Development and validation of a risk model for predicting adverse drug reactions in older people during hospital stay: Brighton Adverse Drug Reactions Risk (BADRI) model

Greg Scutt, Graham Davies, Bala Tangiisuran, Jennifer Stevenson, Juliet Wright, Graziano Onder, Mirko Petrovic, Trish van der Cammen, Chakravarthi Rajkumar

Research output: Contribution to journalArticleResearchpeer-review

Abstract

BACKGROUND: Older patients are at an increased risk of developing adverse drug reactions (ADR). Of particular concern are the oldest old, which constitute an increasingly growing population. Having a validated clinical tool to identify those older patients at risk of developing an ADR during hospital stay would enable healthcare staff to put measures in place to reduce the risk of such an event developing. The current study aimed to (1) develop and (2) validate an ADR risk prediction model. METHODS: We used a combination of univariate analysis and multivariate binary logistic regression to identify clinical risk factors for developing an ADR in a population of older people from a UK teaching hospital. The final ADR risk model was then validated in a European population (European dataset). RESULTS: Six-hundred-ninety patients (median age 85 years) were enrolled in the development stage of the study. Ninety-five reports of ADR were confirmed by independent review in these patients. Five clinical variables were identified through multivariate analysis and included in our final model; each variable was attributed a score of 1. Internal validation produced an AUROC of 0.74, a sensitivity of 80%, and specificity of 55%. During the external validation stage the AUROC was 0.73, with sensitivity and specificity values of 84% and 43% respectively. CONCLUSIONS: We have developed and successfully validated a simple model to use ADR risk score in a population of patients with a median age of 85, i.e. the oldest old. The model is based on 5 clinical variables (≥8 drugs, hyperlipidaemia, raised white cell count, use of anti-diabetic agents, length of stay ≥12 days), some of which have not been previously reported.
Original languageEnglish
Pages (from-to)1-9
Number of pages9
JournalPLoS ONE
Volume9
Issue number10
DOIs
Publication statusPublished - 30 Oct 2014

Fingerprint

Drug-Related Side Effects and Adverse Reactions
Length of Stay
Population
Multivariate Analysis
Sensitivity and Specificity
Hyperlipidemias
Teaching Hospitals
Cell Count
Logistic Models
Delivery of Health Care
Pharmaceutical Preparations

Bibliographical note

© 2014 Tangiisuran et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Cite this

Scutt, Greg ; Davies, Graham ; Tangiisuran, Bala ; Stevenson, Jennifer ; Wright, Juliet ; Onder, Graziano ; Petrovic, Mirko ; van der Cammen, Trish ; Rajkumar, Chakravarthi. / Development and validation of a risk model for predicting adverse drug reactions in older people during hospital stay: Brighton Adverse Drug Reactions Risk (BADRI) model. In: PLoS ONE. 2014 ; Vol. 9, No. 10. pp. 1-9.
@article{fb506fe5081d46ad87217627edc193f6,
title = "Development and validation of a risk model for predicting adverse drug reactions in older people during hospital stay: Brighton Adverse Drug Reactions Risk (BADRI) model",
abstract = "BACKGROUND: Older patients are at an increased risk of developing adverse drug reactions (ADR). Of particular concern are the oldest old, which constitute an increasingly growing population. Having a validated clinical tool to identify those older patients at risk of developing an ADR during hospital stay would enable healthcare staff to put measures in place to reduce the risk of such an event developing. The current study aimed to (1) develop and (2) validate an ADR risk prediction model. METHODS: We used a combination of univariate analysis and multivariate binary logistic regression to identify clinical risk factors for developing an ADR in a population of older people from a UK teaching hospital. The final ADR risk model was then validated in a European population (European dataset). RESULTS: Six-hundred-ninety patients (median age 85 years) were enrolled in the development stage of the study. Ninety-five reports of ADR were confirmed by independent review in these patients. Five clinical variables were identified through multivariate analysis and included in our final model; each variable was attributed a score of 1. Internal validation produced an AUROC of 0.74, a sensitivity of 80{\%}, and specificity of 55{\%}. During the external validation stage the AUROC was 0.73, with sensitivity and specificity values of 84{\%} and 43{\%} respectively. CONCLUSIONS: We have developed and successfully validated a simple model to use ADR risk score in a population of patients with a median age of 85, i.e. the oldest old. The model is based on 5 clinical variables (≥8 drugs, hyperlipidaemia, raised white cell count, use of anti-diabetic agents, length of stay ≥12 days), some of which have not been previously reported.",
author = "Greg Scutt and Graham Davies and Bala Tangiisuran and Jennifer Stevenson and Juliet Wright and Graziano Onder and Mirko Petrovic and {van der Cammen}, Trish and Chakravarthi Rajkumar",
note = "{\circledC} 2014 Tangiisuran et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.",
year = "2014",
month = "10",
day = "30",
doi = "10.1371/journal.pone.0111254",
language = "English",
volume = "9",
pages = "1--9",
journal = "PLoS ONE",
issn = "1932-6203",
number = "10",

