Summary Tuberculosis is a heavy burden in the developing countries. Early diagnosis and adherence to treatment are difficult to achieve by patients. Our qualitative research looked at the paths followed and the barriers experienced by patients at the health care services of Cali, Colombia, while seeking help for pulmonary tuberculosis symptoms. Results show that the cultural-based explanation patients give to the symptoms, the stigma attached to the disease, and the poor quality of health care services (communication skills, organizational structure, attitudes, and knowledge of the tuberculosis control strategy of health care workers) are strong barriers to early diagnosis. Tuberculosis (TB) is a leading cause of morbidity and mortality in developing countries (Raviglione et al. 1995). Its profile has been worsening due to the Human Immunodeficiency Virus epidemic (Shafer & Edlin 1996). Early diagnosis and adherence to treatment are key factors for a successful TB control programme. However, the common nonadherence to treatment, with the consequent low cure rates, seriously weakens control efforts (Chaulet 1987). Early diagnosis is also uncommon: several months of combined patient and health service delay have been reported in several countries (Murray 1994). Patients are usually diagnosed with TB as a consequence of the interaction between their active efforts in seeking help, and the passive case-finding activities of health care workers (HCWs) in health centres (Rieder 1993). Factors affecting the behaviour of patients and HCWs determine the outcome of case finding activities and the speed of TB diagnosis. How soon a patient is diagnosed and receives treatment has obvious implications for the infection risk: the longer the patient is infectious, the greater is the proportion of contacts being infected. This research aims to build up a picture of the pathways that pulmonary TB patients follow and the barriers that they face at the health care services when they seek help from the start of coughing (the epidemic mechanism of TB transmission and the earliest symptom in most cases) until they are diagnosed. This research was part of a larger project to improve the performance of TB control services in this city, and to provide health education to HCWs, patients and the community. Among the several models for explaining health care-seeking behaviour, Kroeger (1983) distinguishes two basic approaches: the pathway model and the determinants model. The former assumes a logical sequence of steps taken by the individual from earliest symptoms until the use of health care services, while the latter focuses on the study of those variables that could explain the choice of a preferred form of health care service. The complexities of the factors which influence the individual's decision-making process serve to complicate the study of health-seeking behaviour. However, the interest of this research was focused mainly on how the perception and treatment of TB symptoms by patients, and the interaction of health care services with patients contribute to explain the delay in diagnosis of pulmonary TB. Therefore the model of health-seeking behaviour proposed by Chrisman (1977) was chosen for underpinning this research. Five different stages are proposed in this model: the 'symptom definition' stage deals with how patients perceive the physical changes produced by the disease; 'illness-related shifts in the role behaviour' refers to the way in which the evolution of symptoms influence how patient relate to their peers; 'treatment actions' refers to the activities undertaken by patients to attenuate the burden of the illness; the 'lay consultation' stage refers to the exploring of peer's opinion about patient symptoms and suggestions for dealing with them; 'adherence' means those activities taken by patients for following treatment and medical advice. This research focused mainly on the first four stages of this model, since the last one was not an objective of this inquiry. The National TB Control Programme of Colombia is a free health care service. The diagnosis is based on direct smear tests and the treatment on short-course chemotherapy. According to the instructions issued by the Programme, the disease must be suspected by the HCWs in every patient with respiratory symptoms (cough, haemoptisis, or shortness of breath) for more than 15 days who seeks health care attention for any reason, or is visiting a medical doctor at a primary health care unit (Ministerio de Salud 1995). A set of three direct smear tests must be ordered in any case that TB is suspected. A chest X-ray must be ordered when the three direct smears are negative and the patient still complains of respiratory symptoms (Ministerio de Salud 1995). However, the cost of thorax X-rays are usually paid by the patient. In fact, thorax X-ray is not part of the case-finding strategy of the National TB Control Programme. The Programme, then, provides a very small quantity of X-ray films to the district hospitals where X-ray facilities are available. This research was carried out in Cali, Colombia, a city of nearly two million people. This is a typical city of a middle-income country, where high inequalities in income coexist with sustained economic growth (Mohan 1994). 