1. Insurance Innovation in mHealth: Results from a Pilot in Uganda
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Insurance Innovation in mHealth: Results from a Pilot in Uganda
Melissa R. Densmore
PhD Candidate
University of California, Berkeley
School of Information
2010 mHealth Summit
Washington, DC,
Tuesday, November 9, 2010
Presented by Melissa R. Densmore,

2. Uganda Mobile GSM Coverage Population: 33,398,682 (cia.gov)
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Population: 33,398,682 (cia.gov)
Landlines: 1/200 people (World Bank ‘08)
Mobiles: 27/100 people (World Bank ‘08)
Uganda
Mobile GSM Coverage
From 16% in 2000 to 100% in 2008
(World Bank ICT At a Glance)
Images from coverage maps available on gsmworld.com
Quick context: this work is based in Uganda, which CIA.gov’s most recent statistics boast a population of 31.3 million and growing, and 162,000 landlines. While cia.gov claims 4.2 million mobile subscribers More recent statistics claim 8.2 million subscribers.
We work specifically in several districts surrounding the urban trading center of Mbarara, Uganda, which has fairly dense mobile coverage, generally with GPRS/EDGE capabilities of varying quality. This particular area, at a junction between Tanzania, Rwanda, and DRC has an especially high incidence of sexually transmitted infections, which is what our partner program seeks to treat.
Internet infrastructure
Uganda only very recently (as of November/December 2009) has direct (I.e. wired) access to the global Internet, being located in East Africa.
Most Internet access ultimately sources from VSAT connections, which are high cost and low bandwidth
Mobile Coverage
8.56 million mobile phones - that’s 1 for every 4 people, in reality, probably about 1 in every people have 2-3 sim cards
This is a map of the GSM coverage for two of the providers
Note specifically rapid increase to 100% GSM coverage of population
Health Care
While health care is provided free in clinics, people don’t’ actually go to public clinics, due to lack of services, doctors, drugs
However – alternatives: tradititional methods are unreliable and private clinics are too expensive
In this project we seek to improve access to private clinics.
Sexually Transmitted Diseases are an especially critical problem in Western Uganda
HIV prevalence: 10% of adult population (15-49 years)
Syphilis prevalence: about 5-7% of adult population
1 in 4 households had at least one phone.
39% reported STI symptoms
only 1/3 sought care
54% of respondents who sought any STI treatment reported using private clinics.
From 2006 Venture Strategies and Mbarara University population survey.
Presented by Melissa R. Densmore,

3. The Uganda Output-Based Aid Project
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
The Uganda Output-Based Aid Project
Goal: Change health-seeking behavior
Purchase 3000 UGX from independent distributors (e.g. pharmacies)
Patients
Provide treatment in exchange for vouchers
Submit claims forms to OBA Management Agency
Health Service Providers
Reviews claims (Fraud detection)
Reimburses existing service providers for services rendered
OBA Management Agencies
claim forms
Providing effective health care in poor countries is an essential component to economic development and poverty reduction. Unfortunately donors supporting this endeavor often find that resources given are not matched by desired gains. The output-based aid (OBA) model of financing seeks to address this by paying healthcare providers directly for services rendered instead of paying for the service provision up front. These “healthylife treatment centers” treat patients in exchange for bar-coded vouchers. The NGO managing the aid money then pays the service providers the cost of the treatment.
You can liken this project almost to a health insurance scheme – except that instead of multiple payers, you have a single payer in the form of an aid agency, and that only two conditions are covered: sexually transmitted infections and maternal delivery services.
Unfortunately, the program management is information intensive, necessitating much paperwork to track and reimburse payment claims. Subsequent delays in payment cause hardship for participating clinics.
In this particular case, the HSPs are paid by bank transfer, usually about $500-$1500USD every two weeks via EFT. As part of participation in the program, providers are required to have a bank account, and wires are initiated via a web interface.
We were asked to investigate the possibility of using mobile phones to improve the efficiency of the program, increasing claims processing speed by using mobile phones to submit the forms from the health clinics to the management agency instead of submitting all of the data on the paper forms by hand.
Once approached with this idea, we went to Uganda to meet our potential partners and do a needs assessment, starting with a survey of the clinics to both better understand the claims process, where their struggles were, and how mobile phones might be of benefit.
But program management is information intensive!
vouchers
Presented by Melissa R. Densmore,

