Developing a Patient-Centered Community-Based Model for Management of Multi-Drug Resistant Tuberculosis in Uganda: A Mixed Methods Study

Background: The advent of all-oral regimens for the management of multi-drug resistant tuberculosis (MDR-TB) makes the implementation of community-based directly observed therapy (CB-DOT) a possibility for this group of patients. We set out to determine patient preferences for different attributes of a community-based model for the management of MDR-TB in Uganda. Methods: The study was conducted at ve tertiary referral hospitals. We used a parallel convergent mixed methods study design. To collect quantitative data, we conducted a discrete choice experiment (DCE) with three different attributes of community-based care (provider type, location of care, and type of support) combined into eight choice sets, each with two options and an opt-out. We collected additional qualitative data by eliciting patient reasons for selection of each choice set. We tted a mixed logit choice model to determine patient preferences for different attributes of community-based care. Thematic analysis using NVivo12 was done to understand the reasons for the choices made.


Background
Multi-drug resistant tuberculosis (MDR-TB), de ned as resistance to both isoniazid and rifampicin, the two major rst line anti-TB medicines, threatens global TB control efforts and remains a major public health concern in many countries. 1 Globally, case detection and treatment success rates for MDR-TB are suboptimal. In 2019, only 57% of patients started on second line therapy were successfully treated. 2 In Uganda, 64% of those started on treatment in 2016 were successfully treated while an estimated 19% died and 15% were lost to follow up. 3 These suboptimal treatment outcomes are a potential risk for the development and spread of further resistance to TB treatment. 4 Uganda currently implements a mixed model of care for MDR-TB characterized by initial hospitalization for two to eight weeks followed by ambulatory directly observed therapy at a public or private health facility near the patient's home. 5 However, delivery of care through a health facility-based directly observed therapy (HF-DOT) model has been documented to signi cantly contribute to poor treatment outcomes particularly in resource limited settings 6 .
Patients who receive care through this model experience various inconveniences (e.g., travel and waiting times) and incur signi cant direct (e.g., transport costs) and indirect costs (e.g., time lost from work) that hinder successful treatment completion. 6 In the management of both drug susceptible and drug resistant TB, community-based treatment support models have been associated with improved treatment outcomes and increased cost effectiveness compared to health facility-based models. 7,8,9,10 Since March 2018, the shorter 9-12 months treatment regimen has been the standard of care for patients with MDR-TB without extrapulmonary TB or resistance to uoroquinolones or second line injectable drugs amongst others. This shorter treatment regimen is better tolerated, and in clinical trials resulted in better treatment outcomes than the two-year treatment regimen. 11 However, this regimen also has lower tolerance for lapses in adherence and there is only one chance to complete the shorter regimen. 5 Provision of effective adherence support for patients receiving the shorter MDR-TB regimen is therefore a priority intervention for the National TB and Leprosy Program (NTLP). Further, the use of community health workers (CHWs) to offer treatment adherence support, particularly in the continuation phase of treatment, is recommended in the Uganda National Guidelines for Programmatic Management of Drug Resistant Tuberculosis 5 but is yet to be implemented. We therefore set out to assess patient preferences for different attributes of community-based care for MDR-TB.

Study setting
The study was carried out at ve tertiary referral hospitals purposively selected because they provide care to about 80% of patients started on second line TB treatment in Uganda every year. Second line TB treatment consists of six drugs; kanamycin (Km), moxi oxacin (Mfx), Ethionamide (Eto), clofazimine (Cfz), pyrazinamide (Z), high dose Isoniazid (H) and ethambutol (E) given in two phases; an intensive four-to-six-month period and a ve-month continuation phase. In the initial intensive phase, all seven drugs are given while in the continuation phase only four drugs are given that include Mfx, Cfz, Z and E. 5

Study Design
We used a parallel convergent mixed methods study design to collect both quantitative and qualitative data on patient preferences for different attributes of community-based care.

