Patient delay determinants for patients with suspected tuberculosis in Yogyakarta province, Indonesia

Authors


Corresponding Author Willem A. Lock, Department of Public Health, Erasmus MC, University Medical Center Rotterdam, PO Box 2040, 3000 CA Rotterdam, The Netherlands. Tel.: +31 107 038 460; Fax: + 31 107 038 475; E-mail: wa.lock@planet.nl

Summary

Objectives  Indonesia has a high incidence of tuberculosis (TB), despite the successful introduction of the directly observed treatment short-course strategy (DOTS strategy). DOTS depends on passive case finding. It is therefore important to identify determinants of patient delay and reasons for visiting a DOTS healthcare provider when seeking care. The aim of this study was to assess these determinants in TB suspects (coughing for at least 2 weeks).

Methods  Cross-sectional data were gathered with a structured questionnaire in which psychosocial determinants were based on an extended version of the theory of planned behaviour (TPB). The study was conducted in five governmental lung clinics of Yogyakarta province. In total, 194 TB suspects that registered at the lung clinics were interviewed.

Results  The median patient delay was 14 days (range 0–145). Ordinal regression analyses showed that visiting a private healthcare provider when first seeking health care, reporting travel distance/travel time as reason for choosing a certain healthcare provider when first seeking health care, discussing the symptoms with family and a reported short travel time, but no factors of TPB, were significantly associated with a shorter patient delay. An important factor negatively associated with visiting a DOTS clinic was the reported travel time.

Conclusion  Accessibility of the healthcare provider was the main determinant of patient delay, but the role of psychosocial factors cannot be fully excluded. Urban and suburban areas have relatively good access to (private) health care, hence the short delay. Thus, future studies should be focussed on extending the DOTS strategy to the private sector.

Abstract

Objectifs:  L’Indonésie a une forte incidence de tuberculose (TB), malgré l’introduction réussie de la stratégie de traitement de courte durée sous observation directe (stratégie DOTS). DOTS dépend du dépistage passif. Il est donc important d’identifier les déterminants du retard du patient et les raisons pour visiter un professionnel de la santé DOTS lors de recours à des soins. Le but de cette étude était d’évaluer ces déterminants chez les personnes suspectées de TB (toux persistante de plus deux semaines).

Méthodes:  Des données transversales ont été recueillies par un questionnaire structuré dans lequel les déterminants psychosociaux étaient basés sur une version étendue de la Théorie du Comportement Planifié (TPB). L’étude a été menée dans cinq cliniques du poumon gouvernementales de la province de Yogyakarta. Au total 194 cas suspects de TB, enregistrés dans les cliniques du poumon ont été interviewés.

Résultats:  Le retard médian des patients était de 14 jours [intervalle 0 à 145]. Des analyses de régression ordinale ont montré que la visite à un prestataire privé de soins de santé lors du premier recours aux soins, le report de la distance/temps du voyage comme la raison du choix d’un prestataire particulier de soins de santé lors du premier recours à la santé, la discussion des symptômes avec la famille, le report d’une durée de déplacement courte, mais pas les facteurs du TPB, étaient significativement associés à un retard de plus courte durée des patients. Un facteur important négativement associé avec la visite d’une clinique DOTS était la durée rapportée du déplacement.

Conclusion:  L’accessibilité du prestataire de soins était le principal déterminant du retard du patient, mais le rôle des facteurs psychosociaux ne peut pas être totalement exclu. Les zones urbaines et suburbaines ont un accès relativement meilleur aux soins de santé (privés), d’où le court retard. Ainsi, les futures études devraient se concentrer sur l’extension de la stratégie DOTS dans le secteur privé.

Abstract

Objetivos:  Indonesia tiene una alta incidencia de tuberculosis (TB), a pesar de la exitosa introducción de la estrategia DOTS. El DOTS depende del hallazgo pasivo de casos. Por ello es importante identificar los determinantes del retraso de los pacientes y la razón por la cual buscan un proveedor de DOTS en el momento de buscar ayuda sanitaria. El objetivo de este estudio era evaluar estos determinantes en sospechosos de padecer TB (tos durante al menos dos semanas).

