Abstract
Delayed diagnosis of active pulmonary tuberculosis (TB) among hospitalized patients is common and believed to contribute significantly to nosocomial transmission. This study was conducted to define the occurrence, associated patient risk factors, and outcomes among patients and exposed workers of delayed diagnosis of active pulmonary TB. Among 429 patients newly diagnosed to have active pulmonary TB between June 1992 and June 1995 in 17 acute-care hospitals in four Canadian cities, initiation of appropriate treatment was delayed 1 week or more in 127 (30%). This was associated with atypical clinical and demographic patient characteristics, and after adjustment for these characteristics, with admission to hospitals with low TB admission rate of 0.2–3.3 per 10,000 admissions (odds ratio [OR]: 7.4; 95% confidence interval [CI]: 3.2,17.5) or intermediate TB admissions of 3.4–9.9/10,000 (OR: 2.3; CI: 1.6,3.2) as well as potentially preventable (late) intensive care unit admission (OR: 16.8; CI: 2.0,144) and death (OR: 3.3; CI: 1.7,6.5]). In hospitals with low TB admission rates, initially missed diagnosis, smear-positive patients undergoing bronchoscopy, late intensive care unit admission (OR: 2.3; CI: 0.1,56), and death (OR: 3.8; CI: 1.2,12.1) were more common than in hospitals with high TB admissions ( > 10/ 10,000); a similar trend was seen in hospitals with intermediate TB admissions. Even after adjustment for workers' characteristics and ventilation in patients' rooms tuberculin conversions were disproportionately high in hospitals with low and intermediate TB admission rates and significantly higher in hospitals with overall TB mortality rate above 10% (OR: 2.5; CI: 1.6,3.7). In the hospitals studied, as the rate of TB admissions decreased, the likelihood of poor outcomes and risk of transmission of TB infection per hospitalized patient with TB increased. Institutional risk of TB transmission was poorly correlated with number of patients with TB and better correlated with indicators of patient care such as delayed diagnosis and treatment and overall TB-related patient mortality.
Keywords: tuberculosis diagnosis; delayed diagnosis; tuberculosis mortality; nosocomial transmission; occupational infections
Delayed diagnosis of active tuberculosis (TB) is an important problem in many general hospitals in industrialized countries because it results in greater patient morbidity and mortality and intrainstitutional spread of TB. This problem has been attributed to the relative rarity of active TB in these hospitals, resulting in a lack of experience and expertise in the diagnosis and management of TB (1-6).
The potentially devastating consequences of delayed or missed diagnosis was demonstrated in the late 1980s and early 1990s, when over a dozen institutions in the United States experienced major nososcomial outbreaks of multidrug-resistant strains with mortality rates as high as 80% (7-10). Implementation of many of the subsequently recommended measures (11-13) were shown to be effective in achieving rapid and substantial reduction of nosocomial transmission (14-17). However, in the vast majority of acute-care hospitals in industrialized countries, the extent and determinants of nosocomial TB transmission are not known, nor are the potential benefits of different control measures (18-20).
In a multicentre population-based cross-sectional survey conducted in 1995–1997, tuberculin conversion among health care workers in 17 acute-care hospitals in four Canadian cities was strongly associated with years and type of work and ventilation of nonisolation rooms (21). We have estimated the occurrence of delayed diagnosis and treatment among patients with TB admitted to these 17 hospitals as well as the associated characteristics and outcomes of these patients and tuberculin conversions among workers caring for them.
Seventeen acute-care hospitals in four Canadian cities—Montreal, Toronto, Edmonton, and Vancouver—were selected for the study. These were classified on the basis of the ratio of TB admissions to total admissions as: (i) very low: < 0.4/10,000; (ii) low: 0.4–3.3/10,000; (iii) intermediate: 3.4–9.9/10,000; and (iv) high: > 10/10,000.
Within each hospital, medical records departments were asked to provide lists of all patients admitted between July 1, 1992 and June 30, 1995, where active TB was listed among the discharge diagnoses. Additional patients were identified from a review of the records of the microbiology, pathology, and infection control departments.
Active TB was considered confirmed if there was a positive culture for Mycobacterium tuberculosis or if there was histologic evidence (necrotizing granulomas) and no other organisms were identified. Patients were considered to have active pulmonary disease if there was a positive culture for M. tuberculosis from respiratory specimens, consistent histology from a lung biopsy or autopsy, or M. tuberculosis was isolated from an extrapulmonary site and the concurrent chest radiograph was interpreted to have active parenchymal disease. Patients were considered to be newly diagnosed if they were not on active TB treatment at the time of admission and had not been transferred from another health care facility with a known or presumptive diagnosis of active TB.
The hospital records of all patients with newly diagnosed active pulmonary TB were reviewed to abstract demographic characteristics, such as age and country of birth, and clinical characteristics, such as cough and sputum production, chest X-ray findings, acid-fast bacilli (AFB) smear status, drug sensitivities, and human immunodeficiency virus (HIV) status, if known. Dates of admission, respiratory isolation, treatment, transfers within the hospital, and discharge were recorded.
