The following blog features an article summary of a working paper I have with Lorna Ferguson examining police missing persons data accuracy across categories used for risk assessment and determining police response. You will also learn a little about my involvement in this research as a police officer — a unique opportunity for a RCMP officer who is used to investigating and not diving deep into research and evidence-based techniques. The challenges and benefits to embarking on such an endeavor, while still balancing a professional and family life, was a project worth sacrifice as it contributes to advancing evidence-based policing in Canada. But, you'll learn a bit more about this throughout this blog post.
First, Some Background on This Research Project:
Multiple inquiries into missing persons — especially missing women and girls — throughout Canada highlight the difficulties with the data entry and quality of the available data regarding missing and murdered persons (see here and here for examples). Missing person reports are a priority for police in Canada, and so being able to properly classify different types of missing persons cases is essential for risk assessments and police responses. Put another way, different types of classifications can impact risk assessment results, which assist in determining the urgency level and subsequent police search and/or investigation. As a police officer with 16 years of experience, I knew that having an evidence-based approach in an area of policing such as this could be beneficial for the police operating environment, especially given that accurate data and data quality can impact the resources allocated to and outcome of each missing person case. To this end, my research colleague, Lorna Ferguson, and I agreed that a study examining the ‘History’ classifications (‘no previous history,’ ‘repeat,’ and ‘habitual/chronic’) of missing person reports that are often used in risk assessment could be beneficial for both academics and police organizations.
Why? Well, because, within existing literature, the ambiguities surrounding what constitutes a missing individual as “repeat” or “habitual/chronic” have been discussed. The main issues being that there is no formal definition for, nor is there any research on, how many times an individual is reported missing before they are provided with each classification. With no formal definition, each police service, and even individual police officers, are left to make their own definition – ultimately resulting in data quality issues. In addition, the very little research on how many times an individual is reported missing before they are provided with these classifications calls into question where the categories come from and if they are useful in such important operational tasks/decisions. This study dives deep into these classifications to determine and reveal distinctions in how missing person cases could/should be classified.
For this, we used closed missing persons reports and predictively analyzed how these cases could/should be classified outside of known data errors. For this, data were extracted from two Canadian municipal police services record management systems, which included all closed missing persons reports between 2014 to 2019.
What we found was very interesting. When it came to data quality issues, it was noticed that at around 10 previous missing occurrences, data issues tended to subside. In my personal opinion, this number is relatively high when you are investigating missing person cases and should be an area of concern. Our study also found that 10.9 percent of cases had no previous history of going missing in these data, highlighting that 89.1 percent of missing person cases are individuals who have gone missing multiple times. In addition, it was noticed that those with zero previous missing events that were classified as ‘repeat’ or ‘habitual/chronic,‘ and those with one to 20 previous missing events that were classified as ‘no previous history’ made up 8.6 per cent of all missing person records. We ultimately discovered that at least around 9 per cent of all missing persons cases are misclassified, with these figures likely being higher. This conclusion comes from that fact that we could not determine if data errors occurred for those noted as ‘repeat’ or ‘habitual/chronic’ with one or more previous missing events due to the absence of formal definitions that would make it clear which number of previous reports should be noted under each category.
We also found that most cases with no previous history of going missing are classified accurately, as 60.1 per cent are marked as ‘no previous history.’ But this means that 39.9 per cent of missing person cases that had no previous history of going missing are classified incorrectly by police as either ‘repeat’ or ‘habitual/chronic,’ or were not assigned a label, regarding those cases with zero previous missing episodes. This misclassification also occurred for persons reported missing more than one time. For instance, those with one previous missing person report were classified as having ‘no previous history’ of going missing in 19.3 per cent of the cases.
Translating such large data errors only in this section of data to every column of missing persons data points to a huge area of concern. Then, to further examine this, ruling out the missing person cases with known data quality issues, we were then able to predict case classifications to see if standardization for these cases could occur at all. When we did this, we found missing person cases classified as ‘repeat’ are significantly more likely to have one, two or three missing person reports. Those with one previous missing occurrence are over three times more likely to be classified as repeat compared to habitual/chronic cases. This highlights the potential for some type of a standardized definition regarding these classifications.
Thoughts on Findings:
As a police officer who has worked many years as a general duty cop before working in specialized sections, I found these results to be compelling. From this study, there are contributions that could be made to existing literature on missing persons, as well as for policing. The first being the descriptive overview that recognized several data quality issues concerning case classifications, documenting potential validity issues regarding police missing persons data. This is critical for frontline officers and specialized sections investigating missing person cases, as we rely on our systems to assist with our risk assessment process. Data quality is of the utmost importance for our day-to-day operations. The second is predictions estimating the increased chances for case classifications across certain numbers of previous missing episodes for each case type. Again, this shows an area in policing and missing persons that could potentially be reviewed for better practices.
This study responds to calls for the development of a formal definition for these case classification types. Based on the estimations offered in this research, the findings lend support for the potential of ‘repeat’ missing persons to be formally defined as cases involving one to three previous missing episodes, and ‘habitual/chronic’ cases to be regarded as those reports involving four or more prior disappearances. Having a formal definition takes the ‘guessing’ from the file and allows a more in-depth, evidence-based approach to risk assessment and police response.
Particularly, standardizing definitions for ‘repeat’ and ‘habitual/chronic’ categories allows for risk assessment to appropriately reflect the urgency- and risk-level of the case (i.e., ensuring the case is labelled adequately to reflect the reality of missingness for each individual), and, as such, suitable police resources can be allocated to locating missing individuals. Further, data accuracy can assist in ensuring that the already strained police resources are not stressed further whereby greater resources are not allocated to, for example, a ‘repeat’ missing person case that is marked as at-risk as a result, but yet has no prior missing events.
To conclude, I want to highlight that this is just one data column of several in missing persons data used for risk assessment and determining police response that requires examination. As an RCMP officer for 16 years, I know how important it is to evaluate our operating environment, as well as conduct and review research to ensure we are performing our duties to the best of our abilities as a police organization. I believe this research highlights an area —data quality — of our policing practices that needs more attention and points to a matter that needs to be taken up as a task amongst evidence-based policing practitioners, scholars, analysts, and so on. I also believe it brings to light an area of concern that most frontline officers don’t necessarily think could have such a profound impact on the way they conduct their search and/or investigation. As a police officer, you learn and apply the policies and practices, which are part of the organization, but don't often dig deep into the reason behind what made these policies and practices the way they are. Classification of missing persons cases is an excellent example of this. The results from this study should create a dialogue and inform action for police organizations, policymakers and academic researchers in regard to missing person reports and their classifications, along with missing persons data.
Writer: Sgt. Wendy Picknell
Wendy Picknell is a Sergeant in the Royal Canadian Mounted Police (RCMP) and is part of the Vulnerable Person Unit, Victim of Crime Section. Wendy has a degree in Police Studies and Psychology. Her research has been focused on research papers related to various areas in policing, while finishing her Bachelor of Arts degree. Wendy was a Virtual Scholar in Residence for Canadian Society of Evidence-Based Policing (CAN-SEBP) and her current research is focused on missing persons.