}

Development and validation of a risk model for predicting adverse drug reactions in older people during hospital stay: Brighton Adverse Drug Reactions Risk (BADRI) model. / Scutt, Greg; Davies, Graham; Tangiisuran, Bala; Stevenson, Jennifer; Wright, Juliet; Onder, Graziano; Petrovic, Mirko; van der Cammen, Trish; Rajkumar, Chakravarthi.

In: PLoS ONE, Vol. 9, No. 10, 30.10.2014, p. 1-9.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Development and validation of a risk model for predicting adverse drug reactions in older people during hospital stay: Brighton Adverse Drug Reactions Risk (BADRI) model

AU - Scutt, Greg

AU - Davies, Graham

AU - Tangiisuran, Bala

AU - Stevenson, Jennifer

AU - Wright, Juliet

AU - Onder, Graziano

AU - Petrovic, Mirko

AU - van der Cammen, Trish

AU - Rajkumar, Chakravarthi

N1 - © 2014 Tangiisuran et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

PY - 2014/10/30

Y1 - 2014/10/30

N2 - BACKGROUND: Older patients are at an increased risk of developing adverse drug reactions (ADR). Of particular concern are the oldest old, which constitute an increasingly growing population. Having a validated clinical tool to identify those older patients at risk of developing an ADR during hospital stay would enable healthcare staff to put measures in place to reduce the risk of such an event developing. The current study aimed to (1) develop and (2) validate an ADR risk prediction model. METHODS: We used a combination of univariate analysis and multivariate binary logistic regression to identify clinical risk factors for developing an ADR in a population of older people from a UK teaching hospital. The final ADR risk model was then validated in a European population (European dataset). RESULTS: Six-hundred-ninety patients (median age 85 years) were enrolled in the development stage of the study. Ninety-five reports of ADR were confirmed by independent review in these patients. Five clinical variables were identified through multivariate analysis and included in our final model; each variable was attributed a score of 1. Internal validation produced an AUROC of 0.74, a sensitivity of 80%, and specificity of 55%. During the external validation stage the AUROC was 0.73, with sensitivity and specificity values of 84% and 43% respectively. CONCLUSIONS: We have developed and successfully validated a simple model to use ADR risk score in a population of patients with a median age of 85, i.e. the oldest old. The model is based on 5 clinical variables (≥8 drugs, hyperlipidaemia, raised white cell count, use of anti-diabetic agents, length of stay ≥12 days), some of which have not been previously reported.

AB - BACKGROUND: Older patients are at an increased risk of developing adverse drug reactions (ADR). Of particular concern are the oldest old, which constitute an increasingly growing population. Having a validated clinical tool to identify those older patients at risk of developing an ADR during hospital stay would enable healthcare staff to put measures in place to reduce the risk of such an event developing. The current study aimed to (1) develop and (2) validate an ADR risk prediction model. METHODS: We used a combination of univariate analysis and multivariate binary logistic regression to identify clinical risk factors for developing an ADR in a population of older people from a UK teaching hospital. The final ADR risk model was then validated in a European population (European dataset). RESULTS: Six-hundred-ninety patients (median age 85 years) were enrolled in the development stage of the study. Ninety-five reports of ADR were confirmed by independent review in these patients. Five clinical variables were identified through multivariate analysis and included in our final model; each variable was attributed a score of 1. Internal validation produced an AUROC of 0.74, a sensitivity of 80%, and specificity of 55%. During the external validation stage the AUROC was 0.73, with sensitivity and specificity values of 84% and 43% respectively. CONCLUSIONS: We have developed and successfully validated a simple model to use ADR risk score in a population of patients with a median age of 85, i.e. the oldest old. The model is based on 5 clinical variables (≥8 drugs, hyperlipidaemia, raised white cell count, use of anti-diabetic agents, length of stay ≥12 days), some of which have not been previously reported.

U2 - 10.1371/journal.pone.0111254

DO - 10.1371/journal.pone.0111254

M3 - Article

VL - 9

SP - 1

EP - 9

JO - PLoS ONE

JF - PLoS ONE

SN - 1932-6203

IS - 10

ER -