Assessment of a clinical score for screening suspected pulmonary tuberculosis cases

Abstracts

OBJETIVO: Avaliar acurácia de escore clínico (sensibilidade) no diagnóstico presuntivo de tuberculose pulmonar em triagem. MÉTODOS: Estudo descritivo-analítico transversal com 1.365 pacientes atendidos no setor de pneumologia em Unidade Básica de Saúde de nível secundário da cidade do Rio de Janeiro, RJ, de 2006 a 2007. Os participantes responderam um questionário padronizado, aplicado por equipe de enfermagem, contendo informações referentes à idade, peso e sintomas clínicos. O resultado presuntivo do diagnóstico de tuberculose pulmonar foi obtido pela soma da pontuação dos dados coletados. Diagnóstico de tuberculose ativa baseou-se nos resultados bacteriológicos e na decisão médica. Foram calculados sensibilidade, especificidade, valores preditivos positivos e negativos para uma prevalência especificada, e intervalos de 95% de confiança para diversos pontos de corte do escore. O desempenho do escore foi avaliado pela curva receiver operating characteristic (ROC). RESULTADOS: Para o diagnóstico de tuberculose, tosse > 1 semana e > 3 semanas mostrou sensibilidade respectivamente de 88,2% (86,2;90,2) e de 61,1% (57,93;64,3), especificidade de 19,2% (16,6;21,8) e 51,3% (48,1;54,5). O escore clínico com 8 pontos mostrou uma sensibilidade de 83,13% (77,8;87,6), especificidade de 51,8% (48,5;55,1), valor preditivo positivo de 91,6% (90,0;83,2) e negativo 32,9% (30,1;35,7). CONCLUSÕES: Tosse (> 3 sem) apresentou baixa sensibilidade e especificidade. Escore clínico com elevada sensibilidade pode ser uma ferramenta alternativa na detecção de tuberculose pulmonar, pois, além de agilizar o atendimento do caso suspeito na unidade, permite padronizar a primeira abordagem pela enfermagem.

Tuberculose Pulmonar; Diagnóstico Clínico; Técnicas e Procedimentos Diagnósticos; Triagem; Sensibilidade e Especificidade

ORIGINAL ARTICLES

Cristiano Bel Alves de Castro^I; Paulo Albuquerque da Costa^II; Antonio Ruffino-Netto^II; Ethel Leonor Noia Maciel^IV; Afranio Lineu Kritski^V

^IPrograma de Pós-Graduação em Clinica Médica. Universidade Federal do Rio de Janeiro (UFRJ). Rio de Janeiro, RJ, Brasil

^IIInstituto de Doenças do Tórax. UFRJ. Rio de Janeiro, RJ, Brasil

^IIIFaculdade de Medicina de Ribeirão Preto. Universidade de São Paulo. Ribeirão Preto, SP, Brasil

^IVLaboratório de Epidemiologia. Programa de Pós-Graduação em Saúde Coletiva. Universidade Federal do Espírito Santo. Vitória, ES, Brasil

^VDepartamento de Clínica Médica. Faculdade de Medicina. UFRJ. Rio de Janeiro, RJ, Brasil

Correspondence

ABSTRACT

OBJECTIVE: To assess the accuracy (sensitivity) of a clinical score for presumptive pulmonary tuberculosis cases during screening.

METHODS: Descriptive cross-sectional study comprising 1,365 patients attending the department of lung diseases at a secondary care outpatient clinic in the city of Rio de Janeiro, Southeastern Brazil, during 2006 and 2007. All respondents answered a standardized questionnaire administered by the clinic's nursing staff. Information on age, weight and clinical symptoms were collected. The presumptive diagnosis of pulmonary tuberculosis was made by summing up the scores of the data collected. The diagnosis of active tuberculosis was based on bacteriological findings and medical criteria. There were estimated sensitivity, specificity, positive predictive value and negative predictive value for a set prevalence, and 95% confidence intervals for different score cutoffs. The score performance was assessed using the receiver operating characteristic (ROC) curve.

