Data Integrity for Record Management

Last night speaking at the DFW Audit SIG one of the topics I wished I had gone a little deeper on were controls, and how to gauge their strength.

As I am preparing to interview candidates for a records management position, I thought I would flesh out controls specific to the storage of and access to completed or archived paper records, such as forms, as an example.

These controls are applied at the record or system level and are meant to prevent a potential data integrity issue from occurring.

Generation and Reconciliation of Documents

 Data Criticality
Unique identifierFor each recordNoNo
Who performs controlled issuanceIndividuals authorized by quality unit from designated unit (limited, centralized)Individuals authorized by quality unit from (limited, decentralized)Anyone (unlimited, decrentalized), often user of record
ReconciliationFull reconciliation of record and pages based on unique identifierFull reconciliation of records and pages based on quantity issuedNo reconciliation
Controlled printYesYesNo
Bulk printingNoYes, by controlled processYes
Destruction of blank formsPerformed by issuing unit, quality oversight required (High level of evidence)Performed by the operating or issuing unit, quality unit oversight requiredPerformed by the individual, quality unit oversight required (periodic walk throughs, self-inspections and audits)

Storage and Access to completed and archived paper records

 Data Criticality
Where StoredClimate-controlled roomClimate-controlled roomOffice retention location
How Removed & ReturnedLimited conditions for removal (e.g. regulatory inspections) method of recording the removal and return of the record(e.g. archive management system, logbook). Most use of documents either in controlled reading area or by scans.Method of recording the removal and return of the record(e.g., archive management system, logbook).Method (e.g. logbook) recording of documents checked-in/checked-out
Access ControlCard key access with entry and exit documented.Card key access with entry and exit documented.Limited key access
Periodic User Access ReviewAnnuallyAnnuallyEvery 2 years

There are also the need to consider controls for paper to electronic, electronic to paper and my favorite beast, the true copy.

For paper records a true copy of a picture of the original that keeps everything – a scan. The regulations state that you can get rid of the paper if you have a true copy. Many things called a true copy are probably not a true copy, to ensure an accurate true copy add two more controls.

 Data Criticality
Review requirementsDocumented review by second person from the quality unit for legibility, accuracy, and completenessDocumented review by second person (not necessarily from the quality unit) for legibility, accuracy, and completenessDocumented verification by person performing the scan for legibility, accuracy, and completeness
Discard of original allowedYes, as defined by quality unit oversight, unless there is a seal, watermark, or other identifier that can’t be accurately reproduced electronically.Yes, performed by the operating unit, unless there is a seal, watermark, or other identifier that can’t be accurately reproduced electronically. Quality unit oversight requiredYes, individual can discard original Quality unit oversight required

Upcoming Data Integrity Virtual Presentation

I will be presenting at the February Audit SIG of the DFW Section of the ASQ on Data Integrity.  Many companies struggle with the concepts of data integrity as it involves both paper and electronic data, dealing with legacy computer systems and the organization culture. This session will lay out the core principles of data integrity:

  • Organizational culture should drive ALCOA
  • Data governance is part of the management review process
  • Data Risk Assessments with appropriate mitigations (full risk management approach)

The Audit SIG webinar is scheduled for Tuesday February 9, 2021 at 6:00 pm.  To sign up RSVP to by February 8 by 6:00 pm.  An email with a link to the webinar will be returned to those that RSVP. 

MHRA 2019 GMP Inspection Data and Documentation observations

Transparency is something that regulatory agencies need to get better at, both in sharing more and doing it in a timely manner. The fact that the 2019 data from the MHRA was released in October of 2020 is pretty poor. As a reference, the FDA releases their data pretty reliably at the end of the calendar year for the given year.

Been evaluating the MHRA’s 2020 data on Chapter 4 Documentation, which is the 2nd largest category of observations in 2019 (and in 2018 before it).

80 different inspections cited comments against the Principles section

Good documentation constitutes an essential part of the quality assurance system and is key to operating in compliance with GMP requirements. The various types of documents and media used should be fully defined in the manufacturer’s Quality Management System.

