Last week I was just thinking back to some interesting scenarios I have worked on over the years where content analytics was the key to both solving a critical business problem, but more importantly in generating real business transformation which is critical to success of any analytics project. These scenarios are not new or surprising, but I thought it might be helpful to intersperse my usually conceptually dense blogs with some light reading :-) So today lets look at a healthcare scenario for content analytics.
For most hospitals, the knowledge that resides within their unstructured text (referral or discharge letters, clinical notes, etc.) is largely ignored by their operational processes. Today when someone enters the Emergency Room it can take hours for their entire patient record to be retrieved from archives. Even for hospitals which have full digital records, there are frequently digital gaps (such as medical correspondence). In addition, even if all the information is complete, it can constitute hours of light reading and not something which is rapidly consumable for an emergency situation. This is clearly unacceptable. It was the opinion of the medical consultants that if they had quick & easy access to the key information encoded in the most recent outpatient and discharge letters (consolidated & prioritized), then that would significantly mitigate risk of mis- diagnosis & treatment in an emergency setting.
This approach to solving what could have been perceived as a simple search problem, proved to demonstrate a number of additional benefits above and beyond digitizing, searching, and retrieving clinical correspondence.
- Firstly, through integrating the clinical documents directly into a PIMS system and converting them into a richly described XML format, you can convert these letters into HL7 records which means that instead of physically posting letters to General Practitioners you are now able to electronically transfer everything and have proper workflow and audit trails around the information.
- Secondly, this new richly annotated content repository, which is linked to the structured PIMS data, provides a hugely valuable set of clinical data that could be used by consultants for research, reporting, tracking, evaluation, etc.
- Thirdly, through integrating content analytics into the content creation process itself, writing of patient letters by the medical staff, you can introduce quality control and auditing at a very low cost to the clinicians creating the content. In fact, the additional time that may be involved in quality control is mitigated by other efficiencies introduced by the system.
- Finally, today most hospitals have Electronic Medical Record (EMR) initiatives which are frequently unpopular, inefficient, and, in many cases, unsuccessful. One issue is that most these programs solve the “unstructured / structured” problem by simply throwing a myriad of huge, complex, and restrictive forms-based processes at clinicians. These forms not only cause huge amounts of frustration but their rigidity frequently means key information may be lost or misrepresented. Text fields are frequently used to workaround this data entry problem, however since the text fields are “black boxes” from an application integration perspective, we are back to square one – information silos. Real-time content analytics allows clinicians to describe patient symptoms, diagnosis, and prognosis in a natural way, while at the same time allowing the key concepts and facts to be captured normalized across different patient groups and medical vocabularies.