As described in detail elsewhere (21), in the selected hospitals all workers in the departments of respiratory therapy, physiotherapy, all clinical care personnel in the emergency rooms, medical intensive care units (ICUs), a sample of medical and surgical wards, plus a sample of nonclinical personnel were invited to participate. In these areas, for a sample of isolation and nonisolation rooms air changes per hour and direction of airflow were measured (21, 22). Participating workers completed self-administered questionnaires and underwent tuberculin skin testing (TST) after verification of prior tests. Workers were included in this analysis if they had at least two tuberculin tests more than a year apart, with the earlier negative (< 10 mm).
Definitions of delays in diagnosis(i) Initially missed diagnosis—failure to initiate respiratory isolation and/or treatment within the first 24 hours after admission to hospital. This was selected as an indicator because most patients undergo repeated detailed medical and nursing evaluations within the first 24 hours of admission. (ii) Delayed treatment: interval from admission until initiation of treatment of seven days or more.
Patient outcomes(i) All cause mortality: during the same hospitalization when the diagnosis of active TB was first made.
(ii) All cause ICU admissions: this appeared to be bimodal; most patients were admitted directly or were transferred to an ICU within 4 days of admission (“Early ICU”). A second group was admitted to an ICU after 7 to 30 days of hospitalization (“Late ICU”).
Worker outcome—tuberculin conversionAn initial negative TST (< 10 mm) followed, after an interval of at least a year, by a reaction to a subsequent tuberculin test of at least 10 mm, with an increase of 6 mm or more, as recommended for Canadian populations (23, 24). Previously, we have shown that use of the more stringent definition recommended for populations in the United States (25) reduced sensitivity without altering findings (21).
The association of patient characteristics with initially missed diagnosis, delayed treatment, and mortality was examined. Multivariate analyses provided adjusted estimates of the association of these delays with outcomes, hospital admission rate, and two indicators of patient care. (i) Percent smear-positive patients bronchoscoped—selected because bronchoscopy poses substantial risk of transmission (29, 30) and may have been avoidable in smear positive patients; and (ii) % initially admitted to a nonmedical ward (other than the ICU)—selected because workers on these units may be less familiar with care of pulmonary TB.
Finally, the association of indicators of patient care with tuberculin conversion among workers was estimated in bivariate and multivariate analyses. These indicators were at hospital level, including the percent of patients that died, were admitted to a nonmedical ward, or were bronchoscoped (smear-positive only), and at unit level including percent of patients with initially missed diagnosis, or delayed treatment, and average air changes per hour in isolation and nonisolation rooms.
Between 1992 and 1995, 837 patients with active TB were diagnosed at the study hospitals, of whom 200 (24%) had extrapulmonary disease only, 208 (25%) had pulmonary disease, but were not hospitalized, or were already known to have active TB when first hospitalized. This left 429 (51%) who were newly diagnosed to have active pulmonary TB following hospitalization, of whom 54% were born in Canada or other low-incidence countries, 60% were AFB smear positive, and 22% had concomitant extrapulmonary TB. The median duration of hospitalization was 19 days (range: 1–384), and the median ICU stay was 7 days (range 1–157).
The diagnosis was initially missed in 193 (45%) of all hospitalized patients, of whom 101 (53%) were smear positive. Treatment was initiated after a week or more in 127 (30%) of all patients, of whom 49 (45%) were smear positive. Almost half of those who were not diagnosed within the first 24 hours had treatment delayed by more than a week. Among the patients with delayed treatment, the median interval from admission until isolation was 12.5 days (range 0–152 days), and from isolation until treatment was zero days (range 0–98 days). Intervals from isolation until treatment of more than 60 days were observed in three HIV-positive patients—initially placed in isolation because of their HIV infection rather than for suspected TB. Overall, 52 (12%) patients died, of whom 15 underwent autopsy. Nine were first diagnosed at autopsy.
As shown in Table 1, initially missed diagnosis, delayed treatment, and mortality were significantly associated with older age and absence of cough. Indicators of atypical manifestations—such as negative sputum AFB smear, noncavitary chest X-ray, and extrapulmonary disease were associated with delays, but not mortality. HIV infection was associated with delays and mortality, although not significantly because only 23 patients were HIV positive (14% of tested). Presence of drug resistance was not associated with any patient outcomes, but only 46 (11%) patients had strains with any drug resistance, and less than 2% had multi-drug resistant strains.
