Copyright is owned by the Author of the thesis.  Permission is given for 
a copy to be downloaded by an individual for the purpose of research and 
private study only.  The thesis may not be reproduced elsewhere without 
the permission of the Author. 
 



 1 

 
 
 

Realism in Synthetic Data 
Generation 

 
 
 
 
 

A thesis presented in fulfilment of the requirements for the degree of: 
 
 

 
Master of Philosophy 

in 
Science 

 
 
 
 
 

Scott McLachlan  

(MCSE, MCT, DipSysEng, GradDipInfSc, GradDipLaw, GradDipBus, MIITP, MBCS) 

School of Engineering and Advanced Technology 
Massey University 

Palmerston North, New Zealand 
 

 
 
 
 
 

Supervised by: 
 

Dr. Kudakwashe Dube 

School of Engineering and Advanced 
Technology 

Massey University 
Palmerston North, New Zealand 

 

Prof. Thomas Gallagher 

Applied Computing and Engineering Technology 
Missoula College 

University of Montana 
Missoula, USA 

 
 

2017  



 2 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Copyright is owned by the Author. Permission is given for a copy to be downloaded by an individual for 
the purpose of research and private study only. This thesis may not be reproduced or disseminated 
elsewhere without the express written permission of the Author. 
 
  



 3 

 
 
 
 

Abstract 
 
 
There are many situations where researchers cannot make use of real data because either the data 

does not exist in the required format or privacy and confidentiality concerns prevent release of the data. 

The work presented in this thesis has been undertaken in the context of security and privacy for the 

Electronic Healthcare Record (EHR). In these situations, synthetic data generation (SDG) methods are 

sought to create a replacement for real data. In order to be a proper replacement, that synthetic data 

must be realistic yet no method currently exists to develop and validate realism in a unified way. This 

thesis investigates the problem of characterising, achieving and validating realism in synthetic data 

generation. A comprehensive domain analysis provides the basis for new characterisation and 

classification methods for synthetic data, as well as a previously undescribed but consistently applied 

generic SDG approach. In order to achieve realism, an existing knowledge discovery in databases 

approach is extended to discover realistic elements inherent to real data. This approach is validated 

through a case study. The case study demonstrates the realism characterisation and validation 

approaches as well as establishes whether or not the synthetic data is a realistic replacement. This 

thesis presents the ATEN framework which incorporates three primary contributions: (1) the THOTH 

approach to SDG; (2) the RA approach to characterise the elements and qualities of realism for use in 

SDG, and finally; (3) the HORUS approach for validating realism in synthetic data. The ATEN framework 

presented is significant in that it allows researchers to substantiate claims of success and realism in 

their synthetic data generation projects. The THOTH approach is significant in providing a new 

structured way for engaging in SDG. The RA approach is significant in enabling a researcher to discover 

and specify realism characteristics that must be achieved synthetically. The HORUS approach is 

significant in providing a new practical and systematic validation method for substantiating and justifying 

claims of success and realism in SDG works. Future efforts will focus on further validation of the ATEN 

framework through a controlled multi-stream synthetic data generation process. 

  



 4 

 
 
 

Publications related to this thesis: 
 
 
 
 
McLachlan, S., Dube, K., & Gallagher, T. (2017). Managing Realism in Synthetic Data Generation. 
Manuscript submitted to JAMIA. 
 
McLachlan, S., Dube, K., & Gallagher, T. (2017). THOTH: The generic approach to and characterisation 
of Synthetic Data. Manuscript submitted to JAMIA. 
 
Walonoski, J., Kramer, M., Nichols, J., Quina, A., Moesel, C., Hall, D., Duffett, C., Dube, K., Gallagher, 
T., & McLachlan, S. (2017). Synthea: An approach, method and software mechanism for generating 
synthetic patients and the synthetic electronic healthcare record. Manuscript submitted to JAMIA. 
 
McLachlan, S., Dube, K., & Gallagher, T. (2017). The Realistic Synthetic Electronic Health Record: 
Challenges, rationale and future directions. Manuscript submitted to JAMIA. 
 
McLachlan, S., Dube, K., & Gallagher, T. (2016). Using the CareMap with health incidence statistics for 
generating the realistic synthetic electronic health record. IEEE International Conference on Healthcare 
Informatics, ICHI’16. 
 
 
 
 
 
 
 

Glossary 
 

ATEN The ATEN framework is an SDG lifecycle incorporating the THOTH, RA and HORUS 
approaches.  
 

AU DoH Australian Department of Health 
 

CPG Clinical Practice Guideline 
 

HiKER Group Heath Informatics and Knowledge Engineering Research Group 
 

HIS Health Incidence Statistics 
 

HORUS Uses the knowledge developed by RA as the basis for validating realism in synthetic 
data and justifying success in SDG. 
 

NZ MoH New Zealand Ministry of Health 
 

RA A systematic approach used to discover realistic elements, characteristics and rules 
necessary to the creation of realistic synthetic data. 
 

PK Primary Key 
 

SDG Synthetic Data Generation 
 

THOTH The generic approach for SDG 
 

  



 5 

 
 

 
Dedicated for Danika, Thomas, Liam and James. 

 
 
 

 
 

Acknowledgements 
 
I acknowledge with the greatest of appreciation the assistance of my supervisors and the wider 
members of the Health Informatics and Knowledge Engineering Research (HiKER) Group who 
supported my development as a researcher in the tradition of the scientific method. The support of my 
proof reader and sometimes editor who every day pointed out when my references were out of order 
and when what I had written didn’t actually say what I thought I had said.  
 