RESULTS: For the diagnosis of tuberculosis, cough for more than one week and cough for more than three weeks showed a sensitivity of 88.2% (86.2, 90.2) and 61.1% (57.93, 64.3), specificity 19.2% (16.6, 21.8) and 51.3% (48.1, 54.5), respectively. The clinical score of 8 had a sensitivity of 83.13% (77.8, 87.6), specificity of 51.8% (48.5, 55.1), positive predictive value of 91.6% (90.0, 83.2) and negative predictive value of 32.9% (30.1, 35.7).

CONCLUSIONS: Cough for more than three weeks showed low sensitivity and specificity. A highly sensitive clinical score can be an alternative tool for screening pulmonary tuberculosis as it allows early care of suspected cases and standard nursing care approach.

Descriptors: Tuberculosis, Pulmonary, diagnosis. Clinical Diagnosis. Diagnostic Techniques and Procedures. Triage. Sensitivity and Specificity.

RESUMEN

OBJETIVO: Evaluar la exactitud de escore clínico (sensibilidad) en el diagnostico presuntivo de tuberculosis pulmonar en pesquisa.

MÉTODOS: Estudio descriptivo-analítico transversal con 1.365 pacientes atendidos en el sector de pneumonologia en Unidad Básica de Salud de nivel secundario de la ciudad de Rio de Janeiro, Sureste de Brasil, de 2006 a 2007. Los participantes respondieron un cuestionario estandarizado, aplicado por equipo de enfermería, conteniendo informaciones referentes a la edad, peso y síntomas clínicos. El resultado presuntivo del diagnostico de tuberculosis pulmonar fue obtenido por la suma de puntuación de los datos colectados. Diagnostico de tuberculosis activa se basó en los resultados bacteriológicos y en la decisión medica. Se calcularon sensibilidad, especificidad, valores predictivos positivos negativos para una prevalencia especificada, e intervalos de 95% de confianza para diversos puntos de corte del escore. El desempeño del escore fue evaluado por la curva receiver operating characteristic (ROC).

RESULTADOS: Para diagnostico de tuberculosis, tos > 1 semana y > 3 semanas mostró sensibilidad respectivamente de 88,2% (86,2;90,2) y de 61,1% (57,93;64,3), especificidad de 19,2% (16,6;21,8) y 51,3% (48,1;54,5). El escore clínico con 8 puntos mostró una sensibilidad de 83,13%(77,8;87,6), especificidad de 51,8% (48,5;55,1), valor predictivo positivo de 91,6% (90,0;83,2) negativo 32,9% (30,1;35,7).

CONCLUSIONES: Tos (>3 sem) presentó baja sensibilidad y especificidad. Escore clínico con elevada sensibilidad puede ser una herramienta alternativa en la detección de tuberculosis pulmonar, pues además de agilizar la atención del caso sospechoso en la unidad, permite estandarizar el primer abordaje por la enfermera.

Descriptores: Tuberculosis Pulmonar, diagnostic. Diagnóstico Clínico. Técnicas y Procedimientos Diagnósticos. Triaje. Sensibilidad y Especificidad.

INTRODUCTION

The World Health Organization (WHO) Stop TB Program was launched in 2006 as part of the Global Plan designed to scale up tuberculosis (TB) control. It intends to develop and assess new tools and strategies for TB control, diagnosis and treatment that would effectively allow to eliminating TB.¹⁵

It has become a priority to find and treat early active (bacilliferous) cases so that infection control measures can be implemented, thus preventing the contact of these patients with children, elderly and immunosuppressed individuals in waiting and examination rooms.ª

The WHO and other international organizations have suggested the implementation of TB control measures in so-called "risk environments." They defined high-risk environments for Mycobacterium tuberculosis transmission among patients, healthy individuals for the patient or the patient for health professionals.^6,16

In recent years infection control measures in Brazil have been mostly recommended for and implemented in hospital or inpatient settings that require longer and more intensive care.^6,

