2.1 Study overview
We undertook a case-control study of the association between MSI and poverty, time use and quality of life in post-conflict Myanmar. Cases were recruited from two physical rehabilitation centres, prior to receipt of rehabilitation services. One age (+/- 5 years of the case’s age) and sex matched control was recruited per case, living in the same community as the case and having no physical impairment. All cases and controls underwent in-depth interview using a structured questionnaire.
2.2 Sample Size Calculation
There is no existing data on the possible association between MSI and poverty on which to calculate a case-control study design sample size (21). Consequently, this study followed Norman et al.’s recommendation that a sample size of 64 per group would detect a medium effect size of 0.5(22). Accounting for prospective drop out of up to 40% at one year post follow up, a sample of 100 cases and 100 controls were recruited.
2.3 Participant Recruitment
Cases were recruited from two physical rehabilitation centres in Myanmar, which were the main providers of prostheses in the country in 2015: The National Rehabilitation Hospital in Yangon (NRH, operated by the Myanmar Ministry of Health), and the Hpa-An Orthopaedic and Rehabilitation Centre in Hpa-An (HORC, operated by the Myanmar Red Cross Society in collaboration with the Ministry of Health and the International Committee of the Red Cross).
Participants were eligible for enrolment in the study if they:
- were ≥18 years old,
- had never previously been fitted with a prosthetic or orthotic assistive device,
- were determined by a trained physiotherapist to need to be fitted with either a prosthetic or orthotic device due to MSI,
- were able to communicate independently or via translator,
- did not plan to migrate outside of Myanmar within the following twelve months.
Clients at NRH and HORC meeting the above criteria were provided oral and written information about the study and requested to formally consent to participate. All clients were assured that they had the right not to participate and that this would not affect the services they received.
For each client who met the eligibility criteria and agreed to participate (“cases”), one matched control was identified from the same local community as the case. Controls were identified as follows: the same sex as the case, +/- five years of age, able to communicate independently or via translator, not planning to migrate outside of Myanmar within the following twelve months and did not have an MSI.
To identify controls, data collectors accompanied cases to their homes or were provided with information from the cases to identify their home independently. The data collector spun a bottle outside the case’s house and walked in the direction of the bottle to the nearest house to identify a control matching the above criteria. If an eligible control was available, the data collector provided relevant study information and asked the control if they wished to participate before taking written consent and beginning the interview.
If no eligible control was identified within the household, or the eligible control chose not to participate, the data-collector returned to the case’s household, re-spun the bottle and continued the process until an eligible control was identified.
2.4 Data Collection
. Cases were assessed using the Rapid Assessment of Musculoskeletal Impairment (RAM) tool to identify MSI presence, severity and aetiology according to pre-validated algorithms (4, 23). The RAM was developed and validated for use in LMICs, and has been previously used in Kenya, Rwanda, Cameroon and India (23-25).
Physical functioning was assessed using two standardised tools: the Physical Performance Test (PPT) and the Two Minute Walk Test (TMWT (26). The PPT comprises 9 items based on the time it takes them to complete each task. A score of 0 relates to inability to complete a task, with higher scores for quicker completion rates. The PPT has not previously been used in low income settings.The TMWT is a widely validated test of aerobic capacity and endurance in post-stroke rehabilitation, spinal cord injury and amputation (27-29). The TMWT measures the distance ambulated in two minutes on flat ground.
Time use was measured using the ‘Stylised Activity List’ developed by the Living Standards Measurement Study (30). The tool contains thirteen broad activities comprising areas of personal care (e.g. sleeping, bathing/dressing and medical care), productive activities (both paid and non-paid activities including household tasks), leisure (in and outside the household) and time spent resting (no activity). The number of hours spent undertaking each activity on the previous day is recorded, alongside whether or not assistance was needed in undertaking each activity. This tool has previously been used in assessing the long term impact of cataract surgery in Bangladesh, Philippines and Kenya(31).
We used the WHOQOL-BREF, developed by the World Health Organisation (WHO) to assess quality of life. The WHOQOL-BREF comprises 26 items related to physical, psychological, social and environmental domains of quality of life, and uses Likert scale responses ranging between very poor/very dissatisfied/not at all, and very good/very satisfied/an extreme amount. The WHOQOL-BREF has shown excellent reliability and validity in more than 20 countries (32).
