INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Ensuring the completion of treatment for patients with tuberculosis (TB) is the highest priority of TB control (1). The use of directly observed therapy (DOT) is the most effective case management strategy to achieve this objective since a health care worker observes every dose of medication. A recent review of the literature suggests that the use of DOT reduces treatment non-completion by 77% (2). Many experts have argued that all patients with TB should be placed on DOT because, compared with self-administered therapy (SAT), DOT not only cures more patients but also generates net cost-savings (3, 4).

The cost-effectiveness of DOT in the United States has been examined in two studies (5, 6). These studies demonstrated that in populations at high risk for treatment failure (21%) or default (40%), DOT is more cost-effective than SAT. These results support advocacy for greater and perhaps universal use of DOT in the United States.

Despite arguments for universal DOT, only 30% of reporting areas in the United States reported that more than half of their patients received DOT throughout their course of treatment (7). In California, where in 1995 only 31% of patients received DOT for the full course of treatment (8), most health departments are targeting DOT to persons at higher risk for treatment nonadherence. Advocates of this approach argue DOT is more expensive to administer than SAT, and they have limited resources. Some also argue that universal DOT is not needed to achieve high treatment completion. In fact, the degree to which DOT is used does not always correlate with treatment completion rate (9). Many health departments, including for example the program in San Francisco, have very high treatment completion rates while selectively using DOT (9, 10).

In California, DOT is recommended for persons at high risk for treatment nonadherence (11). In many health departments, a patient without risk characteristics such as substance abuse, homelessness, or incarceration is not routinely provided DOT. Increasing the use of DOT in these persons is clearly indicated if their treatment default rate is high. But we recently estimated that only 1.7% of these patients defaulted from treatment (California Department of Health Services, Tuberculosis Control Branch, unpublished data). With such a low default rate, we questioned whether DOT should be expanded in this low-risk group.

To inform the decision on expanding DOT to persons at low risk for treatment default, we sought to determine the cost-effectiveness of this approach. First, we determined the incremental cost per additional case cured from using DOT in a patient population at relatively low risk for treatment default. Then we identified the major determinants of DOT cost-effectiveness. In essence, our analysis allowed us to model the cost-effectiveness of moving from selective DOT (defined as DOT for patients at high risk for patient default) to universal DOT (including DOT for patients at low risk for patient default).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We determined the number of additional cases cured and the incremental cost per additional case cured for the DOT strategy in comparison with SAT. Our analysis was restricted to 1,377 patients reported in California during 1995 with the following characteristics: at least 15 yr of age; no history of HIV infection; disease without documented resistance to isoniazid, rifampin, and pyrazinamide; antituberculosis treatment was entirely self-administered; no history of injection-drug use, non-injection-drug use, homelessness, and incarceration as reported on the Report of Verified Case of Tuberculosis (RVCT) (12). The latter set of risk characteristics formed our definition of low risk for treatment default. The default rate for this cohort was 1.7%.

Outcome Probabilities

To estimate the number of cases cured by each treatment option, we adapted a decision-tree model developed by Moore and colleagues (5). The model requires data on the number of patients who completed therapy, defaulted, or died prior to therapy completion, stratified by therapy administration type (DOT or SAT). For our analysis we used treatment outcomes on the RVCT (12). We used actual treatment outcomes for the 1,377 patients who received SAT, then estimated what their treatment outcome would be had they received an effective DOT plan. We assumed that DOT would not reduce the risk of death while receiving therapy. Among patients who survived initial therapy, DOT would reduce the risk of defaulting by 76.7% (2).

The probability of patient cure, relapse (drug-sensitive and multi-drug-resistant [MDR]), and death after default was determined from the model developed by Moore and colleagues (5) (Figure 1). RVCT data were analyzed with SAS (13). The cost-effectiveness analysis was developed on a Microsoft Excel spreadsheet (14).

View larger version (22K): [in this window] [in a new window]   Figure 1.   Decision-tree of therapy outcomes. The first branching node after therapy initiation is completion of therapy, default, or death before therapy completion. The next branching node for those who survive initial therapy is cure versus relapse. Probabilities of initial and final therapy outcome, stratified by therapy administration type, are in parentheses. SAT = self-administered therapy; DOT = directly observed therapy.

Cost Analysis

All costs are reported in 1998 U.S. dollars. Costs were adjusted to their 1998 values using the medical care services component of the U.S. Consumer Price Index. Costs are analyzed from the perspective of the health care system; patient costs and productivity losses are not included. We included only costs expected to differ by therapy administration type, which included the variable cost of DOT, initial drug regimens, incentives and enablers, diagnosis and treatment of relapse, contact tracing and treatment of contacts. The cost of initial diagnosis and hospitalization for TB was not included because it is assumed that these will not differ by therapy administration type.

