all documentations on the internet were updated by early January, including those for the 2010 round yet to start.
From among the myriad of improvements made here are six highlights:
1. The data entry software, once regarded as full of bugs, and near impossible to maintain has been tamed by the two Gijima software developers, and nowadays, new developments and changes can be made with confidence, bugs get fixed promptly, and generally the system is robust and reliable.
2. We have developed a web-based Unified Report, as a management tool to provide a consolidated real time database-linked tracking and progress report for surveillance activities from printing, document distribution and control, data collection, fieldwork quality control, data entry and archiving (both physical and digitally). This reports shows operations staff hour by hour, where every form is – where delays and bottlenecks are occurring and has been turned into a reliable modern web-based reporting tool, covering not only Individual and Household surveillance, but also Verbal Autopsies.
3. The time between data collection and data entry for both the Household and Individual Surveillance has been substantially reduced over years as shown in the graph below.
4. The reduction in time between data collection and data entry has allowed us to adopt a real time data cleaning approach which in turn allows us to make datasets available to scientists in just under a month from closure of a surveillance round. For example both round 21 for Household Surveillance and round 7 for Individual Surveillance closed data capture on 10th December 2009 and new datasets were issued out to scientists on 7th January 2010, irrespective of two weeks festive holidays shutdown of the Centre.
5. The mountain of 12m paper forms in our archive, growing by about 1.2m per year – is starting to shrink as we now scan every sheet into a digital archive. So far about 1.4m have been scanned, quality controlled and destroyed and the process is now running smoothly. The digitized forms take up no space, are held much more securely than any paper forms, and can be searched and viewed on screen easily.
6. The consistency and quality of the data in the database has improved greatly, as we tighten up the data entry rules and improve the software. Hence scientists now get data that, while not perfectly clean, is at least mostly consistent. No longer do they have to spend a lot of time carrying out their own data cleaning and editing, but now they can start their analyses more or less immediately.
7. All forms (over 20) have had a substantial makeover for 2010, making their style consistent and modern, and easier to read and fill in.
8. The number of documented datasets available has continued to grow and improve. In the last couple of years we have produced consistent Household Socio-Economic datasets. No longer do we rely on one of our collaborators, Cally Ardington in Cape Town, producing them for us. We have also produced datasets for the Verbal Autopsy system, and for our vaccinations data. There are now over 60 datasets listed on our internet, all with up-to-date, standardised, freely-downloadable documentation.
9. The biggest and most important advance in what we provide to scientists is the “Demography Dataset”. The consistency of our database has made it possible to develop a single dataset of short episodes of exposure, detailing births, deaths and migrations. This allows scientists to carry out demographic analysis in a much simpler way than before. This development is the first time we produce a substantially processed dataset (as opposed to essentially raw data) and its impact on making ACDIS data comprehensible to and analysable by a much wider audience is immense. No longer does someone have to spend months understanding the details of the data structure, forms, questions learning how to combine datasets etc. Our ambition for the future is to develop the documentation of this dataset further by providing worked ‘how to’ examples.