technology
Threat alert
Algorithm predicts health
practitioners who pose
risks to patient safety.
By Dallas Bastian
H
ealth profession regulators have
their eye on an algorithm that can
identify practitioners at high risk
of complaints.
Called PRONE-HP (Predicted Risk of
New Event for Health Practitioners) and
developed by a team from the University
of Melbourne and Stanford University, the
tool sought to fix a challenge regulators
face: identifying health practitioners who
pose substantial threats to patient safety.
Among the characteristics and markers
that the model uses to estimate risk are
demographics like age, sex and practice
locations; profession and speciality; and
information of the number and nature of
prior complaints.
Associate Professor Matthew Spittal,
ARC Future Fellow at the University of
Melbourne, said if a practitioner had a high
score then it would be “a ‘red flag’ to look
more deeply to see, for instance, if there
was a pattern of troubling behaviour or an
imminent risk to patients”.
In their paper, the study’s authors said:
“Because the score can be re-calculated
each time a complaint is lodged, it could
be useful for flagging cases that warrant
deeper review – for example, reviewing
previous complaints against a practitioner
or other information held by the regulator
to ascertain if there are troubling patterns
of care or imminent risks to patients.”
They said this may be the most
appropriate way to use the model.
Still, Spittal noted that the score is not
perfectly predictive of risk.
“There may be practitioners that have
a high score, for example, because of
the area of medicine that they work in,
but do not pose a risk to patient safety,”
he explained.
“For this reason, a score alone will never
be sufficient to classify a practitioner as
risky or not.
“A score isn’t a substitute for a thorough
investigation, but it may assist in prioritising
notifications for investigation.”
So will the tool be used for the nursing
profession in the near future?
Spittal said the model is not likely to be
useful for all professions.
While the research team found that
the tool showed promise for identifying
high-risk doctors and dentists, and
showed potential for identifying high-risk
chiropractors, psychologists, pharmacists
and podiatrists, it was less successful
when it came to other professions
that the Australian Health Practitioner
Regulation Agency regulates, like nurses
and midwives, due to the number of
false positives.
There were far too many instances
where practitioners were identified as high
risk but did not have another notification,
Spittal explained.
“Partly, this is because the number
of notifications relative to the overall
number of nurses and midwives is
If a practitioner had a
high score, it would be a
‘red flag’ to look more deeply
to see if there was a pattern
of troubling behaviour or an
imminent risk to patients.
low – much lower than for doctors
and dentists.”
The team said profession-specific
algorithms might perform better.
AHPRA was interested in seeing how
the tool could be used in practice, Spittal
said, but he added that will require
more work. ■
nursingreview.com.au | 31