| Characteristics | Diagnosis Missed in First 24 Hours | Treatment Delayed by 1 Week or More | Mortality | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Odds Ratio | (95% CI) | Odds Ratio | (95% CI) | Odds Ratio | (95% CI) | |||||||
| Age | ||||||||||||
| 1–64 (reference group) | 1.0 | — | 1.0 | — | 1.0 | — | ||||||
| 65–79 | 2.0 | (1.2, 3.4) | 2.1 | (1.3, 3.6) | 4.4 | (2.1, 9.4) | ||||||
| 80–96 | 1.7 | (0.9, 3.5) | 2.2 | (1.1, 4.3) | 8.4 | (3.6, 19.8) | ||||||
| Male sex* | 0.9 | (0.6, 1.4) | 1.0 | (0.6, 1.6) | 0.6 | (0.3, 1.1) | ||||||
| Born in Canada or other low TB incidence country (versus born in intermediate + high TB incidence) | 2.1 | (1.4, 3.3) | 1.7 | (1.1, 2.6) | 1.6 | (0.8, 3.0) | ||||||
| Absence of cough | 1.7 | (1.1, 2.7) | 1.6 | (1.0, 2.6) | 3.4 | (1.7, 6.7) | ||||||
| Concomitant extrapulmonary disease | 2.0 | (1.2, 3.4) | 1.4 | (0.8, 2.4) | 1.2 | (0.5, 2.5) | ||||||
| HIV | ||||||||||||
| negative (reference group) | 1.0 | — | 1.0 | — | 1.0 | — | ||||||
| positive | 2.3 | (0.9, 5.9) | 1.6 | (0.6, 4.1) | 2.7 | (0.7, 10.1) | ||||||
| unknown | 0.9 | (0.6, 1.5) | 0.6 | (0.4, 1.0) | 0.5 | (0.3, 0.98) | ||||||
| Noncavititary infiltrate on chest X-ray | 1.2 | (0.7, 1.9) | 1.3 | (0.8, 2.2) | 1.4 | (0.6, 3.2) | ||||||
| Sputum smear negative | 1.5 | (1.0, 2.4) | 2.0 | (1.2, 3.2) | 0.9 | (0.4, 1.7) | ||||||
As shown in Figure 1, as the interval from admission until initiation of treatment lengthened, mortality increased. Mortality was substantially and significantly higher in all HIV-negative age groups (p < 0.0001 from chi-squared test) and also in HIV-positive patients if treatment was delayed 1 week or more (Figure 2). After adjustment for all significant patient characteristics, initially missed diagnosis and delayed treatment were strongly associated with death and late ICU admissions, as shown in Table 2.
Fig. 1. Association of mortality with interval in days from admission until treatment. All patients admitted within 4 days to an ICU (early ICU admission) were excluded from this analysis.
[More] [Minimize]
Fig. 2. Association of mortality with interval from admission until treatment initiated, among patients of all ages and HIV status admitted to an ICU within 4 days, all other HIV-positive patients, and all other HIV-negative patients categorized into three age groups. Gray bars, treatment within less than 1 week of admission; black bars, treatment initiated after 1 week or more of hospitalization.
[More] [Minimize]| Diagnosis Missed in First 24 Hours | Treatment Delayed by 1 Week or More | Mortality | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Odds Ratio | (95% CI) | Odds Ratio | (95% CI) | Odds Ratio | (95% CI) | |||||||
| Indicators | ||||||||||||
| Treatment delayed by 1 week or more | ||||||||||||
| Unadjusted | 14.5 | (8.9, 24) | — | — | — | — | ||||||
| Adjusted | 14.5 | (7.8, 26) | — | — | — | — | ||||||
| Bronchoscoped (all patients) | ||||||||||||
| Unadjusted | 1.4 | (1.0, 2.1) | 2.1 | (1.4, 3.1) | 1.6 | (0.9, 2.8) | ||||||
| Adjusted | 1.7 | (1.1, 2.6) | 2.4 | (1.5, 3.7) | 1.8 | (0.96, 3.5) | ||||||
| Bronchoscoped (smear positive only) | ||||||||||||
| Unadjusted | 1.6 | (1.02, 2.7) | 2.5 | (1.4, 4.6) | 2.0 | (0.9, 4.3) | ||||||
| Adjusted | 1.7 | (0.98, 3.1) | 2.6 | (1.4, 5.1) | 2.0 | (0.8, 5.1) | ||||||
| Admitted first to nonmedical† | ||||||||||||
| Unadjusted | 4.3 | (2.8, 6.5) | 2.3 | (1.5, 3.5) | 2.7 | (1.4, 5.3) | ||||||
| Adjusted | 4.0 | (2.5, 6.0) | 1.8 | (1.2, 2.8) | 3.1 | (1.5, 6.6) | ||||||
| Outcomes | ||||||||||||
| Early ICU admission | ||||||||||||
| Unadjusted | 2.9 | (1.4, 5.9) | 0.9 | (0.4, 2.0) | 6.2 | (3.2, 12.3) | ||||||
| Adjusted | 2.8 | (1.3, 6.3) | 0.8 | (0.3, 1.8) | 8.2 | (3.3, 20.2) | ||||||
| Late ICU admission† | ||||||||||||
| Unadjusted | 12.6 | (2.5, 64) | 23.0 | (5.4, 99) | 10.6 | (3.7, 31) | ||||||
| Adjusted | 7.3 | (0.9, 60) | 16.8 | (2.0, 144) | 10.0 | (2.2, 46) | ||||||
| Death | ||||||||||||
| Unadjusted | 4.9 | (2.6, 9.2) | 4.3 | (2.4, 7.6) | — | — | ||||||
| Adjusted | 4.3 | (2.0, 9.1) | 3.3 | (1.7, 6.5) | — | — | ||||||
Of the 35 patients admitted directly or transferred to an ICU within 4 days of hospitalization, 14 (40%) died, compared with 5 of the 10 patients (50%) admitted to an ICU after a median of 18 days (range 7–52). Compared with the 384 never in the ICU, or to the 35 with early ICU admission, the 10 patients with late ICU admission were somewhat more likely to have been older, female, Canadian born, with less cough, and sputum that was AFB smear negative (data not shown in tabular form). After adjustment for these potentially confounding characteristics, both early and late ICU admission were associated with initially missed diagnosis and mortality (Table 2). However, only late ICU admission was associated with delayed treatment.
