Identification of practitioners at high risk of complaints to health profession regulators.
Dentists
Doctors
Patient complaints
Quality and safety
Risk prediction
Journal
BMC health services research
ISSN: 1472-6963
Titre abrégé: BMC Health Serv Res
Pays: England
ID NLM: 101088677
Informations de publication
Date de publication:
13 Jun 2019
13 Jun 2019
Historique:
received:
26
08
2018
accepted:
03
06
2019
entrez:
15
6
2019
pubmed:
15
6
2019
medline:
24
9
2019
Statut:
epublish
Résumé
Some health practitioners pose substantial threats to patient safety, yet early identification of them is notoriously difficult. We aimed to develop an algorithm for use by regulators in prospectively identifying practitioners at high risk of attracting formal complaints about health, conduct or performance issues. Using 2011-2016 data from the national regulator of health practitioners in Australia, we conducted a retrospective cohort study of 14 registered health professions. We used recurrent-event survival analysis to estimate the risk of a complaint and used the results of this analysis to develop an algorithm for identifying practitioners at high risk of complaints. We evaluated the algorithm's discrimination, calibration and predictive properties. Participants were 715,415 registered health practitioners (55% nurses, 15% doctors, 6% midwives, 5% psychologists, 4% pharmacists, 15% other). The algorithm, PRONE-HP (Predicted Risk of New Event for Health Practitioners), incorporated predictors for sex, age, profession and specialty, number of prior complaints and complaint issue. Discrimination was good (C-index = 0·77, 95% CI 0·76-0·77). PRONE-HP's score values were closely calibrated with risk of a future complaint: practitioners with a score ≤ 4 had a 1% chance of a complaint within 24 months and those with a score ≥ 35 had a higher than 85% chance. Using the 90th percentile of scores within each profession to define "high risk", the predictive accuracy of PRONE-HP was good for doctors and dentists (PPV = 93·1% and 91·6%, respectively); moderate for chiropractors (PPV = 71·1%), psychologists (PPV = 54·9%), pharmacists (PPV = 39·9%) and podiatrists (PPV = 34·0%); and poor for other professions. The performance of PRONE-HP in predicting complaint risks varied substantially across professions. It showed particular promise for flagging doctors and dentists at high risk of accruing further complaints. Close review of available information on flagged practitioners may help to identify troubling patterns and imminent risks to patients.
Sections du résumé
BACKGROUND
BACKGROUND
Some health practitioners pose substantial threats to patient safety, yet early identification of them is notoriously difficult. We aimed to develop an algorithm for use by regulators in prospectively identifying practitioners at high risk of attracting formal complaints about health, conduct or performance issues.
METHODS
METHODS
Using 2011-2016 data from the national regulator of health practitioners in Australia, we conducted a retrospective cohort study of 14 registered health professions. We used recurrent-event survival analysis to estimate the risk of a complaint and used the results of this analysis to develop an algorithm for identifying practitioners at high risk of complaints. We evaluated the algorithm's discrimination, calibration and predictive properties.
RESULTS
RESULTS
Participants were 715,415 registered health practitioners (55% nurses, 15% doctors, 6% midwives, 5% psychologists, 4% pharmacists, 15% other). The algorithm, PRONE-HP (Predicted Risk of New Event for Health Practitioners), incorporated predictors for sex, age, profession and specialty, number of prior complaints and complaint issue. Discrimination was good (C-index = 0·77, 95% CI 0·76-0·77). PRONE-HP's score values were closely calibrated with risk of a future complaint: practitioners with a score ≤ 4 had a 1% chance of a complaint within 24 months and those with a score ≥ 35 had a higher than 85% chance. Using the 90th percentile of scores within each profession to define "high risk", the predictive accuracy of PRONE-HP was good for doctors and dentists (PPV = 93·1% and 91·6%, respectively); moderate for chiropractors (PPV = 71·1%), psychologists (PPV = 54·9%), pharmacists (PPV = 39·9%) and podiatrists (PPV = 34·0%); and poor for other professions.
CONCLUSIONS
CONCLUSIONS
The performance of PRONE-HP in predicting complaint risks varied substantially across professions. It showed particular promise for flagging doctors and dentists at high risk of accruing further complaints. Close review of available information on flagged practitioners may help to identify troubling patterns and imminent risks to patients.
Identifiants
pubmed: 31196074
doi: 10.1186/s12913-019-4214-y
pii: 10.1186/s12913-019-4214-y
pmc: PMC6567559
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
380Subventions
Organisme : Australian Research Council
ID : FT180100075
Références
Stat Med. 2000 Jan 15;19(1):13-33
pubmed: 10623910
Health Care Manag Sci. 2000 Sep;3(4):269-77
pubmed: 11105413
JAMA. 2002 Jun 12;287(22):2951-7
pubmed: 12052124
Am J Med. 2005 Jul;118(7):773-7
pubmed: 15989912
Ann Intern Med. 2006 Jan 17;144(2):107-15
pubmed: 16418410
Acta Paediatr. 2007 Mar;96(3):338-41
pubmed: 17407452
JAMA. 1991 Oct 16;266(15):2093-7
pubmed: 1920697
BMJ Qual Saf. 2013 Jul;22(7):521-4
pubmed: 23576772
BMJ Qual Saf. 2013 Jul;22(7):532-40
pubmed: 23576774
Stat Med. 2013 Dec 30;32(30):5381-97
pubmed: 24027076
BMJ Qual Saf. 2014 Feb;23(2):147-52
pubmed: 24101575
BMC Med. 2015 Jan 06;13:1
pubmed: 25563062
Aust Health Rev. 2015 Sep;39(4):483-485
pubmed: 25796534
BMJ Qual Saf. 2015 Jun;24(6):360-8
pubmed: 25855664
N Engl J Med. 2016 Jan 28;374(4):354-62
pubmed: 26816012
BMC Med. 2016 Dec 2;14(1):198
pubmed: 27908294
Br J Psychiatry. 2017 Jun;210(6):387-395
pubmed: 28302700
Arch Gen Psychiatry. 1983 Mar;40(3):249-57
pubmed: 6830404
JAMA. 1994 Nov 9;272(18):1421-6
pubmed: 7933423