As of February 2010

Research Design

The project was designed as a cluster randomised trial.  Treatment and control clusters were chosen among the population using statistically robust techniques, creating a sample of 1,960 households (in 28 clusters) to test the effect of the interventions on both insured and uninsured households.  

Sampling the uninsured populated presented a number of challenges. During the project design phase, VimoSEWA expected to use government census data, which was not always be available. With no accurate maps or censuses for slums or rural villages, researchers armed with an area house listing had to systematically follow a health worker through the study areas to document each uninsured household she would potentially work with.  While the challenges in identifying the control group resulted in a delay, they also provided a number of lessons for individuals conducting research in urban slums.

Household survey

The household survey was planned for five rounds over the two year project to track illness and health-seeking behaviour. Health workers and researchers developed and tested the survey tools over several months.  Client surveys underwent at least five revisions, in which modifications were made to questions and language.  The final tools reflect VimoSEWA’s grassroots experience on which types of questions elicit accurate responses, rather than standard health survey formats.   In addition, the surveys were designed to capture the entire range of treatment-seeking behaviour over a 2 year period, rather than the more commonly used method of one-time recall.

SEWA Academy, the research agency partner, trained a team of 16 surveyors in December 2009.  Following the training, the baseline survey was initiated in January 2010, about two months later than anticipated due to the additional time needed to finalize the research design.  During the initial survey, researchers encountered a number of problems including locating households without an exact address, finding survey times that were convenient for women, and locating workers in the informal economy who migrate for employment. 

Health Intervention

The health intervention consisted of group health education sessions with women by SEWA’s trained local community health workers (CHWs) on the most claimed-for illnesses (waterborne illness and malaria/fever), and the most common treatment (hysterectomy).  Fourteen CHWs implemented the intervention in Ahmedabad city and district. The intervention consisted of the following:

1. Group education sessions on diarrhoea, malaria, and hysterectomy, with groups of 10-15 women conducted 3-5 times a month by each CHW in her work area.  In addition, Mahila SEWA Housing Trust, a SEWA sister agency that specializes in housing and sanitation, conducts group sessions on sanitation infrastructure with intervention area CHW groups. 

2. Communication aides for the education sessions were developed by an advertising agency:

1. Two tailor-made ‘snakes and ladders’ participatory games on diarrhoea and malaria
2. A 15-minute film on hysterectomy with case studies from SEWA members, Illustrations, interviews with local gynaecologists, and information on potential side effects
3. Posters and flip charts with illustrated diagrams
4. Take-home illustrated pamphlets for participants

3. Community media:  Community wall paintings on prevention and treatment of diarrhoea and malaria.

4. Monthly refresher training:  Conducted for CHWs on the specific health education topics and general health communication techniques.

The intervention is an add-on to SEWA’s existing programme:

CHW activities in control and intervention areas

Communication aides were developed to be relevant and interesting for participants.  Various media, e.g. wall paintings or posters, group education aides (flip charts, games, and films), and individual handouts (leaflets and samples) were created in partnership with an advertising agency.  The communication messages for each session were developed with the health worker team, based on existing evidence in public health literature and the team’s suggestions on feasibility, local acceptability, and perceived effectiveness. The 14 CHWs were broken into three small groups for workshops to decide, with inputs and review from a physician and external experts, 1) how to identify the illness 2) treatment options and 3) prevention messages for communication material for waterborne disease, malaria/fever, and hysterectomy.

 The implementation team participated in two training sessions. The first training focused on designing the treatment and communication protocols for three common illnesses: waterborne illness, malaria, and fever. The second training, which was conducted by a senior gynaecologist from New Delhi’s largest public hospital, focused on appropriate use of hysterectomy, which occurs at a much higher frequency and at a lower average age than medically indicated. The trainings will continue on a monthly and then quarterly basis.