Excluding the 35 patients with early ICU admission, 151 (30%) were admitted first to a surgical or other nonmedical ward or were held in the emergency room for more than 24 hours. Even though the characteristics of these patients were similar to those of patients admitted to a medical ward, they were significantly more likely to have initially been mis-diagnosed, had delayed treatment, undergone bronchoscopy, or died (see Table 2).
As shown in Table 3, characteristics of the patients with TB were similar in the three groups of hospitals. However, initially missed diagnosis, admission to nonmedical wards, bronchoscopy, and delayed treatment were all significantly more common in patients admitted to the hospitals with low and intermediate TB admission rates. The interval from admission until treatment was significantly longer in hospitals with low and intermediate admission rates than in hospitals with higher TB admission rates, as shown in Figure 3. When adjusted for potentially confounding differences in the characteristics of patient populations served at different hospitals, indicators of care and outcomes were significantly worse in hospitals with low and intermediate TB admission rates, particularly among smear-positive patients (Table 4).
| TB Admissions per 10,000 Admissions Annually | ||||||||
|---|---|---|---|---|---|---|---|---|
| Low (0.2–3.3) | Intermediate (3.4–9.9) | Moderate to High (> 10.0) | p Value (from chi-squared) | |||||
| Hospitals, number | 7 | 7 | 2 | |||||
| Ventilation | ||||||||
| Nonisolation rooms, ACPH | 3.62 | 3.34 | 5.06 | — | ||||
| Isolation rooms, ACPH | 7.48 | 6.79 | 7.78 | — | ||||
| All patients, number | 79 | 223 | 127 | |||||
| Characteristics | ||||||||
| Mean age, yr | 55 | 52 | 46 | 0.01 | ||||
| Born in Canada or other low incidence | 41% | 60% | 49% | 0.006 | ||||
| HIV positive, % of tested | 15% | 15% | 5% | — | ||||
| Cavity on chest X-ray, % | 22% | 35% | 30% | — | ||||
| Smear positive, % | 62% | 65% | 51% | 0.04 | ||||
| Extrapulmonary disease, % | 20% | 26% | 17% | — | ||||
| Indicators | ||||||||
| Initially missed diagnosis, % | 56% | 58% | 16% | < 0.0001 | ||||
| Delayed treatment, % | 43% | 35% | 12% | < 0.0001 | ||||
| Initial admission to nonmedical, % | 43% | 58% | 3% | < 0.0001 | ||||
| Bronchoscoped, % | 51% | 40% | 34% | < 0.0001 | ||||
| Outcomes | ||||||||
| Early ICU admission, % | 11% | 7% | 8% | — | ||||
| Late transfer to ICU, % | 1% | 4% | 1% | — | ||||
| Died, % | 23% | 12% | 6% | < 0.0001 | ||||
| Smear positive patients, number | 47 | 145 | 65 | |||||
| Indicators | ||||||||
| Initial mis-diagnosis, % | 51% | 50% | 6% | < 0.0001 | ||||
| Delayed treatment, % | 34% | 28% | 5% | < 0.0001 | ||||
| Initial admission to nonmedical, % | 38% | 53% | 0 | < 0.0001 | ||||
| Bronchoscoped, % | 53% | 41% | 29% | 0.04 | ||||
| Outcomes | ||||||||
| Early ICU admission, % | 11% | 8% | 8% | — | ||||
| Late transfer to ICU, % | 0 | 2% | 0 | — | ||||
| Died, % | 19% | 12% | 5% | 0.05 | ||||
| Workers, number | 515 | 540 | 212 | |||||
| Years of work | 9.2 | 10.4 | 11.1 | 0.001 | ||||
| TST conversion, % | 20.4% | 21.3% | 7.55% | < 0.0001 | ||||
| Average annual TST conversion, % | 2.22% | 2.05% | 0.68% | |||||
Fig. 3. Interval from admission until treatment in hospitals categorized by their TB admission rate. Solid line, high admission rate; long dashed line, intermediate rate; short dashed line, very low and low rates (see text for definition of rates).
