Factors determining household expenditure for tuberculosis and coping strategies in Tajikistan

Authors


Corresponding Author Raffael Ayé, Swiss Tropical and Public Health Institute, Swiss Centre for International Health, Socinstr. 57, Basel, Switzerland. E-mail: raffael.aye@birding.ch

Summary

Objective  To investigate factors influencing expenditure levels and the use of potentially detrimental coping strategies among tuberculosis (TB) patients. For the purpose of the present study, potentially detrimental coping strategies included borrowing money and selling assets.

Method  Questionnaire survey with an initial and a follow-up interview of each adult new pulmonary TB case registered over a period of 4 months, conducted in 12 districts with DOTS in Tajikistan, one of the poorest countries in the world.

Results  Patients and their households faced mean expenditures of US$ 396 related to a TB episode. In multivariate mixed-effect regression models, the main determinants of out-of-pocket payments–either over the whole course of the disease or after enrolment in DOTS treatment–were ‘complimentary treatment’ besides the anti-TB drugs, duration of hospitalization and treatment delay. Complimentary treatment mainly consisted of vitamins and rehydrating infusions. Sex showed no association with expenditure. To cope with the costs of illness, two-thirds of patients employed a potentially detrimental coping strategy. TB patients raised on average US$ 23 through loans with interest, US$ 57 through loans without interest and US$ 102 through selling assets.

Conclusion  The catastrophic out-of-pocket payments faced by TB patients are correlated with receiving complimentary treatment, delay to treatment and duration of hospitalisation. The widespread use of potentially detrimental coping strategies illustrates that TB constitutes a substantial risk of impoverishment. More parsimonious use of complimentary treatment and hospitalisation could reduce illness-related costs for patients and should be carefully considered.

Abstract

Objectif:  Investiguer les facteurs influençant les niveaux des dépenses et l’utilisation de stratégies de survie potentiellement préjudiciables chez les patients tuberculeux (TB). Pour la présente étude, les stratégies de survie potentiellement préjudiciables comprenaient l’emprunt d’argent et la vente de biens.

Méthode:  Enquête par questionnaire avec une interview au début et au cours du suivi de chaque nouveau cas adulte de TB pulmonaire enregistré sur une période de quatre mois, menée dans 12 districts appliquant la stratégie DOTS au Tadjikistan, un des pays les plus pauvres du monde.

Résultats:  Les patients et leurs familles faisaient face à des dépenses moyennes de 396 $ US liées à un épisode de TB. Dans les modèles de régression multivariée à effets mixtes, les principaux déterminants de paiements directs de la poche, soit sur l’ensemble du cours de la maladie ou après l’inscription dans le traitement DOTS étaient «traitement gratuit” en plus des médicaments anti-TB, la durée d’hospitalisation et le retard du traitement. Le traitement gratuit consistait principalement en des vitamines et perfusions pour réhydratation. Il n’y avait pas d’association entre le sexe et les dépenses. Pour faire face aux coûts de la maladie, deux tiers des patients ont utilisé une stratégie de survie potentiellement préjudiciable. Les patients TB ont eut recours en moyenne à un prêt de 23 $ US avec intérêt, un prêt de 57 $ US sans intérêt et à 102 $ US obtenus en vendant des biens.

Conclusion:  Les paiements catastrophiques directs auxquels sont confrontés les patients TB sont en corrélation avec les traitements gratuits, le retard dans le traitement et la durée d’hospitalisation. L’utilisation généralisée des stratégies de survie potentiellement préjudiciables montre que la TB constitue un risque important d’appauvrissement. Une utilisation plus parcimonieuse de traitement et d’hospitalisation gratuits pourrait réduire les coûts liés à la maladie pour les patients et devrait être attentivement considérée.

Abstract

Objetivo:  Investigar los factores que influencian los niveles de gasto y el uso de estrategias potencialmente dañinas para afrontarlos en pacientes con tuberculosis (TB). En el presente estudio, las estrategias potencialmente dañinas incluían el pedir dinero prestado y la venta de activos.