The actions implemented at primary care units (PCU) are part of a major TB control strategy as they have an impact on two main aspects of the disease: early detection and adequate flow of care for patients with pulmonary TB. They reduce the risk of Mycobacterium tuberculosis infection among providers and other patients or family members who attend the health unit.¹⁵

However, a major challenge remains the early detection of patients with pulmonary symptoms and potentially suspected cases among users of PCUs. Thus, preventive measures can be more quickly and effectively implemented, especially in priority municipalities with high rates of TB, as is the case of the city of Rio de Janeiro, Southeastern Brazil. ⁹

This study aimed to assess the sensitivity, specificity and accuracy of a clinical score for presumptive diagnosis of pulmonary TB.

METHODS

A descriptive cross-sectional study was conducted. The study population consisted of patients cared at the pulmonology department of a secondary care outpatient clinic of a hospital in the city of Rio de Janeiro, from September 4, 2006 to July 3, 2007. The hospital is located in an area with approximately 1.1 million inhabitants and provides outpatient care to the districts of Guadalupe, Deodoro, Costa Barros, Pavuna, Acari, Anchieta, and Ricardo de Albuquerque (about 330,000 inhabitants). In 2000, the incidence rate of pulmonary TB reported in this area was approximately 100 cases per 100,000 inhabitants per year.

The questionnaire developed for this study was based on Santos et al¹³ recommendations. Factors associated with pulmonary TB in patients treated at a university hospital identified by multiple regression analysis were assessed. A panel of experts consisting of two pulmonologists and a nurse selected the easiest questions to be included in the questionnaire so that no embarrassment would be caused to patients during the screening interview. It was set a zero-to-six score where six was the highest clinical evidence score for these variables, generating a scoring system for use in patients with suspected pulmonary TB (Table 1).

All interviews were conducted on alternate days by two nurses who were trained on questionnaire administration and standard operating procedure. Data collected were analyzed using SPSS 13.0.

In addition to their clinical evaluation, all symptomatic patients had their diagnosis of pulmonary TB confirmed by bacteriological examination, bacilloscopy with Ziehl-Neelsen staining and mycobacterial culture. Culture isolation of mycobacteria was done using the Löwenstein-Jensen medium in the Mycobacteriology Laboratory at the Clementino Fraga Filho University Hospital of Universidade Federal do Rio de Janeiro and Fundação Oswaldo Cruz Evandro Chagas Institute of Clinical Research.ª HIV testing was offered only to those patients diagnosed with pulmonary TB following the Brazilian Ministry of Health guidelines.

Other study variables included: chest pain (pain, pressure or discomfort in the chest); cough (involuntary, spasmodic and particularly audible expulsion of breathed air caused by foreign bodies in the larynx or irritation of the mucosa in the trachea and bronchi); sputum expectoration (coughing up discharge from the lungs); hemoptysis (expectoration of bright red blood); fever (body temperature above 37.2ºC or feeling hot without checking the temperature); shortness of breath (difficult breathing with the sensation of an incomplete breath); night sweats (excessive sweating during the night soaking bedclothes and/or bedding); weight loss (loss of 3 kg in individuals weighing up to 70 kg and loss of 5 kg in those over 70 kg); and alcohol abuse (identified using the Cut down, annoyed, guilty, eye-opener - CAGE tool).²

The incidence rate of pulmonary TB was estimated by dividing the number of patients diagnosed with TB by the total number of people living in the area (about 330,000 inhabitants).

For the univariate analysis, there were assessed the frequency distribution, measures of central tendency and association between variables using the chi-square test for categorical variables and Student's t-test for continuous variables. The probability of statistical significance was estimated by dividing patients into two groups: pulmonary TB and non-pulmonary TB patients. For hypothesis testing a 5% level of significance was used.

