IT'S been hard to avoid discussing nurses over the last few weeks – quite apart from the fact that our daughter, who like many "twentysomethings" is still living at home, always wants to talk about her work as a nurse when she finishes her shift as a way of unwinding.

As a criminologist, rather than a dad, I’m now also used to being asked about Charles Cullen, the American serial killer who’s recently been portrayed by Eddie Redmayne in the film The Good Nurse. I found this drama much more satisfying than the Netflix series Dahmer, largely because it didn’t prioritise the story of Cullen the serial killer, but instead the investigation that brought his killing cycle to an end, and the part played in his unmasking by his friend Amy Loughren – the “good nurse” of the title.

Frankly I never want to make the murderer the heart of the narrative either in my research, or in my more popular writing where I have always tried to prioritise the victims, or the story of how the killer was caught.

Cullen’s case also raises the broader issue of that very small group of nurses who abuse their position of trust and kill their patients – a subject that I researched with my colleague Professor Elizabeth Yardley in 2014. Why did nurses behave in this way, and what could hospital administrators and law enforcement do to ensure that this rare type of killer did not become more common? What “red flags”, as we called them, about the nurse or about their work performance give cause for concern, and how many of these red flags needed to be present before action would have to be taken?

Our research was based on a sample of nurses who had been convicted of murdering their patients in Europe and North America, and we used a 22-point checklist of personality traits and work behaviours that were associated with that sample. We then discovered how a median of six of these 22 points tended to cluster in the backgrounds of more than half of our sample. This cluster of red flags were higher incidences of death on his/her shift; has a history of mental instability and/or depression; makes colleagues anxious; moves from one hospital to another on a regular basis; is found to be in possession of drugs (both legal and illegal) at home/in their work locker; and, appears to have a personality disorder.

The most prevalent red flag that we identified was to have had higher incidences of death on his/her shift. In other words, where there were a number of deaths over a specified time period that exceeded those that were expected when compared to the usual number of deaths for that ward or the hospital, and shift patterns were then used to determine a suspect. Attendance data revealed the presence or absence of particular members of staff during, or around the time of these unusual death rates.

However, and this was the important finding, we discovered that there was an uneasy fit between scientific and legal principles in this context, as attendance data didn’t actually establish guilt on the basis of the “similar fact” principle of evidence where there needed to be a direct association between specific actions and specific events.

So, we discovered that what might appear sound and convincing from a policing point of view had great deficiencies when employed as evidence in court, and there were practical consequences as a result. Attendance data had been used to convict, for example, the Dutch nurse Lucia de Berk of killing seven of her patients in 2003, although she was acquitted seven years later when it was accepted that attendance data alone could not prove her guilt.

As a result, our research suggested that having just this one red flag should never be used as a basis for conviction, as a higher than average number of deaths over a given period of time may have various explanations – of which an active serial killer is only one – and therefore could in our view only be used as a basis to convict when found in combination with other red flags.

This nursing preamble leads me inexorably to Colin Norris. Born in Glasgow, and trained in Dundee, Norris was convicted in 2008 of killing four of his elderly patients by injecting them with insulin in Leeds, where he worked as a nurse at St James’s University Hospital and Leeds General Infirmary. It was claimed that Norris hated elderly, female patients and that was why Doris Ludlam, Bridget Bourke, Irene Cookes and Ethel Hall had lost their lives. However, Norris was an "outlier" – just like Lucia de Berk, and that encouraged us to look a little more closely at how he had come to be convicted.

We were disturbed by what we found. It was soon clear that the case against Norris was entirely circumstantial but through the police interrogating attendance data he became the supposed “common denominator” in this cluster of deaths, although there was no direct evidence whatsoever to link him to what had happened to Doris, Bridget, Irene and Ethel.

There was no “good nurse” like Amy Loughren who had worked out what Cullen was doing, and reported her anxieties to the police. Instead there was simply rumour and tittle-tattle that was used to spin a motive for Norris, who was jailed for a minimum of 30 years, but who has never stopped maintaining his innocence. To this day both Elizabeth and I believe that there has been a miscarriage of justice, and it is good to see that his case has now been sent back to the Court of Appeal.

Nurses who kill breach our trust at a time when we, or our loved ones are in greatest need. Thankfully they are very, very few in number, but if we want to catch this type of health care serial killer before they repeatedly kill, the lesson from our research was not to simply interrogate attendance data but to vet more carefully who becomes a nurse – despite all the pressures to fill vacancies, and to always take up references when a nurse applies for a post, but who wants to leave his or her previous job having only been there for a very short time.

• In Search of the Angels of Death: Conceptualising the Contemporary Nurse Healthcare Serial Killer by Elizabeth Yardley and David Wilson was published in the Journal of Investigative Psychology and Offender Profiling in 2014.