Documentation may exist in a variety of forms, including paper-based, electronic or photographic media. The main objective of the system of documentation utilized must be to establish, control, monitor and record all activities which directly or indirectly impact on all aspects of the quality of medicinal products. The Quality Management System should include sufficient instructional detail to facilitate a common understanding of the requirements, in addition to providing for sufficient recording of the various processes and evaluation of any observations, so that ongoing application of the requirements may be demonstrated.

There are two primary types of documentation used to manage and record GMP compliance: instructions (directions, requirements) and records/reports. Appropriate good documentation practice should be applied with respect to the type of document.

Suitable controls should be implemented to ensure the accuracy, integrity, availability and legibility of documents. Instruction documents should be free from errors and available in writing. The term ‘written’ means recorded, or documented on media from which data may be rendered in a human readable form.

Principles, Chapter 4 Documentation

The Principles section then goes on to lay out the required document types.

I would love to see more. Is this 80 companies who don’t known what a SMF is? Good documentation practices? Don’t have SOPs and batch records? Have errors in their documents? Don’t approve them? More transparency would be valuable here.

We can learn more by drilling down in the document.

  • There are 87 inspections with 4.1 in section “Generation and Control of Documents”. 1 is critical and 25 are major. Here we see failures in understanding types of documents and controlling them, or maybe just having them in the first place.
  • The 82 against 4.2 (1 critical and 20 major) are more about having the manufacturing and testing process defined (and matching the filing).
  • 103 inspections with observations against 4.3 (23 major) show companies that do not have appropriate approval and release controls
  • 14 for 4.4 (6 major) means there are 14 companies out there who can’t write a good process and procedure. 4.4 has one of my favorite requirements “written in an imperative mandatory style”
  • 60 against 4.5 (13 major) demonstrates a lack of review and keeping documents up-to-date.
  • 12 companies (6 major) have terrible handwriting and cannot stick to ballpoints, yes in fact 4.7 states “Handwritten entries should be made in clear, legible, indelible way.”
  • 103 against 4.8 (1 critical and 28 major) is ALCOA focused on contemporaneous, attributable and accurate.
  • 18 for 4.9 (6 major) is for not correcting data correctly. That’s right 18 companies do not know how to comment correctly.
  • 22 for 4.10 (1 critical and 9 major) is for not clearly laying out the manufacturing records and keeping them for the retention period.
  • 19 for 4.29 (5 major) is a lack of process and procedure for a grab-bag of quality processes from change control to equipment management to cleaning

There are more, but we are in single digit observation territory.

Useful things to be evaluating in your own organization. As a good place to start, here are some questions to ask when contemplating data integrity.

2020 483s on data integrity

Data integrity continued to be a focus of the FDA, though the reduced inspections definitely led to fewer 483s.

Reference NumberShort DescriptionLong Description2020 Frequency2019 Frequency2018 Frequency
21 CFR 211.194(a)Complete test data included in recordsLaboratory records do not include complete data derived from all tests, examinations and assay necessary to assure compliance with established specifications and standards.  Specifically, , ***153833
21 CFR 211.194(a)(4)Complete Test DataLaboratory records are deficient in that they do not include a complete record of all data obtained during testing.  Specifically, ***102428
21 CFR 211.68(b)Backup data not assured as exact and completeBackup data is not assured as [exact] [complete] [secure from alteration, erasure or loss] through keeping hard copy or alternate systems.  Specifically, ***6671
21 CFR 211.194(a)(4)Data secured in course of each testLaboratory records do not include a complete record of all data secured in the course of each test, including all [graphs] [charts] [spectra] from laboratory instrumentation, properly identified to show the [specific component] [drug product container] [closure] [in-process material] [lot tested] [drug product tested].  Specifically, ***4128
21 CFR 211.68(b)Written record not kept of program and validation dataA written record of the program along with appropriate validation data has not been maintained in situations where backup data is eliminated by computerization or other automated processes.  Specifically, ***1671
483s related to data integrity

Human Performance and Data Integrity

Gilbert’s Behavior Engineering Model (BEM) presents a concise way to consider both the environmental and the individual influences on a person’s behavior. The model suggests that a person’s environment supports impact to one’s behavior through information, instrumentation, and motivation. Examples include feedback, tools, and financial incentives (respectively), to name a few. The model also suggests that an individual’s behavior is influenced by their knowledge, capacity, and motives. Examples include training/education, physical or emotional limitations, and what drives them (respectively), to name a few. Let’s look at some further examples to better understand the variability of individual behavioral influences to see how they may negatively impact data integrity.

