Article Text
PDF
Abstract
Objective The objective of this study was to measure the costs of people living with HIV (PLHIV) as well as active tuberculosis (TB/HIV), latent tuberculosis infection (LTBI/HIV) or without TB (HIV/AIDS).
Methods We analysed the costs through the entire pathway of care during the prediagnosis and treatment periods from the Brazilian public health system perspective. We applied a combination of bottom-up and top-down approaches to capture and estimate direct medical and non-medical costs. We measured the mean cost per patient per type of care (inpatient, outpatient and emergency care) and disease category (HIV/AIDS, HIV/AIDS death, TB/HIV, TB/HIV death and LTBI/HIV).
Results Between March 2014 and March 2016 we recruited 239 PLHIV. During the follow-up 26 patients were diagnosed and treated for TB and 5 received chemoprophylaxis for LTBI. During the prediagnosis and treatment period, the mean total costs for HIV or AIDS and AIDS death categories were US$1558 and US$2828, respectively. The mean total costs for TB/HIV and TB/HIV death categories were US$5289.0 and US$8281, respectively. The mean total cost for the LTBI/HIV category was US$882.
Conclusions Patients with TB/HIV impose a higher economic burden on the health system than HIV/AIDS and LTBI/HIV. Patients with LTBI/HIV were the lowest cost group among all disease categories, indicating that preventive TB treatment can avoid the further costs treating active TB.
Trial registration number RBR-22t943, Results.
- aids
- economic analysis
- tuberculosis
Statistics from Altmetric.com
Introduction
The AIDS epidemic has affected 37 million people worldwide and caused 1.1 million deaths in 2015. Among these deaths, one in three was due to HIV-associated tuberculosis (TB). Additionally, one-third of people living with HIV (PLHIV) presented latent tuberculosis infection (LTBI).1 The lack of, or delays in, diagnosis and treatment of active TB can explain the high mortality rate among coinfected patients.2 3 The reduction of AIDS-related deaths is a milestone established by the United Nations through Sustainable Development Goal 3 (SDG 3),1 4 and tackling TB coinfection is a key vehicle for reducing AIDS-related deaths. The treatment of LTBI is also crucial for the reduction of TB incidence in PLHIV. Mathematical models have predicted that the scaling up of LTBI treatment, together with the diagnosis and prompt treatment of active TB, can sharply reduce TB incidence.5 6
However, the funding available to address AIDS/HIV and TB/HIV coinfection is insufficient and might affect the potential success of SDG 3. Recent estimates indicate that US$8.3 billion is needed annually to combat TB in low-income and middle-income countries. In 2015, US$6.6 billion was invested and only 6% of this fund was allocated in TB/HIV coinfection actions.7 Therefore, costing analyses can play a key role in supporting the achievement of the SDGs in a sustainable way, especially with the reduction in the amount of funding available to combat both epidemics in recent years. In addition, the majority of high TB/HIV burden countries face both financial and human resource constraints. Thus, cost analyses are essential to inform better allocation of resources and to support economic evaluations and budget impact studies for decision making.
The objective of this study is to estimate the costs of PLHIV with or without active or latent TB, from the symptomatic phase until the first year of treatment from the perspective of the Brazilian public health system. We thus aim to contribute to the literature on costs of interventions for the control of TB and HIV/AIDS.
Methods
Study location
The study was conducted in the city of Recife, capital of the state of Pernambuco. We conducted the data collection in the Correia Picanco Hospital (CPH). The hospital provides care for approximately 60% of all individuals with HIV/AIDS in the state, carrying out almost 3000 outpatient appointments a month, including emergency and inpatient care.8
Study population, inclusion and exclusion criteria
The cost study was conducted alongside a pragmatic clinical trial designed to evaluate the cost-effectiveness of a protocol for TB diagnosis in PLHIV. The costing study followed the trial criteria: we included newly diagnosed HIV-infected patients, aged 18 years or over. Participants who were being treated for TB at the time of enrolment or had been treated for TB in the previous 3 months were excluded, as the trial aimed to test a protocol for TB diagnosis. We also excluded patients who were treated in the private sector and only visited the hospital to collect medicines. Further details of the clinical trial can be found in the online supplementary file 1.
