Prevalence and clinical predictors of pulmonary tuberculosis among isolated inpatients: a prospective study


  • This study was presented at the 20th ECCMID, Vienna, Austria (10–13 April 2010), abstract 2717.

Corresponding author:M. Lagrange-Xélot, Department of Infectious Diseases, Hopital Saint-Louis, 1 avenue Claude Vellefaux, 75010 Paris, France


Clin Microbiol Infect 2011; 17: 610–614


Guidelines help to prevent the transmission of Mycobacterium tuberculosis in healthcare settings, but may also result in the unnecessary isolation of many patients. We performed a prospective study to assess the prevalence and identify clinical predictors of culture-proven tuberculosis among inpatients isolated for suspected pulmonary tuberculosis (PTB) at our hospital. We also wished to validate a pre-existing clinical decision rule to improve our isolation policy. From August 2005 to January 2007, 134 patients isolated on admission to the ward for suspicion of PTB were prospectively enrolled. The admitting team made the decision to isolate patients on the basis of clinical and radiological findings, without the use of the clinical decision rule, and graded the overall suspicion of PTB. Twenty-six of the 134 isolated patients had PTB (prevalence: 19.4%), as well as one patient not isolated at admission. Univariate analysis revealed that PTB was significantly associated with young age, lack of human immunodeficiency virus (HIV) infection, weight loss, night sweats, fever, upper lobe disease and, especially, cavitary lesions on chest X-ray (adjusted OR 25.4, p <0.0001). Low suspicion of PTB by the admitting team and low clinical decision rule score had negative predictive values of 98.5% and 95.8% for PTB, respectively. Use of the clinical decision rule in addition to the team assessment would have led to the isolation of the patient with PTB not isolated on admission, and avoided 16 (14.8%) unnecessary isolations. In conclusion, the prevalence of PTB among isolated inpatients was high, and the use of a clinical decision rule in addition to clinical impression might improve isolation decisions.


Transmission of tuberculosis (TB) in healthcare settings has become a growing concern in the era of human immunodeficiency virus (HIV) infection and multidrug-resistant TB [1,2]. Guidelines based on the early detection and isolation of patients with suspected pulmonary TB (PTB), recommending the use of single-patient rooms with environmental and respiratory-protection controls until three sputum smears for acid-fast bacilli (AFB) are negative, have led to a decrease in the risk of acquiring TB for both healthcare workers and patients [3–6]. However, these measures may also result in the unnecessary isolation of many patients at low risk for PTB, with significant organizational constraints for the ward, psychological pressure on the isolated patients and their visitors, and an overall increase in hospital costs.

A number of clinical decision rules (CDRs) based on clinical and radiological findings on admission to the hospital have been developed to help identify patients who are at high or low risk of PTB, and thus improve decisions about respiratory isolation of these patients [7–12].

We performed a prospective study to assess the prevalence and identify predictive clinical factors of culture-proven TB among inpatients isolated for suspected PTB in our ward, and assessed whether the CDR developed by Wisnivesky would improve the selection of inpatients for respiratory isolation [9,10].

Materials and Methods

Patient population

The study was conducted in the Department of Infectious Diseases of the Saint-Louis Hospital, a 563-bed tertiary-care institution in Paris. The infectious disease ward is a 30-bed unit. The incidence of TB in Paris is 19.7/100 000, more than twice the national rate, with 1.3% of cases being multidrug-resistant TB. The institutional review board of the hospital approved the study.

From 1 August 2005 to 31 January 2007, all patients admitted to the ward and isolated because of suspected PTB, as recommended by French guidelines, were prospectively enrolled in the study. The decision to isolate patients was not based on Wisnivesky’s CDR, but made by the admitting team, who graded the suspicion of PTB as high, low or intermediate. The admitting team comprised infectious diseases specialists with good clinical experience in TB. Patients were not enrolled in the study if they were receiving anti-TB medication at the time of admission, or had a known positive AFB sputum smear.