[More] [Minimize]| Hospitals with Very Low and Low TB Admission Rates versus High† | Hospitals with Intermediate TB Admission Rates versus High† | |||||||
|---|---|---|---|---|---|---|---|---|
| Odds Ratio | (95% CI) | Odds Ratio | (95% CI) | |||||
| All patients | ||||||||
| Indicators‡ | ||||||||
| Initial mis-diagnosis | 17.0 | (7.1, 41) | 3.7 | (2.6. 5.1) | ||||
| Delayed treatment | 7.4 | (3.2, 17.5) | 2.3 | (1.6, 3.2) | ||||
| Bronchoscoped | 1.7 | (0.9, 3.3) | 1.2 | (0.9, 1.5) | ||||
| Outcomes | ||||||||
| Late transfer to ICU | 2.3 | (0.1, 56) | 2.3 | (0.7, 7.1) | ||||
| Death | 3.8 | (1.2, 12.1) | 1.5 | (0.96, 2.4) | ||||
| Smear positive patients | ||||||||
| Indicators‡ | ||||||||
| Initial mis-diagnosis | 37.3 | (9.2, 151) | 4.7 | (2.6, 8.4) | ||||
| Delayed treatment | 13.7 | (3.1, 60) | 3.0 | (1.5, 5.7) | ||||
| Bronchoscoped | 2.1 | (0.9, 5.2) | 1.2 | (0.9, 1.7) | ||||
| Outcomes § | ||||||||
| Death | 2.2 | (0.4, 11.7) | 1.6 | (0.8, 3.3) | ||||
Crude TST conversion rates (Table 3) and unadjusted odds of conversion (Table 5) were paradoxically greater in hospitals with low to intermediate TB admission rates than in hospitals with the highest TB admission rates. After adjustment for workers' characteristics, odds of TST conversion remained significantly higher in workers in hospitals with low to intermediate TB admission rates, hospitals with higher mortality, or units with more frequent delayed diagnosis or treatment of patients with TB. TST conversion was not associated with indices of patient disease severity (data not shown). Because exposure to airborne pathogens can be substantially modified by air exchange rates (29), odds of TST conversion were adjusted for ventilation in isolation and nonisolation rooms. After this adjustment, TST conversion remained significantly associated with the overall mortality of patients with TB, but odds of conversion was substantially lower in hospitals with very low and low TB admission rates. This suggests that the higher TST conversion rates in these hospitals were attributable, at least in part, to lower air exchange rates in nonisolation patient rooms.
| Unadjusted | Adjusted for Worker Characteristics* | Adjusted for Workers and Hospital Factors† | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Indicator of Exposure | Odds Ratio | (95% CI) | Odds Ratio | (95% CI) | Odds Ratio | (95% CI) | ||||||
| Unit level | ||||||||||||
| More than half of TB patients are undiagnosed in first 24 hours‡ | 1.5 | (1.1, 2.0) | 1.5 | (1.1, 2.1) | 1.0 | (0.7, 1.5) | ||||||
| More than 30% of TB patients are treated only after 1 week or more‡ | 2.4 | (1.5, 3.6) | 1.7 | (1.04, 2.9) | 0.7 | (0.2, 1.2) | ||||||
| Hospital level | ||||||||||||
| More than half of TB patients are first admitted to nonmedical ward‡ | 3.6 | (2.6, 5.0) | 2.9 | (2.0, 4.3) | 1.1 | (0.6, 2.1) | ||||||
| More than one third of smear positive TB patients are bronchoscoped‡ | 3.3 | (2.3, 4.7) | 2.4 | (1.6, 3.7) | 1.4 | (0.8, 2.3) | ||||||
| Mortality rate of TB patients ⩾ 10%‡ | 2.9 | (2.1, 4.0) | 2.0 | (1.4, 2.8) | 2.4 | (1.6, 3.7) | ||||||
| Hospital TB admission rate | ||||||||||||
| Very low (< 0.4/10,000) | 1.7 | (0.5, 5.4) | 0.8 | (0.2, 2.8) | 0.1 | (0.1, 0.4) | ||||||
| Low (< 0.4–3.3/10,000) | 3.1 | (1.8, 5.3) | 2.4 | (1.3, 4.6) | 0.5 | (0.2, 1.3) | ||||||
| Intermediate (3.4–9.9/10,000) | 3.3 | (2.0, 5.6) | 1.7 | (1.2, 2.3) | 0.9 | (0.6, 1.3) | ||||||
In the hospitals studied, initially missed diagnosis and delayed treatment of patients with active TB were common and strongly associated with late ICU admissions and in-hospital mortality. Moreover, as the number of TB admissions decreased, patient outcomes were worse and delays in diagnosis and treatment more frequent—increasing the risk of exposure per patient. Tuberculin conversion among workers was associated with indicators of suboptimal care of patients with TB, particularly mortality. This association was modified by ventilation in nonisolation rooms—where exposure to undiagnosed patients presumably occurred.
In agreement with our findings, population-based studies (i.e., nonoutbreak situations) in hospitals in industrialized countries consistently demonstrate that initially missed or delayed diagnosis and treatment are associated with older age (1, 5, 6, 31, 32), HIV infection (4, 33-36), noncavitary chest X-ray (1, 2, 4), absence of cough and sputum (4, 36), and negative AFB smears (1, 2, 5, 6, 36). All these are, directly or indirectly, atypical features of active TB. Delayed diagnosis of such patients might be prevented through improved awareness by health care workers of the potential atypical manifestations of TB. However, during the 3 years reviewed, the 429 new tuberculosis patients represented less than 0.05% of the estimated 890,000 patients admitted to the 17 participating hospitals. Because this was even less in the hospitals with lowest rates, where delays were greatest and patient outcomes worst, prompt and accurate diagnosis of patients with TB, particularly in the presence of atypical features, is likely to remain a major challenge.