SES was measured in three different ways, each in accordance with World Bank recommendations of reliable and comparable collection of household SES data in LMICs(33): (i) Household income was measured directly as reported average monthly income in the household; (ii) Household expenditure was measured across 85 pre-validated, pilot-tested items related to expenditure on food (including value via home production, received in kind or as gifts), education, health, household and personal items and rent (34); and (iii) Asset ownership was measured using a pre-tested asset list (33 items) to assess the number and type of assets owned by the household (e.g. furniture, vehicles, cattle) and key characteristics of the household structure (e.g. building materials, number of rooms).
All questions related to socio-economic status were asked directly to the person in the household with primary responsibility for the household’s finances.
2.5 Training and field work
Mid-level rehabilitation professionals (e.g. orthopaedic technicians, physiotherapists or physiotherapist assistants) at NRH and HORC were provided training to assist data collection through recruitment and physical assessement of eligible clients.
In addition, six full-time data collectors were recruited from local universities. A two-week training course was held in July 2015 incorporating modules on disability sensitisation (led by a local disabled persons’ organisation), project protocol and data collection tools, informed consent and ethics, study logistics and recruitment, safety and security.
Ten volunteers were recruited from NRH, alongside ten community volunteers as part of the training programme, to pilot-test the tools and study approach. Data was collected, stored and managed using a bespoke Android application,built onPython coding and deployed using Google Nexus tablets
2.6 Statistical Analysis
Data were cleaned and analysed in Stata 14.0 (35). Perfect matching between cases and controls was not achieved, excluding paired analysis approaches.
Chi-squared tests of association and age-sex adjusted logistic regression analyses were used to measure differences in socio-demographic characteristics between cases and controls, whilst descriptive statistics were used to describe case service-centre details.
PPT scores were divided into categories based on crude thirds (0-12, 13-24 and 25-36). PPT category and TMWT average distance were compared between cases and controls using Chi-squared and student t-tests of association/difference respectively.
Household monthly income was divided by household size to estimate Per Capita Income (PCY). Similarly, Personal Consumption Expenditure (PCE) was calculated by dividing household expenditure by household size. Both PCY and PCE were converted into US dollars for ease of interpretation. The assets list was used to derive a household-level relative index indicating SES, via Principle Components Analysis (PCA) and categorised into tertiles (36). PCA involves a statistical calculation of the relative weight of different assets, producing a total score per household.
Due to the skewed nature of income and expenditure variables, raw PCE and PCY results were logged, and exponentiated regression coefficients were derived using linear regression, accounting for age and sex. Age-Sex adjusted Logistic Regression was used to derive odds ratios for the proportion of cases and controls experiencing catastrophic health expenditure (≥ 10% monthly per capita expenditure(37)), below the international Lower Middle Income Country Poverty Line (3.20 USD, adjusted for Purchasing Power Parity), in each PCA tertile, and experiencing an income gap (PCE>PCI).
Time-use allocation was aggregated and any responses totalling less than 19 or greater than 29 hours were removed from the analysis. Age-sex adjusted logistic regression was used to compare participation in different activities amongst cases and controls. Logged linear regression was undertaken, accounting for age and sex, to assess differences in the proportion of time spent in different activities between cases and controls.
Quality of Life scores were aggregated and transformed into scores out of 100. Mean scores were compared using a student t-test.
Multivariate logistic regression analyses were undertaken amongst cases to ascertain associations between:
- case quality of life scores (general quality of life score, general health quality of life score, physical heath quality of life score and psychological health quality of life score respectively)
- and age group, work status, proportion of the day spent resting, proportion of the day spent in productive activities, physical functioning score, PCA tertile, PCE quartile, PCI quartile and proportion experiencing income gap.
2.7 Ethical Approval
Ethical approval for the study was granted by the Research Ethics Committee (ref 9292) at the London School of Hygiene & Tropical Medicine and the Myanmar Ministry of Health Ethical Review Board.