INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Since World War II, all adult immigrant applicants to Canada have undergone chest radiographic screening for tuberculosis (TB). The goals of screening are, first, detection of prevalent cases of active TB, and second, identification of radiographic abnormalities consistent with inactive TB, because of the increased risk of future active disease. Over 250,000 prospective immigrants are now screened annually. Yet the cost-effectiveness of these programs has never been formally evaluated.

The reported prevalence of active TB among screened immigrants has varied from 0.05% to 0.8% (1, 2). Despite screening, the incidence of active TB among immigrants from high prevalence regions remains elevated for up to 20 yr after arrival (3). Foreign-born persons now account for 58% of all cases of active TB in Canada (4).

Tuberculin screening of new arrivals has been suggested because of: (1) greater sensitivity than chest radiography for the detection of tuberculous infection; (2) fewer X-rays, if tuberculin-negative persons are not further evaluated; and (3) detection of more candidates for chemoprophylaxis. Potential disadvantages are false-negative tuberculin tests in patients with active TB or human immunodeficiency virus (HIV) infection, false-positive tests related to bacillus Calmette-Guérin (BCG) vaccination or nontuberculous mycobacteria, and massive program costs related to widespread chemoprophylaxis.

We compared the costs and outcomes associated with mass X-ray screening, mass tuberculin screening, and no screening.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The model was constructed using DATA 3.0 for Windows (TreeAge Software Inc., Williamstown, MA). A 20-yr time frame was used, corresponding to the longest placebo-controlled isoniazid (INH) chemoprophylaxis trial (5). The perspective was that of third-party payers, which in Canada are the federal and provincial governments. Three populations of 20-yr-old immigrants were considered (Table 1). This age was selected because of the potential benefits of chemoprophylaxis for young tuberculin reactors.

Outcomes and Discounting

Three outcomes were compared: expected TB cases, expected years lived, and expected costs (including direct costs of future cases) in 1997 Canadian dollars ($1 Canadian = $0.65 U.S.). For the two screening interventions, the cost per case prevented was also calculated. No quality adjustment for life years was used, because no valid estimates of TB-related health utilities exist. A 3% discount rate was used for all future expenditures and outcomes. The half-cycle correction was used for all Markov processes (6).

Screening Strategies

For all strategies, the individual screened began the simulation in (a) one of three tuberculosis-related states: active TB, tuberculous infection without active disease, or no tuberculous infection; and (b) one of three HIV-related states: HIV-seronegative, HIV-seropositive, early stage (imputed CD4 lymphocyte count of 1,000 × 10⁶/L), or HIV-seropositive, later stage (imputed CD4 lymphocyte count of 500 × 10⁶/L).

Strategy 1: Chest X-ray screening. As shown in Figure 1, active TB cases missed by screening are diagnosed only when symptomatic, and hence more frequently sputum smear-positive. These cases are associated with equivalent case fatality but increased cost, because of the more frequent need for hospitalization.

View larger version (29K): [in this window] [in a new window]   Figure 1.   Chest X-ray screening strategy. Screened individuals have active TB (A), inactive tuberculous infection (B), or no tuberculous infection (C ). Active and inactive TB are diagnosed only if the chest X-ray is read as abnormal. p[Event] indicates the probability of the corresponding event modeled. CXR = Chest X-ray; AbN = abnormal; N = normal; W/U = workup; INH = isoniazid; Rx = treatment.

The mortality rate associated with active TB varies with HIV status. Survivors face the same annual probabilities of death as others of equivalent HIV status (Figure 2). Background age- and gender-specific death rates were obtained from Canadian life tables (7).

View larger version (19K): [in this window] [in a new window]   Figure 2.   Active TB subtree. Persons with active TB at the time of screening either die as a result, or recover. The probability of immediate death (pDieTB) depends on age and HIV status (Table 2). Persons who recover then enter a two-state Markov process whereby each year they may die of other causes (pDieOther)including HIV, where infectedor survive. The symbol # indicates the complementary probability, i.e., 1the probability of the event described by the branch above.