As of July 2011

Health prevention and promotion interventions

In April 2010, the 14 CHWs held their first education sessions prior to the start of the monsoon season, the high season for waterborne disease. The first part of the intervention period therefore focused on water borne disease and malaria. The hysterectomy sessions were postponed allowing CHWs to focus on one topic during their first few months as trainers. In the initial three to four months, implementation was refined, with additional trainings and practical support provided as required.  There have been, on average, 42 health sessions per month (with 20-25 women per session) since the start of the intervention.  They will continue at the same frequency until May 2012.   As the quality of the health education sessions is a critical component of the project, VimoSEWA ensures that health workers attend regular, ongoing trainings and monthly meetings to review issues and education techniques, and improve message delivery. 

To improve consistency, guidelines for the education sessions were developed following observation of ten meetings (six urban and four rural):

1. Each education meeting should cover only one of the three topics.
2. CHWs are to inform the members in the community about the session in advance.
3. The CHWs should start a game as soon as four to five members arrive. By the time all the women arrive, the game should be over and the CHW can begin the discussion using posters.
4. The CHWs should focus on key messages during the meeting and have members repeat the messages at the end of each session.
5. To ensure that meeting participants memorize the key messages, each participant should repeat the messages before taking the soap given out at the end of the training.
6. If the education topic is diarrhoea, the CHW should do a practical demonstration of preparing oral rehydration salts (ORS). The demonstration involves boiling water, cooling it, measuring the correct amount using the measuring cup that was given to each CHW and mixing ORS powder. The CHW should remember to carry a packet of ORS.

The guidelines were discussed during regular monthly meetings to review issues and share best practices.

Action research implementation

From February 2009 to March 2010, the baseline study design went through several iterations, pilot testing, and feasibility analyses. The survey was completed in March 2010 and successfully reached 1,934 households out of the planned 1,960 households. If a household could not be located, the reason was carefully documented. The researchers have prepared a detailed process description, which will be used as part of a process evaluation paper at the end of the project.  A second-round survey was conducted six months after the intervention began (November 2010), and a midline survey was initiated in June 2011.  Further, qualitative research was conducted to understand what influences treatment-seeking behaviour and women’s responses to the intervention.

Preliminary analysis of baseline data

The baseline data cover 11,287 individuals in 1,934 households. The data were entered into a database by SEWA Academy over two months and presented to VimoSEWA in August 2010. Data cleaning was completed after three rounds of internal consistency checks in October and November 2010. Next, descriptive statistics of the baseline study were generated and shared with the VimoSEWA and SEWA Health teams in January 2011. Baseline findings included:

• 47 per cent of households reported an illness in the last month, with primary illness (e.g. fever, diarrhoea, and malaria) most commonly reported.
• The proportion of households who reported hospitalization in the past six months was higher for rural than urban households, and higher for insured versus non-insured families (i.e., rural: 18.6 per cent for insured, 12.3 per cent for uninsured; urban:  14.6 per cent for insured, 10.5 per cent for uninsured)
• Private hospitals accounted for 62 per cent of all rural hospital admissions compared with 45 per cent for urban admissions
• Hysterectomy was the most common reason women were admitted to hospital (exceeding deliveries), and occurred more frequently with rural women.  In fact, 9.8 per cent of rural women surveyed had had a hysterectomy, with nearly one third being performed on women younger than 35 years of age.

Publications and supporting documents

Recently, VimoSEWA published a more detailed paper in Reproductive Health Matters on the prevalence and implications of hysterectomy, among rural and urban women with and without insurance in Gujarat. At least three more paper topics have been identified including women’s health-seeking behaviour in Gujarat, the impact of insurance status on health care utilization and survey design issues.

Links to documents on the baseline study:

• Presentation/overview 
• Survey instrument

As of January 2013

Further baseline analysis

VimoSEWA conducted further statistical analysis on the baseline survey to compare morbidity, outpatient treatment-seeking, and hospitalization patterns amongst insured and uninsured adult women.  Controlling for the limited demographic differences that emerged, it found that morbidity and outpatient treatment-seeking are similar – yet insured women are significantly more likely to be hospitalised.    The illnesses that led to hospitalization were not confined to common illness or hysterectomy:  the insured are more likely to be insured across all health conditions.   

