Cost Measurement.

Costs will be estimated for each of three scenarios: non-intervention (no added study intervention, based on usual care in the study site), control (the study's control condition I, with physician-initiated nutritional interventions), and the study intervention condition II (with "system" initiated interventions). For each condition, costs will be measured on an annualized basis for each pt. These costs will encompass any initial costs of setting up the condition as well as on-going monthly operating costs. Intervention system, professional, and induced services costs associated with additional hyperlipidemia intervention will all be included.

The non-intervention group will consist of patients of the same study physicians who, in the year prior to the study intake period, would have met the other study eligibility requirements and received an LDL-C measurement, as part of regular clinical care, which was in the highest quartile. While this is not a randomized group and is subject to possible selection bias by the provider and patient in receiving a cholesterol test, it provides a reasonable estimate of a comparable group. We considered the inclusion of a true control group, but this would have introduced an entirely different (and possibly more important) set of biases relating to willingness to be in a study and the effect of obtaining study measurements.

To develop the cost of each of the 2 study conditions and the non-intervention group, we will build on the approaches of Dr. Shepard and co-workers in an ongoing cost-effectiveness study of substance abuse treatment (R01DA08739) and on previous work. (132) All 3 groups entail the cost of identifying patients with high LDL-C as well as determining follow-up services. The cost of identifying a person with high LDL-C will be derived by dividing the total cost of testing by the number of patients eligible for the trial, recognizing that several patients must be tested to find each one with an elevated value.

Both treatment conditions (I & II) require sensitizing physicians to appropriate steps in the detection and management of elevated lipids, performing work-ups to test for hyperlipidemia and providing nutritional counseling.  We will estimate direct cost (the personnel, supplies, materials, and education-related costs) by first determining the cost for delivering and receiving sensitization and training and counseling.  This entails identifying all staff and personnel involved in study-related tasks, and determining salary information and percentage of time performing these tasks. (133)

The costs of counseling services will be determined by multiplying the frequency of each service times its unit cost per contact hour. Frequencies will be obtained from clinical and study records. Unit costs will be based on patient contact and preparation time for each session. For group sessions, the cost of a session will be prorated over the average number of participants, using procedures similar to those in the above-mentioned behavioral health study (DA08739).  Costs of supplies and materials will be estimated by recording the average use of each item per service (e.g., recipes given out at each session).  We hypothesize that, while unit costs will remain identical across groups, the increased frequency of testing and counseling in the study conditions will increase total cost. The cost of the system intervention has two components:  1. The initial cost required to set up and "debug" the system (time and materials), and 2. the ongoing cost each month (time and materials) (entering the appropriate data, ensuring that the resulting medical orders are accurate, properly transmitted, and filed).  Personnel costs of each component will be recorded through logs and time sheets.  

Indirect costs associated with clinical care will determined by employing the existing Transition Systems, Inc. (TSI) accounting system used by UMMC. This system can supply detailed cost data on both an individual and an aggregate basis (see appendix P for output examples). Indirect costs of services supported under research funds will be determined by applying the institution's current indirect rate approved by the U.S. Department of Health and Human Services.

Cost-effectiveness analysis.

The second modeling step will use the risk model of Taylor, Pass, Shepard and Komoroff, which now operates on a personal computer. (93) It estimates the change in discounted quality-adjusted life expectancy as a function of the change in the risk factors for CHD (total cholesterol, HDL-C, smoking and blood pressure), using available data from the literature relating risk factor change to clinical endpoints. The change in the first two factors will be estimated as the difference between the actual levels for a patient at the end of the study, compared to the values predicted if they had been in a different study group. For no intervention, the change in risk factors will be small (less than 2 mg per dl), about equal to the change in cholesterol observed in the control group in WATCH. Thus in the comparison of no intervention versus condition II, the net cost is the difference in annualized cost per pt of no intervention (initial detection and often minimal subsequent follow up) compared to the annualized cost per patient of condition II (initial detection and expected substantial subsequent follow up).  The net gain in QALYs is the difference in the average expected gain under no intervention (based on the minimal change in risk factors) compared to the expected gain in condition II (which is hypothesized to be large). The cost-effectiveness ratio represents the cost of achieving a year of good health. The lower this ratio, the more cost-effective the intervention.  In previous studies of primary care, Dr. Shepard has estimated that nutritional counseling was likely to be reasonably cost-effective, whereas cholesterol lowering drugs, because of their very high cost, were not very cost-effective. (93)