Initial treatment cost. Costs of initial treatment that vary by therapy administration type are the incremental costs attributable to the provision of DOT (Table 1). We assume that for ambulatory therapy, patients receiving SAT will ingest drugs daily for 24 wk, whereas patients receiving DOT will receive a daily regimen for 2 wk, followed by 22 wk of a twice-weekly regimen (15). Because incentives and enablers have been shown to be important to ensuring DOT effectiveness (16), we assumed that patients receiving DOT need an additional $25 per week for transportation and food expenses (20). Among patients who complete therapy, DOT initially costs $1,386 more than SAT, and for patients who do not complete initial therapy (whom we assume receive one-third of the recommended treatment), the initial incremental cost of DOT is $462 (Table 1).

Cost of diagnosis and treatment of relapse. The components involved in the cost of diagnosis and treatment of relapse include diagnosis, outpatient treatment, hospitalization cost, excluding physician charges, and cost of physician visits while hospitalized (Table 2). We stratified costs by drug-susceptible versus MDR cases. As did Moore and colleagues (5), we assumed that 50% of relapse would occur in the first year after initial treatment, and 50% would occur in the second year, and discounted these costs at 4% annually. We assumed that relapse cases have the same proportion hospitalized and the same cost of treatment as newly reported TB cases. Using data from the CDC Cost of Hospitalization Study (Dr. Zachary Taylor and Suzanne Marks), we determined that 37% of patients with low-risk TB in California had a TB-related hospitalization.

Cost of contact tracing and treatment. We assumed that contacts to relapse cases had the same prevalence of TB infection and disease as contacts to newly reported cases in California. Costs were stratified by whether the relapse case was drug-susceptible or MDR (Table 3). On the basis of Program Management Reports from local health jurisdictions in California, we estimate that the average TB case has 10 contacts, of whom 35% require treatment for tuberculous infection and 1% require TB treatment.

Contacts to drug-susceptible cases with tuberculous infection are assumed to need a 6-mo course of daily isoniazid (300 mg) (15). TB infected contacts to cases with MDR-TB receive daily pyrazinamide (2,000 mg) and ofloxacin (400 mg) for 6 mo (21).

Sensitivity Analysis

We conducted a one-way sensitivity analysis to determine which parameters substantially impacted our results. The following parameters were varied from their baseline values: probability of default on SAT (0%, 40%), DOT effectiveness in preventing default (50%, 100%), the relapse rate after completing therapy on SAT (from 1.5% to 6%), the proportion of contacts with active disease (0%, 1.5%), the proportion of patients with a TB-related hospitalization (10%, 100%), and the cost of hospitalization for MDR-TB ($20,000, $200,000). We also estimated cost-effectiveness from the program perspective, which excludes hospitalization costs from the analysis. A threshold analysis was conducted on the percent defaulting SAT and relapsing after completing SAT to determine the rate at which the DOT strategy becomes cost-saving.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Actual and Expected 1995 Outcomes for Patients Receiving SAT and DOT

Of the 1,377 patients who received SAT and included in our analysis, 1,279 (92.9%) completed therapy, 23 (1.7%) defaulted, and 75 (5.4%) died prior to therapy completion. Our model indicates that had these 1,377 patients received DOT, 1,297 (94.2%) would complete therapy, five (0.4%) would default, and 75 (5.4%) would die prior to completing therapy.

Expected Final Outcomes of Therapy

Using the SAT strategy, of the 1,377 patients who began therapy, 1,250 (90.8%) would achieve cure, 45 (3.3%) would relapse with drug-sensitive disease, and seven (0.5%) would relapse with MDR-TB. With the DOT strategy, 1,281 (93.0%) would be cured, 19 (1.4%) would have drug-sensitive relapse, and one (0.07%) would relapse with MDR disease. With the DOT strategy, 31.1 additional patients would be cured as compared with the SAT strategy.

Incremental Cost-effectiveness of DOT

For the 1,377 patients treated, the DOT strategy would initially cost $1,834,276 more than the SAT strategy, or an average of $1,332 per patient treated. However, DOT would save a per patient treated average of $347 in treatment of relapse cases and $66 in contact investigation and treatment, for a net incremental cost of $919 per patient treated, and $1,265,063 for the entire cohort. The net incremental cost per additional case cured is $40,620 (Table 4).

