Improving Data Quality by Establishing Good Governance Practices
By Preeya Beczek
Improving Data Quality with better Data Governance cross-functionally is a “hot topic”. Companies are investing in people and technology to get more out of their information assets, realising this through large system and process modernization, plus the ability to connect data across functions, and the bottom line is “the quality of information in systems has to improve – period”. Great things happen when there is high confidence in the information we use as regulatory professionals every day – speed, better decisions, efficiency, and trust.
Data Quality – The magic wand
If we could wave a magic wand for efficiency and low costs when it comes to Data Quality (DQ), I am pretty sure the wish list would look something like this: automated collection of data, prepopulated data fields and transformation in structured format, and data in compliance with data standards. Data would be only entered once, be easy to find, easy to share, easy to analyse; goes to the relevant user in a seamless fashion, user can quickly confirm acceptance of accurate data. Sweet dreams.
Data Quality – where to start eating this elephant?
As we move towards a world of data sharing, structured data transmission, and integrated data with automation, DQ is and will increasingly become an integral part of regulatory submissions in the life sciences industry.
So let me answer some questions that I think organisations and teams are asking:
Data Driven – what does that mean for your organisation?
The main thing is the submission and with the growing structure data submission requirements by many health authorities, most regulatory organizations are exploring ways to become much more data focused and less document driven. The impact of this shift is that regulatory data has to be mastered between Regulatory, R&D, and Commercial systems (Master Data Management) and have good Reference Data Management principles too. Fundamentally, the information being submitted meets the regulations and standards, however, there is growing complexity from health authorities on the “interpretation” of some of the data standards such as IDMP. We are quickly moving to a future where data is expected to be real-time with clear authoritative sources – where there will no longer be any questions on “authoritative information”.
Firstly, lets connect what we already know.
DQ relies on good business policies and processes, authoritative systems of use for those business processes, with defined responsibilities, identified data standards, metrics and KPI. This collection of activities is only possible with a good governance framework. The challenge today is that each department has their own terminology, data silos in several systems managed by several different people, and poor data governance (if it exists at all). So where is the connection, the master data management? Where is the governance and continuous oversight?
There is significant effort and focus being made across the life sciences industry to establish and embed DQ governance councils / offices cross-functionally to substantially improve the quality and speed of their information resources. The challenge will be if can we keep up with the technology. Both industry and health authorities are on the same race. The trick is to sprint before you run.
What are the tangible actions for a Data Quality Council / Office?
- 1) Select and understand the domain – start with one e.g., Safety or chemistry manufacturing and control (CMC). Maybe the choice comes from an area that has some deviations or efficiency gains by looking at DQ and getting a first-time right focus.
- 2) Engage with the stakeholders in the chosen domain to understand the areas that they are interested in, to achieve DQ in their domain. Ask: “What is considered as “noncritical data” to prioritize, as this may help with management of inevitable mistakes. E.g., DQ challenges resulting from transcription/interpretation errors submitted in an IND/CTA or IMPD submission is less critical compared to a case where, if manufacturing operations has evolved / changed the method and there is inconsistency of what was submitted vs. what is happening on the ground. This would be considered a DQ issue and process issue.
- 3) Identify the data standards already defined / adopted and select the relevant ones for the chosen domain that need focus. Some of these standards may be regulated and some may be internally set standards.
- 4) Define the desired outcomes of the DQ program. Often, it’s to gain accuracy, transparency, reusability, compliance, reproducibility, etc. Usually, the outcomes are directly matching the DQ principles the organisation may have chosen to follow (e.g., ALCOA+ and FAIR) at an enterprise level.
- 5) Ensure the processes are applicable to the domain and have the relevant roles and responsibilities identified to which the DQ responsibilities apply. Some organisations weave in the DQ responsibilities to existing roles, others decide to hire in dedicated staff. Ensure that the staff know their responsibilities towards DQ. Often in teams, staff will undertake DQ actions, but these may not be formalised in job descriptions and processes.
- 6) Work with the stakeholders to collect baseline information from the relevant systems of use for the key business processes. Some data may be in the system and some data may be in the documents within a system and it’s a good idea to obtain a report / sample of both types of data when checking for DQ. When data is entered directly into system fields, there can be errors or gaps, but this can be trended easily via system reports and audit trails. When data is contained in documents, often it is a narrative, interpretation, presentation, and justification of several data sources which the health authority is receiving for review and assessment. Companies often have review and quality control of submission documents against source files, to avoid DQ issues.
- 7) Review the reports and outcomes of the monitoring activity undertaken. Share the findings with the stakeholder team sharing trends, gaps, and any errors. Look at what data gaps and issues are critical vs noncritical to fix.
- 8) Make recommendations for improvements; set time intervals to run another round of reports and monitor actions to check if any corrective actions were successful.
This above approach is an iterative approach and only a handful of organisations are at this stage in their DQ and governance journey. The DQ Office is instrumental in facilitating the above approach recommending and prioritizing improvement areas. This is a cross functional and cross domain effort where stakeholder engagement is key. Typically, over a 6 to 12-month period, a strong and focused DQ Office can have impactful insights of DQ issues and challenges at a business process and functional level. Small and simple insights can often lead to speed, compliance, and quality.