Métodos:  Encuesta mediante un cuestionario, realizada durante una entrevista inicial y otra de seguimiento a cada adulto con un nuevo caso de TB pulmonar que se registró durante un periodo de cuatro meses. El estudio fué conducido en 12 distritos con DOTS, en Tajiquistán, uno de los países más pobres del mundo.

Resultados:  Los pacientes y sus hogares tenían un gasto medio, relacio nado con un episodio de TB, de US$ 396. En modelos de regresión multivariada con coeficientes mixtos, los principales determinantes de gastos de bolsillo - bien durante toda la duración de la enfermedad o después de iniciar el tratamiento con DOTS– eran los ‘tratamientos complementarios’ además de los medicamentos anti-TB, la duración de la hospitalización y el retraso en el tratamiento. El tratamiento complementario consistía principalmente en vitaminas e infusiones rehidratantes. El sexo no estaba asociado con el gasto. Para hacer frente a los costes de la enfermedad, dos tercios de los pacientes utilizaron una estrategia potencialmente dañina para afrontar los gastos. Los pacientes de TB consiguieron recaudar en promedio US$ 23 mediante préstamos con intereses, US$ 57 mediante préstamos sin intereses y US$ 102 mediante la venta de activos.

Conclusión:  Los pagos de bolsillo catastróficos enfrentados por los pacientes con TB están correlacionados con el recibir tratamiento complementario, el retraso en el tratamiento y la duración de la hospitalización. El amplio uso de estrategias potencialmente dañinas para afrontar los gastos ilustra el hecho de que la TB constituye un riesgo sustancial de empobrecimiento. Un uso más frugal del tratamiento complementario y de la hospitalización podría reducir los costes relacionados con la enfermedad para los pacientes y debería considerarse cuidadosamente.

Introduction

Tuberculosis (TB) patients often encounter high costs during the course of their disease despite TB chemotherapy offered for free in most settings (Saunderson 1995; Kamolratanakul et al. 1999; Wyss et al. 2001; Habib & Baig 2006; Jackson et al. 2006; Aspler et al. 2008). TB drugs are indeed provided for free in most areas of the former Soviet Union, but patients have to pay for other services such as additional medicines, X-rays and laboratory services (Mosneaga et al. 2008). In Tajikistan, household costs of an episode of TB amounted to US$ 4900 purchasing power parity (PPP), of which US$ 1840 PPP (38%) were TB-related expenditure (Ayéet al. 2010a). Expenditure and loss of income associated with TB may push already poor households deeper into poverty (Russell 2004) and have been identified as major barriers to TB services (Wei et al. 2009). Several studies observed differing costs depending on socio-economic status and hospitalization (Saunderson 1995; Kamolratanakul et al. 1999; Russell 2004); others have shown that the provision of treatment observation influences costs to the patients (Saunderson 1995; Floyd et al. 1997). The only study so far that statistically tested such observed differences found clinic-based observation of treatment, longer patient-delay and male sex to be the main determinants of higher household costs in urban Zambia (Aspler et al. 2008). Overall, the factors associated with high TB-related costs are still poorly understood.

Several authors have pointed out that the investigation of household costs of disease is incomplete if coping strategies employed by households are ignored (Russell 2004; McIntyre et al. 2006). The most immediate response to illness-related costs is to use cash income and savings, but this option is not available to all households (Russell 1996; McIntyre et al. 2006). Further common coping strategies are borrowing and selling assets. Borrowing can increase the level of debt; sale of assets decreases the resilience to future economic shocks (Russell 2004). This is particularly true for the sale of productive assets, because this reduces future household income. While there are many more coping strategies (Sauerborn et al. 1996; Obrist et al. 2007), available studies emphasize that selling productive assets and taking out loans often negatively affect future income and can lead to impoverishment (Russell 1996, 2004; McIntyre et al. 2006; Leive & Xu 2008). For the purpose of this article, the term ‘detrimental coping strategies’ refers to selling assets and taking out loans. The present study aimed at identifying determinants of TB-related out-of-pocket payments and the use of detrimental coping strategies in Tajikistan.