4. Information Constraints
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Information Constraints
Providers often have questions about claim summary reports
4/12 didn’t know how many claims have been rejected
3/12 had not gotten feedback
4/12 have computer training
Obscure and Infrequent Feedback
“I don’t know. I don’t know how we are performing. I don’t know how we are faring… and of course it takes a lot of time.
12/12 own a mobile phone
The 11 here indicates that the service provider made an error on the form, indicating a price for the drug that was higher than what the management agency actually reimburses. This is more of a clerical error than anything else. But other common errors also occur, and the service providers often do not understand the numbered codes. While error codes explanations were provided soon after the program launched, most providers had lost that sheet – often buried under a pile of papers. And when asked, they told us that they felt that it was futile to inquire for more feedback or to try to dispute any rejections.
In the photograph seen in this slide, it’s written at the bottom, “All above denied b'cos rest of P[atien]t mgt n[ot]. Ethical [i.e. not compliant with reimbursement guidelines]”. This is a sample medical advisor review of a claims summary, often occurring weeks after the original mistake had been made several times, before it could be caught and corrective measures could be made, as noted in the first line: “Cipro pricing b4 C[ontinuing] M[edical] E[ducation]”
Timeliness in claims processing is a critical issue – not just because healthcare businesses need cash flow, but because errors that are not caught early are repeated until a medical advisor has a chance to review the claims. In this case, cipro is an expensive drug, and this service provider has lost the cost of five doses of cipro, and probably the lab reagents for the accompanying consultations when repeating the same mistake. If the error had been caught within the first few days, they might have only lost one or two doses.
Sequential Rejections
“Cipro pricing b4 C[ontinuing] M[edical] E[ducation]… All above denied b'cos rest of P[atien]t mgt n[ot]. ethical”
Often an entire month’s worth of claims might be rejected at once for the same error
Presented by Melissa R. Densmore,

5. What kind of mHealth Application fits your context?
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
What kind of mHealth Application fits your context?
Communications prioritized over claims processing
Complicated forms exceeding 150 characters
Voice/IVR/SMS-based Communication Systems
Smartphone-based Applications
Can make use of advanced capabilities of newer phones
Application Flexibility
Information Management capabilities
Uses user’s existing hardware
Ease of deployment
Little training required
Larger rollout required: 110 clinics + 80 distributors + sales
Small rollout: Initially planned for clinics
1.5 Phones owned per clinic
8 SMSes sent weekly
Expert Audience: English Speaking clinicians
Partner Buy-In: RFP initiated by MA and developed by local agency
Once we had evaluated the context, it was time to make a decision about what kind of application to build to address the issues and the context identified by the needs assessment.
Here I categorize phone-based mHealth applications into two categories: Voice/SMS-based communications systems, that use the users’s existing phones, and smartphone applications that typically require some advanced application software.
I think it is important to acknowledge the fact that a good percentage of the phones out there are still the nokia candy bars – chosen for their durability and battery life. the applications in the former category will have a potential for scale – but may have a limited capability, due to the limits of the medium. We like smartphone or java-based applications because we as developers feel empowered to manage and visualize information, to add features to phones like cameras and sensors, and gps locations. However, the price and reliability of the hardware will ultimately limit the scalability and replicability of our experiments.
Low-Hanging Fruit: Smartphone based applications is more appropriate for targeting the service providers, and/or smaller groups of users, where voice/sms systems are better for targeting larger groups of people.
Example:
<click 1>
Claim Mobile
<click 2>
Bulk SMS
Partner Buy-In: Interest in mobile claims processing from MA and donors
Confirmed mobile coverage area
SMS is more reliable than GPRS, if more expensive
Presented by Melissa R. Densmore,

6. Claim Mobile from paper form… to phone
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Claim Mobile
from paper form… to phone
Using mobile phones as a platform for facilitating information management
Dynamic, self-verifying forms reduce errors and provide ongoing training
GPRS-based form submission and approval
Phone-based clinic data management
Improved communications loop
August 2008 Pre-Pilot
Rural Clinic
12/86 claims via CM
Discrepancies noted on paper claims that would have been avoided via CM
Non-monetary errors go unreported
Urban Clinic
18/18 claims via CM
5 following study
Usability
Onscreen keyboard is preferred
“Qwerty” keyboard is acceptable
We first tested Claim Mobile in two health facilities in August The “computer” on the phones could do pre-verification, reducing errors on the clinic side before they were submitted, and the communications capabilities and the data entry aspect would decrease the amount of time it would take to process the claims once they arrived.
We tested the full system as we were designing it, deploying Claim Mobile and submitting claims from the mobile phone to the web application from two clinics, a rural high volume clinic and an urban low volume clinic.
Over and over their refrain was that “we can learn” – that even with very little experience with computers or qwerty keyboards they were happy to learn new things and new technologies that this was something they were happy to work with.
However, as a caveat, when I returned the following year, I found the rural clinic had entered a number of claims without submitting them, for fear of making a mistake. When I visited, the phone was entirely discharged and unused. At the urban clinic, however, the clinician was using the phone for distance education, using its Internet capabilities to study for her public health program. She asked to purchase the phone rather than to retuern it ot the research project.
Presented by Melissa R. Densmore,

