|Cost-Effectiveness Tool for Evaluating Interventions to Prevent Mother-to-Child Transmission - Manual and Model (UNAIDS, 2000, 94 p.)|
Variable costs are costs that vary with the volume of service. These are primarily the cost of personnel and disposable supplies. Variable costs are likely to constitute at least 50% of total VCT cost. Using the detailed costing approach, resource consumption is determined by identifying, measuring and valuing all cost elements of the variable cost portion of VCT.
Personnel. This is the amount of staff time required per client for each task. The costing of personnel resources entails calculating the average length of time spent on each VCT activity. This time estimate is then multiplied by the wages of the personnel needed to provide the specified activity. For services provided to groups, such as most pre-test counseling, the personnel costs are divided over the average size of the group. Activities to be costed include: seeking consent for pre-test counseling for the HIV test; entering clients name and demographic information into project records; administration of the HIV test; pre- and post-test counseling; and laboratory work. Remember, even if the client drops out of the process before the intervention itself, the time staff spends on that client is a real cost. The CET performs the arithmetic entailed in these calculations. The users task is to enter reliable information on parameters such as the wage rates, length of time required for each task, and on the proportion of women who drop out of the VCT cohort at each stage of the process.
Supplies. The HIV test kit is usually the most important item in this category. This includes the kit for the initial test (e.g., ELISA) and the kits for confirmatory tests, which may be more expensive (e.g., Western blot). These should be costed according to the price the agency actually pays. In most cases this will be a bulk purchase price available through a non-profit agency, not the local retail price for this item. To the bulk price should be added the cost of transporting it to the point of use.
This table provides space for entering detailed information on the items needed to calculate the variable costs. It includes information on each significant cost item and the number of women who receive each item. Several of the information items ask for very specific figures on the percentage of women who drop out of the treatment group at various points in the VCT process. Column E provides figures that convert the percents that you enter into actual numbers of women. In this way you can match the percentages entered in the CET with the actual numbers of women observed in your program.
Double-check cell C5: External benefits of VCT
Remember to check the value in cell C5 for external benefit of VCT. This number adjusts the cost of VCT used to calculate cost-effectiveness. We recommend that you use the conservative default value of 0% for the base-case analysis.
Influence on cost-effectiveness: High. Variable costs are likely to be 50% or more of total VCT costs.
Expected effort of data collection: High. Since variable cost estimates have an important influence on cost-effectiveness and because relevant data should be readily available from project documents and personnel this should be considered a high priority.
C35. Prior testing. In cell C35 enter information on the percentage of women who register for VCT who are known to be HIV-positive from previous testing. Only those women who do not require further counseling and testing should be included. Data sources: Document indicating prior positive test results.
C36 and C37. Test characteristics. Enter the false positive rate for the initial and confirmatory tests you plan to use. The false positive rate is 1- specificity. The ELISA test has a specificity of 99.5% (Phillips, 1994). We have entered 0.5% in both C36 and C37 as the default values. Mislabeling of test results and other errors of clerical or lab technician are probably a greater source of error than the technical limits of the tests, this is hard to quantify and will vary from setting to setting. Program managers should in any case conduct periodic assessments to ensure that the error rate is very low.
C40 - C42. Wages per hour including benefits. Enter the average hourly compensation rates for clerical staff, counselors and lab staff. This should include both wages and benefits calculated on an hourly basis. Since pay is usually provided on a monthly basis, the hourly rate can be derived by dividing the average monthly wage by the number of days worked per month multiplied by the average number of hours per day. Other benefits such as health insurance and paid holidays should be added to the monthly wage. The formula for calculating the hourly wage would then be:
(Monthly financial compensation + monthly value of benefits)
(Average days worked per month x average hours worked per day).
Clerical staff here refers to those who assist with managing the paperwork involved in processing the daily caseload. This would be primarily registration and other patient record management and perhaps collecting patient fees, issuing receipts, etc. It would not include higher-level financial, administrative, or supervisory activities. Data sources: Project financial documents.
C45 - C48. Test kits and other cost items. Enter the per-unit cost of the initial test kit; the confirmatory test kit; and blood-draw tubes in cells C45, C46 and C47 respectively. In C48, enter the cost per woman of ancillary service such as transportation and meals if these services are needed to maintain the VCT caseload. Data sources: Project expenditure documents.
C50 - C51. Cost sharing. Enter average per-patient revenues from registration fees or other sources (if any). This figure can be obtained by dividing the monthly cost-sharing revenue by the number of women who register for VCT. Data sources: Project expenditure documents.
C58 - C61. Screening to identify high-risk women. Rather than counsel and test all women some programs may conduct an initial screening to identify high-risk women. The purpose of this screening is to reduce the resources used for pre-test counseling and testing of those who are HIV-negative. This strategy is most sensible in areas with very low prevalence of HIV. Since it inevitably means that some HIV-infected women will be inadvertently screened out, careful planning is required to hold these occurrences to a minimum. If this program does conduct an initial screening enter Yes in cell C55. Entering Yes unmasks cells C58 through C61.