Methods

Study setting and data collection

WHO (2009) estimated TB incidence in Tajikistan for the year 2007 at 231 cases per 100 000. Roll-out of the internationally recommended strategy for TB control, DOTS, was ongoing and coverage reached 100% by the end of 2007. However, hospitalization rates remain high: 58% in a study in 10 districts (Thierfelder et al. 2008). This study, which received ethical approval from the Ministry of Health of Tajikistan, made use of cost data collected alongside two studies investigating the timing of costs of illness and delay to TB treatment (Ayéet al. 2010a,b). The studies were conducted in 12 districts representing urban, rural, lowland and mountainous settings. All new adult (≥15 years) pulmonary TB patients who were registered in the 12 districts between 1st December 2006 and 31st March 2007 were eligible. The study participants received TB treatment according to the national guidelines of Tajikistan. This involved an intensive phase of treatment lasting 2 months and a continuation phase lasting 4 months. If sputum smear conversion was not achieved after 2 months, the intensive phase was prolonged by 1 month.

The research team visited patients in hospital or at home during the intensive phase of treatment, obtained written informed consent and administered the first questionnaire. Three to 4 months later, a follow-up questionnaire interview was conducted with the same patients. The questionnaires included detailed questions about the costs incurred, economic coping strategies and household assets. The first questionnaire asked about the period from first symptoms to onset of treatment and the intensive phase. The follow-up questionnaire asked about the continuation phase. For analysis, we extrapolated costs from the day of the interview until the end of the respective phase.

Statistical analysis

Three statistical models were built in order to achieve the objectives of the present study and are described below. The three outcome variables were: (i) total TB-related expenditure, (ii) TB-related expenditure after onset of treatment and (iii) TB-related detrimental coping strategies.

All analyses were conducted in Stata IC/10.1 (Stata Corporation, USA, 1985–2008). We applied multiple imputation to our dataset to deal with missing observations as recommended by Manca and Palmer (2005) for similar problems and in order to avoid the problems of complete-case analysis (Schafer 1997). We built a wealth index based on 18 asset variables (Filmer & Pritchett 2001). Details on these two steps have been presented elsewhere (Ayéet al. 2010a).

The factors influencing direct costs to households were identified through a mixed-effects linear regression. We reduced the number of repeated significance tests by identifying a basic model at the outset and by considering only a limited number of predictor variables (Grafen & Hails 2002). We built our basic model around five variables. A variable identifying individual DOTS centres had to be included in the model based on the design of the survey and was included as a random effect. Three variables which influence household costs were identified from the literature, namely sex (Aspler et al. 2008), socio-economic status (Kamolratanakul et al. 1999; Russell 2004) and hospitalization (Saunderson 1995; Russell 2004). Socio-economic status was included as a continuous variable, the wealth index described above. Hospitalization, one of the case management factors of main interest, was also included as a continuous variable, namely the log-transformed duration of the hospital stay in days. The fifth variable, the second of the main variables of interest relating to TB case management, was a binary denoting patients that had received complimentary treatment, including medication for perceived iatrogenic problems. Two further variables were considered in addition to the basic model: the sputum smear result and the total delay from onset of symptoms to treatment–both as proxies for the severity of disease and as continuous variables. Both the length of hospital stay and the delay until onset of treatment were log-transformed because the relationship was expected to have the shape of a saturation curve rather than a linear increase. The full model was run, its validity was checked by visually examining the residual errors and then the model was simplified by excluding factors with P > 0.2. However, the five factors defined from the outset were retained even if their P-values were above the threshold (Grafen & Hails 2002).

In order to test more specifically the influence of a case management decision on the costs incurred subsequent to this decision, a similar mixed-effects regression model was built, but the outcome variable included only those expenditures that were incurred after the onset of anti-TB chemotherapy (referred to as ‘during treatment’ below). This model was built in order to test the influence of hospitalisation and the use of complimentary treatment on TB-related expenditures. The decision whether to hospitalise or whether to use complimentary treatment is usually taken at the beginning of TB-treatment and cannot have any influence on expenditure made prior to treatment. Also, clinicians may want to know what influence different strategies of case management have on the patient’s household economy. Explanatory variables and criteria for model simplification were identical to the previous model.