Individuals labeled as TB-infected with chest radiographic abnormalities but no active disease are prescribed isoniazid, if physicians follow standard practice (Figure 3A).

View larger version (21K): [in this window] [in a new window]   Figure 3.   Inactive TB infection subtree. Infected persons falsely classified as having active disease are treated accordingly, while those correctly classified receive isoniazid prophylaxis, and those whose infection is missed receive no therapy (A). Those who survive treatment (where given) enter a three-state Markov process (B). They may subsequently reactivate (probabilities defined in the text and Table 2), die of TB (probability pDieTB), or die of other causes (pDieOther), including HIV, when infected.

Annual reactivation probabilities are determined by the presence and extent of HIV disease, the presence or absence of X-ray abnormalities, and the completion or noncompletion of prophylaxis (Figure 3B).

Persons without TB infection face background mortality risks, including those related to HIV; those who are mislabeled as having TB infection or active disease also face the risks of therapy given accordingly (Figure 4).

View larger version (16K): [in this window] [in a new window]   Figure 4.   Uninfected subtree. Uninfected individuals who are incorrectly classified as having either active disease or infection face the risks of treatment. Those who survive treatment and all those who are correctly identified enter a two-state Markov process which incorporates background mortality (including HIV-related mortality where relevant).

Strategy 2: Tuberculin skin test screening. As shown in Figure 5, detection of active TB requires a positive tuberculin test (induration  10 mm), followed by an abnormal X-ray and positive microbiology. Anergic individuals with active TB (who deny symptoms) or tuberculous infection are consistently missed. Prophylaxis is potentially administered to individuals with false-positive tuberculin reactions caused by BCG or atypical mycobacteria.

View larger version (29K): [in this window] [in a new window]   Figure 5.   Tuberculin skin test screening strategy. Screened individuals have active TB (A), inactive tuberculous infection (B), or no tuberculous infection (C). Only tuberculin reactors are referred for chest X-rays, with further evaluation dependent on the results of the chest X-ray. The branches ultimately lead to the same three Markov processes illustrated in Figures 2-4. PPD = tuberculin skin test using purified protein derivative; CXR = chest X-ray; AbN = abnormal; N = normal; W/U = workup; INH = isoniazid; Rx = treatment.

Strategy 3: No screening. In the absence of screening, all active TB cases are diagnosed passively and there is no systematic chemoprophylaxis. The number of expected active TB cases depends on the initial prevalence and the expected reactivation rates. Mortality relates to HIV, active TB, and background causes.

Base Assumptions

Model assumptions are listed in Table 2.

Natural history of HIV infection. To simplify calculations, HIV- related mortality was derived using the declining exponential approximation of life expectancy (DEALE) method (8). The estimated median survival of later stage HIV-infected persons was 7 yr, yielding an annual mortality of 9.3%. For early-stage HIV-infected persons, estimated median survival was 12 yr: 90% survival for the first 6 yr (annual mortality 1.7%), followed by 55% survival for the following 6 yr, as in later stage HIV disease (annual mortality 9.3%) (9, 10). We assumed that 25% of HIV-infected persons had advanced disease at screening (9, 10). The remainder progressed to late-stage disease after 6 yr.

Prevalent cases of active TB. The estimated prevalence of active TB was 2.3% among HIV-seronegative immigrants with tuberculous infection (1). Because no HIV-associated prevalence data are available for immigration screening, we arbitrarily assumed that the baseline prevalence of active TB among persons with tuberculous infection was doubled (4.6%) with concomitant early-stage HIV disease and quadrupled (9.2%) with advanced HIV disease.

TB-associated mortality. We used age-specific Canadian TB-associated mortality rates (Table 2) for HIV-seronegative and early-stage HIV-seropositive individuals (11). For later-stage HIV-seropositive persons, an additional 5% mortality risk was attributed to active TB.

Reactivation of tuberculous infection. For younger adults, the risks of reactivation remain relatively constant over the time frame used. The annual reactivation risk was 0.1% for HIV-negative individuals with tuberculous infection but normal X-rays (1, 12). For HIV-negative tuberculin reactors with abnormal chest X-rays, the annual risk was 0.66% (1).

