A total of 333 patients were recruited during consultation at each of the six selected facilities for exit interviews, and 259 patients were successfully administered an exit interview. The final rate of follow-up (for patients participating in the full costing study) was 84% and is shown in . No patients or clinicians refused the initial consultation observation. At the 5% significance level, patients in RDT facilities were no more likely to attend the exit interview than those in control facilities; after stratifying by the portion of the patient population more than five years of age, differences were less significant. Patients who were lost to follow-up were not significantly different from those who were successfully re-interviewed based on several demographic measures ( and ). No significant differences were found for age distribution, patient sex, method of travel to and from the health facilities, or occupation of the patients' head of household. Additionally, we found no significant differences for the same set of measures between facilities that offered RDTs and those facilities that did not offer RDTs ( and ). Unfortunately, patients who were more than five years of age were significantly more likely (29% versus 15%) to leave the health facility before completing the exit interview.
Losses to follow-up during the study, Tanzania.
Comparability of control and experimental populations, and those lost to follow-up, Tanzania*
Comparability of control and experimental populations, and those lost to follow-up, Tanzania*
Of the 259 patients who were administered exit interviews, 178 were interviewed at experimental facilities (with RDTs) and 81 were patients at control facilities (no RDTs). Within the RDT facilities, patients were significantly less likely than in control facilities to receive results for a laboratory test for malaria (84% versus 95%; P = 0.009, by Fischer's exact test), a difference that was also significant in patients ≥ 5 years of age (86% in RDT facilities versus 98% in control facilities; P = 0.04) but not in children less than five years of age (82% in RDT facilities versus 92% in control facilities; P = 0.13). Patients in RDT facilities were also significantly less likely to have a positive test result for malaria parasites (14% versus 43%; P < 0.001). Although in control and RDT facilities large fractions of all patients received laboratory diagnosis, clinicians in the RDT facilities appeared to be more parsimonious in their use of tests, at least among adults.
Adults were significantly more likely than children less than five years of age to have positive results for malaria in control facilities, but not in RDT facilities (55% versus 28% in control facilities; P
= 0.016 and 15% versus 13% in RDT facilities; P
= 0.69). These results clearly confirm the problem with microscopic examination, and the low quality of routine microscopy was confirmed in more detailed studies in the same facilities.3
Patients within RDT facilities were significantly less likely to receive the first-line anti-malarial drug artemether/lumefantrine (ALU) (Coartem™; Novartis, Basel, Switzerland) (ALU) compared with patients in control facilities. This finding was seen when the analysis was restricted to patients who left the facility with any drug prescription and all patients observed: 12% versus 52%; P < 0.001 (with any prescription) and 10% versus 48%; P < 0.001 (all patients). This difference remained highly significant regardless of the age of the patient.
Patients in RDT facilities were also more likely to receive ALU in correct correspondence with the results of their diagnosis. When only patients with a laboratory diagnosis were examined, those in RDT clinics received ALU in correspondence with the laboratory diagnosis 95% of the time versus 82% in control facilities (P = 0.002). Because a related study has shown that most positive blood slide results in control facilities are false-positive results (Kahama J and others, unpublished data), it follows that high clinician compliance with microscopy results leads to overuse of anti-malarial drugs. Patients less than 5 years of age appeared no more likely to be correctly prescribed ALU than patients ≥ 5 five years of age (93% for patients < 5 years of age versus 88% for patients ≥ 5 years of age; P = 0.17). Differences remained statistically insignificant when restricted within either RDT facilities or within control facilities.
Implementation costs of RDT program (provider).
Cost data on implementation was collected over a 14-month period. During this period, approximately 435,400 RDTs were issued to implementing facilities, and use data indicated that approximately 330,000 RDTs for malaria had been performed. Because of this high volume of tests, the cost of implementation training and support for RDT rollout was relatively low when considered per test. The total cost of the RDT intervention over this period (not including the test kits) was estimated to be $16,946 in 2008 USD or $1,883 USD per implementing facility. Thus, we estimated that the cost of implementation per RDT (excluding the test kits themselves) was between 0.04 USD and 0.05 USD. The test kits themselves were estimated to cost USD 0.66 each. When calculating the cost per patient in RDT clinics, we include the cost of RDT implementation.
The bulk of the expenses went to staff salaries for the implementation of the RDT rollout (72%) and for training and quality control at the implementing facilities (22%). The only other substantial line item cost was transport, which accounted for 3% of the total cost of implementation.
Patient perspective: Direct costs (expenditure).
Patient costs consist of two main parts: direct costs due to expenditure on medicines, transport, diagnostics, or other health services, and indirect costs, such as lost productivity or the opportunity cost due to time spent seeking care. We attempted to measure direct costs and indirect costs.
Patient expenditures were directly reported by patients. shows arithmetic mean expenditure per patient in RDT or control facilities arising before and during the first consultation, and after the first consultation for the subset of patients with follow-up. Expenditures have been subdivided into several categories, and are reported in TSH and USD.
Patient expenditures, Tanzania*
shows that significant differences in reported expenditure were found between patients at RDT clinics and those at control clinics. Patients' mean total expenditures were lower in RDT clinics (USD 1.02) compared with control clinics (USD 1.33), and were significantly different by the Kruskal-Wallis test for equality of populations. Patients' mean expenditure on drugs was 0.36 USD lower in RDT clinics than in control clinics.
shows bootstrapped means and bias corrected confidence intervals for each of the parameters shown in . Each estimate is based on 10,000 re-samples of the observed data.