Further, we built a mixed-effects model on the amount of funds mobilized through detrimental coping strategies. Taking out loans (with or without interest) and selling assets were considered detrimental coping strategies, but not the use of household income, savings and funds received as donations from family, friends and other well-doers. For the analysis, we did not differentiate between productive and non-productive assets, because the pilot study had shown that by far the largest part of asset sales concerned productive assets, most commonly lifestock. The explanatory variables in the model on detrimental coping strategies were the same as in the models on expenditure, but an additional variable was considered, a binary identifying patients who had worked in Russia until falling ill. This was included because it was thought that it could have a major influence on the amount of cash available at the household level.

For all three models, we plotted the residuals and checked visually whether they were approximately normally distributed. Residuals were close to normally distributed when expenditure data (first two models) were log-transformed and when funds mobilized through detrimental coping strategies (third model) were root-transformed.

Results

We identified 282 eligible patients. Reasons for drop-out and demographic characteristics of the 204 patients who consented and were interviewed are listed in Tables 1 and 2, respectively.

Table 1.   Reasons for drop-out
ReasonNumber of cases (% of drop-outs)
Wrong or insufficient address18 (23)
Not found at home or in hospital when visited17 (22)
Died5 (6)
Initial defaulter3 (4)
Already in continuation phase when visited25 (32)
Did not consent10 (13)
Total78 (100)
Table 2.   Characteristics of the 204 study participants
Age (years)15–2425–3435–44≥ 45  
  1. *These patients had returned after labour migration to Russia.

Women34231517  
Men44401516  
Total78633033  
ProfessionHouse-wifeMigrant worker*FarmerPetty tradeUn-employedOther
Women521113914
Men03013192032
Total523124222946
SettingUrbanPeri- and semi-urbanRural  
Women8378  
Men18106  
Total911184  

Expenditure over the whole course of disease

As reported previously (Ayéet al. 2010a), mean expenditure for an episode of TB amounted to US$ 396 (±standard deviation US$ 357; median US$ 282). The full multivariate linear regression model for total expenditure was run and simplified as described in the Methods section. The factor ‘sputum smear result’ was excluded. The final multivariate regression model showed that delay until onset of treatment and complimentary treatment were the main determinants of expenditure (Table 3). On average, doubling of delay increased total expenditure by 17%. Receiving complimentary treatment increased total expenditure by a factor of 2.10. Complimentary treatment most commonly included Aloe vera 1% 1 ml, glucose 5% infusion, vitamin C 1% 1 ml, vitamin B1, B6 and B12 (in single and compound formulation), calcium gluconate 10% 10 ml, and Polyvinylpyrrolidone 6% infusion. There was a statistical tendency for increased costs with longer hospital stay (P = 0.073).

Table 3.   Results of the mixed-effects linear regression on total expenditure (in US$, log-transformed); n = 204
FactorCoefficient/slopeSE†P-valuefmi‡
  1. *Denotes significance levels.

  2. SE = standard error.

  3. ‡fmi = fraction of missing information. The fmi is a measure of the information contained in the missing data in a multiply imputed dataset (Schafer 1997).

Sex0.1470.1320.2630.160
Wealth index0.04490.03470.1960.130
Days in hospital (log-transformed)0.01630.009100.0730.131
Delay in days (log-transformed)0.2310.604<0.0005***0.151
Complimentary treatment0.7440.2800.008**0.128
Intercept3.910.392<0.0005***0.139
DOTS centre (estimate for variance-covariance matrix)−1.170.420N/A0.178

A slight departure from linearity was observed in the influence of the sputum smear result on total expenditure. A sensitivity analysis with sputum smear result as a categorical variable showed fully consistent results.

Expenditure during anti-TB therapy

The regression model on expenditure during treatment could not be simplified. Delay was not significantly associated with direct costs during treatment. However, receiving complimentary treatment and duration of hospital stay were associated with increased expenditure during treatment and this was highly significant (Table 4). Receiving complimentary treatment led to 2.12 times higher expenditure during treatment. Patients hospitalized for 1 week had a 50% higher expenditure than patients who were not hospitalized. Patients hospitalized for 2 months had 5.6% higher expenditure than patients who were hospitalized for 1 week only.

