OpenMRS and National Reporting
Issues for discussion:
- Centralised data warehouse with anonymized patient data
- Data validation and auditing
- How to engage the ministry? (to be discussed in later session)
- Harmonizing reporting systems
- Data ownership/access
- Importance of local data collection and use/ feedback down to the local level
- National accountability to citizens
- Lots of levels of information use; many needs up to the national level - need to interact with local and regional decisionmakers
- Reporting definitions
Centralised Data warehouse
o One central place where data can be stored; a master copy of OpenMRS for access
o A structure to allow many secondary uses of that data
- Currently not very good at supporting large volumes of data from disparate sites
- Wanted to automate a process of data extraction and automation
- MoH usually only receives aggregate data, but would like to actually have patient level data in the warehouse
o Health Metrics Network has identified that typically when moving up the chain of data collection, you get fewer data elements
- How much of the original individual level data can be stored?
o where should data be stored?
o what are processes for getting data out?
o informed consent? Documentation as to which patients are included and the triggers for inclusion
o what elements should be included? Entire record or subsets of data?
- Would like to use this to generate MoH reports, but then could also use analytical tools for data mining
o What can be attached, eg SNOMED maps
- SA in 1995 - created a data warehouse from Joburg Gen, Somerset Hosp HIV clinics. Data mining allowed demonstration of the value of Vit B supplementation in delaying AIDS onset. This was later prospectively proven
Individual vs. aggregate data:
- Normally shy away for exporting whole patient record due to issues of confidentiality
- Should data be analysed locally or at government level
o Electronic TB register - patient level data off a TB register, compiled into a district register, then exported to the national level.
o Although can only see de-identified, anonymised data at higher levels, all patient data is still included.
o Could use similar architecture.
- TB - represents a small slice of the population, but can you actually implement across all facilities in the primary health care system without computers?
- Could do in a stepwise manner, but will probably still need aggregate systems for a long, long time
- At the district level, the important thing for a manager is to have full coverage of all his services - in many places trying to have EMRs in all facilities is not realistic
- The paper data collection systems of aggregate data are important for the manager to get that coverage across the whole district.
- Balancing the need for data collection with the need to protect data from unwarranted use is difficult - report only what is necessary
- For some diseases like HIV, need to follow individual level data
- De-identified data that can still be linked back to individual records is one way
- Going to need both systems for some time to come.
- In places like the US, patient level data is usually managed by the health care facility - they are responsible for maintaining the privacy of those patients. Need to justify getting access
- Default behaviour would be to pass aggregated data only, but could have some way to link back to the individual record - don't actually pass the whole record on
- Need to justify why you would need the individual record
- Holding all patient records in one central location is not really a standard that is upheld anywhere
- SA TB register - patient data is carried within the warehouse, can drill down to patients.
o Very useful because started project long before HIV was being treated in SA - allowed several publications about HIV and TB links
- Tension initially about doing research on systems like OpenMRS
- Begin with aggregated, then patient level - seems like a sensible roadmap
- Aggregated data doesn't replace the data warehouse project
- Issue of architecture - where does that data live? Being able to get down to the patient level data has immense value, but better to keep patient level data out of the warehouse and link back instead - then have permissions in place, don't store all patient level data in the warehouse
- Public health surveillance is all about building cohorts of data. Public health rationale for collecting this data.
- US system pays for datasets to be extracted from individual level data - aggregate data doesn't serve this purpose
- Don't confuse patient level data with identified patient level data - need identifiers at the clinic level, but this is about the need for role-level data (anonymised) vs. aggregated data. Aggregated data can't replace patient level data.
- Need to validate data quality - how can you do this without holding individual records? Need to be able to trace back to the original site.
- Could be paper records but need to be well designed so can be computerised and sampled in different ways.
- May be valuable to de-identify higher up the tree.
- Other tools which collect data in the primary healthcare facilities - lots of data needed in the warehouse, not all is collected by the EMR e.g. HR data, data on medicines - stock-levels etc, facilities, medical equipment
- Policy issue about collecting individual records above the facility level - need to design who has access. Should be a policy governing how to drill down
- How to convince the ministry that what you're doing is what they want?