Sensitivity was defined as the percentage of patients identified by the score as having pulmonary TB among those with positive bacteriological examination or meeting clinical criteria for pulmonary TB. Specificity was defined as the percentage of patients identified by the score as non-pulmonary TB among those with negative bacteriological examination and no clinical evidence of TB. There were excluded those patients who were initially not considered as having TB but were diagnosed with pulmonary TB within two years from the date of their enrollment in the study according to the Sistema de Informação de Agravos de Notificação (SINAN - Notifiable Diseases Surveillance Database). Positive predictive value †(PPV) and negative predictive value (NPV) were defined as the probability of a patient with suspected pulmonary TB have or do not have active pulmonary TB, respectively, by the proposed score. Accuracy was defined as the proportion of correct predictions (sum of true positives and true negatives) and expresses the degree of certainty to make decisions on patient selection based on the score. Accuracy ranges from 0% to 100% and the higher the accuracy the greater the likelihood of identifying patients with pulmonary TB using the proposed score. The related 95% confidence intervals were also estimated.

Duration of cough was used to define symptomatic patients. Its accuracy was assessed based on the bacteriological examination. Then a scenario different pulmonary TB prevalences was used to assess the predictive power (PPV and NPV) of the clinical score at a cutoff >8.

The score performance was assessed using the receiver operating characteristic (ROC) curve. The ROC curve provides the cutoff values with more optimal sensitivity (vertical axis) against the complement of specificity (horizontal axis) which corresponds to a point on the ROC curve nearest to the upper left corner of the ROC graph as true-positive rate is 1 and false-positive rate is 0. The area under the curve was used to test its discriminative capacity: an area under the curve (AUC) lower than 0.6 indicates poor fit; 0.7, reasonable; 0.8, good; and higher than 0.9, excellent.¹

To validate the cut-off score for the diagnosis of pulmonary TB the bootstrap method was used. This method generates samples with the same size n of the study sample by drawing and replacement of all elements and comparison of the samples generated.

The study project was approved by the Research Ethics Committee of the Clementino Fraga Filho University Hospital, Universidade Federal do Rio de Janeiro (protocol no. 067/06).

RESULTS

During the study period 1,365 symptomatic patients sought care at the health unit, of which 1,148 were screened with the study questionnaire. Of these, 243 (20.9%) were diagnosed with pulmonary TB, 905 (77.7%) did not have pulmonary TB and 16 (1.4%) were first diagnosed as non-TB and then reported as TB cases in the SINAN within two years of their enrollment in the study. The incidence rate of TB in the sample studied was 73.6 per 100,000 inhabitants per year. Among 1,148 symptomatic patients, 134 (16.7%) were alcohol dependent, 673 (58.6%) were smokers and 272 (23.6%) had history of prior TB treatment. Among those diagnosed with pulmonary TB, 160 (65.8%) were HIV tested, of which 18 (11.2%) tested positive for HIV.

The comparison between screened and non-screened patients showed that 50.3% of screened and 48.8% of non-screened individuals were males (p=0.734). The prevalence of pulmonary TB was 20.9% and 25.4% (p=0.215) in the screened and non-screened patients without any statistically significant differences between these variables.

Table 2 shows the variables included in the clinical score according to TB and non-TB patients. All variables showed a statistically significant association, except for shortness of breath.

Table 3 shows the association of the symptom cough with the diagnosis of pulmonary TB due to its clinical importance. The prevalence of cough in the study sample was 85% (976/1.148). The highest sensitivities were found when there was cough for more than one week.

The clinical score for pulmonary TB showed sensitivity of 3.3% to 100%, specificity of 1.0% to 99.6%, and accuracy of 21.9% to 79.1%. The cut-off >8 showed the highest sensitivity and accuracy (Table 4).

The AUC for the pulmonary TB score was 0.734. The choice of a cutoff of 8 ("reasonable" fit in this model) favored a higher sensitivity compared to a cutoff of 9.

The internal validation score >8 assessed using the bootstrap method with 40 samples showed that in 39 (97.5%) the mean was within the mean range estimated for the sample studied (0.5263 to 0.5793).

The test of normality of 40 samples using the Shapiro-Wilk test showed a p-value = 0.17, which confirms the null hypothesis of a normal distribution and validation of the clinical score.