Among HIV-infected tuberculin reactors, the estimated annual incidence of active TB ranged from 2.6% (CD4 > 350 × 10⁶/L) to 13.3% (CD4 < 200 × 10⁶/L) (13). These patients were not characterized by their baseline chest X-rays. For early-stage HIV patients, we thus assumed an aggregate annual risk of 2.6%. For later-stage HIV patients, we used a risk of 6.5%, as reported for patients with CD4 counts of 200 to 350 × 10⁶/L. We then incorporated the same relative risk estimate (6.6) and prevalence estimate (11%) for abnormal X-rays previously reported (1, 14). This yielded annual reactivation risks ranging from 1.6% to 26.6% for TB- and HIV-coinfected persons.

Completion of 6 mo of INH was assumed to produce a 65% reduction in the annual risk of reactivation, regardless of HIV infection status, which persisted for the duration of the simulation (5, 15, 16).

Secondary transmission. Estimates of secondary transmission rates were obtained from Salpeter and Salpeter (17): passively diagnosed TB cases were therefore assumed to produce an average of 3.5 infected contacts per case. In Montreal, 71% of such cases are smear-positive. Among actively diagnosed cases among screened immigrants and refugee claimants in Montreal, only 14% are smear-positive. Smear-negative cases are estimated to transmit infection 0.22 times as often as persons with smear-positive disease (18), so we assumed that actively diagnosed cases would produce a mean of 1.5 infected contacts per active TB case. In the base case analysis, we further assumed that thorough public health investigation would result in identification and prophylaxis of all infected contacts; the costs are described subsequently.

The same model (17) was used to estimate the number of secondary active TB cases that would occur during the 20-yr simulation, in the absence of contact investigation and prophylaxis. We estimated that during the 20 yr after diagnosis, passively detected cases would each generate 0.7 secondary active cases, whereas actively detected cases would each generate 0.3 secondary active cases (incorporating a 3% discount rate)based on the proportions of smear-positive index cases in each category. In one sensitivity analysis, expected incidence and costs were recalculated using these parameters.

Diagnostic tests. The reported proportion of culture-proven pulmonary TB cases with abnormal chest X-rays ranges from 90% to 100% (19, 20). We assumed that the chest X-ray would lead to further investigation in 95% of active TB cases.

We assumed that 11% of infected individuals had abnormal X-rays; in the U.S. Public Health Service trial in psychiatric institutions, 10.6% of placebo-treated tuberculin reactors had abnormal X-rays, as did 11.6% in the INH treatment arm (12). Furthermore, in Montreal (1996-1997) 6.0% of 8,800 screened immigrants and refugee claimants from high-prevalence countries were found to have chest X-rays suspicious for possible active TB; if the overall prevalence of tuberculous infection was 50% in this group, the estimated prevalence of significant X-ray abnormalities among the TB-infected was virtually identical to that previously reported. The assumed specificity of the chest X-ray for tuberculous infection was 97.5% (21).

In Montreal, 90% of pulmonary TB cases are culture-positive (22). The assumed specificity of sputum microbiology was 99%, owing to rare laboratory contamination (23).

In the absence of later-stage HIV infection, tuberculin testing was assumed to be highly sensitive (99%). Fifty-five percent of the HIV-TB cohort (13) with CD4 counts from 200 to 350 × 10⁶/L were labeled anergic, so sensitivity of tuberculin testing must be less than 45% in such individuals, when a cut point of 5 mm is used. Because a cut point of 10 mm is used in the absence of known HIV seropositivity, we estimated the sensitivity of tuberculin testing to be 20% among TB- infected persons with late-stage HIV (24).

Specificity of tuberculin testing is limited by widespread use of BCG vaccination in most countries outside North America. We assumed that 50% of immigrants (including those uninfected with TB) received BCG vaccination in primary school or subsequently (25). Among such individuals who are immunocompetent, 25% will have  10 mm induration on tuberculin skin testing, in the absence of tuberculous infection (25). Hence the false-positive rate is 12.5%, and the specificity is 87.5%. Among persons with late-stage HIV who have been BCG-vaccinated, only 20% of the expected false-positive reactions will be observed. The false-positive rate is therefore 2.5%, and specificity 97.5%.

