At the AAMI 2022 eXchange in San Antonio, Texas, a panel discussion “How Data Analytics Can Improve the Efficiency of Your HTM Department” was moderated by Doug Brown, vice president of enterprise development at reLink Medical. Other expert panelists included: Grant Smith, CRES, BSET, director of radiological engineering at Duke University Health System and Glenn Schneider, CRES, MHSA, director of clinical engineering at Cincinnati Children’s Hospital.
The discussion began with an overview of the topics that would be covered and the anticipated takeaways. Those takeaways included education, identifying important metrics, ideas for utilizing those metrics, all identified through a panel discussion with industry thought leaders.
It also looked at impediments that might be holding an HTM department back from implementing the suggestions. Those impediments included the time required to mine data, the knowledge required to extract data from CMMS, the condition of the data, the availability of the data, implementing changes based on what was yielded in discovery and benchmarking or comparing the results to other known data sets.
Big data sets can provide all kinds of useful information that can streamline any number of tasks and decisions made in an HTM department. That data may be just sitting there unused after substantial input efforts. The trick is to know what can be done with it and how to go about extrapolating the right data to achieve new levels of efficiency and cost savings.
Depending on an HTM department’s size, some have a dedicated data analyst who can make the best use of the available data.
Brown says that the average hospital is about 48 percent accurate and most employees don’t know how to mine the data or make use of it. He says that data must be both normalized and aggregated.
Diving into the Data
During the panel discussion, Smith said that his department currently uses analytics to evaluate travel, labor utilization, in-house teams to determine staffing models. He says that when clinics are added to a system, the analytics help understand travel time to help add staff effectively.
How does an HTM department transition from making little use of data sets to having a robust program of parsing data to bring new efficiencies and cost savings to bear?
“Historically, HTM departments have made decisions based upon anecdotal opinions, not empirical data. The easiest way to begin to utilize the data is to begin by understanding what it tells you. Each HTM department will have different needs and expectations, but in all cases, they can be validated by data,” Brown says.
He says that instead of departments making decisions based upon assumptions, they can now make them based upon facts.
“For example, how many technicians do you need? Are your current technicians efficient? How does your departmental performance compare to other, similar hospitals?” Brown asks.
He says that as frequently stated, the “devil is in the details.” He says that by finding answers to the questions raised, the HTM department must then create a plan to promote change in the department.
“Figuring out where you stand in terms of performance does no good unless you translate that into action that corrects deficiencies; and therein lies the most frequent problem. The HTM departments often stopped at identifying issues and didn’t progress to solving them even though they had the data to do so. Once the data indicates that their PM performance is slower than others in their category, it only helps if they develop an action plan to become more efficient. The issue with what we were doing is that we were able to show them where the issues were, but it was up to management to develop a plan to overcome them,” Brown says.
To prepare for putting that existing data to good use, there is a preparatory step that must occur.
“First the data needs to be scrubbed (standardized) – either hire someone or use your own resources to establish standard names for the elements of the data (manufacturer, model, description, etcetera). Then, review and correct/eliminate duplicate or non-standard data for your inventory,” Schneider says.
He says that once the data has been cleaned, put safeguards in place to prevent it from re-occurring.
“Minimize the number of people that have access to making changes to specific naming schemes. If possible, establish a person or group who approves any recommended changes to the data controls, then implements those changes, and performs a verification. Perform audits to ensure the team is entering data accurately and completely,” Schneider says.
He says to determine what data points are important to leaders and try to benchmark against similar organizations.
“Most use very similar data points, however, your organization may focus more on the financial aspect, while another may lean towards safety. Evaluate and present the desired data,” Schneider adds.
Employing benchmarks is an important part of the process. Schneider says that several can be used, including “Cost to Service ratio – aka CSR,” “Actual PM time versus book time,” “Contract effectiveness – are we at the correct level of contract based on usage” and “Scope of service” – should we take on more in-house repairs and reduce/drop contract coverage or is it more cost effective to contract more items for repair.”
“Benchmarking can be done on a host of measurements. The key to it is that the SA database has millions of work orders that contribute to their industry averages which makes it inherently more accurate due to the ‘law of large numbers’ … the same strategy employed by life insurance companies when determining life insurance rates,” Brown says.
Data-Driven Staffing Decisions
How do you use analytics to create manpower planning? How do you best utilize data for making an add-to-staff decision or justifying more hours? Where does that data come from and how is it presented to decision-makers?
