Mobile-based Enrolment of Catchment Populations: Experience from rural Tamil Nadu in southern India.

[Deepak Rajanna]

ICTPH is a non-profit organisation providing innovative solutions in the provisioning of primary health-care in rural India. With its on-ground partner, Sughavazhvu, ICTPH implements its designs in three pilots that have been set up in Thanjavur district in rural Tamil Nadu. For more information on our health systems design read our publication.

Each of our pilots, referred to as a Rural Micro Health Center or RMHC, caters to a catchment population of about 10,000 people in a radius of 3-4 kms. This post describes the identification of the catchment and enrolment of its population using mobile devices. The problems that our mobile enrolment exercise is trying to solve are

1. Accurately mapping the households in the catchment and validating the logical placement of the RMHC. This data then serves as the seed data for the electronic health record system that's in place at the RMHC, with a record created for each member of every household surveyed.

2. Identifying the patient (and her household) when she visits the RMHC seeking care so that electronic health records are mapped to the right individual.

3. Enabling our community outreach activities with accurate location information of our target populations.

4. Building a strong landscape-epidemiological database by accurately correlating geographies with disease determinants, and studying care-seeking behaviour and disease occurence patterns in their geographical context.

The solution for enrolment consisted of recruiting temporary workers to perform an enrolment exercise that included distributing a household ID card to every household in the identified catchment and collecting information such as names, age, and gender of family members, and the lat-long co-ordinates of the household using a smart phone running Android OS and a data-collection tool called ODK Collect.

Preparation:

 1. Tool selection: After experimenting with several options we selected an open source suite of products called ODK (open data kit) as the tool for capturing enrolment information. The mobile phone selected for this exercise was Samsung Galaxy Pop running Android v2.1, each costing about Rs 8500. A GPRS connection, costing Rs 99 per month, was enabled on each of the phones. At the backend, we used ODK's in-built integration with Google Fusion Tables to aggregate all the information. This was then linked to the database of our home-grown Health Management Information System (HMIS).

2. Mapping: ICTPH employees performed a ground study and obtained information regarding the estimated population sizes in the identified location, based on the government census. This information was used to identify unique clusters of households that would determine the number of enrolment officers. Logical boundaries were drawn so that enrolment officers do not have overlaps in their allocated areas.

3. Recruitment: Using ads in local newspapers and announcements in the village, we recruited 10 enrolment officers. The pay was performance-based; Rs 15 per enrolment if the officer did fewer than 15 enrolments a day, and Rs 20 if she did 15 or more. Since many of the enrolment officers were not used to smart-phones we trained them on using the new devices by performing supervised mock-enrolments with each other. The training also included soft-skills orientation in order to ensure uniformity in the messaging when the officer communicates with the catchment population.

4. Kit: A kit was prepared to be sent with each enrolment officer. This consisted of a mobile phone, charger, set of bar-coded ID cards with unique numbers, a map consisting of the allotted area, and a bag to carry all this. The ID card contains the address of the RMHC and also mentions the services offered there.

Execution

1. Enrolment: The enrolment process consisted of the following steps

* The enrolment officer introduces herself to the household member and briefly talks about Sughavazhvu and its mission, and ensures that the household has not already been enroled.

* She then hands over a card (at random) and captures the barcode using the ODK form. 

* She then captures household information; address, phone number, GPS co-ordinates.

* For each, individual in that household she captures the name, gender and relationship (to the respondent).

* She then thanks the respondent before proceeding to the next household.

At the end of each day, the enrolment officer proceeds to the RMHC to charge the phones and also to send information through Wifi if she wasn't able to send it through GPRS.

2. Monitoring: At the backend, ODK was integrated with Google Fusion Tables so that all the data collected was monitored real-time. The dashboard captured the relative performance of each of the officers and also helped us to review the quality of the captured data. Mapping the enrolment information on to a realtime map - also provided by Fusion tables - also ensured that no officers strayed outside their own allotted boundaries.

3. Audit: A small portion of the previous day's enrolments - about 5% - were selected for a daily audit that was performed by a full-time staff of Sughavazhvu. The audit questionnaire was designed to catch any falsification of data and also to ensure that the officers were courteous.

Challenges

* One of the challenges is for the auditor to find the right household picked for audit. This is due to the non-standard address formats in use in Indian villages.

* It was initially not straightforward to set the incentives for the enrollment officer. In hindsight we have significantly altered the rates per -form.

* The stack of technologies involved - ODK, Google Fusion Tables, Google AppEngine -  has a bit of a learning curve.

Conclusion

The entire enrolment exercise was completed in under 15 days with fairly good quality, capturing information on about 3000 households consisting of nearly 12500 individuals. This exerience proved to be repeatable when we replicated this activity at another pilot, Alakudi, in less than 10 days, at nearly half the cost (Rs. 10 per household enrolled in Alakudi, as opposed to Rs.20 at Andipatti). The data collected from the exercise was fed into the HMIS before the RMHC operations began and the usage of ID cards immediately proved an improvement over the earlier method of identification which was based on name search. The systematic enrolment process also served as a good marketing exercise since it reached every household in the catchment.

The enrolment process is documented in greater detail in a document available on our website and we'll be happy to share our experiences with anyone wishing to replicate this.

Regards

Deepak Rajanna

VP, Technology Solutions

ICTPH (www.ictph.org.in)

[rahul dhar]

Thanks, Deepak Rajanna

a very informative and interesting pilot study you have discussed.