Supplementary file 1
Procedures
The cost study was conducted from the health system perspective, during the first 2 years of the trial (March 2014 to March 2016). We applied a mix of bottom-up and top-down approaches to capture and estimate the direct costs.9–11 We obtained drug prices from the Brazilian Ministry of Health (MoH) database12 and test costs from health system records.13 Staff wages, hospital productivity and values of contracts and utility bills were collected from the CPH administrative division. Further details on data collection and cost estimates are given in the online supplementary file and online supplementary tables S1 and S2. Costs were calculated in local currency (real, 2015 prices) and converted to US dollars using an average exchange rate for the period of study as calculated by OANDA (R$1=US$0.34765).14
Interviews with patients were conducted by trained technical nurses during preadmission at CPH. The interviews were intended to collect data on the use of medical resources at emergency and outpatient care during the HIV prediagnosis period, from the onset of the disease until diagnosis. Interviewers also collected data on demographic, socioeconomic characteristics and lifestyle habits of individuals. For those who were diagnosed with TB or LTBI during the 2-year data collection period, we considered the TB/HIV prediagnosis period as the time between TB first symptoms and its diagnosis. As patients with LTBI/HIV do not present TB symptoms, we considered prediagnosis the period between the first HIV symptoms and LTBI diagnosis.
Subsequent interviews were conducted at every patient appointment at CPH. Besides checking for TB or LTBI diagnosis, the interviewer also collected data on the use of medical resources at outpatient and emergency care and on hospitalisations at CPH and other health services sought by the patients during the treatment period. To assess the use of resources at inpatient care in both the prediagnosis and treatment period, we reviewed patients’ medical notes at CPH and other health services. Details of the drug scheme for TB and LTBI treatment can be found in the online supplementary file.
Data cleaning and analysis
Questionnaires were double-entered in an Excel spreadsheet. The cost estimates were produced in Excel and statistical analyses in Stata/IC V.14. The main outcomes were the cost per type of care (emergency, outpatient and inpatient care) and total costs (prediagnosis + treatment period) per patient per category of disease (HIV or AIDS, AIDS death, TB/HIV, TB/HIV death, LTBI/HIV). The mean was reported for all cost estimates as measures of central tendency, as well as the associated SD. To test difference in proportions, we used the χ2 test for categorical variables or Fisher’s exact test when one or more cells had a frequency of five or less observations. For continuous variables with non-parametric distribution, we used the Wilcoxon-Mann-Whitney test. Differences in cost per category of patient were analysed using a Dunn’s test with Benjamini-Hochberg adjustment for multiple comparisons. All P values below 0.05 were considered statistically significant. We used the mean imputation approach to handle costing missing data; this assigns the mean cost of each item, at each level of care, to the missing value.15 In order to compare our results with studies conducted in other countries, we also presented our results in international dollars applying purchase power parity (2015 prices) (online supplementary tables S5 and S6).16 Costs of the studies presented in the Discussion section were updated to 2015 using US dollars inflation rate, estimated by the International Monetary Fund17 and converted to international dollars applying World Bank indices.16
Sensitivity analysis
A one-way sensitivity analysis was performed to assess uncertainties related to the parameters used. We varied the mean costs of TB/HIV, HIV/AIDS and LTBI/HIV per type of care by ±50%, as we did not have information regarding the highest and lowest values for each cost item. We varied the following direct medical cost parameters at emergency, outpatient and inpatient care: drugs, tests, medical appointment, bed days, ART and TB drugs. Results are presented in tornado diagrams to demonstrate the impact of each parameter change in the total cost.