Métodos:  Se recogieron datos croseccionales mediante un cuestionario estructurado. Los determinantes psicosociológicos estaban basados en una versión extendida de la Teoría del Comportamiento Planeado (TCP). El estudio se realizó en cinco clínicas neumológicas gubernamentales de la provincia de Yogyakarta. En total se entrevistaron 194 sospechosos de TB atendidos en cinco clínicas neumológicas.

Resultados:  El retraso medio por paciente era de 14 días (rango 0–145). El análisis con regresión logística ordinal mostró que el acudir a un proveedor sanitario privado cuando se buscan por primera vez cuidados sanitarios, el reportar una distancia de viaje / tiempo de viaje como la razón para escoger un cierto proveedor sanitario cuando se busca ayuda por primera vez, discutir los síntomas con la familia, un tiempo de viaje reportado como corto, pero ningún factor de TCP, estaban todos significativamente asociados con un menor retraso del paciente. Un factor negativo importante, asociado con el visitar la clínica de DOTS, era el tiempo de viaje reportado.

Conclusión:  La accesibilidad al proveedor sanitario era el principal determinante del retraso del paciente, pero el papel de los factores psicosociales no puede excluirse totalmente. Las áreas urbanas y suburbanas tiene un acceso relativamente bueno a los cuidados sanitarios privados, y por lo tanto hay un menor retraso. Los estudios futuros deberían focalizarse en extender la estrategia DOTS al sector privado.

Introduction

Indonesia has a high incidence of tuberculosis (TB); it ranks fifth among all countries in terms of TB prevalence (WHO 2010). To control the TB epidemic, WHO recommends the directly observed treatment short-course strategy (DOTS strategy). The implementation of this strategy started in 1995. By 2007, there was a population coverage of 100% in Indonesia with a case detection rate of 51% of all new TB cases (WHO 2009). The main clinical elements of the DOTS strategy are free diagnosis, free treatment and supervised treatment (WHO 2006). The global targets set by WHO for 2015 are to have halted and begun to reverse the incidence of TB by 2015 and to halve TB prevalence and mortality rates of the 1990 baseline. In Indonesia, the mortality rate had, as of 2006, fallen by 50% compared to 1990; however, the prevalence rate was reduced by 32% in 2009, and the incidence rate has basically remained constant since 1990 (WHO 2010).

The DOTS strategy depends on passive case finding; thus, an important factor in the successful diagnosis and treatment of TB is the healthcare seeking behaviour of individuals suspected of having TB. According to the TB/HIV clinical manual, a TB suspect is defined as someone with a cough lasting at least 2 weeks (WHO 2004). A long delay in diagnosis is associated with increased transmission of TB (Golub et al. 2006). To improve the control of TB, it is very important to understand the determinants of patient delay – defined as the time between onset of symptoms and the first visit to a healthcare provider – and the reasons for first visiting a DOTS healthcare provider.

A limited number of studies have reported on healthcare seeking behaviour in the Indonesian context. A qualitative study performed in Yogyakarta province suggested that income and the advice of household members had the strongest influence on seeking health care (Rintiswati et al. 2009). Another qualitative study by Watkins and Plant (2004) identified additional determinants of seeking health care, including knowledge and awareness of TB, availability of the healthcare provider, cost and stigmatisation. Other studies have shown that age, gender, education and place of residence were determinants of seeking health care in patients with TB (Johansson et al. 2000; Cheng et al. 2005; Chiang et al. 2005). However, to date, all studies on determinants of patient delay of patients suspected of TB in the Indonesian population were qualitative; no quantitative studies have assessed the relative importance of various factors, including social cognitive factors, in determining healthcare seeking and patient delay.

Insight into psychosocial determinants of patient delay and visiting a DOTS clinic is important as these determinants can be used as change targets in health promotion programmes. We deployed an extended version of the theory of planned behaviour (TPB) as a theoretical framework for assessing the psychosocial determinants of patient delay (Ajzen 1991). The TPB is a social cognitive model that proposes that the intention to perform a certain behaviour is a direct proxy of action and that this intention is determined by: (i) the attitude towards the behaviour, (ii) the subjective norm, or the perceived social pressure associated with performing the behaviour and (iii) the perceived behavioural control, or the extent to which the patient thinks he/she is capable of performing the behaviour (Ajzen 1991). To enhance the prediction of human behaviour, we extended the model with the following factors: perceived vulnerability for TB, the perceived severity of TB, knowledge of TB, perceived severity of symptoms and perceived stigmatisation of TB. (A detailed description of the extended version of the TPB is given in online-only Appendix S2.)