The respiratory isolation policy in our ward is to isolate on admission patients with suspected PTB in single-patient rooms, with strict respiratory protection of healthcare workers and visitors, who are instructed to use FFP2 masks before entering the room, and with a strong recommendation for the isolated patient to put on a surgical mask when someone enters the room. The ward does not have negative-pressure rooms, air filtration or controlled ventilation, but windows remain open for at least 4 h a day to dilute contaminated air.

Data collection

Data were prospectively collected within 24 h of patients’ admission, from patients’ charts, before the results of the AFB sputum smears were available. Demographic data, the presence of TB risk factors, clinical symptoms and findings from physical examination, laboratory tests and chest radiography were recorded. Demographic data included age, sex and country of origin. Risk factors for TB included a previous history of or exposure to TB, a history of positive tuberculin skin test findings, homelessness, previous stays in social shelters or prisons, and HIV infection. Clinical symptoms included significant weight loss (at least 10% of body weight), chills, night sweats of ≥3 weeks’ duration, persistent fever for ≥3 weeks, a history of shortness of breath, cough, purulent sputum, haemoptysis, chest pain and rhinorrhea. Data from the physical examination included body temperature on admission, oxygen saturation level, the presence of crackles and/or rhonchi/wheezing during chest examination, and the presence of peripheral lymphadenopathy. Laboratory data included neutrophil and platelet counts, and haemoglobin, C-reactive protein, aspartate aminotransferase and alkaline phosphatase levels.

Chest radiographs were also reviewed for the presence of upper lobe involvement, reticulonodular infiltrate, consolidation, cavitations, pleural effusion and the presence of mediastinal lymphadenopathy.

Finally, a score was calculated for each patient according to the CDR defined and validated by Wisnivesky and summarized in Table 1 [9,10]. Patients with a score of −6–0 should not be isolated, and patients with a score of at least 1 should be isolated. This score was not used for the clinical management of the enrolled patients.

Table 1.   Wisnivesky’s clinical prediction rule and point scoring systema
VariablePoints assigned
  1. aWisnivesky et al. [9]. Patients with a score of −6–0 should not be isolated, and patients with a score of at least 1 should be isolated.

  2. bAny of the following: exposure to an individual with tuberculosis, institutionalization (prison, shleter, nursing home) in the past 3 years, homelessness, weight loss ≥10% of body weight, night sweats for ≥3 weeks, symptoms of malaise or weakness for ≥3 months, and persistent fever.

Tuberculosis risk factors or chronic symptomsb4
Self-reported positive tuberculin skin test5
Shortness of breath−3
Temperature (°C)
Crackles on physical examination−3
Upper lobe disease on chest radiographs6

Sputum sample assessment

A patient was considered to have active PTB if at least one sputum culture was positive for the Mycobacterium tuberculosis complex. Sputum samples were concentrated and stained with Auramine O fluorescent stain for smear examination and AFB quantification. For cultures, samples were inoculated in Bactec 960 MGIT and Coletsos Medium. Cultures were maintained for at least 42 days to detect the presence of growing organisms.

Statistical analysis

Data are described as mean (standard deviation) or count (%). Baseline characteristics of isolated patients with or without PTB were compared with the Fisher exact test for proportions and the Welsch modification of the t-test or Wilcoxon rank sum test for continuous variables. The strength of association between baseline characteristics and PTB was expressed in terms of ORs with 95% CIs.

The classification performances of the admitting team and of the Wisnivesky CDR for the diagnosis of PTB were assessed by calculating sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV), with 95% CIs. The potential reduction in the number of unnecessary isolated patients with the use of the CDR was calculated.

All tests were two-sided, and p-values <0.05 were regarded as indicating statistical significance. Analyses were performed using the R 2.6.2 statistical package (R Foundation for Statistical Computing, Vienna, Austria).


