Study design and study population

We performed a cross-sectional study during a three-month period from April to June 2008. The reference population was composed of patients diagnosed with all forms of TB in DOTS centres (21 in total) with more than 100 registered patients. There were 27 DOTS Centres in the Luanda province. DOTS centres (6 in total) with less than 100 patients registered in 2007 were excluded from the study due to the small size and because they were located in difficult to reach peripheral rural areas. Eight of the 21 DOTS centres were located in the suburban districts of the province (Samba, Cacuaco and Viana), where the population density for each district was less than 900 inhabitants per km² and the nearest health facility was more than 3 km away. Thirteen DOTS centres were settled in the urban districts where the density population was higher and the nearest health facility was less than 2 km away.

According to the NTP, smear-positive pulmonary TB were defined the patients with two or more sputum smears for acid fast bacilli (AFB) or one sputum positive for AFB and radiological abnormalities consistent with active TB; smear-negative PTB were defined the patients with three negative sputum smears for AFB and radiological abnormality consistent with active TB or failure to respond to antibiotics trials. Extrapulmonary TB (EPTB) patients were defined the patients with TB in organs other than the lungs proven by histo-pathology or TB based on strong clinical evidence consistent with active EPTB and the decision by a physician to treat with a full course of anti-TB therapy.

Overall 10.574 TB patients were registered in the 21 DOTS centres in the previous year. The sample size of 385 subjects was determined using the Taro Yamane formula for a finite population [19] and then proportionally stratified to the 21 selected DOTS centres. All patients diagnosed with TB of all forms according to the national TB guidelines [6] and coming to the selected DOTS centres were included consecutively and interviewed just before starting treatment, up to reach the sample size defined for the single centre. Patients who started treatment prior to interview were excluded.

Patient delay was defined as the time from symptom onset to first consultation. Symptom onset referred to the time at which the first symptom (i.e. persistent cough, fever, weakness, weight loss or chest pain) of the illness for which a patient seeking care began. Health system delay was defined from the date of the patient’s first contact with the health care service to the date of diagnosis. Total delay was defined as the sum of patient delay and health system delay. The majority of the quantitative studies measured delay as a dichotomus variable, typically as approximately 1 month or more for patient delay [16, 20–22], and 15 days for health system delay [16, 21, 23, 24]. In other studies median value of the observed data was used as a cut-off [24, 25]. Following the consultation with experts of the NTP and with treating physicians, we categorised patient delay and health care system delay by using a cut-off point of >30 days and >15 days respectively. In our study the centre of first contact is the first place where the patient seeks health assistance: the TB disease could be suspected in these places, but the TB diagnosis of confirmation was done in the DOTS centres, where the patients received the TB therapy free of charge.

Survey instrument

A structured questionnaire was developed addressing factors suspected to be involved with health-seeking behaviour based on clinical experience and review of relevant literature (Additional file 1). The questionnaire collected information on the following attributes: i) demographic and socio-economic variables: age, sex, occupation at the time of survey, income before starting treatment, educational level, family size, use of alcohol and tobacco; ii) distance to health facilities, costs of travel to health facility and medical expenditure on treatment of TB, waiting time at the centre (time incurred between arrival at the DOTS centre and receiving therapy); iii) patient knowledge about TB, its cause, gravity and treatment (able to mention bacteria/germ/microbe as a cause of TB, to categorise TB as a transmissible disease, to recognise measures to prevent the transmission, to be able to mention that TB is a treatable disease and the approximate duration of treatment); iv) perception of level and quality of the service received at the DOTS centre (taking into acount the opening time of the centre, the waiting time in the centre, the attitude of the health personnel, do you consider to be; i) definitively satisfied, ii) to some extent satisfied or iii) not satisfied with DOTS centre care received?). The questionnaire was drafted in Portuguese, the official language of Angola. It was pre-tested prior to the start of the study, with 40 patients at the Luanda Lebrosy Dispensary, the province’s biggest DOTS centre. The pre-test allowed the identification of misinterpreted questions. Modifications were then incorporated in the final version. Each interview took place before starting the treatment and lasted approximately 20 minutes. Three trained interviewers carried out the fieldwork under the supervision of a physician. A written consent was secured from every eligible tuberculosis patient before inclusion into the study. Informed consent and responses for children under 18 years of age were obtained from the closest accompanying relative. The study was approved by the ethics committees of the Health direction of Luanda province.