34% of the population live below the poverty line. Although impressive goals have been achieved in terms of infant mortality rates and coverage of health care services, the poorest people are still in need of higher quality and better coverage of health care. The city is divided into 20 districts for administrative purposes, and each district has a public health centre. From three to five smaller health units are accountable to the respective health centre of the district. The Public Health Office aims to cover about 60% of the population, mainly the poorest, with a total of 80 primary health care units, that is, 20 primary health care centres and 60 clinics. Most health care units are strategically located within the city, and the transport system has good coverage. The average distance most patients have to travel from home to the dispensaries is small. However, to visit a health care centre it is necessary to use public transport, with an average cost for a round trip equivalent to 10–15% of the daily minimum wage. This cost is significant for the poorest people because their income is usually below the minimum wage. TB control services were available in only 16 primary health centres when data for this research was collected in 1993. Staff and laboratory facilities for processing and analysis of direct smears were available in 15 only. Consequently, those patients with a direct smear performed at the dispensaries had to travel to their respective health centre to deliver the sputum samples. TB case-finding was carried out by physicians and nurse assistants (HCWs with 18 months of training) when patients attended for medical consultation. When the medical doctor suspects TB, the patient is sent to the nurse assistant in charge of the TB Programme, who instructs the patient on sputum collection and delivery of the sample. More than 90% of patients received 'direct observed therapy' (the patient takes the pills in the presence of the HCW in charge of the Programme) shortly after being diagnosed with TB. Cure rates ranged from 70 to 85% during our research period. The notification rate for positive direct smear TB during the 1990s is around 30 ?? 100.000 according to the local Public Health Office. The Public Health Office reported a combined patient and health care services delay for diagnosis of positive pulmonary TB of more than one month in 92% of the notified cases during the period under study. Previous research in Cali described the basic lay knowledge about TB (Jaramillo 1995). It is widely perceived as a contagious disease caused by different factors such as malnutrition, sudden change from a hot to a cold environment, bacteria, poverty, and bad hygiene practices. Conventional treatment with formal medicine was considered best by patients and several folk healers. TB is a stigmatised disease: the patient inspires fear in others due to the risk of infection. Loss of employment and even divorce after diagnosis are not rare. An attempt at patient isolation during the treatment (avoidance of sharing of meals and dishes, kisses, or sexual relationships) is the commonest approach by the community to deal with the risk of infection. These lay beliefs about TB were shared and promoted by many HCWs in charge of the TB control programme at the time of this research. Several qualitative data collection strategies were used: individual unstructured and structured interviews, observation, and group interviews. Group interview methods are qualitative techniques frequently used in the social sciences. Coreil (1995) distinguishes 4 different types of group interviews: consensus panel, focus group, natural group, and community interview. Since it is not always possible in the field to implement in full any one of these techniques, Coreil (1995) proposes that the researcher provides a detailed description of the strategy applied in the interview, rather than using generic names such as 'focus groups', which sometimes are misleading. This advice is followed in this paper. In the first stage of data gathering, 24 informal and unstructured interviews were carried out with TB patients, 11 women and 13 men, aimed at testing the appropriateness of the chosen theoretical model, and to identify key informants to be interviewed in depth. Fourteen physicians of the health care units and each nurse assistant appointed to the TB programme in the 16 health centres were interviewed. Direct observation was carried out at the health centres of the encounter between patients and HCWs when the patients received a request for a direct smear test, and when they delivered the sputum sample. In-depth interviews were conducted with key informants. There were six nurse assistants responsible for collecting sputum samples, four drug-store attendants, and TB patients with the longest (eight patients, three women and five men, diagnosed after more than six months of respiratory symptoms) and the shortest (three patients, two men and one woman, diagnosed with less than four weeks of respiratory symptoms) delay in diagnosis. Most of the observations on patients and HCWs were overt, that is, the status and role of the observer were known. Group interviews were held once these data were gathered and partially analysed. All data collected were recorded in notebooks after the interview or observation. These texts were reviewed, and the data classified in categories appropriate to each one of the stages of the model of health-seeking behaviour. Group interviews were carried out with patients receiving treatment at each one of the 16 health centres offering TB control services, and conducted either by the principal investigator or a social worker trained in group interviewing techniques. The size of the groups ranged between six and 12 patients. The participants' age range was from 18 to 76, and about two thirds were female (according