The aim of our study was to assess the determinants of patient delay for patients that presented at one of the five governmental lung clinics (with a cough of more than 2-week duration) in Yogyakarta province, Central Java – with a focus on psychosocial determinants. Moreover, we assessed which of these determinants was associated with visiting a DOTS healthcare provider when first seeking health care.

Materials and methods

Design and study population

A cross-sectional study was conducted by interviewing TB suspects with a structured questionnaire. The study was conducted in all five lung clinics (all governmental) in Yogyakarta province during the second quarter of 2008. The DOTS strategy has been implemented in these five lung clinics, and in all healthcare centres (117 in total) and 18 public and private hospitals. (Details about the study setting can be found in online-only Appendix S1.) Patients were eligible for inclusion when coughing lasted for at least 2 weeks. Patients were excluded from the study when they reported coughing for more than 6 months (i.e. to reduce recall bias), when it was not possible to interview the patient (i.e. because of not being able to speak the Indonesian language, severe illness, unconsciousness, severe psychiatric disease, severe intellectual impairment and hearing loss), or when the patient was under 15 years old. The targeted sample size was 200 patients, based on the rule of thumb of a minimum of 15 patients per predictor with an estimated model consisting of 10 predictors. The number of patients included in each lung clinic was weighted based on the number of patients diagnosed with TB in each clinic in the second quarter of 2007. Consecutive sampling was used to include the patients in the study upon registration at the lung clinic. All patients provided informed and signed consent to participate in the study. Because of cultural norms – as it is inappropriate to ask further questions after the patient refused to participate – no basic socio-demographic data and reasons for refusal could be collected on patients who refused to participate. Approximately 15% of the patients refused to participate in the study.

The study was approved by the Ethical Committee of Research in Medical Health of the Faculty of Medicine of Gadjah Mada University, Indonesia.

Questionnaire

The questionnaire was developed in English, translated into the Indonesian language and then translated back into English to check the translation. It included socio-demographic data, questions to assess patient delay – measured in days from onset of coughing until first visit to a healthcare provider – and healthcare seeking behaviour and the items of the extended version of the TPB. The socio-economic status was mainly based on property (see Appendix S3 for more detail). Based on the cumulative score (range 0–46), we classified patients as either lower (score <18) or upper class (score >18) using a median split as this gave the best correlation with patient delay. The items of the extended version of the TPB were mainly derived from a study by Watkins and Plant (2004) conducted in Bali. The TPB items were assessed with questions based on a 5-point scale. The questions were posed to the patients by asking them to consider what they would do if they started coughing again. Thus, patients were asked about their current considerations; they were not asked to recall the considerations made when they first visited a healthcare provider during the current episode of coughing. (A detailed outline of the questionnaire is shown in online-only Appendix S3.)

To assess the internal consistency of the direct measures of the extended version of the TPB, Cronbach’s alpha was calculated (Table 1; Streiner 2003).

Table 1.   Descriptive statistics of the factors included in the extended version of the theory of planned behaviour. = 194 patients with a cough that lasted at least 2 weeks
 Mean (range)*Cronbach’s Alpha†
  1. TB, tuberculosis.

  2. *A score below 3 is not in favour of the seeking health care for coughing, a score of 3 indicates a neutral position, and a score >3 is in favour of seeking health care for coughing.

  3. †The Cronbach’s alpha is a measure of the internal consistency of questions that measure the same factor, and it indicates the validity of the measurement of that factor. Cronbach’s alpha ranges from 0 to 1, where an alpha > 0.60 was considered satisfactory and an alpha > 0.80 was considered good.

  4. ‡Not applicable because only one question was asked.

  5. §Not applicable because the factor measured different aspects of the knowledge of TB, which do not need to be internally consistent with each other.

Intention4.2 (1–5)–‡
Attitude4.2 (1–5)0.89
Subjective norm4.1 (1–5)0.90
Perceived behavioural control3.7 (1–5)0.86
Knowledge score9.1 (0–15)–§
Perceived severity of symptoms3.2 (1–5)0.81
Perceived severity of TB2.0 (1–5)0.80
Perceived vulnerability of TB2.4 (1–5)0.96
Perceived stigma2.1 (1–5)0.64

Data collection

Four trained interviewers performed the interviews in the Indonesian language. After obtaining informed consent, the patient was interviewed in a quiet room where confidentiality could be guaranteed. For patients 15–18 years old, parental consent was also obtained. To assess questions related to an event in the past (e.g. the beginning of the cough), the patient was shown a calendar that indicated important holidays and events of the year. Patients received a small non-financial incentive for participating in the study.