Data analysis

Data were analysed using SPSS for windows version 17. Beyond descriptive statistics, associations between the dependent variables (patient delay and health system delay) and the independent variables were analysed by calculating the ORs test and respective 95% confidence intervals (CI). Responses to questions to assess TB knowledge were analysed by calculating their mean and interquartile scores. Using the mean score as a cut-off, the responses were categorised into good knowledge (above or equal the mean) and poor knowledge (below the mean) [26].

Independent variables that showed marginal associations (p < 0.20) in the bivariate analysis were used in a logistic regression analysis in order to identify independent predictors of both patient delay > 30 days and health system delay > 15 days. The association of predictor variables with the dependent variables was assessed by using 95% Cl and adjusted odds ratio (AOR). A p-value of <0.05 was regarded as statistically significant.

Socio-economic characteristics of study participants

Patients’ delay

Patient delay had a median value of 30 days (IQR 14–60 days) and a mean value of 71.37 days (SD 133.94). The distribution of the reported duration of symptoms prior to seeking help is shown in Figure 1. Factors potentially associated with prolonged patient delay were analysed using bivariate analysis for delays of over four weeks, as shown in Table 3. Factors associated with delay included primary education, volume of DOTS centres ≥ 600 patients per year, no food in the DOTS centre and health centre of the first contact if this was different from the DOTS centre. In the logistic regression analysis, primary education (AOR = 1.75; CI (95%) 1.06–2.88; p <0.029) and health centre of the first contact if this was different from the DOTS centre (AOR = 1.66; CI (95%) 1.01–2.75; p <0.046) remained independent risk factors for patient delay >4 weeks.

Health systems’ delay

The median health system delay was seven days (IQR 5–15) and its mean value was 28.66 days (SD 82.38). The distribution of the reported duration of system delay is shown in Figure 2. Of the factors investigated, these were the three independent predictors of health service delay >2 weeks, after adjusting for other variables (Table 3) namely: living in a suburban area (AOR = 2,32; CI (95%) 1.21–4.46; p = 0.011) waiting time in the centre >1 hour (AOR = 4.37; CI (95%) 1.72–11.14; p = 0.002) and the centre of the first contact differening from the DOTS centre (AOR = 5.68; CI (95%) 2.72–11,83; p < 0,00001). In the logistic regression analysis, the centre of the first contact differening from the DOTS centre showed the following results: hospital AOR = 17.66; CI (95%) 1.1–284,42 (p = 0.043); health care provider AOR = 35.57; CI (95%) 2.37–533.1 (p = 0.01); private health care worker AOR = 1.90; CI (95%) 1.06–3.43 (p = 0.032); self-medication AOR = 1.98; CI (95%) 1.00–3.90 (p = 0.05).

We analysed if the system delay was influenced by the patient delay: the chi-square test was not significant (p = 0.12).

Total delay

Among the 385 subjects, the median total delay was 45 days (IQR 21–97) and the mean total delay was 100.43 days (SD 100.43). For 61 (15%) subjects, the total delay was ≥ 20 weeks. The chi-square test between subjects with a total delay < 20 weeks and subjects with a total delay ≥ 20 weeks resulted not significant for the variables considered.