Mortality of patients with TB was 12%, well within the range of 8–25% of other North American studies (6, 37), and was associated with older age, HIV infection, and ICU admission—consistent with studies conducted in similar settings (6, 31, 32, 38-40). The association of mortality with negative AFB smears and absence of cough contradicts studies from the preantibiotic era (41), but is consistent with recent studies in which such characteristics were associated with delayed diagnosis and related mortality (37, 42, 43). Of the 10 patients with late ICU admission (1 week or more), nine had delayed treatment and five died (50%), despite clinical features consistent with less extensive disease. Previous reports have suggested that diagnosis and treatment of TB among patients already admitted to an ICU may be delayed (38, 39). Among the 35 patients with early ICU admission, TB diagnosis was more likely to be missed in the first 24 hours, but delayed treatment was not more common. Mortality in these patients presumably reflected their more serious disease. On the other hand, delayed treatment appeared to result in late ICU admission and subsequent death. These findings, together with the evidence summarized in Figures 1 and 2, support the contention that delayed treatment was a major cause of, not simply associated with, in-hospital mortality from TB.
The association of selected patient care indicators with patient outcomes and workers' TST conversions deserve comment. The proportion of smear-positive patients bronchoscoped was selected as an indicator of exposure, because bronchoscopy of these patients can result in significant exposure (29, 30). This may also be considered an indicator of serendipitous diagnosis of patients with more advanced disease in whom the diagnosis could have been made from a sample of spontaneous sputum—had it been considered. Similarly, the proportion of patients with active pulmonary TB admitted to nonmedical wards may be an indicator of initial misdiagnosis, as well as potential mismanagement by caregivers unfamiliar with TB.
The association of TST conversion with overall mortality rate of patients with TB might have reflected the more extensive disease in patients who died or more likelihood of missed diagnosis and delayed treatment. Regardless of the reason, hospital mortality rate may be a better summary indicator of nosocomial transmission, because it combines extent of patients' disease, undiagnosed TB deaths, and rapidity of diagnosis, isolation, and treatment of patients. An important and striking finding was that these indicators of patient care and patient outcomes improved and TST conversion decreased as the rate of TB admissions increased, even after adjustment for potentially confounding differences in patient characteristics. Greater experience in management of TB may not only reduce delay in diagnosis, as suggested elsewhere (1, 2, 43), but also improve patient outcomes and reduce nosocomial transmission.
A major strength of this study was that it was a population-based study of a large number of health care workers and patients with TB in 17 university and community hospitals. No center had detected an outbreak, and the number of TB cases per center ranged from 1 to 135 over the 3 years reviewed. Therefore, our results should be applicable to acute-care hospitals in low-incidence countries with a similar wide range of experience with TB. However, because we studied only hospitalized patients with newly diagnosed pulmonary disease (our primary goal was to estimate nosocomial transmission), these results may not be applicable to out-patients nor to patients with extrapulmonary TB.
To study a phenomenon such as missed or delayed diagnosis, a retrospective design must be used, because a prospective study would likely substantially increase providers' awareness of TB and result in changed behavior. However, a retrospective design has a number of limitations. Patients with active TB would not have been included if they were completely missed—because they were discharged or died without performance of an autopsy. Certain clinical information, such as results of HIV testing, were not available on all patients. Clinicians may have selectively tested sicker patients because mortality was lower in untested patients than in HIV-negative patients. Delayed diagnosis and treatment were defined on the basis of written orders for respiratory isolation and/or treatment. In some cases, the diagnosis may have been suspected but not sufficiently to take action, while in others the orders may have been written but not performed.
Accurate measurement of nosocomial transmission is like trying to see the wind, because undiagnosed patients are the most important sources of transmission (29, 42, 44-46). Contrary to current North American guidelines (11, 12, 23) that risk of nosocomial transmission is proportional to the absolute number of TB admissions, in this study risk was lower in centers with highest TB admissions, because occurrence of missed diagnosis and delayed treatment was lowest, resulting in lower risk of exposure per hospitalized patient. Risk estimation should consider, in addition to absolute numbers, other proxies of exposure. These include initially missed diagnosis, delayed treatment, admission to nonmedical wards, smear-positive patients undergoing bronchoscopy, and mortality. All are relatively simple to measure. Exposure from undiagnosed patients seems inevitable in low incidence settings, but could be mitigated by better ventilation in nonisolation rooms where these patients will likely be cared for. The findings of this study suggest that a much greater reduction of transmission, plus improvement in patient outcomes could be achieved through greater awareness by physicians of the protean manifestations of TB. However, as the number of TB cases steadily diminishes across North America, maintenance of the necessary knowledge and skills will likely be a major challenge.
The authors thank the administration, infection control, and employee health departments of the participating hospitals and particularly wish to thank the many participating health care workers for their time, help, and collaboration.