Results
In a 2-year study period, 315 PLHIV were recruited, 15 patients were excluded at the randomisation stage, 25 were transferred to another health service and 37 were considered lost to follow-up. The final sample for this study was 239 PLHIV, with 79 patients in the control arm (72 HIV or AIDS, 7 TB/HIV, none LTBI/HIV) and 160 in the intervention arm (136 HIV or AIDS, 19 TB/HIV and 5 LTBI/HIV). In total, during the follow-up period, 208 patients were treated for HIV or AIDS, 26 were diagnosed and treated for active TB/HIV coinfection and 5 were diagnosed and treated for LTBI/HIV (online supplementary figure S1). Table 1 and online supplementary tables S3 and S4 show the baseline characteristics of the patients included in our analysis. In the HIV or AIDS and TB/HIV categories, most patients were male while most of the patients with LTBI/HIV were female (P=0.007). For all disease categories, most patients were in the 18–39 years age group and most patients were literate with a minimum of 4 years of study (90% of the total sample). When compared with HIV or AIDS and LTBI/HIV, patients in the TB/HIV category presented higher rates of alcohol dependence (P=0.009) and use of illicit drugs, crack (P=0.002) and glue (P=0.009).
Socioeconomic, demographic and lifestyle characteristics of the study population
Treatment characteristics
Patients in the TB/HIV category were seen more frequently at emergency care in both prediagnosis and treatment period, but frequency of use only differed from that of the HIV/AIDS category in the prediagnosis period (P<0.001). The average length of hospitalisation during the treatment period was twice as high as that during prediagnosis period for both disease categories. Patients in the TB/HIV category presented lower CD4 counts (<200 cells/m3) at first appointment than the other categories (P=0.009). LTBI/HIV and TB/HIV coinfected patients started ART later than patients with HIV/AIDS (P=0.003). Also, the proportion of deaths was higher among patients with TB/HIV than patients with HIV/AIDS (31% vs 6%; P=0.001) (table 2).
Characteristics of treatment and health outcomes for patients treated at the Correia Picanco Hospital, Recife/Brazil
Cost per site of care
During the prediagnosis period, the TB/HIV category had the highest costs at emergency, outpatient and inpatient care. The difference in the mean total costs was higher between the TB/HIV and HIV/AIDS categories: emergency care—US$419 vs US$109; outpatient care—US$269 vs US$64. There was no hospitalisation among patients with LTBI/HIV during the prediagnosis period, and the mean total cost was higher for patients with TB/HIV than patients with HIV/AIDS (US$532 and US$1710, respectively).
During the treatment period, patients in the LTBI/HIV category did not have emergency care and only one patient was hospitalised (US$549). At outpatient care, all patients presented similar costs, with patients with HIV/AIDS presenting a slightly higher mean total cost: HIV or AIDS—US$777; TB/HIV—US$687; LTBI/HIV—US$609. At inpatient care, the mean total cost for TB/HIV was almost double that of patients with HIV/AIDS (US$4372 vs US$2850) (table 3).
Mean direct medical costs (US$) per cost category for prediagnosis and treatment from the public health system perspective, Correia Picanco Hospital, Recife/Brazil
Costs per patient category
The category TB/HIV death presented the highest mean costs (prediagnosis, treatment and total) when compared with other categories. The mean total direct cost (prediagnosis + treatment period) of TB/HIV death was almost three times that of the mean cost of HIV or AIDS death: US$8281 vs US$2828. When compared with the LTBI/HIV category, the mean direct cost of the TB/HIV category was more than five times higher: US$5289 vs US$882 (table 4). Statistical significance in the total costs (prediagnosis + treatment period) was found for HIV or AIDS versus TB/HIV (P<0.0001), TB/HIV versus LTBI/HIV (P=0.0015) and AIDS death versus TB/HIV death (prediagnosis period only) (P=0.0016). Online supplementary table S7 shows the complete statistical analysis for the prediagnosis, treatment period and total cost.
