By: Robert B. Kuller

Virtually everyone would be in agreement with the statement that operational analysis starts with answering the questions of “who, what, where, when, how and why.” It’s all about patterns of individual behavior and the related comparison of that person’s behavior to his/her peers. These comparisons can be made on a departmental, organizational or industry-wide basis. At its very core, all behavior is contextual. A perfectly explainable behavior in one situation can be completely inexplicable in another. Combine this with the concept of machine learning and you have the makings of a predictive analytics model, whereby patterns of behavior are recorded, interpreted, analyzed and forecast. This technology can be applied to privacy and insider threat risk mitigation at hospitals, and its use can also have a profound impact on hospital operational analysis, providing actionable recommendations for improvement.

In the hospital setting, a large number of employees are authorized to use the electronic medical record (EMR) system. Attempts to limit the numbers of authorized users have generally been unsuccessful since hospital administrators do not want to inhibit patient care activities and the EMR, like it or not, is the hub of patient care coordination. So, at its heart, patient privacy protection is both behavioral and contextual. Some of the typical questions that hospitals need to answer in this area are:



• For what reason did the authorized user access the EMR? Are authorized users accessing patient records for whom they are care team members, or are they just snooping for information? Is this an isolated incident or a problematic pattern of behavior?

• Is the user’s behavior, in terms of tasks performed, typical of the role he/she plays on the care team? For instance, a nurse who does not typically complete discharge summaries performs eight of them on his/her shift. Completing a discharge summary by itself is not an anomalous event, yet this behavior, in the context of the fact that this particular individual rarely, if ever, performs this specific task, may necessitate further investigation.

• Are users’ location and timeline consistent with their previous behavior patterns? For instance, why would a pharmacist be accessing the patient’s record from the maintenance department at 3 a.m. when the normal shift is 7 a.m. to 4 p.m.? Is there a legitimate explanation? Has someone stolen their login credentials, or is something even worse going on? How are any of these questions going to be answered without extensive use of behavioral analytics in a contextual setting? Luckily, through the use of multiple search engines, the EMR’s metadata can be analyzed concerning the who, what, where, when and how of these events. The privacy office is still charged with determining the why.

Taking this model one step further, the same behavioral tools being used for privacy analysis can be easily deployed for operational analysis. A hospital executive can get a quick picture as to what all their employees are actually doing within their institution. For example, it has been recently reported by Health Data Management that physician time is rapidly being consumed by administrative tasks. Physicians routinely state that they spend an enormous amount of “overtime” related to EMR activities, and in the process, job dissatisfaction and professional burnout are on the rise. How valuable would it be to quickly determine exactly who are the anomalous physician users who are spending 3 hours rather than 30 minutes in the EMR at the end of their workday? Data like this can help target training resources toward the specific physicians who need it, thereby increasing physician satisfaction and retention.

Another scenario that comes to mind is getting a deeper understanding of what specific provider behavior patterns are causing bottlenecks and decreases in patient satisfaction. Through the tracking of ordering behavior (what, where and when), physicians who are not issuing timely pharmacy, radiology or other orders while with their patient can cause loss of income for the institution, an increase in patient calls, and ultimately, patient dissatisfaction.

Finally, identifying the totality of patient care activities classified by diagnosis can inform the institution where it is making money and where it is losing money. Although hospitals must treat all patients presenting themselves for treatment, it would certainly make sense not to launch a magnet advertising campaign for a diagnosis/department that consistently loses money for the hospital.

Many midsized and smaller institutions do not have access to the large, expensive data analysis systems in the marketplace that require the support of a cadre of data analysts to extract data, and which place a heavy burden on their IT departments. Behaviorally-based systems allow these institutions to answer key operational questions without high expense, staffing requirements or extensive IT involvement. This is where the future of operational efficiency improvements lies for mid- and small-sized hospitals which have the same actionable information requirements as larger institutions, but not the budget to get them there.

About the author: Robert B. Kuller has spent the majority of his 35-year career in the health care industry working for companies such as Solvay, Siemens, Keynomics, Liquent, Real Media, Kamine Technology Group and Haystack Informatics, in senior and top executive capacities. In addition, he has successfully founded, managed and sold two businesses.