Supported by the National Health and Development Program (NHRDP) of Health Canada, Grant No. 6605-4437-502, the Association Pulmonaire du Québec, a Chercheur Boursier Clinicien award from the Fonds de Recherche de la Santé du Québec 1993–1998 (D.M.), and a Medical Scientist award from the Medical Research Council of Canada (D.M.).
| 1. | Venkatarama KR, Iademarco EP, Fraser VJ, Kollef MHDelays in the suspicion and treatment of tuberculosis among hospitalized patients. Ann Intern Med1301999404411 |
| 2. | L'Ecuyer PB, Woeltje KF, Seiler SM, Fraser VJManagement and outcome of tuberculosis in two St. Louis hospitals, 1988 to 1994. Infect Control Hosp Epidemiol191998836841 |
| 3. | Liam CK, Tang BGDelay in the diagnosis and treatment of pulmonary tuberculosis in patients attending a university teaching hospital. Int J Tuberc Lung Dis11997326332 |
| 4. | Moran GJ, McCabe F, Morgan MT, Talan DADelayed recognition and infection control for tuberculosis patients in the emergency department. Ann Emerg Med261995290295 |
| 5. | Mathur P, Sacks L, Auten G, Sall R, Levy C, Gordin FDelayed diagnosis of pulmonary tuberculosis in city hospitals. Arch Intern Med1541994306310 |
| 6. | Counsell SR, Tan JS, Dittus RSUnsuspected pulmonary tuberculosis in a community teaching hospital. Arch Intern Med149198912741278 |
| 7. | Fischl MA, Uttamchandani RB, Daikos L, Poblete RB, Moreno JN, Reyes RR, Boota AM, Thompson LM, Cleary TJ, Oldham SA, et al.An outbreak of tuberculosis caused by multiple-drug resistant tubercle bacilli among patients with HIV infection. Ann Intern Med1171992177183 |
| 8. | Pearson ML, Jereb JA, Frieden TR, Crawford JTNosocomial transmission of multi-drug resistant Mycobacterium tuberculosis: a risk to patients and health care workers. Ann Intern Med1171992191196 |
| 9. | Edlin BR, Tokars JI, Grieco MH, Crawford JT, Williams J, Sordillo EM, Ong KR, Kilburn JO, Dooley SW, Castro KG, et al.. An outbreak of multi-drug resistant tuberculosis among hospitalized patients with the acquired immunodificiency syndrome. N Engl J Med326199215141521 |
| 10. | Menzies RI, Fanning A, Yuan LTuberculosis among health care workers. N Engl J Med33219959298 |
| 11. | Centers for Disease Control and Prevention. Recommendations and reports. Guidelines for preventing the transmission of mycobacterium tuberculosis in health-care facilities. MMWR 1994;43(RR13):1–132. |
| 12. | Health Canada. Guidelines for preventing the transmission of tuberculosis in Canadian Health Care facilities and other institutional settings. Ottawa: Canadian Communicable Disease Reports; 1996. p. 1–50. |
| 13. | IUATLD and WHO. Control of tuberculosis transmission in health care setting: a joint statement of the International Union Against Tuberculosis and Lung Disease (IUATLD) and the Tuberculosis Programme of the World Health Organization. Paris, France: IUAT; 1993. |
| 14. | Blumberg HM, Watkins DL, Jeffrey PA-C, Antle A, Moore P, White N, Hunter M, Green B, Ray SM, McGowan JEPreventing the nosocomial transmission of tuberculosis. Ann Intern Med1221995658663 |
| 15. | Maloney SA, Pearson ML, Gordon MT, Del Castillo R, Boyle JF, Jarvis WREfficacy of control measures in preventing nosocomial transmission of multidrug-resistant tuberculosis to patients and health care workers. Ann Intern Med12219959095 |
| 16. | Barnhart S, Sheppard L, Beaudet N, Stover B, Balmes JTuberculosis in health care settings and the estimated benefits of engineering controls and respiratory protection. J Emerg Med391997849854 |
| 17. | Bangsberg DR, Crowley K, Moss A, Dobkin JF, McGregor C, Neu HCReduction in tuberculin skin-test conversions among medical house staff associated with improved tuberculosis infection control practices. Infect Control Hosp Epidemiol181997566570 |
| 18. | Scott B, Schmid M, Nettleman MDEarly identification and isolation of inpatients at high risk for tuberculosis. Arch Intern Med1541994326330 |
| 19. | Kellerman S, Tokars JI, Jarvis WRThe cost of selected tuberculosis control measures at hospitals with a history of Mycobacterium tuberculosis outbreaks. Infect Control Hosp Epidemiol181997542547 |
| 20. | Raad I, Cusick J, Sherertz RJ, Sabbagh M, Howell NAnnual tuberculin skin testing of employees at a university hospital: a cost-benefit analysis. Infect Control Hosp Epidemiol101989465468 |
| 21. | Menzies RI, Fanning A, Yuan L, FitzGerald JMHospital ventilation and risk of tuberculous infection in Canadian Health Care Workers. Ann Intern Med1332000779789 |
| 22. | Menzies RI, Schwartzman K, Loo V, Pasztor JMeasuring ventilation of patient care areas in hospitals: description of a new protocol. Am J Respir Crit Care Med152199519921999 |