Medication use and side effects. All persons diagnosed with active TB were prescribed full therapy. However, we used a probability of 73% that INH prophylaxis is prescribed for suitable candidates (26). Probabilities of completing therapy and of medication-related side effects are listed in Table 2. We assumed that persons who developed major side effects (hepatitis) during INH prophylaxis would not receive further therapy, while minor side effects could lead to substitution of rifampin, with equivalent cost and effect. Furthermore, we assumed that under program conditions, discontinuation of prophylaxis owing to side effects or noncompliance would occur early ( 3 mo), precluding any significant or long-lasting protection (15).

Costs and Probabilities of Hospitalization

We used cost estimates listed in Tables 3 and 4. Costs were derived from annual reports of the Montreal Chest Institute and Royal Victoria Hospital, and physician fees set by the Quebec health insurance board. We included physician and personnel costs, equipment and supplies, medications, hospital bed costs, and overhead. All estimates were expressed in 1997 Canadian dollars. The Canadian hospital expenditure price index was used to adjust for inflation.

Based on data from six Montreal-area hospitals during 1992-1995, we considered that 80% of active TB cases diagnosed passively require admission to hospital. The median length of stay was 19 d. These figures are similar to those reported in a U.S. survey of 16 states with more than 45% of the total population (27), where there were 0.79 hospital separations with a primary diagnosis of TB for every reported case, and the mean duration of hospitalization was 19.9 d. We assumed that only 50% of active TB cases diagnosed by screening required hospitalization, because patients are generally asymptomatic and more often smear-negative.

Data from the Montreal Chest Institute and the Montreal public health unit yielded cost estimates for contact investigation of $694 per index case, plus $365 per infected contact placed on prophylaxis. The total assumed costs of contact investigation and follow-up were thus $1,970 per passively diagnosed case, and $1,241 per actively diagnosed TB case. When secondary cases of active disease were included in the analysis, these were assumed to be diagnosed passively, with the attendant costs largely related to hospitalization, as described previously.

Sensitivity Analyses

The results were subjected to one-way sensitivity analyses of all assumed probabilities and costs. Ranges are shown in Tables 2 and 3. Published cost estimates from other centers were used to frame the sensitivity analyses, following conversion to 1997 Canadian dollars. For sensitivity analyses, the prevalence of HIV infection was assumed to be 10% and the prevalence of tuberculous infection 50% (Population 1), except where indicated otherwise.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Base Case Analysis

Health outcomes. In all three simulated cohorts, tuberculin skin test screening prevents the most active TB cases over the following 20 yr (Table 5). The number of cases prevented by tuberculin or radiographic screening increases with the risk of future active disease, and hence with the prevalence of HIV infection as well as TB infection. As anticipated, screening does not appreciably alter life expectancy, which is determined primarily by HIV seroprevalence. The gains in expected survival over the 20-yr simulation are minimalless than half a day per person screened in the highest risk group. This endpoint will not be discussed further.

Costs and cost-effectiveness. Tuberculin screening is considerably more expensive than chest X-ray screening (Table 5). Program costs increase with the prevalence of tuberculous infection, indicating that they stem largely from subsequent management of screened individuals, rather than from the screening tests themselves.

For the high TB prevalence populations (1 and 2), chest X-ray screening costs $3,943 and $10,627 per active case prevented. In these groups, replacing the chest X-ray with tuberculin skin testing as the initial screening modality leads to more than double the number of cases prevented, but at greater cost. For the low-prevalence population, both strategies are extremely costly relative to their impact on TB prevention.