Data analyses

The data were double entered in Epi Info software, version 3.4.3 (CDC, http://www.cdc.gov/epiinfo/). Analyses were performed with SPSS (version 16.0 for Microsoft Windows).

Patient delay was not normally distributed, as the Shapiro–Wilk test showed a P-value < 0.0001. Therefore, patient delay was categorised into five categories with approximately equal numbers of patients per category. The determinants of patient delay were analysed with ordinal regression, which takes into account the order of the categories of the dependent variable. Ordinal regression assumes that the effect of the independent variables is the same across the different categories of the dependent variable. The test of parallel lines showed a P-value > 0.05, reflecting that use of ordinal regression was justified (Norušis 2009). Logistic regression was used to identify factors that determined the choice of visiting a DOTS clinic when first seeking health care. For both outcome measures (patient delay and first visit being a DOTS clinic), univariate analysis was conducted to estimate the crude effects of the factors. Unless otherwise indicated, factors of the extended version of the TPB were analysed continuously. Variables with a P-value < 0.20 were then selected for the multivariate analyses. A likelihood ratio backward selection procedure was used to select the variables for the model; variables with a P-value < 0.10 were maintained. Based on the multivariate analyses, adjusted odds ratios were presented for the variables maintained in the model.

Results

Patient characteristics

From the end of April to the end of June 2008, 194 patients who had been coughing for >2 weeks were interviewed. The interviews lasted on average 45 min to an hour. For 56 patients (29%), this interview represented their first time seeking health care. Patients had sought health care a median of two times and a maximum of nine times. Eighty-eight patients (45%) visited a private healthcare provider when first seeking health care. The mean patient age was 39 years (14.1 SD, range 16–77), and 104 (54%) patients were men. On a scale of 0–46, the mean SES score was 21 (5.1 SD, range 9–38), which corresponds with middle class. (More details on patient characteristics can be found in online-only Appendix S1.)

Determinants of patient delay

The median patient delay was 14 days (range 0–146), with a mean of 19.1 days (1.52 SD). Patients who visited a public healthcare provider when first seeking health care had a median patient delay of 17 days (range 0–146); patients who visited a private healthcare provider had a median patient delay of 7 days (range 0–112). Patients generally had high intentions, favourable attitudes, subjective norms in favour of visiting a healthcare provider and positive perceived behavioural control, with mean scores above 3.5 (Table 1). This means that patients had the intention to seek health care for coughing and perceived no barriers to perform this behaviour. The direct measures were considered good by the Cronbach’s alpha scores.

Univariate analyses showed that the factors in the extended version of the TPB (direct and indirect measures) were not significantly associated with patient delay (Table 2). Patients who reported many symptoms (5–10) had a shorter patient delay than those that reported few (1–4) symptoms. The factors marital status, occupation, residence (urban/rural), highest education, SES score, the use of self-medication, the use of traditional medicine, visiting a traditional healer or the serious symptom of haemoptysis were significantly associated with patient delay.

Table 2.   Ordinal regression of determinants of patient delay in individuals that sought health care for a cough that lasted >2 weeks in lung clinics in Yogyakarta province, Indonesia
VariableNMean patient delay (days)Crude OR (95% CI)P-valueAdjusted OR (95%CI)P-value
  1. OR, odds ratio; CI, confidence interval; HP, healthcare provider; TB, tuberculosis.

  2. *Factors are coded in such a way that a high score is theoretically in favour of the behaviour.

  3. †Based on a median split of the continuous variable.

  4. ‡Includes two patients who did not discuss their symptoms at all.

  5. §The lung clinics, healthcare centres and governmental hospitals were categorised as public systems.

