Clinical search engine for instant access to data at The San
The Sydney Adventist Hospital (The San) plans to implement a clinical text search platform developed by Health Language Analytics (HLA) to speed up access to a range of health data that can be used for clinical as well as research purposes.
Based on HLA's text analysis and natural language processing (NLP) technologies, the new CliniSearch functionality will allow clinical as well as administrative staff and researchers to quickly find, retrieve, code and use the vast amount of health data that is stored as text and normally has to be manually retrieved and analysed.
HLA CEO Jon Patrick said the technology was essentially a clinical search engine for health data that is captured as text in unstructured documents such as pathology reports, radiology reports, discharge summaries, nuclear medicine reports, theatre reports and clinical progress notes.
Professor Patrick said CliniSearch uses a concept-based search mechanism that can understand the idiosyncrasies of clinicians’ writing styles and language usage to accurately collect the data required.
It can be used to find records with certain characteristics – such as all patients with prostate cancer – identify patients for case studies, find specific cases without having to remember the patient and to search for radiology, pathology and other text notes in preparation for oncology multi-disciplinary team meetings, for example.
Professor Patrick said the technology could handle many different communities of users of clinical documents.
“You've got a community of people who need to identify cohorts,” he said. “For ED discharge summaries, first and foremost you are going to want to know what the diagnosis is and what the plan is.
“If you go into an oncology setting, there will be administrative staff who need to find all of the documents for a patient for multidisciplinary team meetings. Another community again is cancer registries, or any sort of registry for that matter.”
San CIO Chris Williams said the idea was to provide clinicians with a larger volume of more accurate information faster so they can deliver improved patient care.
“We had our department heads, clinicians and technology staff asses the HLA platform and perform various tasks to ensure that it was robust, flexible and had the capacity to expand its library automatically as new terminology and expressions were found in documents,” Mr Williams said.
“Tests on 57 cancer queries showed an accuracy of 96 per cent, and we know that can be improved with tuning.
“In one of the tasks we set, we provided 8000 histopathology records to HLA and requested that all prostate diagnoses be found and retrieved. The automatic process took just a few minutes with greater than 95 per cent accuracy which usually would have taken many staff weeks to compile.”
The San plans to implement CliniSearch for clinical documentation hospital-wide, Mr Williams said.
“We expect that the HLA text search analytics solution will be made available to all San staff as well as visiting clinicians and researchers. This will provide them with instant access to valuable data not automatically available in any other hospital and it will continue to ensure the San is a leading health technology entity.”
While CliniSearch resembles a normal search engine, underneath the hood is a data warehouse containing clinical documentation that has already undergone natural language processing in advance.
Professor Patrick said clinical text in the document is converted to SNOMED codes and indexed. When a search for liver cancer is undertaken, for example, the system will turn that into a SNOMED concept and search for records that refer to liver cancer.
CliniSearch can also be used to recruit a common cohort or phenotypes for clinical trials, to target particular disease classes for research, to alert staff when a pertinent record is created in the storage system, to check the accuracy of clinical coding across disease classes, to semi-automate codification for billing and to identify and understand the case mix of a particular organisation.
“It is powerful stuff, but it has taken us years to get it up to the sort of accuracy we are getting now and to convince people that it is worthwhile having,” Professor Patrick said.
Posted in Australian eHealth