Table 4.   Results of the mixed-effects linear regression on expenditure during anti-TB chemotherapy (in US$, log-transformed); n = 204
FactorCoefficient/slopeSE†P-valuefmi‡
  1. *Denotes significance levels.

  2. SE = standard error.

  3. ‡fmi = fraction of missing information. The fmi is a measure of the information contained in the missing data in a multiply imputed dataset (Schafer 1997).

Sex−0.05860.1280.6460.141
Wealth index0.03440.03570.3350.204
Sputum smear result0.07960.06130.1960.260
Days in hospital during treatment (log-transformed)0.02560.009450.007**0.192
Delay in days (log-transformed)0.09940.06010.0990.178
Complimentary treatment0.7520.2990.012*0.261
Intercept3.980.419<0.0005***0.236
DOTS centre (estimate for variance-covariance matrix)−0.9470.332N/A0.123

There was no statistical relationship between sex and expenditure during treatment; neither did the sputum smear result in either model.

Coping strategies

The most common coping strategies were the use of household income, donations received and selling assets (Table 5). The highest amounts were raised through household income and selling assets, namely on average US$ 226 (± 23.8) and US$ 102 (± 13.3). Two-thirds (65.7%) of patients relied on a detrimental coping strategy to handle the costs of TB. The mean total amount for all detrimental coping strategies was US$ 182 (± 20.8).

Table 5.   Frequency and extent of use of different coping strategies by TB patients; n = 204
 Proportion of users among patients (%)Mean (in USD)SEMedian among users of strategy (in USD)
Detrimental coping strategies
Credit8.4122.687.80195.28
Loan29.9756.9315.4478.57
Asset sales49.26101.9313.29190.29
Total detrimental strategies65.70181.5420.80207.81
Non-detrimental coping strategies
Use of household savings30.4663.2011.08145.78
Household income67.39226.3823.78275.35
Donations48.0138.855.5348.91
Total non-detrimental strategies92.28328.4424.34281.80
Total all strategies99.10509.9731.05438.33

The mixed-effects linear regression did not show any statistically significant relationships (Table 6). There was a weak statistical tendency (P = 0.096) for patients with higher wealth indices to raise less money through detrimental coping strategies. Patients who had returned from Russia due to their TB showed a statistical tendency (P = 0.069) for raising more funds through detrimental coping strategies than patients who had not been to Russia.

Table 6.   Results of the mixed-effects linear regression on detrimental economic coping strategies (in US$, root-transformed); = 204
FactorCoefficient/slopeSE†P-valuefmi‡
  1. SE = standard error.

  2. ‡fmi = fraction of missing information. The fmi is a measure of the information contained in the missing data in a multiply imputed dataset (Schafer 1997).

Sex−4.823.090.1200.286
Wealth index−1.220.7330.0960.200
Sputum smear result8.194.490.0690.365
Days in hospital during treatment (log-transformed)0.2690.1960.1700.273
Complimentary treatment3.315.670.5590.135
Intercept16.55.910.0050.143
DOTS centre (estimate for variance-covariance matrix)2.830.048N/A0.427