NOTE: Data entry for cells C58 through C61 is only required if this program conducts an initial screening to identify women at high risk of HIV. This is only indicated if Yes is entered in C55. If this program dies not conduct an initial screening enter No in C55 and proceed to cell C64.
C58. Percent who decline initial high-risk screening and therefore drop out of VCT cohort. This is the attrition (drop-out rate) of women who refuse the high-risk screening and are therefore no longer in the counseling and testing cohort. Data sources: By examining patient records, project staff can tabulate the number of women who reported for antenatal services and received a referral for the initial high-risk screening for MTCT. The percent who decline the screening is then calculated by comparing this figure (number of women available) with the number who actually registered for the screening. This comparison should be carried out over five randomly selected days in order to ensure an adequately precise estimate.
C59. Minutes per woman needed to conduct the initial screening. If the screening is conducted in a group, this would be the length of time required to process each group divided by the average number of women per group. Data sources: Direct observation on three randomly-selected days.
C60. Percentage of women who were in fact found to be at high risk. This is the portion of women of all women screened who are determined by the screening criteria to be high risk. Only these women would then proceed to VCT. Estimates should be based on staff observation of the average number of minutes required for each high-risk screening session and the average number of women in each session. Data sources: Direct observation on three randomly-selected days.
C61. Number of women who were in fact HIV-infected but were excluded by the initial screening. The proportion of undetected positives inadvertently screened out of treatment in each program can be determined by periodic sampling of those screened out and offering to test them. In general, as the value in cell C60 goes down (screening criteria more stringent and higher proportion of positives in cohort screened in for services), the value in cell C61 goes up (more undetected positives screened out of services). Apart from its value in calculating cost-effectiveness, this figure is very important in ensuring that the screening process is not screening out an unacceptably large number of HIV-infected women. Data sources: Estimating this input requires periodic random sampling of those screened out and offering to test them to determine how many are in fact HIV-infected. Since this is an important parameter for reasons additional to the cost-effectiveness analysis, we recommend that this check should be carried out on an ongoing basis on one randomly selected day each month. Ten women should be randomly chosen on each of these days.
Cell C64 - C67. Registering for VCT. What appears in these cell range varies according to whether initial screening (C55) is offered or not. Please follow the instructions that correspond with your situation. Use the instructions in the first box Registering for VCT-I (below) if no initial screening for high-risk women is offered and the second box Registering for VCT-II on page 32 if initial screening is provided.
Registering for VCT - I
C64. Percent of those known to be HIV-infected but who decline to register. Another point of possible attrition, this is the proportion of women known to be positive from previous testing who arrived at the clinic for the MTCT intervention but who then declined to register. We expect that this number would be very small, particularly since, being already documented as HIV-infected, they would not have to be tested again. Using patient registry data, calculate the percentage of women referred for VCT who (1) are known from previous tests to be HIV-infected, but who (2) decline, for whatever reason, to register for VCT. Data sources: This information is probably available from client registry data. Since the number of women with known HIV+ status is likely to be small, a sample of 50 known HIV-infected women who received antenatal services should be adequate. If possible, this sample should be dawn from two or more time periods separated by 1-2 months.
C67. Time in minutes required to register women of unknown HIV status for VCT. Data sources: As in C66, observe patient flow directly. A sample of 15-20 women is adequate.
Registering for VCT - II
C70. Percent of women receiving ancillary services. Percent of women who register for VCT who received ancillary services. If no ancillary services are provided, set this to 0%. This is the default value. Data sources: Review project records of the number of clients who receive ancillary services in a month. If family members of a client receive ancillary services such as transportation this should also be counted as services to that client. Calculate the percentage of all clients seen during that month that this constitutes. Since ancillary services may vary by time of year, collect this information during different seasons if possible. Average the results over the number of months that you collected data. A total sample of 100 clients should be adequate.
Cell C73 - C79. Pre-test counseling.
C73. Percent registered for VCT but who do not accept pre-test counseling. Data sources: Requires direct observation of the number of clients who drop out before pre-test counseling. Compare the number of registrations in a given day with a count of those who complete pre-test counseling. Completing this count over five randomly selected days would give a reasonable estimate. In most cases this number will be quite small and zero is the default.
C74. The number of women who receive pre-test counseling in one session. For individual counseling this will of course be 1; but usually group counseling will be employed and the number should reflect the average size of the pre-test counseling group. In most cases we expect that a maximum number is set by policy. However, if patient flow is lower than expected, sessions may be conducted with less than this maximum. Data sources: Counting the actual number in each session over five randomly-selected days would provide an adequate basis on which to take an average. This may be an input that is sensitive to seasonal fluctuations. If possible, repeat this exercise during normal, low and high caseload seasons.
C75. Minutes required for pre-test counseling session. Data sources: Whether it is individual or group, time the actual length of the counseling sessions on five randomly selected days and take the average of all sessions.
