INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

As the United States moves back on track toward elimination of tuberculosis (1), an increasingly important priority is the use of screening for tuberculosis infection and preventive therapy in persons at high risk for developing tuberculosis (2). But with limited health budgets, public health officials need to determine the most cost-effective prevention programs. To accomplish this, it is necessary to identify high-risk populations that are most amenable to tuberculosis prevention and methods for effective implementation of prevention programs.

Injection-drug users are at increased risk for being infected with Mycobacterium tuberculosis, human immunodeficiency virus (HIV), or both (3). Because persons dually infected with these pathogens have an accelerated rate of tuberculosis (4- 7), and tuberculosis in turn may accelerate the progression of HIV disease (8, 9), it is especially important to prevent tuberculosis in injection-drug users. Although the screening of injection-drug users for tuberculosis infection and administering preventive therapy have been recommended, the successful implementation of these prevention activities has not been well documented.

One venue for implementing a prevention program is the methadone maintenance clinic. According to federal regulations, screening for tuberculosis infection is mandated for all clients upon admission to a methadone maintenance program (10). However, the percentage of clients completing the screening process and, more importantly, a course of preventive treatment has generally been low. This may partly be due to the fact that injection-drug users are more likely to be nonadherent with medical interventions, perhaps because they have a variety of sociobehavioral problems (11). However, there is frequently a lack of coordination between the drug treatment program, which screens for tuberculous infection, and the public health program, which conducts medical evaluation and administers preventive therapy. This lack of coordination is sometimes caused by separate administrative and funding systems for these programs and likely contributes to the failure of tuberculosis prevention efforts among injection-drug users accessing methadone maintenance programs.

In 1989, the Centers for Disease Control and Prevention (CDC) funded a pilot program in 25 state and local health departments to conduct screening for tuberculosis infection and directly observed preventive therapy (DOPT) in correctional facilities and drug-treatment centers (12). In 1990, this program, called the HIV-related Tuberculosis Prevention (HRTP) Project, was implemented at five methadone maintenance clinics in San Francisco.

Using program data, we evaluated the program effectiveness of screening for tuberculosis infection, DOPT, and tuberculosis prevention. We then determined program cost-effectiveness in preventing tuberculosis. Finally, we described the program factors that we believe were essential to the program's success. Our results show that when properly administered, screening for tuberculosis infection and DOPT in methadone maintenance clinics can be a highly effective and cost-effective program to prevent tuberculosis.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Program Description

The HRTP Project in San Francisco began with administrative planning between the tuberculosis control program and the methadone maintenance clinics. A jointly developed interagency agreement specified each agency's commitment of resources and administrative responsibilities for the expansion of services under the HRTP Project. In addition, a Qualified Service Organization Agreement was developed, which fulfilled federal regulatory requirements for confidentiality when sharing client-identifying drug treatment information between the two agencies (13).

Staff at the methadone maintenance clinics were responsible for tuberculin skin testing, HIV counseling and testing, tuberculosis and HIV education, and DOPT. Staff at the tuberculosis clinic were responsible for performing medical evaluation on all clients referred from the methadone maintenance clinics and making the decision on who should receive preventive therapy.

All clients without a prior history of a positive tuberculin skin test (TST) were skin tested by the Mantoux method using 5 tuberculin units (TU) of purified protein derivative; results were read after 48 to 72 h. All clients with a TST  5 mm as well as those with a prior history of a positive TST were referred to the tuberculosis clinic for medical evaluation. In addition, all HIV-seropositive clients with a TST < 5 mm were encouraged to have a baseline chest radiograph. At the tuberculosis clinic, clients were classified as TST reactors if they had one of the following: a prior history of a positive TST, a TST  10 mm and they were HIV-seronegative, and a TST  5 mm and they were HIV-seropositive or their HIV serostatus was unknown. All TST reactors were further evaluated for tuberculosis.

Once tuberculosis was excluded, clients were recommended for isoniazid preventive therapy if there was no contraindication (14). When the baseline liver function tests were more than three times above the upper limit of normal, isoniazid preventive therapy was not recommended unless the client was a recent TST convertor or had a chest radiograph consistent with old healed tuberculosis (i.e., Class 4). These patients were closely monitored by the tuberculosis clinic while they were receiving preventive therapy and received monthly liver function tests and twice monthly symptom review.

Isoniazid (300 mg) and pyridoxine (50 mg) were provided daily at the methadone maintenance clinics concurrent with the methadone dosage. The methadone program nurse observed the ingestion of isoniazid and pyridoxine and recorded each dosage and date. The duration of isoniazid therapy was 6 and 12 mo for HIV-seronegative and HIV-seropositive clients, respectively.