Results of non-parametric bootstrap for confidence interval estimation of patient expenditures, Tanzania*
Arithmetic mean patient expenditures, when reduced into smaller component parts, failed to show significant differences in all but the line item expenditure for drugs at the first health facility visit, which was highly significantly different in RDT clinics (TSH = 464 [USD = 0.38] versus TSH = 902 [USD = 0.74]; P = 0.002, by Kruskal-Wallis test). However, bootstrapped confidence intervals showed that the difference was only close to statistical significance.
The similarity of expenditure across the two types of facilities helped to support our assumption that the populations of patients in control and RDT facilities were similar because the cost of transportation and actions taken before attending the health facility would not be expected to be significantly different between the two groups. Furthermore, it supports the argument that effects on patient expenditures were largely limited to those on drug purchases. Expenditures on drugs at the health facility accounted for the largest single component of patient expenditure, followed by laboratory fees and travel costs.
Patient perspective: Indirect costs.
Additionally, patients incurred indirect costs through lost income, reduced productivity, and the opportunity cost of lost time caused by attending the facility either as patients or as caretakers of patients. One hundred eight (42%) patients or caretakers reported missing work to attend the health facility. Of that group, 85% reported lost income as a result. Neither result was significantly different at the 10% level between RDT and control facilities (P = 0.16, degrees of freedom = 1 and P = 0.66, degrees of freedom = 1). Among those reporting lost income, mean lost income was reported as 7,175 TSH (5.87 USD), a figure that was not significantly different between control and RDT groups (P = 0.16, by Kriskal-Wallis test). This figure is significantly larger than patients' expenditures on all other categories. For patients who lose income to attend the facility, the opportunity cost of facility attendance is far larger than the direct costs of health care and such large opportunity costs might prevent significant numbers of persons from accessing care.
Total time per visit, including transportation time, was measured by adding estimates of time at which patients or caretakers left their home or work place to attend the facility to the time they spent at the health facility (determined by the time of the start of their exit interview), with an additional time factor added for their estimated time to return home. In control clinics, mean time per visit was estimated to be 4.7 hours, and in RDT clinics, it was estimated to be only 4.0 hours (t = 2.8703, P = 0.005). Thus, being a patient in an RDT clinic in our sample was associated with approximately 42 minutes shorter total visit time. Although a certain amount of this variation can be attributed to slightly shorter travel times to RDT facilities (mean travel time = 35 minutes) compared with control facilities (44 minutes; P = 0.058), we observed a mean difference of approximately 9 minutes of travel in each direction, or a total of 18 minutes. Reduced waiting times and total visit times might help to reduce the opportunity costs of facility attendance and thus could improve access to care, although reductions seen are small (< 10%) in relation to total visit time.
In this analysis, we focus on gross provider economic costs and not net costs, which would account for the collection of user fees by health facilities. lists the costs that were included in the analysis.
Costs included in provider perspective analysis, Tanzania
shows the results of the provider perspective analysis for the RDT and control facilities. The table shows the results of non-parametric tests for each of three sub-divisions of total provider costs. These costs are analyzed either within control or experimental facilities. Drug costs represent the cost to the provider of all drugs and prescription provided to a given patient. Facility cost is the cost of the commodities whose use is measured at the facility level but not linked to specific patients (overhead, staff costs, equipment, and general consumables, excluding drug costs). Thus, there are only six observations corresponding to the number of facilities in our study. Total costs include drug costs, facility costs, and other marginal costs, including RDTs and consumables, which are patient specific rather than general consumables used for all patients. Significant differences were found for all costs except facility cost, although the latter result is compromised by the extremely small sample size.
Provider costs per patient, Tanzania*
When patients who attended RDT clinics were compared with patients who attended control clinics, drug costs were significantly lower for patients who attended RDT clinics (USD 1.28 versus USD 1.71; P = 0.014). However, total provider costs were higher for patients who attended RDT clinics (USD 3.63 versus USD 2.32; P < 0.001 for total cost).
We were again confronted with results that were highly non-normally distributed, including in some cases a large zero mass and in all cases a significantly right-skewed distribution. Thus, we estimated confidence intervals by using non-parametric re-sampling (bootstrapping) with 10,000 re-samples for each outcome, excepting facility cost ().
Results of non-parametric bootstrap for confidence interval estimation of provider economic costs, Tanzania*
Bootstrapped confidence intervals generally confirm the results of the Kruskal-Wallis tests. However, there are important differences between the two results. The confidence intervals for drug costs between RDT and control facilities show a large overlap when analyzed for all age groups. However, when the sample is stratified by those < 5 years of age and those ≥ 5 years of age, a significant difference exists for patients ≥ 5 years of age. This finding may be the result of a combination of high rates of ALU prescriptions in control facilities and the higher cost of this drug for adults in relation to other adult drugs. Once facility costs are included, the total cost of treating a patient, including all provider costs, is significantly higher in RDT facilities than in control facilities.
Summary of results.
The results indicate that in the presence of RDTs, drug cost savings are likely to accrue to patients, and may also accrue to the providers, especially for adults. However, whether these savings translate into overall cost savings is unclear. For patients, it is likely that there is some reduced overall spending when RDTs are available. However, the savings is small (USD 0.36) and it represents only a small component of the total economic costs to patients.
For providers, the drug cost savings is of a similar order (USD 0.43) as a result of RDT introduction. Unfortunately, these savings appear to be too small to offset the entire cost of RDT introduction and use. Thus, it appears that RDTs may increase the cost of treatment per patient in public facilities, despite reducing anti-malarial drug use and creating drug cost savings for the health system. Additionally, the cost savings arise largely from reduced anti-malarial use among adults who are most likely to be charged a user fee for drugs. Thus, the resulting reduction in user fee revenue caused by reduced patient drug expenditure will reduce the financial incentives for RDT implementation.