A Program Breakdown of Benchmark Indicators
For Pharmacy Staffing and Drug Costs

Kevin W. Hall

Introduction

For over fifteen years, the Hospital Pharmacy in Canada Survey has been producing data that has helped pharmacy managers to respond more knowledgeably to questions concerning their staffing and drug costs. The usefulness of the data has been improved over the years by creating subsets of the overall data, that grouped hospitals based on size, teaching status and the type of drug distribution system in use. However, the usefulness of the data for interfacility comparisons remained limited by the fact that data from hospitals within the subsets were combined, regardless of the mix of patient services and types of pharmacy services being delivered. This meant that a diverse group of facilities, such as hospitals providing primarily acute care services, hospitals providing significant amounts of long-term care services, pediatric hospitals, and psychiatric hospitals were all included in the same pool of data. Similarly all of the data from within a single hospital was being rolled up despite the fact that many hospitals, particularly the larger ones, operated a number of quite distinct programs to serve different patient groups within their facility.

In the last two Hospital Pharmacy in Canada Survey Annual Reports, the results of a new approach to developing benchmark indicators for pharmacy departments in Canada were first reported. The new benchmarking section was an attempt to identify the staffing and drug costs incurred by pharmacy departments in the delivery of services to specific inpatient and outpatient programs, such as critical care, pediatrics, outpatient prescription services, and mental health. In addition, it attempted to capture the resources committed to those indirect patient care programs that are operated by some but not all pharmacy departments, such as a regional drug information centre or an investigational drug service.

An important premise underlying this benchmarking approach is that there should be a reasonable degree of consistency when the pharmacy resources required to service the same type of patient group, or to deliver a similar type of pharmacy service, are compared between facilities. The assumption underlying this premise is that a similar standard of care is being delivered at the facilities being compared; an assumption that is probably not always correct. Nonetheless, the argument has been made by pharmacy managers that only by breaking down a pharmacy department into its sub-component parts, and identifying the resources committed to each, is it possible to create more refined and useful data on which to attempt inter-facility resource-utilization comparisons. In the future it would be ideal if this type of program-specific benchmarking analysis could be combined with accepted outcome measures, so that both quality of care and efficiency of service delivery could be assessed.

Although justification of the pharmacy resources is an important reason for developing improved pharmacy benchmark indicators, it is not the sole use for program-specific pharmacy data. Planning for any new or expanded patient care program is greatly facilitated when there is some knowledge of the pharmacy staffing and/or drug costs being incurred at other facilities to service the same type of patient group, or to operate the same type of pharmacy program. Likewise, with all of the program consolidations, program transfers and downsizing that are occurring as a result of the regionalization of health care, it is important to have data on which to base the amount of resources that will be removed from one site and transferred to another. Applying the existing overall departmental average for staffing (e.g. paid hours per patient day), or drug costs (e.g. drug cost per patient day), will underestimate the resources associated with pharmacy resource-intensive programs, such as oncology, critical care and pediatrics. On the reverse side, such an approach would overestimate the resources required for programs that are less pharmacy resource-intensive, such as long term care or mental health. Thus the availability of data on the pharmacy resources required to service defined patient groups was felt to be an important planning tool that could be made available to pharmacy managers as a result of this programmatic benchmarking analysis.

The new program-based benchmarking analysis, utilized in the last two surveys of hospital pharmacy in Canada, created new program-specific subsets of the overall data. The results demonstrated that the variability in pharmacy resource utilization was reduced when this program-specific methodology was used. It was therefore decided to continue this analysis in the 2001/2002 report.

Methods

The benchmarking section of the survey consisted of five sections. In Section I, survey respondents were asked to provide data on total pharmacy staffing for their entire pharmacy operations, total inpatient and outpatient drug costs, total hospital beds, total patient days and the type of drug distribution system in use at their hospital. In Section II respondents were asked to provide similar data for subgroups of inpatients such as critical care patients, bone marrow transplant patients, and long term care patients. The survey instructions for this section requested that the pharmacy resources associated with each distinct inpatient program be reported separately. In Section III, respondents were asked to provide data for any outpatient pharmacy programs operated by their hospitals, such as outpatient prescription dispensing and home parenteral nutrition. In Section IV, respondents were asked to provide data on any other unique pharmacy services that they operated, such as a regional drug information service, investigational drug services and contract services to external organizations. Section V was intended to represent the remaining pool of inpatient acute care patient groups such as family medicine, internal medicine, and general surgery. If the survey was completed correctly, the inpatient beds, inpatient days, pharmacy staffing resources and drug costs in Section V would equal those in Section I minus those in Sections II through IV.