to the data available at the health centres, around 60% of patients receiving treatment were women). Most participants did not know one another before meeting in the group interviews. Overall attendance was nearly one third of the total patients receiving treatment at each health centre. Nurse assistants attended as observers at four meetings. The fact that this was against the best interests of this research and that their role as gatekeepers may be compromising for other concurrent research projects called for careful consideration of their interest in attending the meetings. Since the way in which the patients participated in these four meetings and since the contents of the information collected was not substantially different to the other meetings, it was decided to include them in the analysis for this paper. Group interviews lasted about one hour, and were carried out early in the afternoon on workdays. The basic agenda of the meeting consisted of a presentation and subsequent discussion of patients' experiences at each stage of the model, ending with a presentation of medical facts about TB. Usually the story of one participant was presented early on in the meeting as a vignette, helping to develop the discussion in an open and relaxed way. In several meetings the participants were surprised when they discovered how common TB is, on the basis of the numbers and diversity of people attending. Participants showed great interest in sharing and discussing their experiences. However, in some cases it was not possible to involve the group in a full discussion because the conductor was not able to overcome either the shyness, passivity or lack of interest of the participants. In those cases the format of the meeting approached that of a community interview, that is, a question and answer session rather than a discussion meeting. Information about TB prevention, diagnosis and treatment was presented at the end of the meeting. It aimed to clarify misconceptions that arose during the group interview, particularly those which added to the stigma attached to the disease. Table 1 provides a summary of the main findings of this inquiry. It must be taken into account that the steps proposed by the health-seeking behaviour model of Chrisman (1977) do not follow a rigid and logical sequence. Although each stage is intertwined with components of the other stages, they are presented in a logical sequence for a better understanding of this complex process. Most of the patients spontaneously reported that coughing was one of the first symptoms of the disease. However, for patients suffering from some other ailments as well it was not so easy to identify the first symptom, and backpain or dizziness were occasionally mentioned as the first one. Several patients reported only cough as their initial symptom. At this stage almost everyone identified the cough as part of a flu–like syndrome. Because patients initially defined TB as flu, drugs for relieving the 'flu' symptoms were commonly purchased from pharmacies. They were successful in controlling fever and a feeling of illness, but not always the cough. Then, stronger home-made remedies and a second visit to the pharmacy for anticough drugs were the following step. During their first visit to the pharmacy the patients asked directly for any brand of 'antiflu' drug. During the second visit patients presented chronic cough as a specific complaint and requested a good anticough syrup. Most of the patients attributed partial or definite relief from coughing either to the anticough syrups or to the home-made remedies. Though self-medication and pharmacy attendance were the commonest behaviours for dealing with the initial symptoms, several patients said that they attended private and public health centres. At the health centres they received a similar prescription to that obtained by those who attended pharmacies. Unremitting symptoms, in spite of home-made remedies and drugs obtained from the pharmacy, called for a review of 'symptom definition'. At this point, a folk complex called gripa pasmada was the most common explanation. Those patients interviewed generally assumed that flu has several phases before a full recovery. These phases consist of prodromes, full blown syndrome with fever, and a final phase with cough and some malaise. Gripa pasmada (unripened flu) is the name given to a flu syndrome stuck in the second or third phase, with cough as a prominent symptom. Reasons given by the patients for getting a gripa pasmada were lack of good nutrition during the peak of symptoms, noncompliance with drugs prescribed or self-medicated, exposure to sudden changes in temperature (such as moving into an air-conditioned room after staying out in the sun), and unprotected exposure to rain or mist (particularly at night or after a heavy working day). Sudden exposure to cold weather or becoming drenched by rain received the name of desmande. This was for the patients the most important factor contributing to the development of gripa pasmada. In some cases coughing for more than 6 months was seen as a sequel derived from this complex. 