During the study period, 1207 patients were admitted to our ward, 716 of whom were HIV-infected (prevalence: 59%), and 134 patients (11.1%) were placed in respiratory isolation on admission for suspected PTB. Baseline characteristics of these isolated patients are shown in Table 2. Most patients were male (70%), with a mean age of 43 years; 47% originated from sub-Saharan Africa; 22% of them had a history of TB; and 45% were HIV-infected. Cough was reported by 62% of patients, and was the most frequent clinical symptom. Also, 46% of patients had a pulmonary infiltrate on chest X-ray, involving upper lobes in 43%.

Table 2.   Comparison of baseline characteristics between patients with and without culture-proven pulmonary tuberculosis (PTB)
n = 134
Patients with PTB
n = 26
Patients without PTB
n = 108
p-ValueOR (95% CI)a
  1. ASAT, aspartate aminotransferase; HIV, human immunodeficiency virus; SD, standard deviation; TB, tuberculosis; TST, tuberculin skin test.

  2. aFor continuous variables, ORs are given for a cut-off value at the sample median except otherwise stated.

  3. bObtained by exact logistic regression.

 Age (years), mean (SD)43 (14)37 (12)45 (14)0.00750.62/10 years (0.42–0.91)
 Male sex, N (%)94 (70)19 (73)75 (69)0.811.19 (0.46–3.11)
 Country of origin
  Europe35 (26)6 (23)29 (27)0.891 (reference)
  Sub-Saharan Africa63 (47)12 (46)51 (47) 1.14 (0.39–3.35)
  Other36 (27)8 (31)28 (26) 1.38 (0.42–4.49)
 TB risk factors, N (%)49 (39)9 (39)40 (38)>0.991.03 (0.41–2.60)
 Previous history of TB, N (%)30 (22)3 (12)27 (25)0.190.39 (0.11–1.41)
 TST >5 mm, N (%)6 (4)1 (4)5 (5)>0.990.82 (0.09–7.37)
  HIV infection, N (%)60 (45)7 (27)53 (49)0.0490.38 (0.15–0.98)
Clinical symptoms
 Weight loss ≥10%, N (%)47 (35)14 (54)33 (31)0.0382.65 (1.11–6.35)
 Chills, N (%)27 (20)4 (16)23 (21)0.780.70 (0.22–2.23)
 Night sweats >3 weeks, N (%)33 (25)11 (42)22 (21)0.0402.83 (1.14–7.03)
 Fever >3 weeks, N (%)35 (27)11 (42)24 (23)0.0502.51 (1.02–6.17)
 Shortness of breath, N (%)41 (31)5 (19)36 (34)0.160.46 (0.16–1.33)
 Cough, N (%)82 (62)16 (62)66 (62)>0.990.99 (0.41–2.40)
 Purulent sputum, N (%)30 (22)7 (27)23 (21)0.601.36 (0.51–3.63)
 Haemoptysis, N (%)11 (8)3 (12)8 (7)0.451.61 (0.40–6.56)
 Chest pain, N (%)25 (19)5 (19)20 (19)>0.991.04 (0.35–3.08)
 Rhinorrhea, N (%)3 (2)0 (0)3 (3)>0.991.05 (0–9.95)b
 Body temperature (°C), mean (SD)37.6 (1.1)38 (1.0)37.5 (1.0)0.024 
  N (%) ≥38°C49 (37)16 (64)33 (31)0.00264.04 (1.62–10.1)
 Percentage oxygen saturation, mean (SD)96.9 (2.4)97 (1.5)96.9 (2.6)0.811.01 (0.42–2.46)
 Crackles, N (%)29 (22)5 (19)24 (22)>0.990.82 (0.28–2.41)
 Rhonchi/wheezing, N (%)23 (17)5 (19)18 (17)0.771.19 (0.40–3.57)
 Peripheral lymphadenopathy, N (%)44 (33)7 (27)37 (34)0.640.71 (0.27–1.83)
Laboratory values
 Neutrophil count, ×109/L4.4 (3–8)6.0 (3.4–8.2)3.9 (2.8–8.0)0.152.66 (0.98–8.97)
 Haemoglobin (g/dL), mean (SD)12.2 (2.0)12.1 (1.8)12.2 (2.1)0.751.17 (0.49–2.75)
 Platelet count, ×109/L276 (212–359)328 (246–380)265 (204–341)0.142.19 (0.90–5.35)
 C-reactive protein (mg/L)50 (14–104)57 (20–112)43 (10–104)0.301.43 (0.58–3.49)
 ASAT (IU/L)26 (19–38)28 (20–43)26 (19–37)0.591.17 (0.49–2.75)
 Alkaline phosphatase (IU/L)79 (60–113)86 (63–126)76 (60–111)0.411.72 (0.72–4.14)
Chest radiographs
 Upper lobe disease, N (%)58 (43)18 (69)40 (37)0.00393.82 (1.52–9.60)
 Pulmonary infiltrate, N (%)62 (46)15 (58)47 (44)0.271.77 (0.74–4.21)
 Consolidation, N (%)33 (25)8 (31)25 (23)0.451.46 (0.57–3.75)
 Cavitation, N (%)14 (11)11 (42)3 (3)<0.000125.4 (6.35–101.7)
 Pleural effusion, N (%)17 (13)5 (19)12 (11)0.321.88 (0.60–5.93)
 Mediastinal lymph nodes, N (%)11 (8)2 (8)9 (8)>0.990.91 (0.18–4.48)

