European Geriatric Medicine - Supplement 1 - Volume 7 - October 2016

184 / 290

Methods: Retrospective descriptive study with a sample of 527 urine
cultures of patients admitted to the Geriatric Hospital of Getafe from
2009 to 2015. Using database WASPSS Project.
Results: Between January 2009 and December 2012, 299 urine
cultures were isolated with >100,000 CFU of E. coli. Antibiotics with
less resistance rate intravenously administrated were Cabapenem
with 0% and Piperacillin/tazobactam had 10%. Within orally
administration available, Fosfomycin has a resistance rate of 8%, 30%
Cefuroxime and Ciprofloxacin 67%. Between January 2013 and
December 2015, 228 urine cultures were isolated with >100,000 CFU
of E. coli. Antibiotic with less resistance rate intravenous admini-
strated were Imipenem and Meropenem. There is a case of Ertapenem
resistanece out of 217, and piperacillin-tazobactam presents 7%.
Within orally administration available, Fosfomycin continues with a
resistance rate of 8%, 23% Cefuroxime and Ciprofloxacin 53%.
Key conclusions: First clinical guidelines recommended antibiotic
resistance is very high based on our study. – Carbapenems remain
effective for the treatment of E. Coli although in the last three years
resistance begin to appear.
P-565
Are statistical natural language processing models for pneumonia
surveillance in elderly patients generalizable across acute care
hospitals?
C. Rochefort 1,2, A. Verma2, T. Eguale 2, D. Buckeridge 2, A. Forster3.
1University of Sherbrooke, 2McGill University, 3The Ottawa Hospital
Research Institute
Background: Natural language processing (NLP) models are increas-
ingly used for adverse event (AE) surveillance in acute care
hospitals but limited information is available on their generaliz-
ability across institutions, which is important for valid bench-
marking. We examined the generalizability of a statistical NLP model
predicting pneumonia from electronic health record (EHR) data; an AE
associated with significant morbidity, mortality and costs in elderly
patients.
Methods: We randomly sampled 4,000 narrative chest X-ray reports
performed at a university health network (UHN) in Quebec (Canada)
between 2010 and 2014. We manually identified pneumonia within
each report, which served as our reference standard. We used a nested
cross-validation approach to train and validate a support vector
machine (SVM) model predicting pneumonia. This model was then
applied to a random sample of 2,281 narrative radiology reports from
another UHN in Ontario (Canada), and accuracy was measured and
subsequently compared to that of an alternative model recalibrated on
Ontario data.
Results: The SVM model predicting pneumonia on Quebec data
achieved 83% sensitivity (95%CI: 78– 88%), 98% specificity (95%CI:
97 – 99%) and 88% PPV (95%CI: 83–94%). When applied to Ontario data,
this model achieved 57% sensitivity (95%CI: 51–63%), 99% specificity
(95%CI: 98 –99%) and 86% PPV (95%CI: 80 –90%). In comparison, the
recalibrated model achieved 76% sensitivity (95%CI: 70 –82%), 98%
specificity (95%CI: 97 –99%) and 86% PPV (95%CI: 82–91%).
Conclusion: A statistical NLP model predicting pneumonia has limited
generalizability when directly applied to EHR data from another
institution. However, good prediction performances can be achieved
after model recalibration on local data.
P-566
Prognosis of the elderly admitted in an Intermediate Medical Care
Unit with community acquired respiratory infection
C. Rodrigues 1, A. Pissarra1, M. Miguéns1, M. Malheiro1, D. Cardoso1,
D. Alberca 1, A. Massena1, V. Batalha2, A. Martins1, L. Campos 2. 1Medical
Oncology Unit, Centro Hospitalar Lisboa Ocidental, 2Internal Medicine
Unit 4, Centro Hospitalar Lisboa Ocidental, Lisbon, Portugal
Introduction: Respiratory infections are an important cause of
morbidity and mortality in the general population and particularly,
in the elderly. These patients have an incremented risk of infection, but
if diagnosed and treated promptly are not, necessarily associated with
higher complication or mortality rates. The admission of patients aged
65 and older in intensive and intermediate care units is a controversial
subject, and therefore, as the life expectancy increases, this poses as a
health care challenge.
