Larry Fennigkoh 500

Larry Fennigkoh, PhD

Dennis Minsent’s insightful article from last fall, “Big Data, Big Benefits: Optimizing Inventory Management (November 2015) makes an excellent case and timely call for HTM departments to start thinking about how to produce meaningful information from their masses of big data. He also emphasizes the importance of keeping your CMMS database current and accurate.

While doing so is no trivial task, especially given the highly dynamic nature of equipment inventories, the benefits can be substantial, increasing the credibility of your queries and associated reports. A reasonably current and accurate database that spans multiple years and includes not only equipment inventory data, but ‘cradle-to-grave’ repair and service history results (ie, service dates, labor hours, materials costs, failure modes, etc), may also possess a veritable gold mine of information. Extracting this information, however, requires a focused and scientific study with special analytical tools—which we have.

The HTM community has a treasure trove of decades of potentially useful data, but data alone does not become information (or better yet, knowledge) without analysis. Such analysis typically involves the appropriate and correct use of classical inferential statistical tools and methods, such as analysis of variance (ANOVA) and multiple and logistic regression. In contrast to simple descriptive statistics (ie, those that are limited to providing measures of central tendency or dispersion within data), inferential statistics allow for the testing of hypotheses and identify relationships. Most importantly, they answer questions, leading us to new knowledge (or frequently, more questions).

Here is just a sampling of the types of questions that could be addressed, for either an individual device category (such as infusion pumps or ventilators) or a particular department (eg, radiology):

  • What are the effects of scheduled PM inspections on unscheduled repair costs, or alternatively, on unscheduled repair labor hours or work order volume?
  • Is the apparent upward trend in added equipment inventory volume statistically significant?
  • Is there a statistically significant difference in balloon pump maintenance costs between manufacturers (or between different facilities in the same healthcare system)?
  • What effect does equipment age have on unscheduled repair costs, unscheduled repair labor hours, or work order volume?
  • What effect does device run-time hours (eg, for ventilators) have on unscheduled repair costs, unscheduled repair labor hours, or work order volume?
  • What factors affect a department’s same-day service response or its mean-time-to-repair for a given device category, department, or facility?

All these questions and more can be addressed from an individual database or through a meta-analysis by combining like fields from multiple databases. Although posing such questions is often referred to as data mining, what we’re really seeking is knowledge discovery—patterns or relationships within the data. Such discovery is how a discipline advances and matures. And it is with such knowledge that clinical engineering and the entire HTM community can best defend itself and demonstrate its tremendous worth and value to healthcare.

To continue where Dennis’ article left off, I would like to challenge and invite the HTM community to start thinking about how to extract meaningful information—or better yet, new knowledge—from our collective maintenance management systems. In order to do so, we are going to need at least a few departments (or a single independent service organization) with reasonably large and clean databases that are able and willing to share their CMMS data, or at least some subset of it. These data, which would be completely anonymized, would then be used as the basis for a statistically rigorous and exploratory data mining research study that seeks information and new knowledge regarding what cause and effect relationships, if any, surround the profession’s basic equipment repair, maintenance, support, and management tasks.

Despite the likely and large variability across HTM departments and their respective CMMS data entry policies, practices, and accuracy, the goal and challenge of doing good science is to discover patterns and meaning in data despite the presence of such variability. We have a host of established and powerful statistical tools to help us do just this. Additionally, and most importantly, these statistical software tools can be used to produce readily usable, easily digestible results that are not buried within or dependent upon an understanding of complex mathematical theory.

A large part of the difficulty the profession experienced a few years ago in defending its actions to the Centers for Medicare & Medicaid Services, and now potentially being revisited by the FDA’s recent call for comments on the repair and refurbishment of medical devices, is not due to our collective lack of data but to a lack of knowledge which may be hidden within the data. We need meaningful studies examining our actions, results, and contributions to the control and management of medical equipment. With only (largely spotty) anecdotal accounts, case studies, or descriptive claims of cost savings, it is no wonder that our nearly 50 years of contributions to healthcare are again being challenged.

If your department or organization has the data and if you would like to collaborate on such a CMMS study, let me know and we could explore the options further!

Larry Fennigkoh is a professor of biomedical engineering at the Milwaukee School of Engineering. Contact him at [email protected].