Patients’ delay

In our setting the patient delay is 30 days and four times higher than the system delay. In a recent review Sreeramareddy et al.[18], found that among the low and middle income countries patient delay varied from 4.9 days in Gambia [27] to 162 days in Tanzania [20]. The average patient delay in low and middle income countries was 31.7 days. It is likely that most studies underestimate the duration of illness/disease before diagnosis, since patients often do not clearly recognise symptoms at the onset of the disease. This has been demonstrated in several recent TB prevalence surveys, in which less than half of the cases fulfil the criteria of being a “TB suspect” and many have only vague symptoms [28]. One can debate if such delay is a patient delay (if a person does not feel sufficiently ill, why take action?) or a delay in diagnosis caused by the nature of the disease. In the current study, only one socio-demographic characteristic appeared to correlate with patients’ delay, namely primary education. Primary educated patients experienced much longer diagnostic delays than those fewer years of education. However, this data is contradicted in patients with secondary education. Perhaps in a proportion of the population primary education rates do not translate into greater awareness of TB. Although higher levels of education can be considered an indicator of TB knowledge, recent evidence suggests that TB treatment initiation is more influenced by perceptions than by knowledge [29]. Lay perceptions may explain why people continue and others may stop taking treatment and this regardless of their educational level [29, 30]. The acknowledgement of social, economic and geographical context is for sure necessary to understand the impact of traditional beliefs and perceptions of illness [31, 32]. Several studies for example show that people have traditional beliefs about TB [32–34]. Khan et al., (2000) found no difference in treatment outcome related to educational levels in Pakistan [35]. In Ethiopia, high default rates were related to a lack of family support, inadequate knowledge and medication side effects [36]. In addition, sex and gender differences have been reported [33, 37].

Since 2007 the Angolan NTP has been implementing a plan with the main objective to increase TB awareness through advocacy, communication and social mobilisation activities [7]. This must be strengthened in the future, improving both attitudes and TB perceptions.

Another factor associated with prolonged patient delay was having a first point of contact that was different from the DOTS centre. One could hypothesize that those who go to a centre different than the DOTS centres are less educated people or people with less knowledge of the disease: the chi-square test did not confirm this assumption (p = 0,3448 for knowledge and p = 0,08 for education respectively). With the onset of symptoms, patients initially practised self medication, visited traditional healers, or resorted to non-prescribed medications from pharmacies. Meanwhile, they would also seek care at a health care provider. These different actions varied across countries [5].

In the present study, the use of private sector, including traditional healers, remains relatively minor as a point of first contact (15%), although its progression should be monitored. The use of traditional healers as the preferred source of care was also low, only 2%. However, this percentage may be reasonable considering the urban-suburban focus of the study. Previous works conducted in sub-Saharan Africa had suggested a higher preference for traditional healers, particularly for people who live in rural areas [20, 21, 38]. We found that only 2% reported self medication as first action, a low percentage compared to other studies. The fact that hospitals were the first point of care sought by approximately one third of patients reflects the poor quality of the primary health care system in Luanda.

In contrast 33% said they went directly to a DOTS centre, which is high compared to other studies [5, 27]. A number of patients have probably been elsewhere before coming to the DOTS centre, but they are not reporting it. This may be due to patients’ reluctance to express their preference freely in a DOTS setting, despite interviewers’ efforts to investigate this aspect in an open manner [39]. Collaboration with hospitals, health care providers, and private health workers is therefore important to reduce enrolment delays in DOTS programmes in Luanda. The free services of the DOTS programmes should probably be made known more widely across the community.

Health system’s delay

In the current study the system delay was very short, suggesting that referrals and diagnosis were swift whenever a person with TB interacted with the health system. However, as we have pointed out, there might be problems validating information about the place of first contact with the health system, especially when it was made outside the public sector. Despite this positive result, rapid diagnostic tests have to be used in the care process. One such test, the newly developed Xpert MTB/RIF, needs implementation studies in low income countries [40]. Similarly, the effectiveness of existing tools needs to be assessed [41].