Intention score*†
 Low16118.61  
 High3321.31.14 (0.59–2.20)0.71  
Attitude*19419.11.67 (0.92–3.00)0.10  
Subjective norm*19419.10.73 (0.34–1.55)0.41  
Perceived behavioural control*†
 Low7922.71  
 High11516.60.80 (0.48–1.33)0.39  
Perceived severity of symptoms*19419.11.12 (0.83–1.50)0.45  
Perceived severity of TB*†
 Low4121.71  
 High15318.40.85 (0.46–1.56)0.60  
Perceived vulnerability for TB*†
 Low9418.21  
 High10020.01.19 (0.72–1.94)0.51  
Perceived stigma†
 Low10321.41  
 High9116.51.29 (0.78–2.13)0.31  
Knowledge score19419.11.11 (0.95–1.28)0.19  
Age (decades)19419.11.11 (0.93–1.32)0.27  
Gender
 Female9013.911
 Male10423.62.39 (1.42–3.97)0.0011.58 (0.92–2.71)0.10
Occupation
 Others8718.11  
 Private sector10719.91.16 (0.70–1.91)0.56  
Residence
 Rural5720.51  
 Urban13717.20.75 (0.43–1.30)0.30  
Highest education   0.28  
 Elementary school or lower5021.71  
 Junior high school3713.50.62 (0.29–1.32)0.22  
 Senior high school6822.01.15 (0.60–2.19)0.67  
 University3915.90.69 (0.33–1.46)0.34  
Socio-economic status
 Lower class5521.51  
 Upper class13918.10.84 (0.48–1.45)0.52  
Number of symptoms
 1–48515.41  
 5–101022.01.57 (0.94–2.59)0.083  
 9     
Discussed symptoms with   <0.001 <0.001
 Spouse1219.111
 3     
 Other family members4714.20.45 (0.25–0.83)0.0110.44 (0.23–0.85)0.014
 Non-family members‡2428.53.00 (1.35–6.68)0.0072.70 (1.19–6.16)0.018
First HP visited§
 Public system1023.611
 6     
 Private sector8813.70.30 (0.18–0.51)<0.0010.51 (0.29–0.90)0.020
Reasons for visiting that particular HP   <0.001 <0.001
 Quality of service8224.011
 Travel distance/travel time7511.70.22 (0.12–0.40)<0.0010.35 (0.18–0.67)<0.001
 Other3723.20.70 (0.35–1.41)0.320.78 (0.38–1.61)0.35
Travel time to first HP   <0.001 0.054
 0–14 min6912.011
 15–29 min7420.42.77 (1.52–5.05)0.0012.17 (1.17–4.04)0.014
 >30 min5126.85.75 (2.92–11.4)<0.0012.61 (1.18–5.74)0.017

The independent variables for the multivariate analysis included gender, symptoms discussed with others, first healthcare provider visited, reason for choosing that particular healthcare provider and travel time. Men had longer patient delays than women, but this effect was not significant in the multivariate analysis. Patients that visited a private healthcare provider during first healthcare visit had a significant shorter patient delay compared to patients visiting a public healthcare provider. Compared to discussing symptoms with a spouse, discussing symptoms with other family members was negatively associated with patient delay, resulting in shorter delay. In contrast, discussing symptoms with non-family members resulted in more delay (positive association). Among the reasons given for choosing a certain healthcare provider, mentioning travel distance and travel time as reason for visiting a particular healthcare provider was significantly associated with a shorter patient delay than mentioning quality of service. A travel time longer than 14 min was associated with a longer patient delay (P = 0.054).

Determinants of visiting a DOTS clinic when first seeking health care

Of the 194 patients included in the study, 110 (57%) visited a DOTS healthcare provider when first seeking health care. The majority of DOTS healthcare providers consisted of lung clinics (55%) and healthcare centres (34%). A tendency to visit a DOTS healthcare provider was associated with the quality of service as reason to choose a certain healthcare provider (P = 0.078, Table 3). The number of reported symptoms was not significantly associated with visiting a DOTS clinic. Other factors that were not associated with visiting a DOTS healthcare provider when first seeking health care included age, use of self-medication and use of traditional medicine. Having senior high school as highest education seems to be significant, but the factor highest education as a whole was not significant (P = 0.18) and could not be kept in the multivariate analyses.

Table 3.   Logistic regression of determinants for visiting a DOTS clinic as the first healthcare provider by patients with a cough that lasted >2 weeks, who visited a lung clinic in Yogyakarta province, Indonesia
Variablen/N (%)*Crude OR (95% CI)P-valueAdjusted OR (95%CI)P-value
  1. OR, odds ratio; CI, confidence interval; HP, healthcare provider; DOTS, directly observed treatment short-course.