Discussion

Expenditure

The main factors leading to higher expenditure related to TB were receiving complimentary treatment, longer hospital stay and longer treatment delay. The latter two factors were each highly significant factors for higher costs in the relevant and appropriate reference period, respectively, but only showed a statistical tendency (P < 0.1) in the other time period. For hospitalisation, the period during treatment is appropriate for analysis, because hospitalisation usually happens during treatment and cannot be expected to influence expenditures that had happened previously. The influence of hospitalisation was substantial, patients hospitalised for 2 months spending more than 1.5 times as much as patients who were not hospitalised at all. Longer delay was associated with higher expenditure over the whole course of the disease. A similar situation arose in Lusaka, Zambia, where longer patient delay led to higher costs (Aspler et al. 2008). Including two proxies for severity of disease had very little influence on the estimated coefficients or significance levels of different factors–with the exception of the factor hospitalization in the model on costs during the whole episode. It seems unlikely that these findings were confounded by severity of disease. The strong influence of complimentary treatment is most likely related to a tradition among Tajik and other post-Soviet TB doctors to frequently prescribe additional medication, often several types of drugs (Mosneaga et al. 2008). Doctors may supplement their meagre incomes through prescribing and/or selling additional drugs. In other post-Soviet countriest doctors practice in public premises but for private revenue (Mosneaga et al. 2008). The use of additional medication in our study population exceeds international standards by far (WHO 2003) and its rationality is doubtful. For the patients, additional medication is associated with the costs of the drugs themselves, and with travel to the health facility for a prescription and to the pharmacy. Reducing the use of complimentary treatment and of hospitalization could be simple measures to reduce the costs to TB patients in Tajikistan. Expenditure was not associated with sex in either of the two models. The p-values were high and the two estimated coefficients in the opposite direction. Men and women incurred similar expenditure over the course of disease in this study.

Coping strategies

About two thirds of households employed at least one detrimental coping strategy. The mean amount raised was US$ 182. This high amount suggests that costs associated with an episode of TB push affected households deeper into poverty. The proportion of households employing detrimental coping strategies is higher than found in 15 African countries, where among hospitalized patients (independent of diagnosis), about 50% reported borrowing and/or selling assets (Leive & Xu 2008). Most commonly, the assets sold were cattle, sheep or goats. These are productive assets and confirm our choice to treat sale of assets as a potentially detrimental coping strategy.

While none of the investigated factors was significantly associated with detrimental coping strategies, some aspects are worth discussing. The factor ‘migration to Russia’ showed a tendency towards higher use of detrimental coping strategies–rather contrary to expectation. It was assumed that patients who recently returned from Russia would on average have more cash available, because wages are several times higher in Russia than in Tajikistan. However, the data showed a tendency only and should be interpreted with care. The long treatment delays experienced by migrants developing TB in Russia could have contributed to higher costs (Ayéet al. 2010b).

Limitations

We used an analytic descriptive design and hence it is not possible to definitely conclude that a statistical relationship constitutes a causal link. Like other studies investigating household costs of illness, we had to rely on self-reported costs. Consequently, recall and reporting bias cannot be excluded (e.g. Saunderson 1995). We limited recall bias by means of conducting two interviews, which reduced recall time. Interviewers were trained to recognize unusually high costs for specific items and to ask back comparing the reported costs to local prices for comparable items. Another limitation is that we only analysed two coping strategies, while many more exist (Obrist et al. 2007). We focused on those financial coping strategies that bear risks for future impoverishment.

Conclusion

Patients and their households face major expenditure during an episode of TB and two thirds of affected households employ detrimental coping strategies that potentially impair future income. Hence household costs of TB are catastrophic. Both sexes experience similar expenditure in this setting. The main factors associated with higher expenditure are receiving complimentary treatment in addition to anti-TB chemotherapy, longer treatment delay and longer stay in hospital. Complimentary treatment, which is probably questionable on medical grounds in many patients, had the strongest effect – it more than doubled expenditure. Hospitalisation also had a strong influence, mainly on expenditure during treatment. Reducing additional medication and limiting hospitalization are simple interventions at the level of case management that could lead to lower expenditure for patients. In view of the high costs, such mitigation strategies are urgently needed. Similar health system characteristics and the common use of complimentary treatment for TB patients suggest that this factor would also play a role for TB patients in other post-Soviet countries (Mosneaga et al. 2008). Studies on the impact of these measures on treatment outcomes in the local context would be highly informative.

Acknowledgements

We are grateful to all participating patients, to health care staff for their support, to Firuza Qurbonova and Zulfira Mengliboyeva for help with the data collection, to Sandra Alba for statistical advice, to Project Sino and its staff for contributing to successful study implementation and to Constanze Pfeiffer Lesong Conteh and Karin Wiedenmayer for their insightful comments on an earlier version of the manuscript and for reviewing the language.

Ancillary