

Oral presentations / European Geriatric Medicine 6S1 (2015) S5
–
S31
S15
Organisation of care and gerotechnology
O-035
Monitoring primary health care utilization in order to prevent
acute hospitalization of older patients
A. Fournaise
1
, N. Espensen
2
, S. Jakobsen
3
, K. Rønhof
4
,
K. Andersen-Ranberg
5
1
Department of Geriatric Medicine, Odense University Hospital,
Odense, Denmark;
2
Department of Geriatric Medicine, OUH Svendborg
Sygehus, Svendborg, Denmark;
3
Svendborg Sygehus, Svendborg,
Denmark;
4
Rønhof og Kemp, Svendborg, Denmark;
5
University of
Southern Denmark, Odense C, Denmark
Objectives:
High age is associated with increased risk of
hospitalisation. We have previously shown that home care
utilization increases significantly prior to acute short-term
admission. But is there a parallel increase in the use of primary
care physicians?
Methods:
Inclusion criteria: 70+ year old living in Svendborg
municipality admitted acutely to the acute assessment unit (AAU)
of Svendborg Hospital during 01.09.2012–31.08.2013 and discharged
within 48 hours (N = 443). For each individual a linkage was
made to the Danish National Health Insurance Service Registry
retrieving data on the use of primary health care physician.
Using the patients’ unique ten digit social security number, these
data were combined with previously gathered information on
sociodemographics, clinical symptoms, AAU discharge diagnoses,
and use (minutes) of home care. We present data on a subsample
of citizens receiving home care 12 months prior to admission and
who were alive six months after discharge.
Results:
N: 157 patients; mean age 84.0 y (
±
7.2); 73.9% females. We
found a significant increase in the monthly frequency of contacts
to the individual’s primary care physician, from 1.03 three months
prior to 1.57 the day of admission (p
<
0.002). Similar increases were
found for house visits (from 0.27 to 0.53 times/month; p
<
0.01),
and for communication (phone/email) (0.84 to 1.12 times/month;
p
<
0.01). The average number of total services provided including
e.g. blood sampling during consultations, increased from 3.39 to
6.29 services/month; (p
<
0.001).
Conclusion:
By sentineling increased primary health care
utilization timely medical assessment and intervention may
prevent acute hospital admissions of older adults.
O-037
Surveillance of adverse events in elderly patients: A study
on the accuracy of applying natural language processing
techniques to electronic health record data
C. Rochefort
1
, A. Verma
1
, T. Eguale
1
, D. Buckeridge
1
1
McGill University, Montreal, Canada
Objective:
Measuring adverse events (AEs) is necessary for quality
improvements but current detection methods are inaccurate. We
determined the accuracy of a potential alternative, the natural
language processing (NLP) of electronic health record data, for
detecting three highly prevalent AEs in elderly patients: a) deep
vein thrombosis (DVT), b) pulmonary embolism (PE) and, c)
pneumonia.
Methods:
A validation study was conducted at a university health
network in Montreal (Canada). We randomly sampled 6,000
narrative radiology reports performed between 2008 and 2013;
2,000 from imaging studies that could diagnose DVT/PE, and 4,000
from imaging studies that could diagnose pneumonia. We manually
identified DVT, PE and pneumonia within each report, which served
as our reference standard. Using a nested cross-validation approach,
we trained three support vector machine (SVM) models (one for
predicting each of the three AEs of interest), and the average
accuracy of each model was measured.
Results:
On manual review, 324 (16.2%) reports were DVT-positive,
154 (7.7%) were PE-positive and 640 (16.0%) were pneumonia-
positive. The SVM model predicting DVT achieved sensitivity of
0.80 (95% CI: 0.76–0.85), specificity of 0.98 (95% CI: 0.97–0.99) and
positive predictive value (PPV) of 0.89 (95% CI: 0.85–0.93). The SVM
model predicting PE achieved sensitivity of 0.79 (95% CI: 0.73–0.85),
specificity of 0.99 (95% CI: 0.98–0.99), and PPV of 0.84 (95% CI: 0.75–
0.92). The pneumonia model achieved sensitivity of 0.83 (95% CI:
0.78–0.88), specificity of 0.98 (95% CI: 0.97–0.99) and PPV of 0.88
(95% CI: 0.83–0.94).
Conclusion:
Statistical NLP models can accurately identify AEs
from narrative radiology reports. These models could assist with
monitoring and prevention efforts.
Biogerontology and genetics
O-038
Genotypic and phenotypic study of complement receptor type
1 polymorphisms in Alzheimer’s disease
R. Mahmoudi
1
, J.-L. Novella
2
, A. Kisserli
3
, L.A. Bertholon
1
, V. Duret
1
,
M. Dram´e
4
, B. Donvito
1
, S. Colas
1
, C. Talbot-Mahmoudi
1
, D. Jolly
4
,
P. Bach-Nga
1
, J.H. Cohen
5
1
Reims University Hospital, Reims, France;
2
Reims University
Hospital, Reims;
3
University of Reims Champagne-Ardenne, Reims;
4
Reims Teaching Hospitals, Reims, France;
5
University of Reims
Champagne-Ardenne, Reims, France
Objectives:
To analyze the phenotypic structural (length) and
expression aspects of complement receptor type 1 (CR1) in
erythrocytes of Alzheimer’s disease (AD) patients.
Methods:
A case-control study (100 AD, 87 controls). A comparison
between the phenotypic CR1 length polymorphism (LP) and the CR1
LP deduced from CR1 gene analysis was performed using Western
blot at protein level and high-resolution melting at gene level. CR1
sites on erythrocytes (density) were enumerated by flow cytometry.
Finally, a molecular analysis of the CR1 gene to type the rs6656401
and rs3818361 polymorphisms was performed.
Results:
Phenotypic CR1 length and CR1 length deduced from
the molecular analysis were concordant in 98.3% of cases. Our
data suggested the existence of silent CR1 alleles. The CR1
density was significantly lower in AD patients expressing the
CR1*2 isoform compared with controls (p = 0.001), demonstrating
lower expression of CR1 in CR1*2 carriers. Finally, rs6656401 and
rs3818361 were strongly associated with CR1 length polymorphism
(p
<
0.0001).
Conclusion:
These findings indicate that AD susceptibility is
associated with the long CR1 isoform (CR1*2), albeit at a lower
density, suggesting that AD results from insufficient clearance of
plaque deposits rather than increased inflammation. Variations in
CR1 function via the CR1 LP/density polymorphism relationship
deserve consideration in relation to the AD susceptibility
mechanisms associated with the CR1 molecule. Qualitative and
quantitative variations of CR1, depending of CR1 allele expression
or the presence of silent CR1 alleles, and different behaviors of
the CR1 molecule regarding its various functions might lead to an
intricately complex picture of CR1 involvement in AD.