  2. *= number of patients that visited a DOTS HP; = total number of patients in this category.

  3. †Factors are coded such that a high score is theoretically in favour of the behaviour.

  4. ‡Based on a median split of the continuous variable.

  5. §Includes two patients who did not discuss their symptoms at all.

Intention score†110/188 (59)0.57 (0.27–1.18)0.128  
Attitude†‡
 Low70/109 (64)1  
 High40/79 (51)0.57 (0.32–1.03)0.063  
Subjective norm†‡
 Low93/146 (64)1  
 High17/42 (40)0.39 (0.19–0.78)0.0080.25 (0.10–0.59)0.002
Perceived behavioural control†110/188 (59)0.77 (0.48–1.25)0.29  
Perceived severity of symptoms†110/188 (59)0.73 (0.51–1.03)0.0720.57 (0.38–0.85)0.006
Knowledge score  0.70  
 Low21/37 (57)1  
 Average71/117 (61)1.18 (0.56–2.49)0.67  
 High18/34 (53)0.86 (0.34–2.19)0.75  
Gender
 Female47/88 (53)1  
 Male63/100 (63)1.49 (0.83–2.66)0.18  
Highest education  0.18  
 Elementary school or lower24/49 (49)1  
 Junior high school21/35 (60)1.56 (0.65–3.76)0.32  
 Senior high school45/66 (68)2.23 (1.04–4.79)0.039  
 University20/38 (53)1.16 (0.50–2.70)0.74  
Main occupation
 Others41/84 (49)11
 Private sector69/104 (66)2.07 (1.15–3.73)0.0162.77 (1.38–5.53)0.004
Residence
 Rural29/54 (54)1  
 Urban81/134 (60)1.32 (0.70–2.49)0.40  
Socio-economic status score (continuous)110/188 (59)0.98 (0.92–1.03)0.42  
Total number of symptoms110/188 (59)0.90 (0.78–1.05)0.17  
Discussed symptoms with others  0.16  
 Spouse72/120 (60)1  
 Other family members21/44 (48)0.61 (0.30–1.22)0.16  
 Non-family members§17/24 (71)1.62 (0.62–4.20)0.32  
Reasons for visiting that particular HP  0.078  
 Travel distance/travel time35/72 (49)1  
 Quality of service54/81 (67)2.11 (1.10–4.06)0.025  
 Other21/35 (60)1.59 (0.70–3.60)0.27  
Travel time to first HP (h)110/188 (59)15.0 (4.1–54.7)<0.00129.0 (6.4–132)<0.001

Multivariate analyses showed a significantly lower chance of visiting a DOTS healthcare provider when first seeking health care among those reporting a positive subjective norm and perceiving their symptoms as serious. Patients visiting a DOTS clinic were significantly more often employed in the private sector and reported a significant longer travel time. We also tested what effect patient delay would have on visiting a DOTS clinic by including it in the multivariate analysis. Patient delay was significantly and positively associated with visiting a DOTS healthcare provider when first seeking health care. Patient delay only minimally changed the effects of the other determinants.

Discussion

The aim of our study was to assess the determinants of patient delay of patients suspected of TB (defined as a patient with a cough for at least 2 weeks) presenting at a lung clinic in Yogyakarta province, Central Java. This study was the first to apply a theoretical framework to an investigation of patient delay determinants in TB. The median patient delay was 14 days, range 0–145. A shorter patient delay was significantly associated with choosing a private healthcare provider when first seeking health care, reporting travel distance/travel time as reason for choosing a certain healthcare provider when first seeking health care, discussing the symptoms with family and a short travel time to the first healthcare provider. The extended version of the TPB did not appear to offer a further explanation of patient delay. Visiting a DOTS clinic when first seeking health care was determined by a positive subjective norm towards seeking health care for coughing; perceiving symptoms as severe; employment in the private sector; and longer travel time to the first visited healthcare provider.