To improve the adherence of clients to medical evaluation at the tuberculosis clinic, clients being referred were initially educated by methadone clinic staff about the benefits of the evaluation and possible treatment as well as the procedures of the tuberculosis clinic. A community health worker then accompanied each client during the clinic visit, facilitating the process of registration, obtaining a chest radiograph and blood tests. To reduce client-waiting time, the clinic made it a priority to see these clients first. Clients were provided bus tokens or were transported to the clinic by the health department car. Food and juice were provided at the clinic.

To improve the adherence of clients to the course of preventive therapy, the methadone program nurse initially discussed the benefits of completing a course of therapy and the purpose of follow-up in the event of adherence lapse. Patients were encouraged to help develop an individual adherence plan, including the provision of locations where the client could be located in the community. When clients did not adhere to scheduled evaluation or isoniazid therapy, a confidential telephone call was made or letter sent to the client. In addition, the community health worker would look for the client in the community.

Evaluation of Program Effectiveness and Cost-effectiveness

We used data collected by the HRTP Project to evaluate the effectiveness and cost-effectiveness of the project for 6 yr (January 1990 to December 1995). Study participants comprised all clients admitted to the methadone maintenance program's five clinics from 1990 to 1995. Data were collected on a standard report form and entered into a computer database developed by the CDC. To evaluate program effectiveness, we determined completion rates of tuberculin skin testing, medical evaluation, and preventive therapy. We also determined the number of active cases of tuberculosis identified through screening.

To determine the number of preventable cases of tuberculosis resulting over a 3-yr follow-up period in the screened population, we matched a database of TST reactors (in the HRTP Project from 1990 to 1994) who had not completed a prior course of preventive therapy at the time of screening with a database of verified cases of tuberculosis reported to the California Tuberculosis Registry from January 1, 1990 through June 1, 1998 (California Department of Health Services, Tuberculosis Control Branch). Clients with the same first name, last name, and date of birth on both databases were considered a match if they were reported with tuberculosis within 3 yr from the date of screening.

We determined the difference between the number of cases of tuberculosis expected to occur over 3 yr without the screening program, and the number of cases that actually occurred over the 3-yr follow-up period. To estimate prevention effectiveness over 10 yr, we projected expected cases that would result over a hypothetical 10 yr of follow-up, with and without the screening program.

To evaluate cost-effectiveness in preventing tuberculosis, we determined the net average cost per case of tuberculosis prevented over the hypothetical 10-yr follow-up period. We determined costs from the perspective of a local tuberculosis control program, but also included costs of inpatient treatment for tuberculosis. Costs were calculated in U.S. dollars, and adjusted to 1998 values using the medical care services component of the U.S. Consumer Price Index. Our base case analysis discounted future costs and cases of tuberculosis prevented at 3%. Screening program data were analyzed with SAS (15). The cost-effectiveness analysis was developed on a Microsoft Excel spreadsheet (16).

Model to Determine Expected Cases of Tuberculosis

We developed a four-state Markov Process (17) to determine the expected number of cases of tuberculosis that would occur in the TST reactor population over 10 yr of follow-up. In each 1-yr cycle of the Markov Process, TST reactors may either remain well, develop tuberculosis and survive, develop tuberculosis and die, or die of other causes. We did not model separately the prognosis of anergic persons because a recent study in this same population of injection-drug users found that anergic and TST-negative persons had a similar rate of tuberculosis (18).

The cohort for the 3-yr and the 10-yr Markov Process was derived from screening program data (3-yr: years 1990 to 1994 [data not shown]; 10-yr: years 1990 to 1995 [Table 1]). Parameters for the Markov Process were derived from the screening program and review of the literature (Tables 1 and 2). State transitional probabilities and the efficacy of isoniazid preventive therapy were determined by literature review. The annual probability of developing tuberculosis was modified by the actual program usage of preventive therapy and the efficacy of isoniazid preventive therapy. Of 417 TST reactors eligible for preventive therapy, 33 (7.9%) had unknown HIV serostatus. To keep estimates of program cost-effectiveness conservative, we modeled the prognosis of these patients as if they did not have HIV infection.

Costs of the HRTP Project

The net cost of the program was determined by summing the actual fixed and variable costs incurred by the project during the study period and subtracting the costs that would be averted from preventing cases of tuberculosis. Averted costs included the cost of diagnosis, treatment, and contact investigation associated with the active cases prevented in the cohort of clients over a hypothetical 10-yr period, as well as the cost of treating active cases developing among infected contacts over the same time period.