C76. Percent of women who completed pre-test counseling but do not accept blood draw for initial HIV test. Another point of possible attrition. Data sources: This input can be calculated by asking staff to count the women who accept blood draw over the course of a randomly selected day and comparing this with the number who completed pre-test counseling on that day. Repeating this exercise over five randomly selected days should be an adequate basis on which to take an average percentage.
C77. Minutes required for blood draw. Data sources: Direct observation of 10-15 clients who are having their blood drawn on two randomly selected days should give an adequate estimate.
C78. Number of initial tests processed in a batch. There may be a maximum determined by the type of equipment in the lab, but depending on patient flow the number actually processed in a batch could be lower. Data sources: The lab staff can count the number per batch over five randomly selected days. If seasonal variation in patient flow is an issue, the sampled days should include days in different seasons.
C79. Minutes of lab technicians time for each batch of initial tests. Data sources:
Project staff should time the number of person-minutes needed to process each batch of initial tests. This should include the time between delivery of samples to the lab and delivery of the results to the counselors. Observations should cover a full day on two randomly-selected days.
Cell C82 - C85. Confirmatory test
C83. Percent of women who test positive on initial test who do not take the confirmatory test. If no confirmatory tests are provided, enter No. The remaining data entry cells in this section are then masked. In the vast majority of cases however, VCT programs will provide confirmatory tests to those who tested positive on the first test. Data sources: Over five randomly-selected days, staff should compare the number of women who test positive with the number who take the confirmatory test. Because women may return for the confirmatory test either before or after they are asked to return, staff should keep track of those who return by name or ID number over a two week**(?) period. This will ensure that results include all who return, not only those who return at the appointed time.
C84. Number of confirmatory tests processed in a batch. See explanation for C78 above.
C85. Number of minutes of lab technicians time for each batch of confirmatory tests. See explanation for C79 above.
C89 - C92. Post-test counseling. Different lengths of time for post-test counseling and different sized groups can be entered for HIV-negative versus HIV-positive women. The default assumption is that HIV-negative women will be counseled in the same sized groups at post-test as at pre-test but that HIV-positive women will be counseled on an individual basis at post-test.
C89. Number of women HIV-negative women who receive post-test counseling in one session. For individual counseling this will of course be 1. Data sources: If the number per session is variable, counting the actual number over five randomly-selected days would be provide an adequate basis on which to take an average. This may be an input that is sensitive to seasonal fluctuations. If possible, repeat this exercise during normal, low and high caseload seasons.
C90. Minutes required for post-test counseling session for HIV-negative women. See explanation for C75 above.
C91. Number of women HIV-infected women who receive post-test counseling in one session. See explanation for C89 above
C92. Minutes required for post-test counseling session for HIV+ women. See explanation for C75 above.
C95 - C96. Registration for MTCT intervention. These cells record the number who may drop out at the final point of possible attrition, namely at the time of registration for the MTCT intervention itself. We assume that registration for MTCT is part of post-test counseling in the case of those women with previously unknown HIV status.
Registration for MTCT Intervention - I
C95. Percent of women of previously unknown HIV status who completed post-test counseling who do NOT register for MTCT intervention. Final point of attrition. Data sources: This can be calculated by comparing the number of women who complete VCT and are referred to the MTCT intervention(s), with the number who actually register. This input, which determines the uptake rate for the interventions, can have an important effect on cost-effectiveness. We suggest you take the average percentage over 10-15 randomly selected days. Care needs to be taken only to include women who were not known to be positive from previous testing. This could perhaps be done by designating a space in the registration books and forms that record whether each client arrived at the facility with known HIV+ status from a previous test.
C96. Percent of women of previously known HIV-positive status who completed post-test counseling who do NOT register for MTCT intervention. See explanation for C95 above.
Registration for MTCT Intervention - II
C95. Percent of women who complete post-test counseling who do NOT register for MTCT intervention. The final point of attrition. The sheet is now set to reflect the provision of an initial screening to identify high-risk women (C55= Yes). All women who register for MTCT therefore completed this initial screening. We assume that in the low prevalence settings where initial screening for high risk make sense, very few women will have known HIV-positive status, and we do not track them separately.
D101 - F102. Total annual variable costs: Result and ranges. The blue cells E101 and E102 contain the results of the calculation of variable costs based on the data you have provided. (For more detail on how these results were derived, see the calculations in cells I45 - V62). In cells D101 and D102 you are asked to enter low-end estimates for supply and for personnel costs respectively. Cells F101 and F102 require high-end estimates for these same variable. These ranges will be used to define the lower and upper end of the ranges used in the sensitivity analyses to examine how cost-effectiveness varies with the cost of VCT.
C107 - F107. Annual number of women who complete VCT. Given the number of women who give birth in this service area; the portion of these women reachable by the program; HIV prevalence; and the number of women who drop out of the VCT process at various stages, cells C107, D107 and E107 display the number of HIV+, HIV-, and combined HIV+/HIV- women who complete VCT and register for the ARV/substitute feeding intervention. F107 shows the percentage of all women who register for VCT who do not complete the VCT sequence and register for the MTCT intervention itself.