Some limitations were inherent to our study design. First, patients who did not visit one of the lung clinics were not interviewed, including those that did not seek health care; this probably resulted in an underestimation of patient delay. Also, patients with TB that visited a lung clinic with a cough that lasted <2 weeks were not interviewed, which in turn could have resulted in an overestimation of patient delay. Second, patients were asked about their beliefs regarding visiting a healthcare provider after they had performed this behaviour. Had the patients been asked to recall their past beliefs regarding visiting a healthcare provider, there answers might have been altered by their current beliefs. Besides that, it is easier for a patient to state his current beliefs instead of recalling past beliefs. Therefore, we assumed that the current beliefs, although they may have been altered by performing the behaviour, still express the differences in beliefs associated with differences in patient delay. However, as the TPB assumes that a certain level of intention is needed before the patient performs the behaviour, these differences in beliefs and intentions might have faded away in the process of healthcare seeking. To avoid these inherent limitations, an alternative study design would be to identify patients suspected of TB at the household level. However, that would have required investigating a prohibitively large population to include an equivalent number of patients with a cough that lasted at least 2 weeks. In this study, we were unable to directly determine reasons why people delayed. When patients were asked for reasons why they delayed, most of them stated it was because the symptoms were not severe. The high frequency of this answer while perceived severity of symptoms was not associated with the delay suggests that this might be a social desirable answer. These limitations may partly explain why this study did not identify any of the TPB factors as determinants of patient delay with the included number of patients.

In this study, the median patient delay was relatively short (14 days) compared to other studies conducted in both developing and developed countries that reported a median delay of over 20 days (Sherman et al. 1999; Pronyk et al. 2001; Demissie et al. 2002; Rajeswari et al. 2002; Odusanya & Babafemi 2004; Farah et al. 2006; Okur et al. 2006; Rojpibulstit et al. 2006; Sreeramareddy et al. 2009). This may be due to the fact that the majority of the included patients lived in urban areas with an extended network of healthcare providers (Mahendradhata et al. 2008; Chiang et al. 2005; Kiwuwa et al. 2005; Ngamvithayapong et al. 2001 who reported median patient delays <12 days). Thus, our study population had good access to health care. Therefore, the generalisability of our study may be restricted to urban and suburban areas with a relatively accessible healthcare system.

The finding that discussing the symptoms with family was associated with shorter patient delay than discussing the symptoms with non-family members confirmed the suggestion by Watkins and Plant (2004) and Rintiswati et al. (2009) that the opinions of people close to patients have an important influence on their healthcare seeking behaviour. Studies conducted in Botswana (Steen & Mazonde 1999), South Africa (Pronyk et al. 2001) and Malawi (Salaniponi et al. 2000) also found a positive relationship between advice from others and healthcare seeking behaviour, but they did not analyse the association with patient delay. The positive influence of the advice of family members on healthcare seeking behaviour highlights the need for population-wide education programmes to inform both the high-risk population and their social environment on adequate healthcare seeking behaviour for coughing >2 weeks. We found that choosing a public healthcare provider when first seeking health care was associated with a significantly longer patient delay. This finding was also reported in a study on TB conducted in south India (Rajeswari et al. 2002). A study conducted in Malawi also found that the distance to a healthcare provider and quality of service were the main reasons for choosing a particular healthcare provider when first seeking health care, but that study did not analyse the effect on patient delay (Salaniponi et al. 2000). Other studies conducted in Ethiopia (Demissie et al. 2002) and south India (Rajeswari et al. 2002) also found that travel time or distance was significant determinant of patient delay. These results suggested that availability, particularly of public healthcare providers, should be the primary concern of policy makers.

It is difficult to explain the finding that working in the private sector was significantly associated with a higher probability of choosing a DOTS clinic when first seeking health care. A possible explanation could be that people working in the private sector had higher incomes, which enabled access to better transportation and reduced the issue of distance to the clinic. Another explanation could be that people in the private sector were more likely to work in shifts; thus, they were more likely to be free for morning visits to a DOTS healthcare provider, which were only open during the morning hours. Other studies found an association between distance and choosing a DOTS healthcare provider in general, but those studies did not analyse the association between distance and visiting a non-DOTS healthcare provider (Demissie et al. 2002; Okur et al. 2006). A qualitative study conducted in Vietnam investigated specific reasons for choosing a private healthcare provider versus a healthcare provider connected to the National TB Program (NTP) (Lonnroth et al. 2001). They found that patients perceived that the NTP provided good health care, but other qualitative aspects were considered negative (e.g. DOTS treatment regimes and longer waiting times). That might also have been the case in our study. We found that patients were more likely to visit a DOTS clinic when quality of service was their reason for choosing their first healthcare provider; however, those patients exhibited a significantly longer patient delay.

