A Recipe for Clinical Decision Support using Streaming Analytics
There seems to be a groundswell in the number of applications being developed for new clinical decision support analytics tools. It was only a matter of time. As an example, computerized 12 lead ECG interpretation was invented roughly forty years ago, and offered nuanced observations on rhythm and morphology with sensitivity and specificity that have made it a useful diagnostic tool to this day. In some cases, research studies that began years ago, produced data that have now matured to become commercial validated clinical decision support algorithms.
With the quantity of measured physiological parameters in an ICU sometimes numbering in the hundreds, it makes sense that a tool be developed to parse, manage, and mine this data as a mechanism to examine and employ algorithms in a clinical production environment. Some of the early scores such as the Modified Early Warning Score (MEWS), only looked at five parameters: systolic blood pressure, heart rate, respiratory rate, temperature, and AVPU Score. By weighting parameters, the algorithm delivers a score that can assist in determination of patient acuity. Other algorithms include Rothman Index, another patient degradation index score. These scores examine minute changes in multiple parameters and forecasts the status of the patient.
Some second generation algorithms such as CleMetric and HeRO use Heart Rate Variability (HRV) to denote patient degradation or susceptibility to sepsis using the predictive work done by Ahmad, Ramsey, et al. Excel Medical’s BedMasterEx platform can interface a GE or Philips patient monitoring network and can stream the waveforms out in a format consumable by other applications as well as embed live waveforms and algorithm scores contextually within the Electronic Medical Record System (EMR) itself.
No matter the EMR vendor, there is a database of unstructured data (nursing notes, etc.), as well as structured data from clinical and lab interfaces, a PDF repository, as well as an ADT system to provide the algorithm with positive patient identification. This provides a rich data set for subsequent drop-in clinical decision support tools. One example is Therapeutic Monitoring Solutions’ ventilator weaning protocol algorithm. These algorithms can help identify patients who are ready to go home, and those who should perhaps not, based upon respiratory rate variability captured via patient monitoring capnography.
Third Generation algorithms such as OBS Medical’s Visensia Index use more sophisticated mathematics and parameters to provide data to support either earlier discharge or closer examination, in an attempt to preclude sentinel events before they occur, an example of leveraging technology that is already in place. Just add the BedMasterEX middleware system, and start adding applications or develop your own using IBM InfoSphere Streams or Unscrambl to allow easy set-up using an intuitive user interface. Unscrambl is also programmable in Python allowing novice researchers a quick way to investigate data. By combining EMR and patient monitoring data using waveform-capable physiological capture software, the state-of-the-art in prognosticative clinical decision support IT infrastructure can be achieved.
Excel Medical will be hosting their annual symposium on clinical and research informatics May 8th through May 10th in Atlanta Georgia at the Atlanta AMA Medical Conference Center.
Want more Information?