I just had one query regarding the type of mobile phone being used in the pilot. Don't you think that the price issue of the mobile (being Rs. 8500)  is also one of the biggest challenge. no doubt the Samsung galaxy is technologically advanced and more useful mobile health tool rather than low end phones but when we talk of step- 2 (i.e  up-scaling) of the pilot how do you plan to contain the cost issue.I suppose cost of the mobile phone is a very detrimental  issue when we talk of  up-scaling of mobile health projects in developing countries. 

--
You received this message because you are subscribed to the Google Groups "ict4chw" group.
To post to this group, send email to ict...@googlegroups.com.
To unsubscribe from this group, send email to ict4chw+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/ict4chw?hl=en.

--
Regards

Rahul Dhar

9872874932

[Deepak Rajanna]

Thanks Rahul for the response. You bring up an interesting point. While it was tempting to go ahead with a low-end mobile phone, there were reasons we finally chose an Android hand-held

* The scanning of barcodes and capturing of GPS is central to our data collection exercise. That alone pretty much ruled out many low-cost phones.

* While making a choice on a technology platform, we felt it was better to stay a step ahead and wait for prices to fall down than invest in a platform that's on its way out. Even among mobile platforms, Android, being open-source and having a robust backing, had IMO the strongest chance to catch on. We are already seeing that happen; we could easily replicate this exercise with Micromax and Spice mobiles that are available at around USD 100 (Rs 5000). We expect that to fall even further.

* We did sense of a lot of vitality in the android eco-system, in terms of open-source applications being written and a strong user community developing. Since the time of our first pilot, we've tried out very specific health-related applications, e-learning modules, mapping tools etc, that would have been possible only with a low-end smart phone. 

-Deepak

[rahul dhar]

Thanks Deepak.

the points you have mentioned are really commendable and an  eye-opener.

regards

rahul dhar

[neal lesh]

Dear Deepak and the rest of the ICTPH team,

 

Thanks for this great and clear post and the discussion so far.    I think your describing a use case that has many novel aspects compared to what we’ve discussed already on this list, including using mapping for a tighter integration of clinic and community ICT systems, and a focus on enrollment.  I have a bunch of questions—just answer the ones you want to!  And sorry if some of these are covered already in the paper—feel free to point me to that for ones that are.

 

That’s really interesting to hear about using this data for validating the logical placement of the RMHC.   Just to confirm, I think you are talking about wanting to put the clinic in the right place—to address the problem that people are often very far from a clinic and thus do not seek care.   Is the goal to just have the most central location or do you weight factors such as how vulnerable the household members are?  Do you tend to have a lot of flexibility on where to place a clinic?  I imagine this is useful just to identify the areas that have the least access.

 

A related question is that you mentioned you used population estimates to divide up the territory for the enrollment officers. How accurate did those estimates turn out to be?

 

You also said it was hard for the auditor to find the right household for a return visit. Did you use the GPS coordinates for the households to help guide the auditor?  Is it still hard?

 

And I’m curious to hear more about the auditing and dashboard monitoring activities? Is it right these happened in real time? Did you end up re-training or firing many of the enrollment officers?    And a question I have asked others-- what did you consider a good level of consistency?  Were you expecting close to 100% match between auditor and enrollment officer?

 

Finally, an bit of an open ended question—which is whether you are developing this process more for service delivery or research purposes?  I don’t think there is a sharp line here but curious if you see for example, mapping every household, as something that would be driven if you want to do research in an area or just as a better way to set up service and provide services?

 

thanks very much!

neal

[Deepak Rajanna]

Dear Neal,

Thanks for the questions. 

Logical placement of the RMHC: This insight actually came when we analyzed the GPS locations of households in an existing RMHC-catchment, and we realized that perhaps the RMHC was not really in the most convenient place for the community as a whole. We now have a separate matrix to determine the placement of the RMHC that includes distance from PHC, bus stop, school, etc. and factors in the "center of gravity" in terms of the households mapped. We'll be happy to share that document with this group in a couple of weeks' time.  

Allocating population patches to enrollment officers: The population estimates from the census data were accurate enough that we could equally distribute the work among the agents. The data was from 2001, so we found more households than estimated, but that difference was uniform across most of the hamlets. The few exceptions were hamlets of migrant workers, where the families would not return until the next harvest cycle and hence were left out of the enrollment. In that case the agent who had to do the enrollments at that village finished her work earlier than others.

Audit: The auditor only received a list of addresses to audit. In hindsight, a map containing the co-ordinates of the households might have helped her significantly. At that time, however, we did not have a system to easily pick a set of households at random and then plot them on a map, because the dashboard co-evolved with the exercise while we were doing it on the field. 

Dashboard: Yes, the dashboard did show real time information. In our first enrollment exercise we had to stop enrollment for a day and re-orient the entire group because we found from the map that they were either crossing outside their allocated boundaries or missing some households on their beat. In the second village where we did this exercise, we preempted this situation by emphasizing it in the training and also by handing out a map clearly marking their boundaries.

Purpose: The data that we collect is both to enable research and to improve our service delivery. I'll give an example for each. A couple of researchers who work with us have enriched this demographic GIS data with disease determinant information. In this case, they have mapped the various water sources in the village and recorded the contamination levels in each of those and are now in a position to look for correlations between that information and diarrhea outbreak statistics. Another example where this GIS mapping helped our service delivery was when we launched a CVD (cardio-vascular disease) intervention, which was introduced a couple of months after the launch of the RMHC. The intervention involved identifying people with CVD risks and distributing educational material as well as vouchers for related diagnostic tests. Since we capture the risk factors for everyone who comes to the RMHC (BP, BMI, Waist-to-hip ratio on each visit; Smoking history and family history of CVD on their first visit), it was easy for us to map the high-risk individuals who had already visited us before the intervention began and then target them using our health extension workers.

Regards

Deepak