Twenty-six of the 134 isolated patients had culture-proven PTB, with M. tuberculosis identified in 24 cases, and Mycobacterium africanum in two. Fourteen (53.8%) of these patients had a positive AFB sputum smear. The prevalence of PTB among isolated inpatients in our study was therefore 19.4% (95% CI 13.6–26.7), and was 2.2% among admitted patients.

Among the 108 isolated patients who did not have PTB, nine presented with extrapulmonary TB. Three HIV-infected patients had Mycobacterium avium identified in sputum cultures, one of whom had a positive AFB sputum smear.

In the univariate analysis, patients with PTB in our study were significantly younger, less frequently HIV-infected and had more frequently experienced weight loss, night sweats and fever than patients without TB (Table 2). Patients with TB also had more frequent upper lobe involvement and more cavitary lesions on chest X-ray. There were no differences in respiratory symptoms and laboratory parameters between patients with or without PTB.

The performances of the admitting team and of the Wisnivesky CDR for the diagnosis of PTB are reported in Table 3. A high or intermediate suspicion of PTB by the admitting team was significantly associated with PTB (p <0.0001), with a sensitivity of 96.2%, a specificity of 63.8%, a PPV of 39%, and an NPV of 98.5%. Indeed, only one patient with a low suspicion of PTB and negative AFB sputum smear had PTB. This 36-year-old HIV-negative man from India was admitted for a weight loss of 3 kg, with haemoptysis, without fever or night sweats, and with mediastinal lymphadenopathies on chest X-ray without visible infiltrate. He had a suspicion of lymphoma and his Wisnivesky score was 4.

Table 3.   Concordance of clinical impression of the admitting team and of the Wisnivesky clinical prediction rule in isolated patients with and without culture-proven pulmonary tuberculosis (PTB)
Clinical suspicion of PTBPTB
N = 26
N = 108
W < 1W ≥ 1W < 1W ≥ 1
  1. W, Wisnivesky score.


The Wisnivesky score was also significantly associated with the diagnosis of PTB, with a sensitivity of 96.2%, a specificity of 21.3%, a PPV of 22.7%, and an NPV of 95.8%. Use of the CDR would have correctly identified all but one patient with PTB. This 25-year-old HIV-negative man, originating from Mali, was admitted for a 2-week history of low-grade fever, and a thoracic abscess; he had mediastinal lymphadenopathies and rib osteolysis with a left lower lobe consolidation on chest X-ray. He had negative AFB sputum smears, with a Wisnivesky score of 0, and was isolated with an intermediate suspicion of PTB.

