How one hospital network used data to assess the risk of biomedical equipment in anesthetizing areas.
By Ben Lewis, MBA, CHTM, CHOP-A, and Seth Washispack, MBA
Based on a recommendation by international accreditor DNV, HonorHealth, a hospital network with six medical centers in Arizona, performed statistical analysis to inform a risk assessment for operating in anesthetizing areas with a relative humidity environment of between 20% and 60%. This case study presents a novel method for utilizing building automation system and medical equipment maintenance data for evaluating the risk of biomedical equipment and provides evidence for other facilities who choose to operate their anesthetizing areas in relative humidity ranges of between 20% and 60%. Here’s our story.
Introduction
To provide a rigorous and repeatable risk assessment, HonorHealth statistically tested the risk of operating biomedical equipment in anesthetizing areas with relative humidity levels of between 20% and 60% (focusing on the 20%-30% range since the risk of operating in this range has been debated). Using building automation system and corrective maintenance work order data, HonorHealth found an insignificant correlation between biomedical equipment failure and the relative humidity range of 20% and 30%.
Background
In April 2013, CMS issued a memorandum on a categorical Life Safety Code waiver permitting new and existing ventilation systems in critical access hospital’s (CAH) anesthetizing locations to operate with a relative humidity of between 20% and 60%. This decision was made to provide flexibility during construction and to save on operational costs incurred to meet the previous humidity standards of between 30%-60%.
A multidisciplinary group consisting of the American Hospital Association, the American Society for Healthcare Engineering, the Association for the Advancement of Medical Instrumentation (AAMI), the Association of Perioperative Registered Nurses, and others came together in January 2015 to provide new guidance. One month later, CMS communicated that a risk assessment for operating in conditions between 20% and 30% relative humidity was required. Further, CMS-accredited facilities, surveyed by DNV-GL, were cited for operating in anesthetizing areas without a risk assessment under PE.8 (SR.6) / (SR.7).
To develop a robust risk assessment, HonorHealth sought to statistically test the relationship between biomedical equipment failure and relative humidity. Two datasets were collected for this analysis. First, corrective maintenance work order data was collected via HonorHealth’s computerized maintenance management system (CMMS) to account for the equipment failure. Second, relative humidity data was collected from HonorHealth’s building automated system (BAS)—a software and hardware solution that controls and monitors ventilation, humidity, air, power systems, and more. The BAS also senses, tracks, and stores temperature and humidity levels by location.
HonorHealth’s healthcare technology management (HTM) and supply chain operations teams partnered with the facilities management department to compile and analyze the corrective work order and relative humidity data for two medical centers’ operating rooms.
Data Analysis and Results
The analysis focused on understanding the relationship between relative humidity and biomedical equipment failure. For statistical robustness, two tests were used: a negative binomial regression and a Wilcoxon Rank-Sum.
Collection of the humidity percentage data from the BAS occurred at the beginning of each hour (e.g., humidity at 0:00, 1:00, 2:00) of the day. The humidity data was collected for 25 ORs at the HonorHealth Scottsdale Osborn campus and 27 ORs at the HonorHealth Shea campus over one year. Additionally, collection of corrective maintenance work orders from the CMMS occurred during the same one-year period as the humidity data for the OR departments at the respective campuses.
The negative binomial regression results demonstrated an insignificant correlation between relative humidity and corrective maintenance work order counts at both campuses when the humidity is between 20% and 60%. These Osborn results are shown in Figure 1 and Table 1. Figure 1 visually demonstrates the lack of fit between the corrective maintenance work order count and relative humidity through randomly scattered data points. Table 1 statistically corroborates the lack of fit from Figure 1 with a large p-value (generally, p-values of 0.05 and smaller are considered significant).
The Wilcoxon Rank-Sum for the difference in corrective maintenance work order counts for relative humidity above 30% and between 20% and 30% was also insignificant. Table 2 provides the details of the test. The high exact probability value of 0.54 indicates the improbability of the relative humidity range of 20-30% increasing the number of corrective work orders compared to the relative humidity range of 30-60%. For additional robustness, the same tests were run for the Shea campus. Similar results can be found in Figure 2, Table 3, and Table 4.
The two insignificant results provide evidence for a low risk of biomedical equipment failure in a relative humidity environment of between 20% and 30%. These results informed HonorHealth’s risk assessment by providing evidence to conclude an improbable likelihood of biomedical equipment failure due to the relative humidity operating range of the OR rooms (20%-60%).
Key Limitations
There were two major limitations mitigated in the analysis: a lack of data granularity and missing data. The first, lack of data granularity, emerged from the specificity differences of the datasets. The corrective maintenance work orders were compiled at the facility level while the OR humidity percentages were compiled at the facility and room level.