Study Variables
The dependent variable was the preferred community-based directly observed therapy (CB-DOT) model, which was the most common chosen combination of the three different CB-DOT attributes that included principal care giver, location of care, and type of additional support. Independent variables included patient demographics, e.g., age, sex, education, duration on MDR-TB treatment, and HIV status.

Data Collection
We included patients 18 years or older who had completed the intensive phase of treatment for MDR-TB. We used strati ed sampling to determine the number of patients to be selected from each hospital, and simple random sampling to select the patients to be interviewed at each hospital. We collected quantitative data on patient preferences using a discrete choice experiment (DCE), a technique for eliciting individual preferences for a product or service. 12 We chose three attributes (provider type, location of service, and provision of additional support) each with two or three attribute levels ( Table 1).
Attributes used in this study were chosen from an initial exploratory study carried out six months prior. Using a fractional factorial design, we chose eight choice sets. Recommendations in the literature 13,14 have shown that more than eight choice tasks are a cognitive and time burden for participants. We included an opt-out response category so that respondents could choose "neither" choice set to re ect dissatisfaction with either potential CB-DOT model.
The nal design had 96% d-e ciency ( Table 2). To cater for limited literacy levels among the respondents, visual representations of each attribute were developed and used during data collection. Data was collected electronically and entered directly into TB Info, a web-based system.
Additional quantitative data on age, gender, marital status, date of treatment initiation, duration on MDR-TB treatment, and underlying co-morbidities was abstracted from health facility records. Qualitative data was collected through patient interviews carried out after each choice set was presented to nd out the reasons for choices made for each set of attributes.

DCE analysis to determine the patients' preferred MDR-TB CB-DOT model
For the primary analysis to determine patients' preferred attributes for community-based care, we used mixed logit choice model (with 1000 Halton simulation draws). We calculated utility scores by substituting coe cients into the model, and computed ranks to determine the most preferred MDR-TB CB-DOT attributes. We estimated the proportion of respondents that had positive relative preference for a certain attribute, using positive proportion = Φ(β/SD). Where, Φ = standard normal cumulative distribution function, and β is positive coe cients from mixed logit choice model, and SD = standard deviation representing level of preference heterogeneity across participants. The following model was used: Utility model (V) = β0 + β1 c_chw + β2 c_expert + β3 l_home + β4 s_phcall + β5 s_tvouch Where, V is utility derived from respondents' preference β1 to β5 are coe cients indicating relative importance of each attribute. 15 For the secondary analysis to determine if these preferences differed by gender, age groups, HIV status, or duration on MDR-TB treatment, we tted interactions to perform sub-group analysis. The signi cance level for all the analyses was determined at 5%.

Qualitative data
A deductive approach with descriptive thematic coding was used to analyse data using NVivo Version 12. Interview transcripts were reviewed for content related to the research question. Codes were assigned to relevant segments of the text and similar codes aggregated to form themes that were used to develop coherent narratives. Data within and across themes were synthesized to generate an understanding of why certain attribute choices were preferred.
Human subjects' approval   (Table 4) and a positive constant term (coe cient = 0.52) implying that participants preferred at least one of the suggested hypothetical MDR-TB community-based care models to none.
Preference for speci c attributes of community-based care For speci c attributes, the positive coe cients indicate that participants preferred: (1) treatment delivered by CHWs or expert clients rather than a family member; (2) treatment delivered at home rather than workplace; and (3) monthly travel vouchers as the form of additional support rather than phone call or SMS reminders. The estimated proportion of respondents with positive preference for CHWs and expert clients as the principal care giver for MDR-TB treatment, compared to a family member, was 73.0% and 69.2%, respectively. The proportion of patients with positive preference for delivery of treatment at home compared to delivery at workplace was 85.4%, whereas patients with positive preference for phone call reminders and monthly travel vouchers compared to SMS reminders was 99.9% and 99.3%, respectively (Table 4).

Preference for models (combination of attributes) of community-based care
The most preferred community-based care model (i.e., combination of attributes) was a CHW giving treatment at the patient's home and the patient receiving monthly travel vouchers for additional support. The top four preferred MDR-TB CB-DOT models contained travel vouchers or phone call reminders and treatment delivered at home (Table 5).