However, during the study period, one patient not isolated on admission was eventually found to have culture-proven PTB, with a positive AFB sputum smear. This patient had a Wisnivesky score of 6, and would have been isolated had the score been used for the decision regarding isolation. This 39-year-old HIV-infected man from the Democratic Republic of Congo was admitted with a diagnosis of acute bacterial pneumonia of the left lower lobe, and AFB sputum smears were ordered only after 5 days of unsuccessful antibiotic treatment.

Among patients without PTB, 69 had a low suspicion of PTB at admission, and 16 of these patients also had a Wisnivesky score of <1 (Table 3). If the Wisnivesky CDR had been used in addition to the admitting team impression, among the 108 patients isolated without PTB, 16 (14.8%) patients could have avoided isolation.


Guidelines for preventing the transmission of M. tuberculosis in healthcare settings are effective, but delays in recognition and isolation of patients with active PTB may still occur [3–6,13]. Strict enforcement of these guidelines may also result in the unnecessary isolation of many patients at low risk for PTB, especially in low-endemicity areas [7,9]. Therefore, regular assessment of these policies is recommended, including review of new tools to improve isolation decisions. CDRs are based on standardized clinical and radiological findings available at the time of patient admission, and can help to improve isolation decisions [7–12].

In our prospective assessment of our PTB isolation policy, 134 (11.1%) of the 1207 admitted patients were isolated for suspicion of PTB. The prevalence of culture-proven PTB among these patients was 19.4%, much higher than in previous studies [10]. We failed, however, to isolate one patient with active PTB until 5 days after admission, although the CDR would have recommended doing so. This failure to isolate 1/27 (3.7%) patients with active PTB has public health implications, both for other patients and for healthcare workers [14].

When assessing clinical and radiological predictors of PTB among isolated patients in our study, we were able to identify, in the univariate analysis, the presence of cavitations on chest X-ray as the main factor significantly associated with the presence of PTB. This factor is used in most CDRs for the isolation of patients with suspected PTB [7–12,15].

However, we were unable to identify predictive factors associated with the absence of PTB on admission. Although a low suspicion of PTB by the admitting team had an NPV for PTB in our study (98.5%), one isolated patient with a low clinical suspicion of PTB was found to have PTB. The use of the Wisnivesky CDR would have correctly identified this patient as needing isolation. Had the Wisnivesky CDR been applied to the 69 patients without PTB who had a low suspicion at admission, 16 patients with CDR scores below 1 would have avoided unnecessary isolations (Table 3).

However, the Wisnivesky CDR alone would not have performed better than the admitting team, as one patient with TB would have been missed and not isolated. Also, we could not assess in this study the total number of patients who would have been isolated had the Wisnivesky CDR been applied to the whole study population. Indeed, in this study, the Wisnivesky CDR did not perform as well as in others, with a somewhat lower specificity and NPV of 21.3% and 95.8%, respectively [9,10,16]. Decision rules should be used to help in decision-making, not to replace the clinical assessment.

Our study had several limitations. The experience of the admitting physicians in this study might explain the good performance of clinical suspicion as compared with the CDR. The number of patients with culture-confirmed PTB was low, and we should be careful before generalizing our findings. Also, our study focused on patients with culture-proven PTB; only 53.8% of them had positive AFB sputum smears and were therefore highly contagious. Finally, our results may only apply to the patient population that we investigated, i.e. inpatients in a ward with a high prevalence of HIV infection.

In conclusion, we suggest that the use of a CDR in addition to the clinical judgement of the admitting team might improve our decisions regarding the need for isolation of patients admitted with suspected PTB. Such a benefit needs to be further evaluated in a prospective study to assess the adequate isolation of all contagious patients while minimizing the unnecessary isolation of patients without PTB.

Transparency Declaration

The authors declare no conflicting interests.