Fixed costs. Fixed costs were personnel costs that did not vary with the number of clients screened. Fixed costs were allocated on the basis of time spent on project activities, and included a project manager, a community health worker, and a clerk who provided clerical support. Fixed costs (in 1998 dollars) declined from $130,285 in 1990 to $77,858 in 1995. To assess costs for the project after it was fully operational, and not during the start-up period, we multiplied fixed costs for 1995 by the 6 yr analyzed.

Variable costs. Variable costs included cost of performing and reading TST, providing incentives and enablers, conducting a medical evaluation of TST reactors, administering DOPT, and managing side effects of treatment. From the San Francisco Department of Public Health Tuberculosis Control Program, we collected the unit cost of personnel, Mantoux testing supplies, incentives and enablers, chest radiograph, bacteriologic studies, liver function tests, and isoniazid (Table 3).

Averted costs. To determine the cost averted, we calculated the cost of diagnosis, treatment, and contact investigation associated with the cases of tuberculosis prevented by the program. Using our Markov model, we determined the number of additional cases of tuberculosis expected to occur over a 10-yr period among contacts to preventable cases, and used this number to calculate expected costs. For this analysis, we assumed that contacts of the clients with tuberculosis were the same age and had the same prevalence of HIV infection as the HRTP Project population.

The total average cost of tuberculosis treatment was estimated as $20,023, and was determined by summing costs for outpatient treatment, inpatient treatment excluding physician charges, and hospital-related physician charges. Outpatient treatment cost ($3,238) was estimated by determining drug-resistance rates among clients with tuberculosis in California with a history of injection-drug use and using published data (19) on cost of tuberculosis treatment to calculate a weighted average cost.

Inpatient treatment cost ($16,775) was determined by adding the average cost of TB-related hospitalization (excluding physician charges, $20,175) for California patients with tuberculosis and a history of injection-drug use with hospital-related physician charges ($538) (Medi-Cal Policy Division, California Department of Health Services), and then multiplying this figure by the proportion of injection-drug users in California with a tuberculosis-related hospitalization (81%). Data on proportion hospitalized and costs excluding physician charges were derived from a prospective study of patients with tuberculosis (Dr. Zachary Taylor and Suzanne Marks, CDC, unpublished data). These patients were followed from diagnosis to the completion of therapy (or therapy discontinuation) and all TB-related hospitalizations were identified. Costs were derived from hospital charges using Health Care Financing Administration (HCFA) hospital-specific cost-to-charge ratios.

To determine the cost of contact investigation and treatment of tuberculosis infection and active tuberculosis in contacts, we used contact investigation data collected by the San Francisco Tuberculosis Control Program (Table 4). In 1995, patients with tuberculosis and a history of injection-drug use had on average eight screened contacts, of which 39% had tuberculosis infection and 3% had active tuberculosis. Fifty percent of contacts with positive TST received some preventive therapy, of whom 73% completed treatment.

Sensitivity Analysis

We varied program performance, excluded savings from averted contact tracing and treatment, and reduced the percentage of patients with active tuberculosis hospitalized, in order to assess their effect on the number of cases prevented and cost per case prevented (Table 5).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Client Population

Of the 2,689 clients seen by the program from 1990 to 1995, 59% were men, 58% were non-Hispanic white, 27% were African-American, 12% were Latino, 2% were Asian/Pacific Islander, and 1% belonged to other racial/ethnic groups. At initial screening, the median age was 40 yr (range, 18 to 77 yr), with 75% of clients between the ages of 33 and 50 yr; 1,704 (63%) clients were HIV-seronegative, 472 (18%) were HIV-seropositive, and 513 (19%) had unknown HIV serostatus.

At initial screening, 491 HIV-seronegative clients, 111 HIV-seropositive clients, and 82 clients with unknown HIV serostatus had tuberculous infection. After the initial screening, 1,060 clients received at least one subsequent TST. There were 53 TST conversions over 2,243 person-years (PY) of observation, resulting in an overall conversion rate of 2.4/100 PY. The TST conversion rate was 3.1/100 PY, 2.4/100 PY, and 0.4/100 PY in HIV-seronegative, HIV-seropositive, and clients with unknown HIV serostatus, respectively.