Sub-group analysis
Sub-group analysis showed that patients' preference of the CB-DOT model attributes did not differ by gender. More HIV-positive than HIV-negative people with MDR-TB preferred their treatment delivered at home rather than at work (interaction term P-value = 0.04). Older patients aged 35 years and above preferred a CHW more than younger patients (interaction term P-value = 0.02). Those who had been on MDR-TB treatment for a longer duration (6 months or more) preferred a CHW (interaction term P-value = 0.02) or expert client (interaction term P-value = 0.02). Mix logit choice model with continuous assumption of choices. CI -Con dence interval, SD heterogeneity. -standard deviation for preference heterogeneity, SE -standard error. % positive indicates proportion of patients with positive preference for an attribute.

Qualitative results
Several key themes emerged from the data regarding the preferred attributes of community-based model of care.

Provider Type
Respondents reported that a CHW or expert client was preferred because they were trained and knowledgeable. In addition, they had the ability to offer adherence and psychosocial support including guidance on how to take their medications because of their professional or personal experience. Further, patients felt that expert clients could be more empathetic because they have been through similar experiences. They therefore trusted them to maintain con dentiality in the process of offering care.
"A community health worker encourages you to take medicine than any other person. Also provides counselling to the patient and explains the bene t of taking medicines on time." [20 year Family members were least preferred and viewed as unable to offer the support needed.
"...……. family members, they sympathize so much and may discourage you from sticking to your treatment schedules as required and also they lack experience with this type of TB disease." [22 year.,Female,Single] Location of Care Patients preferred to receive care from home citing privacy and a lower risk of TB-related stigma. They also felt that receiving care at home saves them time and money and presents an opportunity for health education for their family.
"The treatment at home also is a bonus, because the community cannot discriminate if con dentiality is kept by the expert client." [29 year., Female, Divorced/separated] Treatment at home and monthly travel vouchers for monthly appointments is my preferred choice because there is time management while taking drugs since no out movement.
[66 year., Male, Married] "It becomes less costly when I get treatment at home." [31 year., Male, Married] The workplace was not preferred by most of the respondents. It was associated with stigma and fear for loss of employment or clients.
"I feel con dentiality is kept if treatment is given at home compared to my workplace where I could easily be seen and get stigmatized." [22 year

Support Type
Participants felt that monthly travel vouchers were the best way to promote appointment keeping as they provided relief from worrying about transport costs and money left over transport fares could be used to meet other needs like food, other household items, payment of debts, savings and farming amongst others.
"Monthly travel vouchers are the best for me because I will be knowing that transport is readily available and even the ticket will be reminding me of the clinic appointment date.