- Completed an audit process, took a long time - painful. What have others done?
- AMPATH - sent reports, where questioned, sent the 'recipe' - here's how we inferred the numbers, if you have suggestions, let us know.
- Data quality assessment tools
o look at data over a period, e.g. clinical data about new patients enrolled
o get a representative sample of a range of facilities, done through MoH staff
o go down to the source documents - patient cards and look over a three month period - need to literally go through every record, but in the end gives a very good picture of the data that is being received at higher levels
- Have a feedback system to report back to the local level - two way understanding of the validation process
- Australia - clinical level through a clinical governance framework, physicians validate the data, then goes to the quality branch of the ministry - they have tools
- Need to provide documentation on medication errors or surgery that went wrong
- NB to set expectations with the ministry - sometimes don't want to know this info - in the past data has been manipulated to look good.
- Ghana - normally data is aggregated. Validating the data is difficult Don't have the patient data to validate. Orginally had patient data entered in CSPro, now are switching to OpenMRS and this how they'll be able to validate - moving down to a patient level to validate data.
Harmonizing reporting systems
- Asked to furnish reports for multiple different stakeholders
- strategies to harmonise, involving a concept dictionary
- do others have the same challenges?
- AMPATH encounters similar problems - sometimes different agencies use different definitions
- When cross referencing reports, the nubers look different - hard to explain why the numbers differ across reports, but this is due to indicator definitions
- How to you respond to changing definitions /indicators when you are working with aggregate data only?
- Need instructions on how to calculate based on patient level data, when you calculate, must provide numerator and denominator, can see how indicator was calculated
- Work closely with ministry - often sit at data revision meetings defining the indicators, HISP never comes up with its own set of indicators
- If OpenMRS comes with standard reporting, which isn't what the ministry requires, there should be a way to export anonymised data so that this can be processed to meet other demands.
- With new reporting framework it is possible to export a variety of different data sets, can be based on calculations you've defined, or can export a table of data, can be done in a de-identified manner if you choose.
- Need to define an indicator within the system, can create more than one for different reporting agencies, can give a different display name in the report
- In some current systems, reports to MoH and others can use some aggregated data based on actual patient records, but data clerks can enter facility level data that is pre-aggregated. Stored in a separate table, in the legacy system
- Summer of Code project to bring the same functionality as a module of Open MRS
- Can have both patient level data and aggregated data that you've added, can export to the same report
- Export in excel
Who owns the data?
- In Australia the data belongs to the person acquiring healthcare services
- Need to get patient consent, esp in rape cases, mental health cases
- Who is able to edit the data? How closely should the ministry be involved in developing protocols around who can access and manipulate the data? Otherwise the ministry may say that this data isn't true
- In India, all districts report aggregate data
- Unless government health staff participate in reporting and start to accept the application, it hasn't worked.
- Must have the ministry's approval and involvement, otherwise not sustainable
- About 2 years ago completed the Kenyan AIDS indicator survey - sensitive data. De-identified, but can still drill down to district
- Data owned by goverment, hosted by Kenyan National Bureau of Statistics
- Have to get authorization at the national level - complete a data access form
- Need to attach an approved research protocal, or explain the programmatic purpose for the data
- Data ownership - legal opinion, distinction between data in paper record, compiled database, data warehouse - those who had compiled the records had right of ownership.
- People who have invested time and money in compiling data do have rights over that data.
- Routine data collected for patient care, health care workers are government employees, so local health facility and host government owns that data
- Concern that foreign donor might have sponsored compilation of data, and would then have rights. When it is a research funder, and data was expected to be used to publish research, courts would uphold these rights?
- Not the case in most European countries and Canada, and parts of the US
- NIH funded projects - data belongs to the public and can be requested
- Facebook was required to delete data
- Orgs have to publish info saying that the data exists, making a copy available to the individual