Sensitivity Analyses

Health outcomes. In all sensitivity analyses, the tuberculin testing strategy prevents more active TB cases than does the chest X-ray strategy. Screening carries the greatest impact when HIV infection and tuberculous infection are most prevalent, or estimated reactivation risks are higher. A higher prevalence of baseline chest X-ray abnormalities is associated with an increased number of active cases, but does not change the ranking of the strategies. The tuberculin screening strategy remains most effective for TB prevention as long as the sensitivity of the tuberculin test among immunocompetent persons exceeds 32% in Population 1, or 42% in Population 2.

Table 6 illustrates the impact of increased adherence to isoniazid prophylaxis. In Population 1, radiographic screening with 100% physician and patient adherence with prophylaxis would increase the number of cases prevented from 1.6 to 4.3 per 1,000 screened. In Population 2, the number of cases prevented would increase from 1.2 to 3.4 per 1,000 screened.

Costs and cost-effectiveness. As the prevalence of HIV infection and/or tuberculous infection increases, the cost per case prevented by either screening strategy falls, although total expected costs rise. The impact of changes in INH prescription and completion rates on incremental cost-effectiveness is shown in Table 6. Improved adherence would result in cost savings, because of the greater number of future TB cases prevented. In Population 1, chest radiographic screening is associated with net cost savings if the INH completion rate exceeds 73%; additional expenditures of  $305 per chemoprophylaxis candidate to ensure 100% prescription and 100% compliance would actually save money relative to the base scenario for radiographic screening, because more active cases are prevented. For Population 2, the corresponding threshold is $222.

Increasing the cost of outpatient TB treatment increases the cost of all strategies, but does not influence incremental cost-effectiveness estimates. However, when the cost for inpatient treatment of active TB is increased modestly ($9,157), the chest X-ray screening strategy is cheaper than no screening (Population 1). At the same threshold, the incremental cost per case prevented is $31,539 for the tuberculin strategy versus chest X-ray (or no screen). For Population 2, chest X-ray screening is cheaper than no screening when the cost per inpatient admission exceeds $11,313.

Substituting a much lower cost estimate for contact investigation and treatment ($608 per index case) increases the costs of both screening strategies, but does not alter their ranking: $10,208 per case prevented by chest X-ray screening, and $33,774 per additional case prevented by tuberculin skin test screening (Population 1).

Table 7 shows the impact of including secondary active cases in the model, assuming that no systematic contact investigation is undertaken. For Populations 1 and 2, chest radiographic screening is net cost saving. Indeed, with 1% HIV seroprevalence, chest radiographic screening is then cost saving above a threshold TB infection prevalence of 39%.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Immigrant and refugee screening for TB is standard policy in Canada, the United States, and many western European countries. These are expensive and complex public health programs, which have not previously been evaluated from the standpoint of cost-effectiveness. The purpose of this analysis was to compare the impact and associated costs of two strategies for TB screening among immigrants and refugee claimants.

In some jurisdictions, these costs are borne by immigrant applicants themselves, imposing a significant financial burden on these individuals and their families. In many such instances the costs are assumed by family members already resident in the host country. However, in Canada the government-funded health care system bears the full cost for screening of all refugee claimants, and of many potential immigrants who apply from within the country.

Our results suggest that chest radiographic screening is a low-cost preventive intervention for young immigrants with a high probability of tuberculous infection. Indeed, for such individuals it appears to be cost saving where inpatient treatment costs only slightly more than our base case estimate, or if adherence to prophylaxis is improved. In an earlier Canadian report, FitzGerald and Gafni estimated that the direct cost of INH prophylaxis for 20-yr-old low-risk reactorsa generally accepted preventive interventionwas $10,475 per active case prevented (1997 dollars), which is similar to or exceeds our base case estimates for chest radiographic screening in the high-prevalence groups we modeled (28).

Tuberculin skin testing prevents additional TB cases, but it is consistently more expensive than radiographic screening, making its routine use difficult to justify. It identifies a large number of low-risk and false-positive reactors as candidates for chemoprophylaxis. For populations at high or intermediate risk, radiographic screening is more cost-effective because it identifies individuals with chest X-ray abnormalities who are at significantly higher risk of reactivation, and who benefit the most from prophylaxis.

