Machine learning classification of surgical pathology reports and chunk recognition for information extraction noise reduction

Background and aims: Machine learning techniques for the text mining of cancer-related clinical documents have not been sufficiently explored. Here some techniques are presented for the pre-processing of free-text breast cancer pathology reports, with the aim of facilitating the extraction of information relevant to cancer staging.

Materials and methods: The first technique was implemented using the freely available software RapidMiner to classify the reports according to their general layout: ‘semi-structured’ and ‘unstructured’. The second technique was developed using the open source language engineering framework GATE and aimed at the prediction of chunks of the report text containing information pertaining to the cancer morphology, the tumour size, its hormone receptor status and the number of positive nodes. The classifiers were trained and tested respectively on sets of 635 and 163 manually classified or annotated reports, from the Northern Ireland Cancer Registry.

Results: The best result of 99.4% accuracy – which included only one semi-structured report predicted as unstructured – was produced by the layout classifier with the k nearest algorithm, using the binary term occurrence word vector type with stopword filter and pruning. For chunk recognition, the best results were found using the PAUM algorithm with the same parameters for all cases, except for the prediction of chunks containing cancer morphology. For semi-structured reports the performance ranged from 0.97 to 0.94 and from 0.92 to 0.83 in precision and recall, while for unstructured reports performance ranged from 0.91 to 0.64 and from 0.68 to 0.41 in precision and recall. Poor results were found when the classifier was trained on semi-structured reports but tested on unstructured.

Conclusions: These results show that it is possible and beneficial to predict the layout of reports and that the accuracy of prediction of which segments of a report may contain certain information is sensitive to the report layout and the type of information sought.

Invisible Victims: Recorded Crime and Children in the UK

While violence against children is a common occurrence only a minority of incidents come to the attention of the authorities. Low reporting rates notwithstanding, official data such as child protection referrals and recorded crime statistics provide valuable information on the numbers of children experiencing harm which come to the attention of professionals in any given year. In the UK, there has been a strong tendency to focus on child protection statistics while children as victims of crime remain largely invisible in annual crime reports and associated compendia. This is despite the implementation of a raft of policies aimed at improving the system response to victims and witnesses of crime across the UK. This paper demonstrates the utility of a more detailed analysis of crime statistics in providing information on the patterns of crime against children and examining case outcomes. Based on data made available by the Police Service for Northern Ireland, it highlights how violent crime differentially impacts on older children and how detection rates vary depending on case characteristics. It makes an argument for developing recorded crime practice to make child victims of crime more visible and to facilitate assessment of the effectiveness of current initiatives and policy developments. Copyright © 2013 John Wiley & Sons, Ltd.

Palliative Medicine:Scoping the level and extent of service provision to people with intellectual disability at end of life Poster Presentation

Background: Concerns exist about the end of life care
that people with intellectual disabilities receive. This population
are seldom referred to palliative care services and
inadequate data sets exist about their place of death.
Aim: To scope the extent of service provision to people
with intellectual disabilities at the end of life by specialist
palliative care and intellectual disability services in one
region of the United Kingdom.
Methods: As part of a larger doctoral study a regional survey
took place of a total sample (n=66) of specialist palliative
care and intellectual disability services using a postal
questionnaire containing forty items. The questionnaire
was informed by the literature and consultation with an
expert reference group. Data were analysed using SPSS to
obtain descriptive statistics.
Results: A total response rate from services of 71.2%
(n=47) was generated. Findings showed a range of experience
among services in providing end of life care to people
with intellectual disabilities in the previous five years, but
general hospitals were reported the most common place of
death. A lack of accessible information on end of life care
for people with learning disabilities was apparent. A few
services (n=14) had a policy to support this population to
make decisions about their care or had used adapted Breaking
Bad News guidelines (n=5) to meet their additional
needs. Both services recognised the value of partnership
working in assessing and meeting the holistic needs of
people with intellectual disabilities at end of life.
Conclusions: A range of experience in caring for people
with intellectual disabilities was present across services,
but more emphasis is required on adapting communication
for this population to facilitate them to participate in their
care. These findings could have international significance
given that studies in other countries have highlighted a
need to widen access to palliative care for this group of
people.