Nosocomial infections: Validation of surveillance and computer modeling to identify pat at risk

Ann Broderick, Motomi (Tomi) Mori, Mary D. Nettleman, Strephen A. Streed, Richard P. Wenzel

Research output: Contribution to journalArticle

88 Citations (Scopus)

Abstract

To estimate the accuracy of routine hospital-wide surveillance for nosocomial infection, the authors performed a validation study at the University of Iowa Hospitals and Clinics, a 900-bed tertiary care Institution, by daily concurrent surveys of all patients' charts. The study extended over a 10-month period from January to October 1987. The sensitivity and specificity of the reported data were 80.7% (95% confidence interval (Cl) 72.2-89.2) and 97.5% (95% Cl 96.4-98.5), respectively. The predictive values of positive or negative reports of an infection were 75.3% (95% Cl 66.3-84.2) and 98.1% (95% Cl 97.3-99.1), respectively. In a separate analysis, the data entry system was reviewed for eight descriptive variables among all patients with infections (n=443) identified over a 2-month period. The data entry was found to be 94-99% accurate. To improve the efficiency of current surveillance, the authors used data gathered during the study to develop a computer model for the Identification of patients with a high probability of having a nosocomial Infection. The use of stepwise logistic regres sion identified five variables which independently predicted infection: age of the patient (years), days of antibiotics, days of hospitalization, and the number of days on which urine and/or wound cultures were obtained. Optimal sensitivity and specificity (8 1.6% and 72.5%, respectively) were found when the model examined patients with an 8% or higher a priori probability of infection; this figure corresponded to a review of 33% of the patients' charts. Increasing the a priori probability would progressively increase specificity and reduce both sensitivity and the number of charts needed for review. If it is prospectively validated, the model may provide a more efficient mechanism by which to conduct hospital- wide surveillance.

Original languageEnglish (US)
Pages (from-to)734-742
Number of pages9
JournalAmerican Journal of Epidemiology
Volume131
Issue number4
StatePublished - Apr 1990
Externally publishedYes

Fingerprint

Cross Infection
Infection
Sensitivity and Specificity
Validation Studies
Tertiary Healthcare
Information Systems
Computer Simulation
Hospitalization
Urine
Confidence Intervals
Anti-Bacterial Agents
Wounds and Injuries

Keywords

  • Computer simulation
  • Cross infection
  • Patient identification systems
  • Risk
  • Sensitivity and specificity (epidemiology)

ASJC Scopus subject areas

  • Geriatrics and Gerontology
  • Epidemiology

Cite this

Nosocomial infections : Validation of surveillance and computer modeling to identify pat at risk. / Broderick, Ann; Mori, Motomi (Tomi); Nettleman, Mary D.; Streed, Strephen A.; Wenzel, Richard P.

In: American Journal of Epidemiology, Vol. 131, No. 4, 04.1990, p. 734-742.

Research output: Contribution to journalArticle

Broderick, A, Mori, MT, Nettleman, MD, Streed, SA & Wenzel, RP 1990, 'Nosocomial infections: Validation of surveillance and computer modeling to identify pat at risk', American Journal of Epidemiology, vol. 131, no. 4, pp. 734-742.
Broderick, Ann ; Mori, Motomi (Tomi) ; Nettleman, Mary D. ; Streed, Strephen A. ; Wenzel, Richard P. / Nosocomial infections : Validation of surveillance and computer modeling to identify pat at risk. In: American Journal of Epidemiology. 1990 ; Vol. 131, No. 4. pp. 734-742.
@article{35b199c735164a8083ba4ea370df8b18,
title = "Nosocomial infections: Validation of surveillance and computer modeling to identify pat at risk",
abstract = "To estimate the accuracy of routine hospital-wide surveillance for nosocomial infection, the authors performed a validation study at the University of Iowa Hospitals and Clinics, a 900-bed tertiary care Institution, by daily concurrent surveys of all patients' charts. The study extended over a 10-month period from January to October 1987. The sensitivity and specificity of the reported data were 80.7{\%} (95{\%} confidence interval (Cl) 72.2-89.2) and 97.5{\%} (95{\%} Cl 96.4-98.5), respectively. The predictive values of positive or negative reports of an infection were 75.3{\%} (95{\%} Cl 66.3-84.2) and 98.1{\%} (95{\%} Cl 97.3-99.1), respectively. In a separate analysis, the data entry system was reviewed for eight descriptive variables among all patients with infections (n=443) identified over a 2-month period. The data entry was found to be 94-99{\%} accurate. To improve the efficiency of current surveillance, the authors used data gathered during the study to develop a computer model for the Identification of patients with a high probability of having a nosocomial Infection. The use of stepwise logistic regres sion identified five variables which independently predicted infection: age of the patient (years), days of antibiotics, days of hospitalization, and the number of days on which urine and/or wound cultures were obtained. Optimal sensitivity and specificity (8 1.6{\%} and 72.5{\%}, respectively) were found when the model examined patients with an 8{\%} or higher a priori probability of infection; this figure corresponded to a review of 33{\%} of the patients' charts. Increasing the a priori probability would progressively increase specificity and reduce both sensitivity and the number of charts needed for review. If it is prospectively validated, the model may provide a more efficient mechanism by which to conduct hospital- wide surveillance.",
keywords = "Computer simulation, Cross infection, Patient identification systems, Risk, Sensitivity and specificity (epidemiology)",
author = "Ann Broderick and Mori, {Motomi (Tomi)} and Nettleman, {Mary D.} and Streed, {Strephen A.} and Wenzel, {Richard P.}",
year = "1990",
month = "4",
language = "English (US)",
volume = "131",
pages = "734--742",
journal = "American Journal of Epidemiology",
issn = "0002-9262",
publisher = "Oxford University Press",
number = "4",