Total mean cost (US$) for prediagnosis and treatment periods from the health system perspective, Correia Picanco Hospital, Recife/Brazil
Sensitivity analysis
During the prediagnosis period, the cost item ‘medical appointment’ had the highest impact on the total cost for all types of care and disease categories. During the treatment period, the cost item ‘ART’ had the highest impact on outpatient care for HIV or AIDS and TB/HIV categories. Medical appointment also had the highest impact on all other types of care. Tornado diagrams are presented in online supplementary figure S2.
Discussion
Patients with TB/HIV cost 3.4 times more than those with HIV/AIDS, and those who died due to TB/HIV coinfection cost almost three times more than those who died due to AIDS. In the meantime, patients with LTBI/HIV presented the lowest cost among all disease categories, indicating the treatment of patients at the latent phase can be cheaper than the active phase. The highest cost for TB/HIV was mainly due to inpatient care. Our findings reinforce the hypothesis that patients with TB/HIV have more complications during treatment and therefore are hospitalised more frequently and treated with more expensive drugs.18 Indeed, in our study, the cost of drugs and tests at inpatient care was higher for TB/HIV than HIV or AIDS. Another study carried out in Sudan found greater costs of hospitalisation among TB/HIV coinfection than TB/HIV-negative patients. Nevertheless, the updated costs of hospitalisation in international dollars (2015) found in the Sudanese study were lower than the costs found in our study: $2847 vs $6801. The mean treatment cost of TB/HIV category was also lower compared with our study: $1568 vs $6341.19
Other studies carried out in low-income and middle-income countries reported similar or lower costs compared with our study. In Burundi, the mean annual cost to treat patients with HIV/AIDS was $3223 vs $2032 in Brazil.20 In Thailand the mean cost to treat patients with TB/HIV was $1535 vs $6341 in Brazil.21 In China, TB was the most expensive opportunistic infection in PLHIV after Cytomegalovirus infection: $647 and $3189, respectively.22 In South Africa, the cost of hospitalised patients with TB/HIV was $3925. In the same study, the treatment of other common diseases in PLHIV presented higher costs, such as endocrine and metabolic disease ($5260) and gastrointestinal disease ($4889).23 Another South African study reported hospitalisation costs of patients with TB/HIV varying from $2541 to $4885 according to type of TB, ART status and ward (adult and paediatric).24
In contrast to ART, the cost of TB drugs was lower at outpatient care. Furthermore, in Brazil, TB drug costs in patients with HIV were lower than in other low-income and middle-income countries: $36 vs $50 in Burundi, $338 in China and $397 in Thailand.20–22 In Sudan, the cost of TB drugs in PLHIV varied from $113 to $380 according to TB outcome and TB drug scheme.19
The higher cost of TB/HIV can be explained by the higher proportion of patients presenting with CD4 counts below 200 cells/m3, which indicates that coinfected patients are arriving at the health service at a more advanced stage of HIV infection for the first appointment. Another hypothesis is the delay in TB diagnosis and treatment, which can cause a deterioration in the patient’s health and consequently increase treatment costs. Also, analysing the lifestyle of patients with TB/HIV, we perceive higher rates of lifestyle vulnerability, such as drug and alcohol dependency. All these aspects can be linked to rising costs and worse outcomes, such as higher mortality and hospitalisation rates, within patients with TB/HIV. UNAIDS states that the achievement of the Fast-Track Target prevention and treatment tools could decrease the number of HIV-related deaths, including TB deaths in patients with HIV, by 81% up to 2030.25
In our study, only 5 out of 213 patients with HIV/AIDS received isoniazid preventive therapy (IPT) to prevent TB. During the study period, the hospital faced a shortfall in the provision of tuberculin Skin Test (TST) reagent, and IPT was based only on the history of contact with patients with TB and a medical practitioner’s decision as to whether treatment was offered. Our estimates showed that the mean cost to treat LTBI/HIV was much cheaper than the treatment of TB/HIV (US$882 vs US$5289). The treatment is strongly recommended by WHO to prevent TB in PLHIV and should be provided to those who are unlikely to have TB, regardless of TST readings.26 27 Thus, the scaling up of IPT for all PLHIV should be adopted to prevent deaths and save costs related to treatment of TB/HIV coinfection.