Discussion
We carried out a parallel convergent mixed methods study to determine patient preferences for different attributes of community-based care. We found that people with MDR-TB preferred care to be provided at home by a CHW or an expert client who is a member of the community and as additional support, to receive travel vouchers to enable attendance at monthly clinic follow-up visits. MDR-TB patients preferred community-based care to the current standard of care, which is HF-DOT. CHWs and expert clients were viewed as knowledgeable, experienced, empathetic, and skilled to properly counsel and guide patients on how to manage side effects. They were also seen as able to maintain con dentiality in the process of offering care. Family members were viewed as lacking adequate MDR-TB related knowledge and patients were skeptical of their ability to offer the support required.
Preference for and patient satisfaction with lay providers has also been observed elsewhere, such as in rural Swaziland where DOT and administration of injectable forms of MDR-TB medication was delegated to trained community treatment supporters. 16 In that study, preference for lay providers was driven by their ability to offer adherence counselling, con dentiality, and perceived lower treatment costs, reasons similar to those given by patients in our study.
Most patients preferred home care noting it provides privacy, safety, and comfort. Further, patients viewed home-based care as being less costly as it saves on time spent accessing care and daily transportation costs to follow-up health facilities. The workplace was perceived as a possible source of stigma that could lead to loss of clientele for small business owners or loss of employment. Similar to our study, ndings from rural northern Uganda 17 showed that home-based care was acceptable to both patients and health providers noting that it is safe, conducive to recovery and time saving. This study further showed that home-based care enabled psychosocial support. In our study, however, psychosocial support was mentioned as a reason for preference for a certain provider type rather than place of care.
In Bangladesh, a decentralized, community-based treatment program for patients with drug-resistant tuberculosis used home-based care DOT to address various needs of MDR-TB patients. It was a patient's preferred approach evidenced by their retention in care resulting into improvement in other treatment outcomes. 18 Similarly, a quasi-experimental study done in India showed home-based care to be associated with low stigma 19 similar to our study ndings.
In rural South Africa, MDR-TB patients preferred to receive MDR-TB and HIV care at home, and this was associated with reduced levels of rejection creating strong emotional bonds between patients, families and communities that is critical to health. 20 The home is seen as a place conducive for recovery and offers both psychological and emotional support needed to enable healing. 21 In our study, patients preferred to receive care at home noting that home is private, is associated with low levels of stigma, saves time and money, and provides an opportunity for health education including infection prevention and control at the family level. Speci cally, patients who were MDR-TB HIV co-infected preferred to be treated from home than those who were HIV negative. This may be related to the fact that in 2017, Uganda rolled out differentiated service delivery models for persons infected with HIV that included community-based drug delivery options. 22 It is possible that clients' positive experience with these care delivery models positively in uenced their choice for home based care. 23 Despite documented evidence that digital mobile technologies are useful in supporting TB care, 24 majority of participants preferred monthly travel vouchers over mobile-based support, such as SMS and phone call reminders. In our study, varying degrees of literacy and hearing loss due to drug toxicity probably decreased the preference for mobile technologies. In addition, the utility of the travel voucher in meeting other household needs besides travel to the monthly hospital appointment served as a main driver for this choice. The majority of our respondents earned only about a dollar a day and were receiving a transport voucher worth about 50 dollars a month from the NTLP under its ongoing "enabler program". The travel voucher therefore shielded them against incurring catastrophic costs during their treatment. In Uganda, a recent study done to examine costs incurred during TB treatment showed that that more than half of households affected by TB experienced catastrophic costs, de ned as spending more than 20% of their annual income on TB and these costs were 30 times higher among MDR-TB patients compared to drug sensitive tuberculosis (DS-TB) patients. 25 The major drivers of cost according to the Uganda TB cost survey were non-medical and included transportation, food, and nutritional supplements. Many respondents in our study reported that they could use the travel voucher to buy food.
The study had several strengths and limitations. We had regional representation across the country, involved MDR-TB treatment centers that treat more than 80% of the MDR-TB cases. The results presented here are therefore applicable to MDR TB patients diagnosed across the country. In addition, we used patients who had been on treatment for MDR-TB. Their opinions therefore are grounded in their lived experiences. However, the retrospective nature of the study may have resulted in some recall bias which could have resulted in under reporting. However, we tried to minimize this by using hospital records for non DCE variables. In addition, DCE experiments are sometimes prone to DCE bias which could result in misrepresentation of preferences. This was minimized by using pictorial questionnaires to aide understanding of choice sets. Further, the study used routinely collected data which is prone to missing data. However, efforts were made to minimize this by training research assistants on quality data collection and respective standard operating procedures prior to data collection. Finally, data collected did not include children and other risk populations, like pregnant women. Future studies could include children and other vulnerable populations so that their views are taken into consideration.

Conclusions And Recommendations
Our respondents preferred to take their medicines at home supported by a member of their community but revealed a critical need for additional support to help mitigate the costs of accessing care. Due to the current COVID-19 pandemic, innovations that reduce the need for health facility visits while still providing additional support to meet client needs will become increasingly relevant. 2 The feasibility and effectiveness of these models of care should be further evaluated. Studies to determine the feasibility and effectiveness of our preferred patient care model are underway.