Program Effectiveness

Of the 2,689 initially screened clients, 538 (20.0%) had a documented history of a positive TST in the past. Of 2,151 clients eligible for initial tuberculosis screening, 2,143 (99.6%) completed tuberculin skin testing (Table 6). Including TST reactors identified during repeat screening, 552 clients were eligible for medical evaluation, of whom 529 (95.8%) were evaluated for possible tuberculosis (Table 7). The screening program identified five new active cases of tuberculosis. Of 417 TST reactors recommended for preventive therapy 378 (90.6%) began a course of chemoprophylaxis (Table 7). Of this group, 285 (75.4%) completed the recommended course of therapy (Table 1).

Prevention Effectiveness over 3 yr

Of 453 TST reactors identified between 1990 and 1994 and eligible for preventive therapy, one (0.2%) was reported on the Tuberculosis Registry of the State of California within 3 yr of initial screening. Our Markov model predicted that without the screening program, 20 (19.4 discounted) cases of tuberculosis would be expected to occur over the 3-yr follow-up period. Therefore, over the first 3 yr of follow-up, as much as 95% of expected cases of tuberculosis were prevented.

Projected Prevention Effectiveness over 10 yr

Our Markov model predicts that without the screening program, 57.7 cases of tuberculosis were expected in the cohort of TST reactors over 10 yr. With the screening program, 27.7 cases of tuberculosis were expected, indicating that the program prevented 30.0 (28.1 discounted) cases of tuberculosis. The program also prevented 7.6 tuberculosis-related deaths.

Program Cost-effectiveness

The program cost of screening and preventive therapy was $771,569 (Table 3), but the program was projected to avert $876,229 for a net savings of $104,660. The net average savings per case prevented was $3,724.

Sensitivity Analysis

The sensitivity analysis revealed that if savings from prevented contact investigation and treatment of disease contacts were excluded from the analysis, the net average cost per case prevented would be $7,431 (Table 5). If 60% of patients with active cases of tuberculosis had a TB-related hospitalization, the average cost per case prevented would be $2,702. Had the completion rate of preventive therapy been 30%, only 27.7% of expected cases of tuberculosis would be prevented, and the average cost per case prevented would be $12,677. In contrast, had the completion rate of preventive therapy been 95%, 56.3% of expected cases of tuberculosis would be prevented, with an average savings of $6,674 per case prevented.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The HRTP Project in San Francisco, based in methadone maintenance clinics, is a highly effective and cost-effective model of tuberculosis prevention among injection-drug users. In a population that is frequently nonadherent to medical interventions, more than 99% of clients completed tuberculin skin testing, more than 99% of TST reactors were referred for medical evaluation if they had never completed a course of preventive therapy, and 95% of such persons were evaluated. Equally important, more than 80% of those who started a course of preventive therapy actually completed it (when those who moved or died during therapy were excluded from the analysis). Nearly half of those who failed to complete a course of preventive therapy did so because of a valid medical reason, and less than 10% were lost to follow-up or refused their medication.

Because of the effective implementation of the HRTP Project, the program was effective in preventing tuberculosis among clients in the program. Over a 3-yr follow-up period, only one case of tuberculosis was identified among the 453 TST reactors screened between 1990 and 1994. We estimate that more than 90% of the cases expected to develop over the 3-yr period were prevented. Over a 10-yr period, we estimate through our model that nearly 50% of the cases would be prevented.

In an era of limited health budget, we have documented that the HRTP Project in San Francisco is a very cost-effective program for preventing tuberculosis in a population at high risk for tuberculosis and HIV infection. When the cost of treatment and contact investigation associated with future cases of tuberculosis was included in our cost-effectiveness analysis, the program actually generated net savings per case of tuberculosis prevented. But even when these costs were excluded from our calculation, the cost per case prevented was still substantially less than the $20,023 needed to treat a case of tuberculosis in the injection-drug-using population.

The HRTP Project in San Francisco serves to highlight program factors that we believe are essential for successful tuberculosis prevention in methadone maintenance clinics. First, there must be coordination between the methadone maintenance clinic and the tuberculosis control program at the administrative, staffing, and service delivery levels. Administrative collaboration, including the joint development of an interagency agreement identifying respective fiscal, administrative, and resource commitments to the program, was critical to the program's success. Collaboration at the staffing level also contributed to the program's success. This included the hiring of a project manager to oversee the project activities and facilitate interface between the two agencies, placement of program nurses into the methadone clinics, and the cross-training of staff to enhance support of newly introduced procedures and staff effectiveness in revised roles and integrated staffing patterns. At the service delivery level, methods for screening for tuberculosis infection and DOPT were carefully integrated into the methadone maintenance program's existing procedures and philosophy while procedures for efficient referral and evaluation of clients in the tuberculosis clinic were established.