This suggests that patients who valued quality of service above other characteristics in the healthcare provider experienced barriers to visiting the healthcare provider of their choice. One important barrier to visiting a DOTS healthcare provider seems to be the travel time as people visiting a DOTS healthcare provider when first seeking health care had a significant longer travel time. Therefore, to improve the utilisation of the DOTS healthcare providers, the focus should be on improving accessibility. A positive subjective norm towards seeking health care for coughing was negatively associated with visiting a DOTS healthcare provider during first healthcare seeking action. It suggests that important others give the patients the advice to visit a non-DOTS healthcare provider as first healthcare visit. This finding further highlights the need not to target the high-risk population only but also the people that are important to them, especially family members.

Another major finding of our study was that none of the factors of the extended version of the TPB was significantly associated with patient delay. This suggested that these factors were not relevant to the control of TB. Nevertheless, the role of psychosocial factors cannot be fully excluded, because we found that discussing the symptoms with others and the reasons of patients for choosing their first healthcare provider were important determinants of patient delay. Thus, conclusions about the importance of psychosocial factors in determining patient delay must await future studies that identify patients with suspected TB at the household level, before they visit a healthcare provider.

Patients who visited a private healthcare provider when first seeking health care had a patient delay sufficiently short to obviate the necessity of further decreasing the delay. In fact, a shorter delay may be counterproductive in the control of TB, because the difficulty in recognising the early symptoms of TB could result in a longer system delay (WHO 2004). Early treatment of symptoms without a proper diagnosis could lead to inappropriate antibiotic use, an increased delay in the diagnosis of TB and unnecessary utilisation of the healthcare system. However, the public system had a considerably longer patient delay. Because most public healthcare providers are DOTS healthcare providers, this indicated that the DOTS healthcare system lacked accessibility. The best way to address this would be to increase the density of public and DOTS healthcare providers, because the travel time to the healthcare provider was also an important determinant of patient delay. However, given the popularity of private healthcare providers and their high density in Yogyakarta province (Mahendradhata et al. 2008), it might be more effective to extend the DOTS strategy to the private sector. This was also suggested in a study that used a mathematical model to estimate the impact of different strategies on tuberculosis mortality and treatment outcome (Ahmad et al. 2009). Although Ahmad et al. (2009) concluded that actively finding cases would provide the best outcomes, that strategy may be less cost-effective than other strategies, including involvement of the private sector in the DOTS strategy. Mahendradhata et al. (2010) conducted a study to evaluate the incremental cost-effectiveness of such an initiative in Yogyakarta province where private practitioners were engaged to refer patient suspected of TB to a healthcare centre for DOTS treatment. They concluded that this strategy could be cost-effective given a well-functioning DOTS programme. Nevertheless, extending the DOTS strategy to the private sector would involve some difficulty. Some characteristics of the DOTS programmes (e.g. supervised treatment) may conflict with characteristics of private healthcare providers (e.g. flexibility and privacy). Also, the broad variety of types of private healthcare providers poses a challenge for successfully extending the DOTS strategy to the private sector. Despite these challenges, some initiatives are currently being implemented to involve the private sector in TB control. The results of our study support the usefulness of these efforts. Yet, some considerations need to be made. A recent study conducted in Yogyakarta province showed that visiting a DOTS healthcare provider during the first healthcare visit did not result in a shorter diagnostic delay, suggested that the focus first needs to be on strengthening the health system and improving diagnostic quality within DOTS services (Ahmad et al. 2011).

We conclude that accessibility of the healthcare provider was the primary determinant of patient delay and that psychosocial factors were not significant determinants. Given the short reported patient delay when a private healthcare provider was chosen in first seeking health care, we recommend that the major concern of policy makers should be improving the utilisation of DOTS healthcare providers, particularly in the private sector.

Acknowledgements

This study received financial support from the KNCV Tuberculosis Foundation, the A.A. van Beek fund and the Beker – La Bastide fund. The authors would like to thank Didi Ispiranto, Lilik Rahmanto, Tanto, Yanto Hidayat, Mas Kris and Yudhistia Sudibya for their help with collecting data. We also thank Prof J. Dik F. Habbema and Dr Jan Hendrik Richardus for their critical review of the design of this study and Dr Caspar Looman for his constructive comments during the analyses.

Ancillary