This analysis is one of few that examine illness patterns of treatment-seeking behaviour when comparing the insured and uninsured. The conclusion is that, while VimoSEWA’s insurance increases utilization of inpatient health services, the implications are mixed for the health system. While insurance may fulfil unmet demand, it may also compensate for poor quality outpatient care that in turn leads to unnecessary hospitalization.  The linked qualitative research explores this question further.   VimoSEWA recommends integration of a health perspective into evaluations of community-based health insurance, to explore if increased hospitalization improves health outcomes. 

Qualitative research

Two qualitative research studies were initiated and conducted from November 2011 to March 2012.  Although not initially planned for in the project proposal, the baseline findings raised two important questions that SEWA wanted to explore further: 1) Why are insured women in urban Ahmedabad hospitalised for common illnesses such as fever and diarrhoea?  2) What are the influences on women’s decision to have a hysterectomy?   

1. In urban Ahmedabad, fever and diarrhoea were the leading reasons for insured women to be hospitalized, but this was not the case for the uninsured.  To better understand these episodes of illness, including women’s perception of their illness and the health care they receive, VimoSEWA conducted qualitative research with women who were recently hospitalized for common illnesses, their families, and local health care providers. The results indicated that women rarely sought hospitalization first – they almost always tried (and spent money on) outpatient care to treat a common illness. Insured women did not appear to seek immediate hospitalization as a substitution for outpatient care because with insurance, they could save out of pocket expenditure. The research highlighted that hospital admission carries considerable opportunity costs for working women who also manage their households, and was chosen only after outpatient treatment repeatedly failed. In summary, insured women appear to seek hospitalization only as a last-resort to cure a persistent illness.

Health care providers felt insured women would more readily agree to be admitted to hospital, as their insurance would cover all or a portion of their health care expenses. The likelihood that a health care provider would be influenced to act in self-interest (moral hazard) by admitting a patient for care in the hospital in order to collect fees paid by an insurer could not be ruled out, however. 

1. The hysterectomy research included 35 women, insured and uninsured, five providers, key informants such as midwives, and local health workers. Gynaecological morbidity, lack of accessible, affordable gynaecological care, a perceived lack of need of the uterus,  provider practice, and women’s demand  emerged as factors to explore further in understanding what drives hysterectomy amongst women under 45. 

Endline survey

An endline household survey was conducted with the same sample of 1,934 households who participated in earlier rounds, with some attrition due to a slum demolition in Ahmedabad city. VimoSEWA is currently conducting in-depth statistical analyses of the cluster randomized trial to ascertain if and how the intervention affected the primary outcomes of illness incidence, treatment-seeking behaviour, hospitalization, expenditure, and claims patterns amongst the insured, as well as compared to the uninsured.   Preliminary analysis of secondary outcomes indicates that the CHW intervention resulted in more interaction and higher satisfaction with SEWA’s community health workers in rural areas.  SEWA’s education intervention improved some preventive health practices amongst urban households, but not in rural areas.   Households in urban intervention areas reported higher use of effective anti-mosquito measures and safer drinking water practices.

Dissemination

Findings from the research have been disseminated through presentations, publications, and sharing workshops.   The baseline survey results were shared with VimoSEWA and SEWA Health leadership, field managers, and grassroots workers in three workshops.   These led to additional hypotheses and support for the initiation of qualitative research.  Two papers have been published in academic journals, and two are currently under review.  VimoSEWA has also presented:  a mixed methods analysis of hospitalization for common illnesses at the University of Twente conference on microinsurance; baseline and preliminary endline findings at the ILO Microinsurance Facility Health Forum in Delhi; a poster on hysterectomy at the WHO Health Systems conference in Beijing; and an analysis of claims, survey findings, and qualitative research on hospitalization at the Munich Re conference in Dar es Salaam.