'Lay consultation' was more common during this phase of the disease. The advice given to patients by friends and relatives ranged from consulting a medical doctor or a folk healer to taking some more powerful home-remedies. It was not uncommon for patients to encounter gentle teasing that related TB with the gripa pasmada in a contradictory way: '? be careful, your pthisis is becoming a gripa pasmada? '. This kind of 'covert' advice is understandable once the stigma attached to the disease is taken into account. Patients found in gripa pasmada a satisfactory explanation for their symptoms until haemoptisis or weight loss were evident. It was at this point that an important 'role behaviour shift' occurred. Everyone involved in the patient's difficulties now felt that there was a serious problem and that the patient should relinquish work and his or her domestic duties in order to seek formal health care. At this stage, those who were reluctant to seek health care, either from folk healers or medical doctors, experienced pressure from family members and peers to do so. Case-finding activities in the health centres were carried out by physicians and by nurse assistants in the waiting room. Physicians and nurse assistants mentioned several case-finding strategies but reported that they were not applied regularly because of lack of time. In fact, most of the physicians interviewed asserted that TB was suspected only when the main complaint of the patient was respiratory symptoms plus fever or weight loss. If a patient visited the medical doctor with only backpain, for example, he or she did not investigate chronic respiratory symptoms. The most frequent activity carried out by nurse assistants was a short exposition for patients who were in the waiting room, inviting those with symptoms such as haemoptisis, cough, weight loss, and fever to attend the 'respiratory airways clinic' (the name of the TB clinics in every health centre). A few nurse assistants mentioned, as a productive tip for case-finding, looking for people with evident signs of poverty in their clothing or appearance. Since TB is widely accepted as a disease of poor people, several HCWs regarded poverty markers as a predictor of TB in those patients with chronic cough. Yet in the neighbourhoods of the health centres, where everybody is poor, only those living in misery were seen as really poor people. Most of the patients stated that they attended medical doctors or folk healers once the self-diagnosis of gripa pasmada was made, where the majority received the same prescription that they were given in the pharmacy for their initial self-diagnosis of flu. A direct smear test was ordered by physicians and some nurse assistants (although for most of the nurse assistants only the physician was entitled to request smears) only when they saw a more typical picture of pulmonary TB. In many cases a chest X-ray was also requested by physicians before the result of the direct smear was available, and even instead of a direct smear. An extreme case was reported by a young woman who was investigated initially with a chest X-ray for her chronic respiratory symptoms and fever. Since the findings were not conclusive, a thorax CT scan was requested. It showed a cavity (not visible in the previous chest X-ray), and only then a direct smear was performed and a positive result obtained. It was common for patients to complain about the high cost of chest X-rays, which compelled them to delay for several weeks the next appointment with the physician. The laboratories of health centres were the facility most frequently used for the direct smear test, even by those patients who received the request from private doctors. However, lab opening times were not appropriate for patients' needs. In almost all the labs sputum samples were accepted only during fixed hours (ranging from two to three hours early in the morning). Lab staff and nurse assistants blamed the patients when they were questioned about causes for delayed diagnosis of TB. Lack of awareness of the importance of the direct smear, bad manners, low quality and irregular provision of the three sputum samples needed for diagnosis were the HCWs' most common complaints about the patients. Although observation of patients delivering sputum samples confirmed these complaints in many cases, the communication skills of most of HCWs were poor. The instructions provided by nurse assistants were not always clear, and it was extremely unusual to find nurse assistants verifying that the patient understood these instructions. In spite of the fact that for some patients the cost of transport was very high, they very rarely mentioned this as an obstacle for delivering the sputum samples. Instead, patients tended to complain about the time lost waiting in the lab queue, either because of loss of earnings in their jobs or because of having to suspend their household duties. Many patients admitted that they were afraid of considering TB as an explanation for their symptoms. HCWs used the folk term enfermo de los pulmones (to be sick of lungs), a socially accepted synonym for TB, to explain to patients the objective of the proposed test. However, many patients were still afraid of the outcome either of the X-ray or the direct smear test because the meaning of enfermo de los pulmones was obvious. Some patients mentioned that deferral of the test was a common way to deal with the fear of being diagnosed tuberculous. Our qualitative research provides an introduction to the pathways followed and the difficulties faced by TB patients of Cali (Colombia) when seeking help for their symptoms at a public health service. The results show that the stages of the health-seeking behaviour model are intertwined and do not develop in a linear sequence. The whole pattern of 'health-seeking behaviour' in the patients interviewed was composed of two cycles determined by 'symptom definition', within which several stages of the model were developed. Patients interviewed in this