7. Claims Submission Processing
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Claims Submission Processing
Health Service Providers
Common errors
Written errors
Invalid client-partner use
Invalid treatment
Submission delays
Batched claims
Transportation ($ + time)
Processing delays
Data entry
Medical & technical review
Reconciling partial
payments and rejections
Payment
30-45 days
Providers travel up to 3.5 hours to submit claim forms
14-90 days
prepare claims
pay claims
batch claims
approve claims
deliver claims
data entry
Most of the processing time is consumed in data entry and validation - to some extent the role of the smart claims is to reduce the data entry process and to streamline the data entry, while pipelining the batch process.
Output Based Aid Management Agencies
Presented by Melissa R. Densmore,

8. Claim Mobile Pilot January 2009 – April 2010
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Claim Mobile Pilot January 2009 – April 2010
Methods: Baseline Survey Participant Observation Controlled Study Management Agency Findings: Delays Resolved via Program management Internal issues w/claim backlog New claims backend pending (delayed until Jan 2010) Claims Forms: From one A4 per visit to 4 pages of A4, plus additional forms. Conclusion: Compare Laptop vs Phone
Health Clinic Findings: Interest in Phones is for secondary usage: patient data management, health education High interest in Computers Low existing training and experience with computers
In Jan 2009 I returned to Uganda for 15 months to expand the deployment to include all the providers in the program. I began the study with a baseline survey, also employing participant observation, in the management agencey and health facilities. My goal was to conduct a controlled study around the deployment of this intervention , to better understand the implications of introduction of ICTs into this context.
Realizing through my observations that the management agency had resolved a lot of delays described before through program management – and that the introduction of a new system would entail a lot of computer training that the agency would niot have the capacity to absorb, I instead chose 8 fa ilities and deployed 8 netbooks and palm treos in each, enabling more effective clinet education and fundraising through better communication of heatlh outcomes.
Presented by Melissa R. Densmore,

9. Bulk SMS: Enabling Broadcast Announcements
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Bulk SMS: Enabling Broadcast Announcements
Having scaled back the smartphone deployment, we also realized that effective communications was still a problem. I worked with my partner organization to commission a Bulk SMS system from a third party organization.
Prices for software devleopment range from “free” applications to $15,000.
We ultimately selected a local software company somewhere in the midlle base on feature set reqs, primarily allowing us to ssend out broadcase announcements to various subgroups, and to use templates to send out payment advices.
150 characters
Addressed from MSIU (on dominant carrier)
Template Support
Two-Way Communication
Automatic Archiving of Messages
SMS Forwarding
Group Addressing
Presented by Melissa R. Densmore,

10. Health Service Provider
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Message Types
Communications Announcements
Hello service provider, please check your post office mail for your payment summary reports and review letter from UOBA.Confirm receipt. Thx. MGT[-]
Confirmation Receipt
am stil wating ad up nw hv nt receivd any leter frm the yo ofice via ma box numbr ZZ ZZ Town. bt i wish u use ma addres its post ofice pple a not eficiet. bt i wl agree wth yo new terms
Program Officer
Program Announcements
Dear svce prvders, a team 'll b coming 2 midwstern UG btn 8th-14th to collect the Dec.claims 4 both ANC and Del.Pse summarize and organise apprtly.[-]
Health Service Provider
Medical Advisor
Medical Protocol Query
does oba for healthy baby cater for cough (rti\'s) in pregnancy? if yes, what drugs are recommended? Clinic.Y
Medical Query Reply
no msiu does not take care of rtis.[-]
This screen highlights a number of actual messages (names and numbers changed) from the syste.
This is actual (Ugandan) user generated content.
Note1 Content is bidirectional and can be conversational: in addition to bulk messages, there are templates, and conversations
Note2 There are “meta” messages (e.g. Confirmation Receipt) about postal communications and s, and especially payments (e.g., pls chk ur mail)
Note3 Use of informal language
Payment Query is a complante: but the ability to complain is a form of empowerment
Prior survey shows that some HSPs were spending $10-$20 a week on looking for payments of medical querys, etc. By being able to document these queries, and having this type of “free” service. HSPs spend less money on airtime and have a better ability to know what is going on with the OBA project.
Finance Officer
Payment Query
Clinic W is asking about the oba payment b\'se it had delayed to be put on the hosp a/c
Payment Advices
Dear FACLILITY,OBA on 22 jan crdted u with u'r facility's latest submn.snd us u'r queries and cnfrm rcpt to mgt on 078XXXXXXX[-]
Payment Confirmation
i acknowledge receipt of paymt 4 nov. '09. thanx. Clinic Y
Presented by Melissa R. Densmore,

