Determinants in the number of staff in hospitals’ maintenance departments: a multivariate regression analysis approach

Abstract

To date, there are no broadly accepted or accurate models to determine appropriate staffing [levels] for clinical engineering departments (CEDs). The purpose of this study is to determine what the determinants of the staffing levels are (total number of full time equivalents (FTEs)) in CEDs in healthcare organisations. In doing so, we used a cross-sectional exploratory approach by using a multivariate regression model over a secondary source of data information from the AAMI Benchmarking Solutions—Healthcare Technology Management database. Two hundred and one healthcare organisations were included in our study. Our study revealed that on average, there are almost 14 biomedical technicians (BMETs) per clinical engineer and one FTE per 1083.72 devices (SD 545.69). The results of this study also revealed that the total number of devices and the total technology management hours devoted to these devices positively affects the number of FTEs in a CED, whereas the hospital complexity, measured by healthcare organisation patient discharges matters inversely. The most important factor that matters in the number of FTEs in CEDs was the total technology management hours devoted to devices. A value of explained variance (i.e. R²) of 85% was obtained, indicating the strong power of the prediction accuracy of our multivariate regression model.

Keywords:

• Maintenance
• multiple criteria analysis; OR in health services
• human resource planning
• clinical engineering

Acknowledgements

The authors wish to thank Oliver Jarvis for his proofreading of this paper. The authors wish to thank our anonymous reviewers for their indirect help in preparing this manuscript. The authors assume overall responsibility for this manuscript.

Disclosure statement

The authors report no conflicts of interest.

Notes

1 We used a combination of the Portney and Watkins [24, p.852] table C.8 for the determination of lambda values (at α < 0.05, with 10 predictors, i.e. k = 10) for regression analysis and the formula n = λ× (1−R²)/R² to estimate our required sample size. Where, n is the required sample size, λ is a parameter depending on the residual degree of freedom (df residual). For our case the determined value of λ is 17.4 at df residual =120, R² is the effect size in a regression model. That for the case of our study design is R² = 0.4. Therefore, according to the formula, our required sample size is 26.1≈26 observations or CEDs were needed.

2 Calculated as the sum of all devices (i.e. imaging and therapeutic radiology, laboratory equipment, general biomedical equipment and other devices such as beds, stretchers, wheelchairs, nurse call systems, patient entertainment systems, general purpose computers, communications equipment, TVs, etc.).

3 We used hierarchical regression models. This method starts with a model containing all the main effects (i.e. individual components) and proceeds sequentially adding statistically significant interactions terms.[32].

4 As a rule we used a “change-of-estimate” rule of 5% or greater to retain a confounder in the model.

5 Concluding there is no effect when, in fact, there is one.