}

TY - JOUR

T1 - Nosocomial infections

T2 - Validation of surveillance and computer modeling to identify pat at risk

AU - Broderick, Ann

AU - Mori, Motomi (Tomi)

AU - Nettleman, Mary D.

AU - Streed, Strephen A.

AU - Wenzel, Richard P.

PY - 1990/4

Y1 - 1990/4

N2 - To estimate the accuracy of routine hospital-wide surveillance for nosocomial infection, the authors performed a validation study at the University of Iowa Hospitals and Clinics, a 900-bed tertiary care Institution, by daily concurrent surveys of all patients' charts. The study extended over a 10-month period from January to October 1987. The sensitivity and specificity of the reported data were 80.7% (95% confidence interval (Cl) 72.2-89.2) and 97.5% (95% Cl 96.4-98.5), respectively. The predictive values of positive or negative reports of an infection were 75.3% (95% Cl 66.3-84.2) and 98.1% (95% Cl 97.3-99.1), respectively. In a separate analysis, the data entry system was reviewed for eight descriptive variables among all patients with infections (n=443) identified over a 2-month period. The data entry was found to be 94-99% accurate. To improve the efficiency of current surveillance, the authors used data gathered during the study to develop a computer model for the Identification of patients with a high probability of having a nosocomial Infection. The use of stepwise logistic regres sion identified five variables which independently predicted infection: age of the patient (years), days of antibiotics, days of hospitalization, and the number of days on which urine and/or wound cultures were obtained. Optimal sensitivity and specificity (8 1.6% and 72.5%, respectively) were found when the model examined patients with an 8% or higher a priori probability of infection; this figure corresponded to a review of 33% of the patients' charts. Increasing the a priori probability would progressively increase specificity and reduce both sensitivity and the number of charts needed for review. If it is prospectively validated, the model may provide a more efficient mechanism by which to conduct hospital- wide surveillance.

AB - To estimate the accuracy of routine hospital-wide surveillance for nosocomial infection, the authors performed a validation study at the University of Iowa Hospitals and Clinics, a 900-bed tertiary care Institution, by daily concurrent surveys of all patients' charts. The study extended over a 10-month period from January to October 1987. The sensitivity and specificity of the reported data were 80.7% (95% confidence interval (Cl) 72.2-89.2) and 97.5% (95% Cl 96.4-98.5), respectively. The predictive values of positive or negative reports of an infection were 75.3% (95% Cl 66.3-84.2) and 98.1% (95% Cl 97.3-99.1), respectively. In a separate analysis, the data entry system was reviewed for eight descriptive variables among all patients with infections (n=443) identified over a 2-month period. The data entry was found to be 94-99% accurate. To improve the efficiency of current surveillance, the authors used data gathered during the study to develop a computer model for the Identification of patients with a high probability of having a nosocomial Infection. The use of stepwise logistic regres sion identified five variables which independently predicted infection: age of the patient (years), days of antibiotics, days of hospitalization, and the number of days on which urine and/or wound cultures were obtained. Optimal sensitivity and specificity (8 1.6% and 72.5%, respectively) were found when the model examined patients with an 8% or higher a priori probability of infection; this figure corresponded to a review of 33% of the patients' charts. Increasing the a priori probability would progressively increase specificity and reduce both sensitivity and the number of charts needed for review. If it is prospectively validated, the model may provide a more efficient mechanism by which to conduct hospital- wide surveillance.

KW - Computer simulation

KW - Cross infection

KW - Patient identification systems

KW - Risk

KW - Sensitivity and specificity (epidemiology)

UR - http://www.scopus.com/inward/record.url?scp=0025234951&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0025234951&partnerID=8YFLogxK

M3 - Article

C2 - 2180283

AN - SCOPUS:0025234951

VL - 131

SP - 734

EP - 742

JO - American Journal of Epidemiology

JF - American Journal of Epidemiology

SN - 0002-9262

IS - 4

ER -