And I can’t leave out Master 4, who recognised that I focus and work better when I have multiple streams 
of input and things to think about. Using this as only a four-year old can; as justification for continually 
distracting me with games, puzzles, stories and an insatiable need for me to join him as he played with 
his vast collection of toy trains. My hope is that I live to see the day when my encouragement of you 
culminates in my receiving a copy of your own thesis. I especially look forward to discussions about the 
distractions you had to deal with. 
 
There are scores of others with whom I have interacted during the eight months spent researching and 
writing this thesis. But for the fact that it would take vast amounts of time and far more space than I am 
given on this page to single you all out, I offer my best wishes and thanks.  
 
 

 
February, 2017.  
Sydney, Australia.  
 
 
 
 
 
 
 
 
 

To the reader; 
 
The fact that you have chosen to pick up or download this thesis is an act that in and of 
itself deserves thanks. If nothing else, and in deference to the content, this single act 
justifies this thesis’ existence.  
 
Thank you.  

 
 
 
 
This thesis is also a tribute to the late bloomers. People like Nikola Tesla, Charles Darwin, Samuel 
Jackson and Richard Adams. To all those who didn’t even begin to realise their vast potential until later 
in life.  
  



 6 

 

Table of Contents 



 7 



 8 

 
 
  



 9 

 

Tables 

Table 1: Established Classifications for Computational Models ........................................................... 25
Table 2: Comparison of Rubin (1993) to Birkin & Clark (1987) ............................................................ 43
Table 3: Characterisation of Synthetic Data Generation Methods........................................................ 46
Table 4: Classification of Synthetic Data .............................................................................................. 48
Table 5: Simplified Generalised Narrative of SDG Articles ................................................................... 63
Table 6: Justification Examples for Part 1 of the Simplified Generalised Narrative.............................. 63
Table 7: Operational examples for Part 2 of the Simplified Generalised Narrative .............................. 63
Table 8: Result examples for the Simplified Generalised Narrative ..................................................... 64
Table 9: Ethnicity Statistics for births at CMDHB in 2012 (expressed as percentages) ....................... 88
Table 10: Age Statistics for births at CMDHB in 2012 (expressed as percentages) ............................ 88
Table 11: Midwifery Patient Database Patient Relational Table Schema extract ................................. 89
Table 12: Formal Concept Analysis for 10 Random Labour and Birth Patients ................................... 92
Table 13: Generalised Relation Table .................................................................................................. 94
Table 14: The qualitative classification rule for Caesarean based on previous mode/s of delivery ..... 94
Table 15: Realism Validation Questions ............................................................................................. 100
Table 16: CoMSER Input Validation Case Study................................................................................ 104
Table 17: Demographic Analysis Table from CoMSER CoMENGINE ............................................... 106
Table 18: Ethnicity Statistics Comparison ........................................................................................... 106
Table 19: Age Statistics Comparison .................................................................................................. 106
Table 20: Synthetic Data Generation Literature ................................................................................. 128
Table 21: Realism in SDG Approaches .............................................................................................. 132
Table 22: Sample gender-specific conditions from the Kartoun (2016) EMR dataset. ....................... 139
Table 23: Ten Random Patients from Kartoun (2016) ........................................................................ 139
Table 24: Documents provided by the Synthea Team ........................................................................ 141
Table 25: Additional Sources for Type2 Diabetes Validation Data ..................................................... 142
 
 

Figures 

Figure 1: The Signpost Diagram used throughout this thesis ............................................................... 17
Figure 2: SDG Literature Search and Categorisation ........................................................................... 33
Figure 3: Distribution of SDG Methods ................................................................................................. 47
Figure 4: Distribution of SDG Domains ................................................................................................. 47
Figure 5: The ATEN Framework ........................................................................................................... 49
Figure 6: Context Diagram for the CoMSER Method (from McLachlan et al, 2016) ............................ 51
Figure 7: CoMSER UML Activity Diagram (from McLachlan et al, 2016) ............................................. 51
Figure 8: The Generic Approach to Synthetic Data Generation ........................................................... 57
Figure 9: The three-step THOTH approach .......................................................................................... 59
Figure 10: The Improved Generic Approach to Validation for Synthetic Data Generation ................... 65
Figure 11: Grounding Validation of the Generic Approach ................................................................... 66
Figure 12: Calibration Validation of the Generic Approach ................................................................... 67
Figure 13: Verification Validation of the Generic Approach .................................................................. 68
Figure 14: Harmonising Validation of the Generic Approach ................................................................ 69
Figure 15: The KDD Process ................................................................................................................ 84
Figure 16: Midwifery Patient Database Relational Schema extract ...................................................... 89
Figure 17: Concept Hierarchy for Child Birth ........................................................................................ 91
Figure 18: Concept Hierarchy for Child Birth with Statistics ................................................................. 91
Figure 19: Concept Lattice example ..................................................................................................... 93
Figure 20: Characteristic Rule from the domain of Midwifery ............................................................... 94
Figure 21: Classification Rule from the domain of Midwifery ................................................................ 95
Figure 22: The HORUS approach embedded into THOTH ................................................................ 102
Figure 24: Synthea Validation Review: Diabetes Prevalence ............................................................. 143
Figure 25: Age at Diagnosis of Type-2 Diabetes Mellitus ................................................................... 145
  



 10

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
“Behind every algorithm there is always a person. A person with a set of personal beliefs that no code 

can ever completely eradicate. You must identify your own personal bias. You need to understand that 

you are human and take responsibility accordingly.”  

 

 (Ekstrom, 2015)