For immigrants from low-prevalence countries, both screening strategies have little impact but very high cost, because of false-positive screening tests, low prevalence of infection, and low risk of disease. If anything, we may even have overestimated the impact of screening in this group, because we used a relatively high figure for the prevalence of active disease among TB-infected immigrants (1).

It might be expected that screening for TB symptoms to select candidates for further evaluation would improve program performance. However, this approach is problematic because applicants may deny symptoms, out of the fear that they will be denied permission to remain in the country. Smear-negative disease may be truly asymptomatic. Moreover, much of the public health impact of TB screening relates to prophylaxis of latent infection, which will be missed by symptom-based screening.

Improving physicians' and patients' adherence to recommended INH prophylaxis represents the most important mechanism for increasing the impact and cost-effectiveness of both strategies. Modest expenditures to improve adherence actually reduce the expected cost of radiographic screening in the long term. This is in keeping with a recent analysis which suggested that directly observed INH chemoprophylaxis among tuberculin reactors and anergic HIV-seropositive individuals attending a methadone clinic would be associated with long-term cost savings, relative to self-administered prophylaxis (29). Proposed alternative "short-course" chemoprophylaxis regimens may also prove cost-effective by increasing adherence, in spite of higher drug costs, although to date there are no published data to support this premise.

Tuberculin screening and INH chemoprophylaxis of foreign-born children were shown to produce net cost savings when the assumed cost of TB treatment was very high (30). However, in analyses that incorporated more modest treatment cost estimates, chemoprophylaxis led to cost savings only for high-risk reactors (28). In the present analysis, tuberculin screening is not cost saving for any immigrant group, while radiographic screening leads to cost savings in immigrant populations at high or intermediate risk only if active TB treatment costs are higher than in the base case examined.

Our analysis involves several uncertainties about the natural history of TB and HIV disease, diagnostic test performance, treatment outcomes, and costs. Although the risks of TB reactivation associated with radiographic and tuberculin skin test findings have been carefully documented in HIV-negative individuals, there is much less information available for HIV-infected persons. However, our results are consistent even when these risks are varied, or when the prevalence of HIV infection is assumed to be low. Likewise, the results are robust to variations in diagnostic test performance.

INH resistance is a growing problem in many areas of the world, and will reduce the effectiveness of INH prophylaxis in individuals from those regions. This will reduce the cost-effectiveness of any TB screening program that leads to INH prophylaxis, but would not change the relative standing of the screening strategies we modeled.

We used cost estimates from one Canadian center. Although generally lower than costs at similar U.S. centers, they fall within the range of figures used in earlier analyses of TB screening and prevention strategies. Substitution of higher U.S. costs for hospitalization for active disease leads to expected cost savings for radiographic screening (relative to no screening) for high- and intermediate-risk groups.

The public health impact of immigrant screening in terms of reduction in secondary transmission is controversial and difficult to evaluate. In our base case analysis, we incorporated local costs of contact investigation and treatment, and assumed that these would effectively prevent secondary active cases. In sensitivity analysis, we made the extreme assumption that there was no contact investigation, under which condition chest radiographic screening was cost saving for the higher prevalence Populations (1 and 2); tuberculin skin testing was somewhat more cost-effective than in the base case analysis. Although it is unlikely that a center would conduct mass immigrant screening without performing contact investigation, this again highlights the relative cost and impact of the two screening interventions.

Compared with tuberculin skin testing, chest radiographic screening for TB among immigrants is consistently cheaper and more cost-effective. Even under current program conditions, the cost of radiographic screening appears easily justified by its potential impact on TB prevention in populations with a high prevalence of tuberculous infection. Greater physician and patient adherence with INH prophylaxis will likely lead to net cost savings in this context. However, for immigrants from low-prevalence countries, any TB screening program provides minimal health benefit and is extremely costly. The money required to screen such individuals would be much better used for other means of TB control and prevention, and for other pressing immigrant health needs. Neither the magnitude of the public health challenge posed by TB nor the shortage of needed funds is likely to change, so it is imperative that resources devoted to combating this illness yield the highest possible returns.