2009 Implementers Group Meeting Program Data Mining

Data Mining – Monday 3 pm

AMPATH

Data use
- research (usually flat files for analysis with SAS)
- MOH, PEPFAR, and USAID indicator reporting
- Data quality reports
- Internal quality improvement programs
- Outreach to patients missing appointments
Need to get indicators defined clearly in terms of database concepts
- Use SQL for queries
- Don't use cohort builder because of complexity of indicators
Would like tools to layout a template and indicate data source for each
OpenMRS DB is replicated, one for transactions, the other for reporting

Uganda

PIH Malawi

Using SQL and BIRT
Same difficulty extracting indicators form DB
Generating data quality reports
Using temporary normalized tables to generate reports faster
Patient-summary tables are update with each DB change
Discussion of Malawi data warehouse
- Based on ART HL7 message
- Malawi did harmonization of data between 2 systems
- Sites can exchange HL7 regardless of system type
- Data warehouse will be merged contents of all sites

Roger Friedman's comments

many of these issues have been dealt with in other workshops
- data quality, indicator reporting
problem of performance is due to EAV data storage by Open MRS
- very common to have separate DBs of different structures for transactions and reporting
- would be better to calculate indicators, store them elsewhere (CRIS?)
- queries would be more straightforward
to improve SQL skills, key is to think in sets rather than procedurally
- query optimizers are good, can perform seemingly hard queries quickly
- self-joins are often important for comparing subsequent visits

Justin Miranda talked about new report framework

Tanzania household survey