Objectives: We set out to compare the expected mortality rate using
the Pneumonia Severity Index (PSI score) with the real mortality rate
in the elderly admitted in an intermediate care unit with the diagnosis
of community acquired respiratory infection.
Methods: Retrospective observational study in an Intermediate
Medical Unit of patients aged 65 and older admitted in an
Intermediate Medical Care Unit with the diagnosis of community
acquired respiratory infection during a period of 3 years. We evaluated
socio demographic parameters, PSI score and clinical parameters
(etiologic agent, antibiotherapy, duration of hospitalization and
outcome).
Results: We identified 52 patients admitted with the diagnosis of
respiratory infection of which 53,8% (n = 28) were aged 65 and older
and 11 were very old (aged 85 and older). The median agewas 83 years
old (minimum 67; maximum 100) with 60,7% (n = 17) men. The mean
PSI score was 147, conferring a expectable mortality of 27–31% (grade
5). The minimum PSI score was 91 and the maximum was 232. An
etiologic agent was identified in 28,6% of the cases (6: streptococcus
pneumoniae; 1: Klebsiella Pneumoniae; 1: Pseudomonas Aeruginosa)
and two cases were complicated with bacteremia to pneumococcus.
The most commonly used antibiotic was amoxicillin/clavulanate acid
in association with clarithromycin. The mean days of hospitalization
was of 13 days. The mortality rate was 14,3% (n = 4). Of these patients
the mean PSI score was 150.
Conclusion: The difference observed between the expectable mortal-
ity using the PSI score and the real mortality is in accordance with the
data stating that, the elderly do not have a worse prognosis than
younger patients when treated accordingly. We conclude that the
specialized care given to these patients in an Intermediate Care Unit
conferred a better prognosis and should, therefore, motivate us to
develop strategies to improve the quality of the criteria to admit
elderly patients in specialized units.
P-567
Incidence of Herpes zoster and its complications in adults ≥ 50
years old: a prospective study in Germany (2010–2014)
R. Schmidt-Ott1, U. Schutter2, J. Simon3, A. von Krempelhuber4,
K. Gopala5, A. Guignard1, E. Espié1. 1GSK Vaccines, Wavre, Belgium;
2Marienhospital Marl, Marl, 3Gesundheitsnetz Osthessen, Fulda, 4GSK,
München, Germany; 5GSK Pharmaceuticals Ltd., Bangalore, India
Objectives: Herpes zoster (HZ) is caused by a reactivation of the
varicella zoster virus and mainly affects individuals aged ≥50 years
and those who are immunocompromised. HZ is usually painful and
may lead to persistent pain (PostHerpetic Neuralgia [PHN]), neuro-
logical sequelae or ophthalmic disease. The objective of this study was
to characterize the HZ incidence and risk of complications in
individuals aged ≥50 years, in Germany.
Methods: A prospective cohort study (e-track: 113206) was conducted
from 2010 to 2014 using three different physicians’ networks covering
a defined geographical region and population. The patients clinically
diagnosed with acute HZ were followed from the initial visit up to 270
days to assess the pain severity using the Zoster Brief Pain Inventory
(ZBPI) questionnaire. We estimated overall, age- and gender-specific
HZ incidence rates per 1000 person-years (PY). The prevalence of PHN
was estimated using the definition of persistence of HZ-related pain
with a ZBPI score ≥3.
Results: The overall HZ incidence was 6.7 per 1000 PY [95% confident
interval (CI), 6.4–7.1], increasing from 4.4 in 50–59 years old to 9.4 per
1000 PY in ≥80 years. Among the 513 enrolled HZ patients, 11.9% [95%
CI, 9.2–15.0] experienced PHN 90 days or more after rash onset, with
the highest prevalence in ≥80 years old (14.3%). Respectively, 4.9% and
2.9% reported PHN 180 and 270 days or more after rash onset.
Poster presentations / European Geriatric Medicine 7S1 (2016) S29–S259 S178