Results of Sensitivity Analysis

The sensitivity analysis revealed that in patient populations with a SAT default rate of 40%, the DOT strategy would be cost-saving and save $2,160 per additional case cured (Table 5 and Figure 2). DOT becomes cost-saving at a SAT default rate of 32.2%. If the relapse rate after completing SAT is reduced to 1.5%, the DOT strategy would cost $307,862 per additional case cured. However, if the relapse rate after completing SAT increases to 6%, DOT would cost $11,182 per additional case cured (Table 5 and Figure 3). DOT becomes cost-saving at a relapse rate after completing SAT of 9.2%.

View larger version (13K): [in this window] [in a new window]   Figure 2.   Plot of the effect of variation of the percent of patients defaulting on self-administered therapy (x-axis) on the incremental cost of directly observed therapy per additional case cured (y-axis).

View larger version (15K): [in this window] [in a new window]   Figure 3.   Plot of the effect of variation of the percentage of patients who relapse after completing self-administered therapy (x-axis) on the incremental cost of directly observed therapy per additional case cured (y-axis).

From the TB control program cost perspective, which excludes hospitalization costs, the DOT strategy has a net cost of $1,168 per patient treated, and costs $51,656 per additional case cured.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our results indicate that, in a population at low risk for defaulting from antituberculosis treatment, the use of DOT does not result in cost-savings. Given a default rate of < 2%, the use of DOT costs more than $40,000 per additional case cured when viewed from the perspective of the health system. By reducing the number of treatment failure and relapse cases, the use of DOT lowers overall hospitalization cost. But since these savings are not usually credited to the TB control program, the cost of DOT from the TB control program's perspective is actually more than $50,000 per additional case cured. These results differ from those in prior studies, which were analyzed in populations with much higher treatment failure or default rates (21 and 40%, respectively) (5, 6).

Our results are robust over a range of treatment cost for TB. Because inpatient cost accounts for much of the treatment cost for TB, we varied the percentage of patients requiring hospitalization when they relapsed as well as the cost of hospitalization for MDR-TB. A large variation in these percentages did not alter our conclusions. Even an increase in the effectiveness of DOT did not result in cost-savings.

Although these results provide an economic justification for using selective instead of universal DOT in California, the following issues should be considered when interpreting and applying the results. First, these results should be considered within the context of available resources to a TB control program. Because we have assumed that DOT does improve treatment quality and completionand most experts agree with thisthe use of DOT will always reduce the number of defaulters and increase the number cured. Therefore, if resources are available, or if new resources can be garnered, health departments should not hesitate to put low-risk patients on DOT. But if TB resources are fixed, the benefits of various TB control activities, including the provision of DOT to low-risk persons, have to be compared. Universal DOT may not be the best use of existing resources.

Health departments should continue to seek funding for DOT whenever possible, especially when such funds are in addition to fixed categorical TB budgets. In this regard, states have the option, under the federal Omnibus Budget Reconciliation Act of 1993, of establishing Medicaid reimbursement for DOT for outpatient TB services for low-income patients with TB satisfactory immigration status (22).

Second, we credit our results to the quality of TB control programs in California. In our analysis, the default rate and the relapse rate after treatment completion for patients receiving SAT were the most important determinants of the cost-effectiveness of DOT. We believe these rates reflect a program's quality. Although the elements of a quality treatment program have not been defined, they likely include training, quality of the therapeutic encounter, accountability, supervision, other non-DOT activities, and the use of incentives and enablers. If the quality of a treatment program declines, both the patient default and relapse rates should increase, and DOT in our low-risk patients becomes more cost-effective. But if the quality of a program improves, default and relapse rates should decrease, and limiting the use of DOT to "high-risk" persons (i.e., selective DOT) becomes a more justifiable option from an economic standpoint.

Universal DOT has been recommended as the means to improve completion of therapy. Our results suggest that improving the quality of a program may be as important as increasing the use of DOT. Results from a randomized controlled trial of DOT versus SAT support our contention (23). The study reported that 40% of patients in each group failed to complete treatment. One interpretation is that adding DOT to a poorly functioning TB control program without addressing other issues affecting program quality will not improve treatment completion. On the basis of our results, we believe a greater emphasis on the quality of a treatment program is needed, not only on whether more DOT is used.

In conclusion, if resources for TB control are fixed and treatment default rate is very low, we believe there is an economic justification for not using DOT in persons at low risk for defaulting. We believe our results are applicable to other areas in the United States with low default rates. Finally, this study, like previous studies in this area, carries the methodologic limitation of not fully measuring the benefits of DOT. Therefore, we may have underestimated the cost-effectiveness of DOT. Future studies should better assess the benefits of DOT and compare the cost-effectiveness of expanding DOT with the strengthening of other TB control activities to determine which approach can more effectively reduce TB morbidity.