The trial design raises some limitations to our study, in spite of advantages that it also brings, such as reduction in potential bias due to the randomisation, practicality of the data collection, and collection of costing data and outcomes at patient level.28 In our study, the gene Xpert test, which is not routinely implemented in the hospital, was used in patients from the intervention arm. However, it was applied at outpatient care and only nine TB suspect patients were able to perform the test due to lack of sputum. Thus, this cost does not seem to influence the final cost of the disease in outpatient care. Furthermore, we cannot generalise our results to countries where costs are likely to be different; thus, our results have a low external validity. The artificial environment created by a trial (patient tracing, for instance) is also another issue to be considered when extrapolating our data, although the pragmatic design is likely to reduce the limitations created by this artificial environment.28 Another limitation relates to the price of laboratory tests collected from the Brazilian MoH database. These values represent amounts paid to providers to partially cover their laboratory costs and do not represent the full costs of the tests. In our sensitivity analysis, these costs did not present an important impact on the cost estimates by disease categories. The small sample size, especially for patients with TB/HIV and LTBI/HIV, could also be a limitation in the generalisability of our results.
Successful experience of TB/HIV control in another Latin American country, Peru, was evaluated. The Community-Based Accompaniment with Supervised Antiretroviral was a cost-saving intervention and decreased the rate of death from 30% to 9%.29 In Brazil, the strengthening of collaborative TB/HIV activities, early and accurate TB diagnosis, IPT and other policies addressing vulnerable populations can save further costs for the public health system and can contribute to the achievement of SDG and UNAIDS Fast-Track goals in a sustainable and consolidated way.
In conclusion, patients with TB/HIV impose a higher economic burden on the public health system than patients with HIV or AIDS and LTBI/HIV in all pathways of care. Further studies should address the costs of scaling up of IPT. It is important that other studies addressing the budget impact of social protection programmes for patients with TB/HIV and costs of intensive TB case finding algorithms and more accurate TB diagnosis among patients with HIV are carried out in the Brazilian context.
Key messages
Patients with tuberculosis (TB)/HIV can cost 3.4 more than those with HIV/AIDS alone. TB/HIV death can cost almost three times more than an AIDS death.
Patients with latent tuberculosis/HIV had lower costs than all the other disease categories.
Results can support policy planning and direct resource allocation for the Brazilian response to the HIV epidemic.
The study could be a reference for economic evaluation for countries with similar socioeconomic and epidemiological characteristics.
Acknowledgments
We thank all participants who consented to take part in this study; the direction of CPH, Dr Ângela Karine Queiroz e Silva, who authorised and supported the implementation of the study; the field work team who performed the data collection: Adriana Barros, Marcela Santos, Perla Serejo, Marina Siqueira, Livia Vasconcelos, Joilson Gonzaga and Erivelton Martins; and Dr Luciana Siqueira, who classified all drugs and tests. We also thank the Institute for Health Technology Assessment (IATS) for the support with travel expenses and scholarships for the field work team.
References
Footnotes
Handling editor Jackie A Cassell
Contributors The study was designed by NS-F, MdFM and ACS. NS-F was responsible for data collection, analysis and writing the manuscript. ACS, LCR, RL and MdFM reviewed the manuscript.
Funding Conselho Nacional de Desenvolvimento Cientifico e Tecnologico (CNPq) (470554/2013-4) and Fundação de Amparo à Ciência e Tecnologia do Estado de Pernambuco (APQ-0184-4.06/13). Some of the investigators received partial support from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq): Scholarship PQ-308491/2013-0 to MdFM; Scholarship 220144/2012-5 to NS-F.
Competing interests None declared.
Patient consent Obtained.
Ethics approval The study was approved by the Fundacao Oswaldo Cruz (no 279.324) and the London School of Hygiene & Tropical Medicine (ref: 7371) ethics committees.
Provenance and peer review Not commissioned; externally peer reviewed.