Second, the use of methadone maintenance clinic staff to observe the ingestion of medications must be a component of the program. Although it is a challenge to make sure drug- using clients adhere to the process of tuberculin skin testing or medical evaluation, it is perhaps most difficult to ensure that clients complete a 6- to 12-mo course of treatment. The use of pill count, urine test for isoniazid, and the Medication Event Monitoring System (MEMS) have been used or studied as a means to monitor adherence to a course of preventive therapy (20); however, they cannot ensure ingestion of medication by the client. DOPT, on the other hand, has been shown to improve adherence to isoniazid preventive therapy in injection-drug users and guarantees that clients ingest the medication (21). Using actual program data, we have further shown that DOPT can prevent tuberculosis in a population of injection-drug users.

Third, the use of incentives and enablers is needed for program success. One of the most obvious incentives in the HRTP Project is the provision of DOPT along with the daily methadone dosing. The program in San Francisco has taken advantage of this built-in incentive to enhance adherence to preventive therapy. Other incentives and enablers in the program include bus tokens or transportation to the clinic, food and juice, as well as shortened waiting-time at the clinic, and the use of a community health worker to facilitate the process of medical evaluation and to rapidly follow up lapses with DOPT. The use of incentives and enablers is believed to improve the effectiveness of directly observed treatment for tuberculosis (22). Our observation suggests that, among injection-drug users, incentives and enablers help improve completion rates of medical evaluation and preventive therapy.

In general, screening for tuberculosis infection should only be conducted if it is possible to ensure proper evaluation of infected persons and completion of preventive therapy for all eligible patients. This approach is supported by our analysis, which showed that all aspects of the programtuberculin skin testing, medical evaluation, initiation and completion of preventive therapymust be effectively implemented or else its cost-effectiveness will be reduced. Nevertheless, the sensitivity analysis revealed that the HRTP program would have remained quite cost-effective even when there were moderate reductions in program effectiveness. This finding is a result of the high rate of tuberculosis and cost of tuberculosis treatment in our client population. Therefore, even though program effectiveness is an important criterion for deciding if a tuberculosis screening program should be instituted or maintained, client population factors are equally important considerations.

Our conclusion regarding the cost-effectiveness of the project is strengthened by our conservative costing approach in two areas. First, we included fixed costs, which accounted for 61% of total project costs. Without fixed costs, the marginal savings per case prevented would have been $20,346 (data not shown). Second, we may have underestimated the total cost of treating active tuberculosis resulting from secondary transmission. We used actual program data from contact investigation to estimate that 0.48 additional cases would result among contacts to each preventable case. But the number of additional cases may be higher since contact investigations frequently fail to identify all infected contacts; this may especially be true for the population using injection drugs since they are frequently involved in transmission of tuberculosis in urban settings (23). Moreover, Salpeter and colleagues (27) recently estimated that 1.2 additional cases would result from each preventable case. Had we used this higher estimate in our analysis, the program would have saved an average of $18,140 per case of tuberculosis prevented (data not shown).

There are limitations to this study. First, our rates of tuberculosis for various populations were derived from the literature and thus may not apply to the injection-drug-using population. For example, the data for HIV-seronegative, TST convertors were derived from studies of persons who were not injection-drug users. Although we assumed the same rate of tuberculosis, the actual rate for injection-drug users may be higher. Similarly, our Markov model may overestimate the reactivation rate among remotely infected HIV-seronegative injection-drug users. Second, the efficacy of isoniazid preventive therapy was derived from controlled clinical trials in HIV- seronegative persons. Similar data do not exist for HIV-seropositive persons in developed countries such as the United States; thus, we may have overestimated the efficacy of isoniazid, especially since 5% of the injection-drug users with tuberculosis in San Francisco from 1993 to 1996 had isoniazid-resistant disease (Source: California Department of Health Services, Tuberculosis Control Branch). Third, as yet no published U.S. data exist on rates of TB-related hospitalization among patients with tuberculosis and a history of injection-drug use. The figure we used in the analysis (81%) may overestimate the actual proportion hospitalized, thus overestimating averted costs and project cost-effectiveness.

In conclusion, screening for tuberculosis infection and provision of DOPT in methadone maintenance clinics is an effective as well as cost-effective approach to prevent tuberculosis. However, key program components such as coordination between public health and methadone maintenance programs and the use of incentives and enablers are likely essential for ensuring program effectiveness. Our study provides a guide for planning, implementing and evaluating tuberculosis screening and DOPT programs in methadone maintenance clinics.