| 23. | Long RE. Canadian tuberculosis standards. 2000/2001 edition. Toronto: Canadian Lung Association; 2000. |
| 24. | Menzies DInterpretation of repeated tuberculin tests. Am J Respir Crit Care Med15919991521 |
| 25. | American Thoracic Society. Diagnostic standards and classification of tuberculosis. Am Rev Respir Dis 1990;142:725–735. |
| 26. | Dawson-Saunders B, Trapp RG. Basic and clinical biostatistics. Norwalk, CT: Appleton & Lange; 1990. |
| 27. | Kleinbaum DG, Kupper LL. Applied regression analysis and other multi-variate methods. North Scituate, MA: Duxbury Press; 1978. |
| 28. | Cox DR. Regression models and life-tables (with discussion). J R Stat Soc 1972;B34:187–220. |
| 29. | Catanzaro ANosocomial tuberculosis. Am Rev Respir Dis1251982559562 |
| 30. | Malasky C, Jordan T, Potulski F, Reichman LBOccupational tuberculosis infections among pumonary physicians in training. Am Rev Respir Dis1421990505507 |
| 31. | Liaw Y-S, Yang P-C, Yu C-J, Wu Z-G, Chang D-B, Lee L-N, Kuo S-H, Luh K-TClinical spectrum of tuberculosis in older patients. JAGS431995256260 |
| 32. | Dahmash NS, Fayed DF, Chowdhury MNH, Arora SCDiagnostic challenge of tuberculosis of the elderly in hospital: experience at a university hospital in Saudi Arabia. J Infect3119959397 |
| 33. | Alpert PL, Munsiff SS, Gourevitch MN, Greenberg B, Klein RSA prospective study of tuberculosis and human immunodeficiency virus infection: clinical manifestations and factors associated with survival. Clin Infect Dis241997661668 |
| 34. | Kramer F, Modilevsky T, Waliany AR, Leedom JM, Barnes PFDelayed diagnosis of tuberculosis in patients with human immuno-deficiency virus infection. Am J Med891990451456 |
| 35. | Pablos-Mendez A, Sterling T, Frieden TRThe relationship between delayed or incomplete treatment and all-cause mortality in patients with tuberculosis. JAMA276199612231228 |
| 36. | Sherman LF, Fujiwara PI, Cook SV, Bazerman LB, Frieden TRPatient and health care system delays in the diagnosis and treatment of tuberculosis. Int J Tuberc Lung Dis3199910881095 |
| 37. | Rao VK, Iademarco EP, Fraser VJ, Kollef MHThe impact of comorbidity on mortality following in-hospital diagnosis of tuberculosis. Chest114199812441252 |
| 38. | Penner C, Roberts D, Kunimoto D, Manfreda J, Long RTuberculosis as a primary cause of respiratory failure requiring mechanical ventilation. Am J Respir Crit Care Med1511995867872 |
| 39. | Heffner JE, Strange C, Sahn SAThe impact of respiratory failure on the diagnosis of tuberculosis. Arch Intern Med148198811031108 |
| 40. | Levy H, Kallenbach JM, Deldman C, Thorburn JR, Abramowitz JAAcute respiratory failure in active tuberculosis. Crit Care Med151987221225 |
| 41. | Rieder HL. Epidemiologic basis of tuberculosis control. 1st ed. Paris, France: International Union Against Tuberculosis and Lung Disease; 1999. p. 1–162. |
| 42. | Naalsund A, Heldal E, Johansen B, Kongerud J, Boe JDeaths from pulmonary tuberculosis in a low-incidence country. J Intern Med2361994137142 |
| 43. | Enarson DA, Grzybowksi S, Dorken EFailure of diagnosis as a factor in tuberculosis mortality. CMAJ118197815201522 |
| 44. | Nicas MAssessing the relative importance of the components of an occupational tuberculosis control program. J Emerg Med401998648654 |
| 45. | Ehrenkranz NJ, Kicklighter JLTuberculosis outbreak in a general hospital: evidence for airborne spread of infection. Ann Intern Med771972377382 |
| 46. | Griffith DE, Hardeman JL, Zhang Y, Wallace RJ, Mazurek GHTuberculosis outbreak among healthcare workers in a community hospital. Am J Respir Crit Care Med1521995808811 |
Members of the Canadian Collaborative Group (in addition to the above-listed authors):
Montreal: Dr. Gilles Blanchette, Hôpital Sacré-Coeur; Dr. Philippe Bolduc, Hôpital Le Gardeur; Dr. Paul Brassard, Unité de Santé Publique; Dr. Marc Charbonneau, Hôpital Notre-Dame; Dr. Serge Déry, Hôpital Ste-Croix; Dr. Michael Libman, St-Mary's Hospital; Dr. Mark Miller, Jewish General Hospital; Dr. Pierre Robillard, Unité de Santé Publique; Dr. Kevin Schwartzman, Montreal Chest Institute and Royal Victoria Hospital; Dr. Terry Nan Tannenbaum, Unité de Santé Publique; Dr. Jacques Tremblay, Hôpital Maisonneuve-Rosemont. Toronto: Dr. Monika Avandano, Westpark Hospital; Dr. John Conly, Toronto Hospital; Dr. Andrew Simor, Sunnybrook Hospital; Dr. Monika Naus, Ontario Ministry of Health. Edmonton: Dr. Geoff Taylor, University Hospital; Dr. Manuel Ma, Dr. Edith Blondel-Hill, Royal Alexandra Hospital. Vancouver: Dr. Julio Montaner, St-Paul's Hospital.