Table 5 shows the expected results of PPV and NPV with the use of the clinical score in scenarios of different pulmonary TB prevalence, at a cutoff >8 as the best performance, with high NPV for pulmonary TB prevalences usually seen at primary care units (5% to 10%).

DISCUSSION

The incidence of reported pulmonary TB cases in the state of Rio de Janeiro in 2006 was 79.1 per 100,000 inhabitants/year,ª slightly higher than that found in the current study.

There was no statistically significant difference of gender and pulmonary TB between screened and non-screened patients, suggesting potentially homogeneous groups.

There was found a proportionally larger number of individuals younger than 59 than among patients diagnosed with pulmonary TB than among non-TB patients, similar to that reported in the literature.^6,b This finding confirms that pulmonary TB affects a greater number of people during their most productive years of life, although some studies have indicated that in Brazil there is a shift in the incidence of pulmonary TB to the elderly.^3,11,12

The association between chest pain and pulmonary TB was also found in other studies that described chest pain as a clinical sign of pulmonary TB^13,14 while others considered it only a typical acute manifestation of extrapulmonary (pleural) tuberculosis.¹⁶

Similar to other studies,^4,7 cough was reported in 94.7% of patients with pulmonary TB. According to the WHO recommendations, the Brazilian Ministry of Health guidelines and several authors, active search for new TB cases should be made among people with cough and/or sputum expectoration for more than two or three weeks (symptomatic individuals), hence the importance of cough as a symptom.^9,d However, the study population was drawn from a referral secondary care outpatient clinic and it is a convenience sample from a department specialized in pulmonary diseases with low prevalence of HIV-TB co-infection (11%). As described in previous studies with populations with similar characteristics, cough for more than one week is a suggestive symptom leading to a presumptive diagnosis of pulmonary TB.¹

Sputum expectoration and hemoptysis showed an association with pulmonary TB in the patients studied, which corroborates the literature.^6,d However, the proportion of hemoptysis was higher than that reported in Macedo studies.⁷

The proportion of fever, night sweats, shortness of breath and weight loss was similar to that reported in other studies.^7,8

Our results suggest that the clinical score can be a useful tool for screening of pulmonary TB cases in primary care settings in areas with low prevalence of HIV infection. In addition, it can be applied by nursing staff. Because it is a simple, effective and efficient tool, no costly complex equipment is required. When used by trained providers, the questionnaire can provide better quality information because screening will not be based exclusively on the criterion cough, as recommended by WHO and the Brazilian Ministry of Health.^9,15,16

Nursing staff are the first providers to see suspected TB cases. They often identify pulmonary symptoms (cough for more than 2 or 3 weeks), and make decisions about exams and referral, taking a subjective approach, which makes it difficult to assess quality of care and compare their performance.

The clinical score can also be used for selecting patients for more costly and complex diagnostic examinations such as mycobacterial culture, chest X-rays and/or other molecular tests. Thus, it is recommended that further research be undertaken in different scenarios of TB prevalence with different cultural backgrounds to better assess the score's applicability and its impact on morbidity/mortality of patients receiving care in the health system.

