The setting for this study was an academically affiliated long-term care facility in Canada with an electronic medical record system including integrated computerized provider order entry. The average age of residents was 86 years and 68% were female. Ten community-based physicians provided regular care to long-stay residents. Units are not assigned to physicians by specialty and there is frequent cross-over among units, as physicians and partners from their medical groups cover for colleagues on nights, weekends, and vacations. Physicians caring for residents had prior experience with CDSS. They provided care for residents in both intervention and control units.
The CDSS for dose and frequency of medication orders for long-term residents with renal insufficiency was developed by a team of physicians, pharmacists, and informatics professionals. Four categories of alerts were developed and implemented: (1) alerts presenting recommended doses; (2) alerts presenting recommended frequencies; (3) alerts recommending that the drug be avoided; and (4) alerts advising the prescriber that information required to calculate creatinine clearance was missing. In a randomized trial of the impact of the CDSS on the quality of prescribing, the 22 long-stay units of the facility were randomly assigned for prescribing physicians to receive or not receive the alerts.8
During the 12 months of the randomized trial, we captured in an audit file each alert that was displayed to a physician when starting to order a drug for a resident of an intervention unit, as well as alerts triggered by initiation of drug orders for residents in the control units where alerts were not displayed. The audit file captured the drug that triggered the alert, with its dose and frequency and identifiers of the physician and patient. We also obtained data with full details on all drug orders that were submitted to the pharmacy so that we could compare each alert with all drugs actually ordered by that physician for that patient on that day. Thus we were able to identify all changes during the process of drug ordering that may have resulted from viewing alerts or second thoughts on the part of prescribers. We also captured information on serum creatinine tests with dates and results.
To estimate the direct and immediate drug- and laboratory-related financial impact of the CDSS, we compared the initial drug orders that triggered alerts for residents with the drug orders actually submitted for these residents on the day of the alert. We also identified orders for serum creatinine tests that were initiated within 1 day of receiving a relevant alert. For each alert, two research pharmacists reviewed the initial and submitted drug orders and used the information on drug name, dose, and duration to assign unit costs based on the US wholesale price at the time.12
Drug orders were conservatively assigned a duration of 30 days of use for a chronic drug and 10 days for antibiotics. As duration of use of many drugs may vary substantially in this setting, we also performed a sensitivity analysis, assigning durations of use of 90 and 180 days for non-antibiotic drugs. When an alert recommending avoidance of a drug led to that drug not being ordered, the pharmacists identified any drugs ordered for the patient on that day that could have served as substitutes for the drugs that were not ordered, and assigned unit costs to these drugs. Costs for recommended serum creatinine test orders were estimated on the basis of Medicare-allowable payments, which ranged at the time from US$7 to US$13 depending on the specific type of test ordered.
As a preliminary estimate of the direct and immediate impact of the CDSS on costs, we compared the costs for the drug orders as they were initiated with costs for the final submitted drug orders after an alert had been received in the intervention units. In the control units, there were also differences between the initial and final submitted drug orders, suggesting that changes in drug orders during prescribing were not always due to receiving an alert. We do not know the circumstances that led to these changes in drug orders in the control units. We suspect that they varied and in some cases involved a purposeful reconsideration of the initial order before finalization, based on recognition of the resident's level of renal impairment. For example, we observed instances where the dosage of H2 antagonist therapy or quinolone antibiotic therapy was reduced without an alert having been displayed. Importantly, we compared costs for the initial and final submitted drug orders in the control units and used these results to adjust the estimates for the intervention units. We summed the resulting differences within alert categories and across the entire CDSS. For alerts advising the prescriber of missing creatinine values, we calculated the cost of creatinine tests within 1 day of the alert. As a sensitivity analysis, we also estimated the costs and savings using 90- and 180-day durations for non-antibiotic drugs.
To assess the relevance of the study findings for the USA, we extrapolated the cost savings from the study site to the US nursing home setting. Data from the most recent National Nursing Home Survey were used to derive the characteristics of US nursing home facilities including average bed sizes and occupancy rates.13
We calculated cost saving per resident-day in the Canadian long-term care study facility and then extrapolated potential savings to the USA by estimating resident-days using bed size multiplied by occupancy rates. Given the large variation in the size of nursing homes in the USA, we estimated cost savings for facilities based on size: fewer than 50 beds, 50–99 beds, 100–199 beds, and more than 200 beds. We therefore report the potential financial impact within bed-size categories and for the average-bed-size nursing home in the USA. To provide a basis for understanding the potential savings related to different rates of renal insufficiency within a nursing home, we calculated the percentage of the residents in the intervention and comparison units who had a creatinine clearance level of <60 ml/min per 1.73 m2
of body surface area at any time during the year of observation.