The Editorial Advisory Board for the survey recognized that the type of survey being proposed would be time-consuming for respondents to complete, and would be time-consuming and challenging for the Board to analyze. As a result the Board decided to limit the distribution of the survey to those hospitals that were most likely to derive benefit from this type of program-based breakdown of their overall hospital data. Specifically, the facilities selected to receive the new benchmarking section of the survey were facilities that were identified in the Board’s database of Canadian hospitals as having over 300 acute care beds, or which were identified as a pediatric hospital. The facilities with more than 300 beds were chosen because it was believed that hospitals in this group were most likely to operate a variety of different programs that could be assessed in terms of their different resource utilization patterns. Pediatric hospitals were selected for inclusion in the survey because the delivery of pediatric pharmacy services was shown in the previous two benchmarking surveys to require significantly greater amounts of pharmacy staffing than the delivery of similar adult care programs. It was also anticipated that specialized pediatric hospitals might show a quite different profile of resource utilization than would a small pediatric service within a primarily adult care facility.

In the 2001/2002 survey, the benchmarking section was distributed to 60 “adult” hospitals with more than 300 beds, and to six pediatric hospitals. All provinces in the country were represented in the distribution list, with the exception of Prince Edward Island which did not have any hospitals that met the distribution criteria.

The returned benchmarking surveys were individually reviewed by the author of this section of the survey. Direct contact was made with many respondents to clarify data discrepancies. Based on the data submitted, indicators such as paid hours per patient day and drug costs per patient day were calculated for each program. To the extent possible, efforts were made to insure that there was consistency in the program data provided by the different facilities. For example, some facilities were able to provide a more detailed breakdown of their programs than were the majority of other facilities. When this was the case, the data for a number of programs reported by that facility were combined to create a program grouping that was similar to that reported by other facilities. An example of this would be the combining of separate data provided for general medicine and general surgery, since most facilities could not provide that breakdown.

The program-specific indicators were then subjected to calculations to determine the mean, median, standard deviation, minimum and maximum values for each indicator. The spreadsheet was set up to calculate this data for all hospitals, as well as for subgroups of hospitals based on their size and the type of drug distribution system reported to be in use within each facility.

Results and Discussion

Thirty individual hospitals or regions, representing a 45% response rate, returned the benchmarking section of the survey. Two of these responses were eliminated from the analysis because inadequate data were submitted to enable the calculation of indicators (one facility), or because the facility fell well below the 300 acute care bed cutoff (one facility). Of the remaining 28 responses, 15 were from adult hospitals with more than 500 acute care beds, 9 were from hospitals with between 300 and 500 acute care beds, and four were from pediatric hospitals. Each province is represented in the results with the exception of Prince Edward Island and Newfoundland.

The completeness of the submitted data varied between hospitals. Each useable data element was included in the analysis, regardless of whether or not the respondent could provide all of the data requested in the survey. For many of the calculated indicators the number of reporting hospitals was high enough to make the data quite meaningful. For other indicators the number of reporting facilities was quite small and caution is warranted with respect to the interpretation and use of that data.

Adult Hospitals

Table H-1 shows the results of the analysis of human resource and drug cost data for the 24 adult hospitals before and after the resources for specialized programs were extracted. The “before adjustment” figures represent the overall pharmacy data submitted in Section I of the survey, and the “after adjustment” figures represent primarily the general inpatient medical and surgical programs that remained in Section V of the survey. Other pharmacy data that would roll up in this “after adjustment” section would include any staffing committed to core pharmacy department functions such as purchasing, inventory management, wardstock distribution, office functions and departmental management. The “after adjustment” figures are felt to represent a fairly homogeneous group of pharmacy services, based on the assumption that respondents have identified the high and low pharmacy resource programs in Sections II to IV of the survey. This grouping should be quite similar to an acute care community hospital that provides services primarily to general medicine and surgery patients. Patient groups requiring specialized pharmacy services and patient groups requiring little in the way of pharmacy services are no longer part of the patient group that is being compared.

