Intrinsic to healthcare prcess and decision analysis is the need to analyze data. This need ranges from being able to determine demand patterns, service patterns, and supply usage patterns, to being able to measure utilization and performance levels. And while the types of analysis needed vary by project, they can include statistical analysis such as looking for and identifying patterns when they exist, hypothesis testing, and distribution fitting so that existing data can be used to identify possible future trends.
Some of the more significant of Phil's data analysis projects include:
This project was commissioned by the research department of what was then the fourth largest shopping center management company in the United States. Their goal was to identify retail chains to which they might lease space in particular shopping centers, their criterion being an expectation of strong sales for the chain's stores in those centers. The approach selected was to first determine the correlation of sales between pairs of chains that had stores in the management company's shopping centers, and to then determine the centers where sales of one of the chains was high. To perform the analysis, Phil obtained monthly sales data for each of the chains in each of the company's shopping centers, loaded that data into a datamart, identified and resolved data issues, and developed software to compute correlations. Phil and an associate then analyzed and summarized the results so that management could use them as part of their sales efforts.
Having recently created and bulk-mailed a large number of catalogs to potential customers obtained from purchased address lists, the e-commerce company Phil performed this work for wanted to determine whether the sales from the catalog paid for itself, and whether and how to do it again. Unfortunately, only a few customers had indicated that they had purchased from the catalog, so the determination of catalog sales had to be done by matching names and/or addresses of purchasers to those who were sent the catalog. Since most of the individuals that bought from the catalog entered their name or address differently than in the address lists, Phil developed a program to match name and address data. Using the results of that program, Phil and another analyst estimated catalog sales by list and identified a list that had ten times the average purchase rate of other lists of potential customers the company had sent catalogs to. Phil then suggested a strategy for taking advantage of this list, with the expectation that this strategy would lead to a higher return on investements in future catalogs.
One of the most common measures of hospital surgical performance is the utilization of operating rooms relative to their staffing levels. And when only a pre-specified set of operating rooms are used in a given day, this utilization is easy to compute. However in the hospital for which Phil did this analysis, there were additional rooms that were not used all the time, and sometimes part or all of individual surgical teams moved to other operating rooms between procedures. To perform this analysis, and to also estimate the potential OR time that could be recovered between late starts and early ends, Phil built a simulation like program to which he inputted actual OR data and from which he outputted statistics such as the OR's utilization and the amount of additional time that could reasonably have been recovered from each OR each day.
To meet Quebec government guidelines, all oncological surgical procedures are required to be completed within 28 days of the time the patient has agreed to have the procedure. Unfortunately, at the hospital for which Phil performed this analysis, significant percentages of these procedures are not performed within those 28 days. This prompted the hospital's director general to ask Phil to determine whether it should be possible to achive the guidelines without changing the way OR time is allocated to individual surgeons. To address this challenge Phil built a simulation like program to determine the feasibility of achieving the guidelines under these circumstances and found that due to the variability in both demand for procedures and surgeon OR time (due to holidays, vacations, stat days, illness) that it would not be possible to do so without providing surgeons more flexibility in how OR time was allocated to them.