Appendix: Research methods
The Deloitte Center for Health Solutions is looking at how much time savings can be achieved over the next 1-3 years using currently available technologies, including automation, AI, generative AI, telemedicine, and combinations of these technologies. We have developed an estimation model that shows. We modeled these savings for her two jobs: revenue cycle and bedside nursing.
To provide model inputs, we use secondary sources that are publicly available by subscription, such as the U.S. Department of Labor's Occupational Information Network and a database that pools operational data from hospital organizations that are members of the largest group purchasing organization in the United States. I used the data. Country (referred to as “hospital database” in this section).
Additionally, we employed the Delphi method.36 Review and establish consensus on the model's inputs, assumptions, and outputs. The Delphi panel for each role was composed of functional experts with recent professional or consulting experience in that role and technical experts familiar with the workflow and the technologies that can be used to optimize that workflow. This included individuals with expertise in clinical informatics, robotic process automation, machine learning, and AI.
The main stages are listed below.
Defining scope
In our estimated model, we chose to represent opportunities within two core hospital functions: Revenue Cycle and NurseThese features were chosen to demonstrate the potential impact of the technology on both clinical and non-clinical operations. The modeling assumed the conditions and staffing of an average community hospital in the United States.
Performing the analysis
To estimate technology time savings in both the revenue cycle and nursing care, we took a four-step approach:
Step 1. Allocate time to spend on individual activities or tasks for both jobs.
For revenue cycle, the revenue cycle department was the level of analysis. Based on the literature review, and with the help of a Delphi committee of revenue cycle experts from Deloitte, we developed detailed workflows for each of the three major revenue cycle stages: patient access, mid-revenue cycle, and patient financial services. We then used operational data from a hospital database on full-time employee (FTE) work hours in revenue cycle functions to estimate the percentage of time spent on each individual workflow activity for an average community hospital.
For nursing, the level of analysis was the individual nurse. We consulted the U.S. Department of Labor's Occupational Information Network to obtain a list of jobs with frequency for acute care nurses.
For both functions, we used the Delphi method to review model inputs and assumptions and reach consensus. There was one Delphi panel of revenue cycle experts and two Delphi panels of functional and technical nursing experts.
Step 2. Estimate the technology time savings for each activity or task in both functions.
For each feature, Delphi's feature panelists evaluated each task or activity and estimated how much of the time spent on this task today could be saved by adopting one or more technologies. Delphi panelists were specifically asked to think about technologies that are mature enough to be implemented within the next 1-3 years.
Step 3. Calculate the total FTE hours for a function or role within an average hospital.
Based on a literature review, we hypothesized that the average nurse works a total of 48 weeks (approximately 1,920 hours) per year.37 Using FTE benchmarks from a hospital database, we estimated that revenue cycle professionals work an average of approximately 1,850 hours per year.
Step 4. Calculate the overall impact in terms of time efficiency or FTE time savings for the function or role.
Estimates from Delphi panelists were averaged to derive total time savings for each role or function. These time savings were then expressed as a percentage of total employee time and FTE hours per year.
To assess the sensitivity of the results to extreme assumptions, we calculated three sets of estimates: conservative, typical, and generous. The typical estimate was the average of all panelists' inputs. For the conservative and generous estimates, we excluded outliers at the upper and lower limits, respectively. If the difference between an individual panelist's estimate and the average (mean) of all panelists was 20 percentage points or more, we treated those individual estimates as outliers. For revenue cycle functions, the conservative and generous range estimates differed from the typical estimates by 8 to 11 percentage points. For nursing functions, the conservative and generous range estimates differed from the typical estimates by 1 to 3 percentage points.
Limitations
We acknowledge that our findings are illustrative and provide directional insights.
This model assumes an existing process and suggests how each element of this process can be improved with technology. The model does not assume or suggest how to optimize the process itself. Therefore, actual savings may be greater as a result of process redesign. On the other hand, this model overestimates savings because it makes neither practical nor economic sense to apply technology to every element of the process. But overall, these two limitations should cancel each other out.
We also recognize that each hospital is organized differently and has different levels of comfort and experience with technology, which can impact the degree to which a hospital can take advantage of its benefits.
Because there are multiple ways to approach and solve a problem, panelists relied on their own expertise, which may have led to variation in their answers, which could have affected the estimates.
Our analysis does not explore the impact on other jobs or adjacent workflows. Our modeling assumes that the overall workflow is the same. But in reality, for many jobs, departments, functions, and entire organizations, job redesign will result in entirely new processes. This will permanently delete the specific task and create a new task or job.
Our modeling does not account for inter-organizational variation due to differences in processes or patient mix, or within-organizational variation due to lack of process standardization or differences in patient complexity.
Finally, the financial and operating metrics for hospitals and health systems in our hospital database are based on organizations that are members of large cooperative purchasing organizations and therefore may not be representative of hospital organizations that are not members.