The data is presented for all 24 adult hospitals and is also reported separately for hospitals with more than 500 beds and for hospitals with 300 to 500 beds. The after adjustment data shows much less variability than the before adjustment data. The variability diminishes even further when hospital size and type of drug distribution system are used to sub-categorize the data. It is worthwhile noting that both the minimum and maximum values are usually brought closer to the mean value by this benchmarking analysis methodology. Hospitals that had reported very low staffing and drug costs in Section I of the survey were usually found to be operating patient care programs, like long term care beds, that consume lower than average amounts of resources. When those low-resource programs were extracted, the paid hours per patient day and the drug costs per patient day increased for the remaining beds. On the other extreme, hospitals that had reported very high paid hours per patient day or very high drug costs per patient day were usually found to be operating a number of specialized resource-intensive pharmacy programs. When those resource-intensive programs were extracted, the paid hours per patient day and drug costs per patient day decreased for the remaining beds.

The decrease in variability can be demonstrated by examining the paid hours per patient day data. The overall data show that the 24 facilities had a five-fold variation in their paid hours per patient day, from 0.24 to 1.31. The breakdown of facilities by hospital size and type of drug distribution system, which has been part of the reported survey data for a number of years, reduces that variability. For example the range for hospitals with more than 500 beds, using a unit dose/IV admixture system of drug distribution, is 0.46 to 1.31 paid hours per patient day, a three-fold variability. The benchmarking adjustment further reduces the range to 0.51 to 0.96 paid hours per patient day, a two-fold variability.

The pharmacy manager who was asked to justify why their pharmacy department appears to have higher staffing indicators than comparable hospitals could use the benchmarking data reported in Table H-1 to determine if that assumption was in fact true. The manager could extract the resources associated with resource-intensive programs and compare the “after adjustment” figure for their hospital with that for the hospital grouping in Table H-1 that is most similar to their own in size and type of drug distribution system. This might well demonstrate that their hospital’s pharmacy staffing was quite appropriate in relation to other facilities when this programmatic approach to benchmarking was used.

Table H-1 also shows before and after adjustment drug costs per patient day. It is clear that there is large variability in drug costs per patient day when overall drug costs are used to create drug cost indicators. Even adjustment for the size of the hospital and the type of drug distribution system leaves a large amount of variability. In contrast, the mean after adjustment drug costs per patient day are more similar, regardless of the size of hospital or type of drug distribution system in use. It is also of interest that the after adjustment drug costs per patient day are, on average, only about one-half of the before adjustment figures. This indicates that specialized inpatient and outpatient pharmacy programs tend to be associated with the most costly drug therapies. This is probably not surprising to most managers who have had to deal with the high cost therapies used in areas such as oncology, critical care, and organ transplant programs.

The impact of the program-based benchmarking approach for 2001/2002 is the same as the results from the previous two benchmarking surveys. Not surprisingly, however, there has been a general upward trend in both pharmacy paid hours per patient day and in drug costs per patient day. With respect to the pharmacy staffing indicator of paid hours per patient day, there was an increase between 5% and 10% in the before and after adjustment figures for all adult hospitals and for hospitals with traditional/mixed drug distribution systems. Interestingly, there was a small decrease in paid hours per patient day for the group of facilities with >90% unit dose and centralized IV admixture systems. On examination, a number of facilities that were in the traditional/mixed drug distribution group in the last survey have moved to the unit dose/CIVA group in the present survey, with only small changes in their staffing indicator of paid hours per patient day. The author is aware that several of these facilities have made the transition to unit dose with the use of automated drug distribution technologies, specifically Pyxis cabinets. It is quite possible that we are seeing a positive impact of these automated technologies on the labour requirements for operating unit dose systems.

Drug costs per patient day were also higher in the 2001/2002 results than they were in the 1999/2000 results. The increases were generally in the 20% to 30% range and were reasonably consistent in both the before adjustment figures and the after adjustment figures.

Pediatric Hospitals

In Table H-2, similar data is provided for the four pediatric hospitals. The numbers of respondents is small and the results must be viewed from that perspective. However, as was the case for the adult hospitals, the program-based benchmarking analysis reduced the variability between facilities with respect to both their staffing and drug cost indicators. As was demonstrated in the last two benchmarking surveys, it is clear that the pharmacy staffing required to service pediatric patients is substantially higher than that required for adult pharmacy services. For both unit dose/IV admixture hospitals, and for traditional/mixed distribution system hospitals, the paid hours per patient day are approximately twice that reported by adult hospitals. As compared to the 1999/2000 results, pharmacy paid hours per patient day have increased approximately 10%.