REFERENCES

• 1. Bastos LG, Fonseca LS, Mello FC, Ruffino-Netto A, Golub JE, Conde MB. Prevalence of pulmonary tuberculosis among respiratory symptomatic subjects in an out-patient primary health unit. Int J Tuberc Lung Dis. 2007;11(2):156-60.
• 2. Castells MA, Furlanetto LM. Validity of the CAGE questionnaire for screening alcohol-dependent inpatients on hospital wards. Rev Bras Psiquiatr. 2005;27(1):54-7. DOI:10.1590/S1516-44462005000100012
• 3. Chaimowicz F. Age transition of tuberculosis incidence and mortality in Brazil. Rev Saude Publica. 2001;35(1):81-7. DOI:10.1590/S0034-89102001000100012
• 4. Job JRPP, Prado PEBS, Vranjac S, Duarte PC. Comparação de dados epidemiológicos da tuberculose pulmonar em Sorocaba, SP, Brasil,em uma década (1986 - 1996). Rev Saude Publica. 1998;32(6):596-7. DOI:10.1590/S0034-89101998000600014
• 5. Johnson RA, Wichern DW. Applied multivariate statistical analysis. Englewood Cliffs: Prentice Hall; 1982.
• 6. Marais BJ, Raviglione M, Donald PR, Harries AD, Kritski AL, Graham SM, et al. Scale-up of services and research priorities for TB diagnosis, management and control: call to action. Lancet 2010;375(9732):2179-91. DOI:10.1016/S0140-6736(10)60554-5
• 7. Macedo LG, Afiune JB, Melo FAF. Características clínicas, radiológicas e bacteriológicas de pacientes portadores de tuberculose pulmonar. J Pneumol. 1992;18(1):118.
• 8. Miller LG, Asch SM, Yu EI, Knowles L, Gelberg L, Davidson P. A population-based survey of tuberculosis symptoms: how atypical are atypical presentations? Clin Infect Dis. 2000;30(2):293-9. DOI:10.1086/313651
• 9. Moreira CMM, Zandonade E, Lacerda T, Maciel ELN. Sintomáticos respiratórios nas unidades de atenção primária no Município de Vitória, Espírito Santo, Brasil. Cad Saude Publica. 2010;26(8):1619-26. DOI:10.1590/S0102-311X2010000800015
• 10. Rodrigues PM, Moreira TR, Moraes AKL, Vieira RCA, Dietze R, Lima RCD, et al. Infecção por Mycobacterium tuberculosis entre agentes comunitários de saúde que atuam no controle da TB. J Bras Pneumol. 2009;35(4):351-8. DOI:10.1590/S1806-37132009000400009
• 11. Moreira TR, Zandonade E, Maciel ELN. Risco de infecção tuberculosa em agentes comunitários de saúde. Rev Saude Publica. 2010;44(2):332-8. DOI:10.1590/S0034-89102010000200014
• 12. Ruffino-Netto A. Tuberculosis: the neglected calamity. Rev Soc Bras Med Trop. 2002; 35(1):51-8. DOI:10.1590/S0037-86822002000100010
• 13. Santos AM, Pereira BB, Seixas JM, Mello FCQ, Kritski AL. Neural networks: an application for predicting smear negative pulmonary tuberculosis. In: Balakrishnan N, Auget JL, Mesbah M, Molenberghs G, organizers. Advances in statistical methods for the health sciences: applications to cancer and AIDS studies, genome sequence analysis, and survival analysis. Boston: Birkhäuser Boston; 2007. p.279-89.
• 14. Santos JWA. Causas pulmonares de dor torácica. Rev AMRIGS. 2006;46(2):25-7.
• 15. Vendramini SH, Santos ML, Gazetta CE, Chiaravalloti-Neto F, Ruffino-Netto A, Villa TC. Tuberculosis risks and socio-economic level: a case study of a city in the Brazilian south-east, 1998-2004. Int J Tuberc Lung Dis. 2006;10(11):1231-5.
• ¹⁶
  World Health Organization. Global tuberculosis control: epidemiology, strategy, financing: WHO report 2009. Geneva: World Health Organization; 2009. (WHO/STM/TB/2009.411).

Assessment of a clinical score for screening suspected pulmonary tuberculosis cases

Evaluación de un escore para rastreo de sospechosos de tuberculosis pulmonar

b

However, it was only in 2010 that the risk of TB infection in primary care units became a matter of concern among health policy makers

^10,11 as a result of increased rates of TB infection reported among community health workers that provide care to TB patients in the Family Health Program in the municipality of Cachoeiro do Itapemirim, state of Espírito Santo, southeastern Brazil.

^c

The normal distribution of samples generated was assessed using the Shapiro-Wilk test. This test calculates a W-statistic that tests whether a random sample of size n comes from a normal distribution.

⁵

Publication Dates

• Publication in this collection
  23 Sept 2011
• Date of issue
  Dec 2011

History

• Received
  23 Oct 2010
• Accepted
  30 May 2011