inquiry do undertake an active search for help once TB symptoms appear, paying high costs in many cases. However, misinterpretation of initial symptoms results in actions (quite understandable once the explanation applied to them is taken into account) (that delay timely diagnosis. TB symptoms understood as folk illnesses leading to lack of search for medical treatment have been reported as contributing to medical diagnosis delay (Ndeti 1972; Lieban 1976; Rubel & Garro 1992). Although the practice of performing chest X-rays together or instead of direct smears is not recommended by the National TB Programme or the World Health Organization, it has been reported in some other places. A study in India found that 95% of TB patients were X-rayed as part of the diagnostic tests, but direct smears were employed only in 85% (Juvekar et al. 1995). A survey among Korean physicians found that up to 47% use X-rays as part of their case-finding strategy (Hong et al. 1995). From the clinical point of view, not in the context of the TB control strategy, a chest X-ray could be one of the appropriate initial tests when studying a patient with chronic respiratory symptoms. However, that is not necessarily the case when the TB epidemiological profile of a community and the economic cost for the patients are taken into account. While patients and HCWs waited until a florid picture of TB appeared, many patients paid high costs in tests, drugs, transport and time while looking for help. Similar economic factors affecting health-seeking behaviour have also been reported in Uganda (Saunderson 1995). Economic difficulties experienced during case holding have been also reported by patients and HCWs in some settings (Pocock et al. 1996; Bevan 1997). In fact, some studies show that those receiving material support or having better economic conditions have significantly higher rates of treatment adherence (Farmer et al. 1991; Barnhoorn & Adriaanse 1992). Meanwhile, those facing a difficult economic situation are more likely to default and to not adhere to treatment (Van der Werf et al. 1990; Nichter 1994; Dick et al. 1996; Johansson et al. 1996). Some studies show that treatment adherence is higher for patients with the most severe and prolonged symptoms. However, the high costs of a delayed diagnosis can leave patients with fewer resources to cope with the economic demands of treatment. A question to be explored further, then, is the extent to which a delayed diagnosis, and the usually high costs that it entails, could increase the risk for patient default and nonadherence to treatment. Lack of communication skills in HCWs, the narrow schedule for lab services at the health centres and the stigma attached to TB contributed to explain the irregular and delayed delivery of the sputum samples by many patients. But HCWs considered the delays as examples of patient unreliability, and employed this fact as evidence to argue that patients are people who need rules in order to behave properly. However, the rules operated by the health centres aimed to ease the duties of their staff rather than to facilitate the access of patients to health services. During the interviews patients were ready to talk openly about TB, and many did not find serious problems in having the disease suggesting that, despite the stigma attached to TB, this is less severe than in other places (Rubel & Garro 1992; Nichter 1994). However, thinking about TB was still fearful for many other patients, producing ambivalent feelings for taking the diagnostic tests. According to the findings of this research, health services-related (communication skills, attitudes, and knowledge about TB control strategy of HCWs), and cultural factors, mainly those associated with 'symptom definition', were affecting negatively the outcome of health-seeking behaviour in those patients interviewed. Although the qualitative nature of this research does not make it possible to attribute causal links between these factors and delayed diagnosis, the data suggest that they are related. Other research findings show that the costs paid by patients for a TB diagnosis is huge compared with the costs they would have paid had they known the implications of chronic cough, and if the TB programme were efficient and effective. Most TB patients in Cali, as everywhere, are among the poorest people, and for them having to buy an 'anticough syrup' or having to pay for an unnecessary chest X-ray can make the difference between having something to eat or not (Dufour et al. 1997). Most TB patients actively seeking help are diagnosed later rather than sooner by the health care system, and in the meantime continue to transmit the bacillus. Correction of these factors through health education of community, improved quality in the health care services, and continued education and supervision of HCWs could help to attenuate the negative impact of delayed diagnosis on the infection risk, the household economy, and the probable negative impact on adherence to treatment. An early version of this paper was presented in abstract format at the Lancet Conference 'The Challenge of Tuberculosis' held in Washington DC, September 14–15, 1995. I am grateful to the Liga Antituberculosa Colombiana y de Enfermedades Respiratorias, Comite del Valle, which supported this research; to Isabel C. Rios for her research assistance; to Colciencias, which provided me a scholarship that enabled me to write this paper; and to Drs Sam Mindel and Eva Gamarnikow for critiques and editorial assistance. All mistakes, of course, are only mine.