Drug costs for the pediatric hospitals were about 30% higher in 2001/2002 than they were in 1999/2000.

It should be noted that one of the three unit dose/IV admixture pediatric facilities is operated as a partially separate component of a larger pharmacy department in a multi-site complex. As such it receives much of its core services, such as overall management, purchasing and inventory control, from the central department. Although the staffing for these services was apportioned to the pediatric facility for the purposes of this analysis, it is uncertain how this would compare to the same services delivered in a completely stand-alone pediatric hospital. It is unknown if this may have contributed to the lower adjusted paid hours per patient day (1.06) that this unit dose/IV admixture pediatric facility demonstrated, as compared to the other two facilities in this group (2.24 and 2.5).

Specialty Programs, Adult Hospitals

In Table H-3, information on staffing and drug costs is presented for a number of inpatient and outpatient specialty programs. The programs included in this table are ones for which enough hospitals provided data to make the program-specific information meaningful. In addition to the mean and median values for the data submitted, Table H-3 also includes the raw data values from all of the reporting hospitals. The raw data is included to provide a better indication of how the individual hospital data was clustered. It provides essentially the same information as a standard deviation but may be more informative for some readers.

In some cases, there is a pronounced clustering of the raw data with very few outliers. This pattern suggests that there is considerable consistency between hospitals with respect to these programs, and that the outliers may represent data reporting errors or inconsistencies on the part of a few hospitals. Examples of this relatively “tight” data include the drug costs per patient day for long term care and mental health, as well as the staffing information for the mental health and investigational drug programs.

For other programs the raw data is scattered over a fairly broad range, without much clustering around the mean. The wide variations in paid hours per patient day for programs like critical care may suggest that there are major differences in the way that these programs are being serviced by pharmacy at different hospitals. For example several of the hospitals operate satellite pharmacy services in critical care that are associated with a very high paid hour per patient day staffing indicator. In contrast other hospitals provide very minimal service to their critical care areas. The variability in pharmacy services to these programs may indicate the need for the profession and the affected clinical programs to establish standards for the pharmacy services delivered to these patient groups.

Wide variations in the drug cost per patient day for some of these specialized programs may also indicate differences in the type of drug therapy being used at different facilities, which would again be a standard of care issue. However it is also possible that the wide variations represent an inconsistency in what is being included in the drug costs for any given program. For example some hospitals start thrombolytics in the emergency room and charge those costs to that area. Other facilities charge those same costs to the critical care areas

The paid hours per concurrent investigational drug study managed was quite consistent between facilities. The mean of approximately 50 hours per study suggests that the human resource costs of managing these studies are substantial. Given that the mean number of concurrent studies was over 100, the average hospital providing this service would be committing approximately 2.5 FTEs to investigational drug study management.

For oncology admixture preparation, the paid hours per admixture cover a fairly broad range, but there is a reasonably close clustering around the mean value of 0.51. This suggests that this figure would be a reasonable benchmark value and that it should be useful for program planning purposes.

It was of interest that the number of facilities that identified staffing resources committed specifically to dialysis increased from 6 to 14. Although there was no useful denominator that would allow a calculated staffing indicator such as paid hours per dialysis patient, the number of FTEs committed to the dialysis program ranged from 0.2 to 5. These were virtually all pharmacists, indicating that the services offered were primarily clinical in nature.

Conclusion

The data provided in the program-based benchmarking survey for 2001/2002 validates the methodology that was used in the two previous benchmarking surveys. The results demonstrate that increases in both staffing and drug costs have occurred over the two year period. Drug cost increases in this program-based analysis appear to be occurring in many different program areas. Staffing for pediatric pharmacy services approaches twice the amount required for adult services. A number of program-specific pharmacy indicators for both staffing and drug costs have been derived from the survey results. Some of these appear to be quite reliable benchmark indicators while others must be interpreted more cautiously. The variability in pharmacy staffing for some specialized programs speaks to the need for the development of standards for the pharmacy services delivered to those programs. If this survey is to be repeated on a regular basis, it would be desirable to better define the data that should be collected and reported for each program area. It is anticipated that this would improve the reliability of the data being collected and thus would result in better benchmark indicators.