Due to their regular travel throughout the OR department, assets are not tracked at room level, requiring the corrective maintenance data to be tracked at the facility level. An assumption was made because of the differences in data collection points: the aggregate of the humidity at each facility affected all biomedical equipment. Data including asset location by room (with equipment assigned to specific locations) could allow for a more detailed study.
The missing data limitation developed from two sources:
- Differing time periods of the data collection from the two databases (based on data availability).
- Installation timing differences of the building automated system controls.
To maximize the data points under evaluation, we chose to exclude the data points in each room at each hour where data was unavailable. Why? Because this method skews the results toward the rooms with more data. The study found an insignificant correlation at the 95% confidence level between biomedical equipment failure and the relative humidity range of 20% and 60%. The correlation became inconclusive when the humidity dropped below 20%. Future studies could build upon this study by investigating the upper and lower humidity values where the biomedical equipment is affected.
Utilizing the Results
The insignificant results of the relationship between humidity and biomedical failures informed HonorHealth’s risk evaluation conclusion: There is an improbable likelihood of biomedical equipment failure due to operating in relative humidity conditions between 20% and 60%.
Moreover, this study demonstrates a method for utilizing building automation system data and corrective maintenance work order data to provide insights into how an environmental factor—humidity—could affect biomedical equipment. Future studies could use similar data sources to determine how other environmental factors, such as temperature and pressure, affect devices. Additionally, further improvements in data analytics could help predict when equipment needs preventative maintenance based on environmental and system factors.
Finally, the statistical testing approach also demonstrates a novel response to a CMS inquiry for a risk assessment. Moving from a more subjective approach (using antidotes) to a more objective approach (utilizing statistical tests with data grounded in the environment where the risk occurs) helped HonorHealth make an informed risk assessment.
Ben Lewis, MBA, CHTM, CHOP-A is the associate vice president of support services at HonorHealth, an Arizona-based hospital network. He is a certified healthcare operations professional through DNV-GL and a certified healthcare technology manager through AAMI. Seth Washispack, MBA, is a supply chain analyst at HonorHealth and has earned his B.S. in biomedical engineering.
References:
- CMS S&C: 13-25-LSC & ASC. Relative Humidity (RH): Waiver of Life Safety Code (LSC) Anesthetizing Location Requirements; Discussion of Ambulatory Surgical Center (ASC) Operating Room Requirements. Baltimore, MD: Centers for Medicare & Medicaid Services.
- CMS S&C: 15-27-Hospital, CAH & ASC. Potential Adverse Impact of Lower Relative Humidity (RH) in Operating Rooms (ORs). Baltimore, MD: Centers for Medicare & Medicaid Services.
- DNV GL National Integrated Accreditation for Healthcare Organizations (NIAHO®). Accreditation Requirements, Interpretive Guidelines and Surveyor Guidance Revision 18-1.Milford, OH: DNV GL Healthcare.
Tables and Figures
Figure 1: Osborn Scatter Plot, CM Work Order Count and Minimum Weekly Median Humidity
Table 1: Osborn Negative Binomial Regression, CM Work Order Count and Minimum Weekly Median Humidity
| Coefficients | Robust Standard Error | z-statistic | p-value | |
| Intercept | 1.92 | 0.39 | 4.91 | 0.00 |
| Min Week Median | 0.00 | 0.12 | 0.34 | 0.73 |
Table 2: Wilcoxon Rank-Sum, Difference in CM Work Order Count Means for Minimum Weekly Humidity Above 30% and Between 20% and 30%
| Above 30%: CM Count | Between 20% and 30%: CM Count | |
| Rank Sum | 1174.5 | 421.5 |
| Expected Sum | 456 | 1140 |
| Observations | 40 | 16 |
| Adjusted Variance | 3019.32 | |
| z-statistic | -0.628 | |
| Exact Probability | 0.54 |
Figure 2: Shea Scatter Plot, CM Work Order Count and Minimum Weekly Median Humidity
Table 3: Shea Negative Binomial Regression, CM Work Order Count and Minimum Weekly Median Humidity
| Coefficients | Robust Standard Error | z-statistic | p-value | |
| Intercept | 3.62 | 0.05 | 70.83 | 0.00 |
| Min Week Median | -0.01 | 0.00 | -1.28 | 0.20 |
Table 4: Shea Wilcoxon Rank-Sum, Difference in CM Work Order Count Means, Minimum Weekly Humidity Above 30% and Between 20% and 30%
| Above 30%: CM Count | Between 20% and 30%: CM Count | |
| Rank Sum | 1174.5 | 421.5 |
| Expected Sum | 456 | 1140 |
| Observations | 40 | 16 |
| Adjusted Variance | 3019.32 | |
| z-statistic | -0.628 | |
| Exact Probability | 0.54 |