“We recently used worked-time data to determine the staffing needs for an addition to our hospital. By running a four-year history of the average time spent on a specific model per year, we could estimate the necessary hours for each new item of the same model. We did that for all equipment that we were adding to our current inventory and used data from others for equipment that we currently didn’t have in our inventory, but were purchasing as a part of the project. Have to have data to justify. It’s often still an estimate, but it’s supported by historical data,” Schneider says.
We also compare PM time spent by our technicians to others through benchmarking with like organizations. This information is shared with the team to better understand differences and possible opportunities for additional education or overall improvement.
“We developed an algorithm to determine how much time (PM or CM) was required to service any device. By doing this for each device in the hospital inventory, we could derive the number of service hours needed to maintain the equipment. We would then apply the hospital’s expectation of labor accountability (usually around 85 percent) to determine the number of hours required – then assume that a 2,080-hour work-year would result in 1,796 useable hours,” Brown says.
He says that simple math allows you to determine from those inputs how many people you need.
“You then have to account for special projects, training time, meetings, travel between hospitals, etcetera, to arrive at the correct number of FTEs required,” Brown adds.
He says that showing the data – and how you reached your conclusions – to management is defensible because of the accuracy of the data which is “washed” by thousands of hospitals covering millions of work orders. As long as the data being captured is accurate (not always the case) the outcome is pretty accurate, too.
“Interestingly, in a survey of HTM managers at one AAMI conference, I asked each one how they got to their labor budget each year. One-hundred percent of them responded that their budgets were always plus or minus percentages based on the previous year. None of the budgeting processes had any relation to the equipment being managed, including the age and condition of it. It was always based on the previous year and was added to or subtracted from based on the hospitals’ economic performance,” Brown says.
He says that is not a viable way to create a budget.
“Consider an IDN with four hospitals. Three of them are older and are maintaining aged equipment. One of them is newer and has predominantly newer equipment, much of which is under warranty. Logic would tell you that the latter hospital probably requires less manpower to maintain their equipment. Older equipment typically requires more attention. But that would rarely be taken into consideration with the average hospital when considering a budget. Rather, the budget would be a percentage of last year’s budget generally ranging from 96 to 105 percent,” Brown says.
Data That is Uniform
The panel discussion also touched on the importance of standardizing terminology.
How much of a problem is simple nomenclature and what are some other variables that cause problems for normalization? How is this cleaned up? How do you limit choices with dropdowns or other approaches?
“Clean data is an oxymoron in our industry. The average hospital was less than 50 percent accurate on OEM nomenclature — model name, model number and device type (modality). Consider the most common portable X-ray machine. We found over 45 permutations, with one hospital that had it in their CMMS 17 different ways. Trying to do analytics with inaccurate data is self-defeating because – on average – you can only get accurate statistical data on less than half of your inventory,” Brown says.
He says that powerful algorithms were employed that compared the customers’ different adaptations of a device to SA vetted device database which contained millions of correct device nomenclature.
“However, not everything can be determined by an algorithm. For example, if the four required bits of information required (OEM, model name, Model Number, Modality) contains things like “GE, N/A, N/A, Unknown” there is little that can be done to determine what the correct device nomenclature is and it would have to be returned to the customer for them to determine what the device was by serial or hospital tag number,” Brown says.
He says that this manual process could be daunting. “Consider a medium hospital organization with 25,000 devices spread over four locations. While the algorithm could typically handle 90 percent of the devices, there would still be 2,500 that would require human intervention to locate the device by serial number and determine the correct nomenclature.
“Hospitals that didn’t have the time or manpower to do this were unable to get high-correlation analytics because we were typically only able to work with 50 percent of their devices,” Brown says. In conclusion, the ability to harness available data can serve many useful purposes and aid greatly with efficiencies and budgets.
“The use of data analytics is a part of our future in HTM. Many hospitals have embraced it, but there are many more waiting to see how it might impact them. There is a reluctance, or fear, by some who are concerned they might be behind the performance curve of the average. But how much better it would be to find out where you stand – how you are performing – and create a program to do even better. How much better to be able to approach the C-suite with a plan for improvement and/or a justification for FTEs than to wait for a budget to be given to you that may not be adequate to do your job properly,” Brown says.
He points out that ignorance is not bliss. “We – as an industry – need to be as well-informed as possible and get in front of new trends that will ultimately change the way we’ve done business over the last few decades. The use of analytics is certainly one of those,” Brown adds.