Living in suburban area was a factor associated with prolonged system delay (median system delay in sub urban area 14 days (IQR 7–30); median system delay living in urban area 7 days (IQR 5–14)). A similar conclusion emerged from other studies [23]. Despite the fact that suburban inhabitants are likely to have poorer access to health care [21, 42–44], interestingly there was no increase in the patient delay in this study compared to urban dwellers. Although the number of positive smear tests was not significantly lower in suburban areas (87/123 in suburban area, vs 190/262 in the urban area, p = 0.71) this suggests that the suburban health centres are poor at diagnostic TB. Management and provision of resources for diagnosis of TB in Luanda province has probably been over-centralised, with the result that suburban dwellers have been disadvantaged.

Although the median waiting time in the health centre was acceptable (15 minutes), 8% of patients declared that they had to wait more than one hour. To the best of our knowledge, this is the first time that waiting times at the centre resulted in system delay. This is probably related to the fact that few studies have analysed the health system factors responsible for delay in diagnosis, frequently using data related to sociodemographic characteristics of patients. With the increase of HIV treatment in sub-Saharan Africa it will be important to facilitate an integrated approach to the management of co-infected persons, creating a “one stop” service in order to facilitate dually infected patients to be treated for both conditions simultaneously. By providing a “one stop” service patients attend one rather than two consultations, which reduces waiting time.

If the centre of first contact differed from the DOTS centre, this was a factor associated both with prolonged patient and system delay. The median patient delay for DOTS centre is 21 days, vs. the 30 days of the non DOTS centre as a whole. Among non-DOTS centres, the median patient delay increased to 33.5 days for traditional healers and up to 37.5 days for the hospital. It seems that the DOTS centres, in comparison to the others centres, are capable of facilitating the first contact with the patient. DOTS centres seem to perform better also in relation to the time needed for diagnosis. The median system delay for the DOTS centres is 5 days which increased to 7 days for non DOTS centres as a whole. The median system delay for hospitals and health care providers is in line with these figures (7 and 8 days respectively), but system delay increased to 14 days for private care providers, 40.5 days for traditional healers and 90 days in cases of self medication.

A final comment on system delay is that it is probably needed to be more inclusive in the definition of a “TB suspect”. Normally, national TB programmes recommend that people with chronic cough (>2–3 weeks) should be tested for TB, but that is missing with many diseases. This is also the norm in Angola.

Total delay

In the present study the total time elapsed between symptom onset and diagnosis of TB (45 days) was similar or lower than the delay reported in the majority of other studies carried out in sub-Saharan Africa [9, 21–24, 27, 38, 39, 45–48]. In these studies total delays ranged from 6 to 16 weeks.

The present study has certain strengths: it involved 21 of the 27 DOTS centres in Luanda province; the questionnaire had been previously tested in a pilot study, which allowed us to improve the instrument. Finally, it is the first on this issue carried out in Angola. However, our study has some limitations. The questionnaire used did not quantitatively address several items such as smoking and alcohol, and did not attempt to explain why the diagnosis was delayed. For measuring the length of patient delay, we depended on the reply of the patient. Consequently, some patients may not remember the exact onset of symptoms. In addition we do not have information on the HIV serostatus of the patients. Although no studies in sub-Saharan Africa have reported a link between HIV and a delay in the diagnosis of TB, this is a condition that needs additional research in the future [22, 46]. Another limitation of the current study is that we are not able to provide data on laboratory turn-around time. So it is not possible to say if the lab plays an important role in the in health system delay. Finally, although for the health centres we have tried to analyse several factors focused more on system delay other than the socio-demographic characteristics of patients, part of the factors analysed are related to socio-demographic and clinical characteristics of patients. As already observed by Finnie et al.[16], these would be considered predisposing factors for system delay in the sense that they may be related to the aspect of system delay in which patient return for diagnosis is required; they also may indicate health care workers’ perceptions of patients that may cause them to be less likely to test and diagnose TB [16]. Anyhow, in future research factors more pertinent to the health system have to be studied in a better way [16, 18].