Kip Wolf “People: The Most Persistent Risk To Data Integrity

Good article in Pharmaceutical Online last week. It cannot be stated enough, and it is good that folks like Kip keep saying it — to understand data integrity we need to understand behavior — what people do and say — and realize it is a means to an end. It is very easy to focus on the behaviors which are observable acts that can be seen and heard by management and auditors and other stakeholders but what is more critical is to design systems to drive the behaviors we want. To recognize that behavior and its causes are extremely valuable as the signal for improvement efforts to anticipate, prevent, catch, or recover from errors.

By realizing that error-provoking aspects of design, procedures, processes, and human nature exist throughout our organizations. And people cannot perform better than the organization supporting them.

Design Consideration

Human Error Considerations

Manage Controls

Define the Scope of Work

·       Identify the critical steps

·       Consider the possible errors associated with each critical step and the likely consequences.

·       Ponder the "worst that could happen."

·       Consider the appropriate human performance tool(s) to use.

·       Identify other controls, contingencies, and relevant operating experience.

When tasks are identified and prioritized, and resources

are properly allocated (e.g., supervision, tools, equipment, work control, engineering support, training), human performance can flourish.


These organizational factors create a unique array of job-site conditions – a good work environment – that sets people up for success. Human error increases when expectations are not set, tasks are not clearly identified, and resources are not available to carry out the job.

The error precursors – conditions that provoke error – are reduced. This includes things such as:

·       Unexpected conditions

·       Workarounds

·       Departures from the routine

·       Unclear standards

·       Need to interpret requirements


Properly managing controls is

dependent on the elimination of error precursors that challenge the integrity of controls and allow human error to become consequential.

Apply proactive Risk Management

When risk is properly analyzed we can take appropriate action to mitigate the risks. Include the criteria in risk assessments:

·       Adverse environmental conditions (e.g. impact of gowning, noise, temperature, etc)

·       Unclear roles/responsibilities

·       Time pressures

·       High workload

·       Confusing displays or controls

Addressing risk through engineering and administrative controls are a cornerstone of a quality system.


Strong administrative and cultural controls can withstand human error. Controls are weakened when conditions are present that provoke error.


Eliminating error precursors

in the workplace reduces

the incidences of active errors.

Perform Work


Utilizing error reduction tools as part of all work. Examples include:

·       Self-checking

o   Questioning attitude

o   Stop when unsure

o   Effective communication

o   Procedure use and adherence

o   Peer-checking

o   Second-person verifications

o   Turnovers


Engineering Controls can often take the place of some of these, for example second-person verifications can be replaced by automation.

Appropriate process and tools in place to ensure that the organizational processes and values are in place to adequately support performance.

Because people err and make mistakes, it is all the more important that controls are implemented and properly maintained.

Feedback and Improvement


Continuous improvement is critical. Topics should include:

·       Surprises or unexpected outcomes.

·       Usability and quality of work documents

·       Knowledge and skill shortcomings

·       Minor errors during the activity

·       Unanticipated workplace conditions

·       Adequacy of tools and Resources

·       Quality of work planning/scheduling

·       Adequacy of supervision

Errors during work are inevitable. If we strive to understand and address even inconsequential acts we can strengthen controls and make future performance better.

Vulnerabilities with controls can be found and corrected when management decides it is important enough to devote resources to the effort


The fundamental aim of oversight is to improve resilience to significant events triggered by active errors in the workplace—that is, to minimize the severity of events.


Oversight controls provide opportunities to see what is happening, to identify specific vulnerabilities or performance gaps, to take action to address those vulnerabilities and performance gaps, and to verify that they have been resolved.