Contributions
Abstract: EP1361
Type: ePoster
Abstract Category: Clinical aspects of MS - Epidemiology
Objectives: The UK MS Register captures real world data about living with Multiple Sclerosis (MS), from 3 sources: People with MS (PwMS) via the internet, the NHS via database capture and by linkage to other routine datasets.
NHS data, though "gold standard", is dependent on clinical staff finding the time and relevant information to enter into clinical systems. Implementations across the NHS are variable, therefore, we looked to other methodologies for ease and accuracy of capture.
Approach: We selected the Clix enrich Natural Language Processing (NLP) software to assess capture of the MS Register minimum clinical dataset, NLP software matches clinical phrases against SNOMED-CT terminology.
40 letters, from 2 NHS Trusts, related to 28 patients were processed by the software against a typical casemix of MS patients, letters were dictated by Neurologists, Specialist General practitioners and Specialist Nurses. The letters were evenly split between docx and PDF formats. Output was parsed by a domain expert for clinical content and scored by data item for sensitivity and specificity. The output was then scored by another researcher for 12 relevant clinical concepts. Finally, a ruleset created to find clinical concepts, run against the data and scored.
Results: 1 letter of 40 failed to load, the rest were analysed for specific data. Date related items were clearly challenging for the NLP software, with only 7% of appointment dates being matched and 22% for diagnosis date.
However, MS Type (93.3%) and EDSS scores 93.75%, recognising the test had been done and the actual value, were both well recognised. Data items that have traditionally been poorly reported such as symptoms were also well recognised, with fatigue being highlighted (78.5%) and gait and walking issues (68.7%).
One item that will need further work was the small number of false positive results in DMT"s with 15% being identified as being on a DMT when this was just being considered or planned.
Conclusion: The NLP pathway could be extremely useful for obtaining and encoding hard to capture clinical data. This has value to researchers and to clinicians looking for insight into their case load. Further work is needed to optimise some concepts. Even with this minimal configuration it´s possible to ascertain MS Type, functional scores, current medication and disabling symptomology, crucially burden on clinical staff is reduced as no additional data tasks are required.
Disclosure: The authors know of no potential conflicts of interest
Abstract: EP1361
Type: ePoster
Abstract Category: Clinical aspects of MS - Epidemiology
Objectives: The UK MS Register captures real world data about living with Multiple Sclerosis (MS), from 3 sources: People with MS (PwMS) via the internet, the NHS via database capture and by linkage to other routine datasets.
NHS data, though "gold standard", is dependent on clinical staff finding the time and relevant information to enter into clinical systems. Implementations across the NHS are variable, therefore, we looked to other methodologies for ease and accuracy of capture.
Approach: We selected the Clix enrich Natural Language Processing (NLP) software to assess capture of the MS Register minimum clinical dataset, NLP software matches clinical phrases against SNOMED-CT terminology.
40 letters, from 2 NHS Trusts, related to 28 patients were processed by the software against a typical casemix of MS patients, letters were dictated by Neurologists, Specialist General practitioners and Specialist Nurses. The letters were evenly split between docx and PDF formats. Output was parsed by a domain expert for clinical content and scored by data item for sensitivity and specificity. The output was then scored by another researcher for 12 relevant clinical concepts. Finally, a ruleset created to find clinical concepts, run against the data and scored.
Results: 1 letter of 40 failed to load, the rest were analysed for specific data. Date related items were clearly challenging for the NLP software, with only 7% of appointment dates being matched and 22% for diagnosis date.
However, MS Type (93.3%) and EDSS scores 93.75%, recognising the test had been done and the actual value, were both well recognised. Data items that have traditionally been poorly reported such as symptoms were also well recognised, with fatigue being highlighted (78.5%) and gait and walking issues (68.7%).
One item that will need further work was the small number of false positive results in DMT"s with 15% being identified as being on a DMT when this was just being considered or planned.
Conclusion: The NLP pathway could be extremely useful for obtaining and encoding hard to capture clinical data. This has value to researchers and to clinicians looking for insight into their case load. Further work is needed to optimise some concepts. Even with this minimal configuration it´s possible to ascertain MS Type, functional scores, current medication and disabling symptomology, crucially burden on clinical staff is reduced as no additional data tasks are required.
Disclosure: The authors know of no potential conflicts of interest