11. Bulk SMS Technical Difficulties
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Bulk SMS Technical Difficulties
Carrier Dependence
SMS service is on dominant carrier
HSPs on other carriers report that they failed to receive SMS messages
10.7% of Warid confirmed non-delivery, other networks have no indicators
Phone Number management
Wrong numbers
Changed numbers
Multiple numbers
Swapped/shifted numbers in spreadsheet
Un-received Texts
HSPs in low coverage areas
HSPs without electricity (phones powered off)
Duplicate and Delayed Texts
System Limitations
Limited Capacity for SMSes on local database
Short Messages req’d hack for long message format
However: Overall perceived as a benefit to the program
There are issues with the new system. Many providers report that they either never received notifications, or used to receive them but stopped getting them after some time. Some providers report that they have been receiving text messages intended for others. Clinic A has reported that Clinic B relayed a message that Clinic A's finance reports were mistakenly being sent to Clinic B. Clinic A notes that they have never received any notices of payment but they do get general announcements from MSIU. Corrections are made in the databases when errors are identified.
While all HSPs could be on any network, SMS systems don’t play well with all networks, and providers using Warid and UTL aren’t receiving the text messages. Zain users receive the messages, but the phone number isn’t automatically displayed as “MSIU”
To overcome the character limit, MSIU links two SMS messages with a continue tag.
Presented by Melissa R. Densmore,

12. mHealth in a vacuum is vulnerable and will not scale
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
mHealth in a vacuum is vulnerable and will not scale
Find a good deployment partner to guarantee sustainability.
100% Mobile Coverage in 2008 (World Bank)
GPRS/Internet Technologies are more and more available
Mobile technology requires less training than computers, and is easier to deploy.
We like to think that “mobile” is the way to leapfrog our way into an easy, scaleable solution for the healthcare, agriculture, etc in the developing would, simply because it seems like everyone and their brother has a mobile phone and knows how to use it.
Just look at the stats - there’s 100% covereage in Uganda! And where there’s GSM, there’s GPRS!
All we have to do is find a good on-the ground partner, and we will have a sustainable project…
Well, the on the ground reality is very different – most phones in use aren’t GPRS enables – and the phones that are in use aren’t on all the time, simply because they are off grid.
Bulk SMS faild us because exchange agreements between providers weren’t settled – resulting in dropped messages.
GPRS and 3G historically fails within a year of deployment due to oversubscription – MTN lost subscribers to Warid, whicih has now lost subscribers to Orange in Uganda.
Sustainability is a fraught question – but simply finding a good partner is not enough – many aid projects are dependent on donors and grants with 3-5 year grant cycles. Until we escape that funding model, we will never be “sustainable”.
And lastly – “mobile” health is a myth implying that our applications and the answer to health solutions resides inside aphone or inside little sensors. None of these solutions truly exist without a good server to back it up.n While admittedly the call of the mobile is lots more sexy than the giant energy sucking server room, we still need to think about what it means to develop and maintain the backends that work with the mobiles. They don’t exist in a vacuum.
Presented by Melissa R. Densmore,

13. mHealth in a vacuum is vulnerable and will not scale
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
mHealth in a vacuum is vulnerable and will not scale
Many AID Projects have 3-5 year grant-dependent funding cycles
Find a good deployment partner to guarantee sustainability.
85% of Ugandans are off Grid
100% Mobile Coverage in 2008 (World Bank)
Exchange agreements between providers are not settled
Mobile deployments are only as good as their backend – and smartphones still need training.
GPRS/Internet Technologies are more and more available
Mobile technology requires less training than computers, and is easier to deploy.
We like to think that “mobile” is the way to leapfrog our way into an easy, scaleable solution for the healthcare, agriculture, etc in the developing would, simply because it seems like everyone and their brother has a mobile phone and knows how to use it.
Just look at the stats - there’s 100% covereage in Uganda! And where there’s GSM, there’s GPRS!
All we have to do is find a good on-the ground partner, and we will have a sustainable project…
Well, the on the ground reality is very different – most phones in use aren’t GPRS enables – and the phones that are in use aren’t on all the time, simply because they are off grid.
Bulk SMS faild us because exchange agreements between providers weren’t settled – resulting in dropped messages.
GPRS and 3G historically fails within a year of deployment due to oversubscription – MTN lost subscribers to Warid, whicih has now lost subscribers to Orange in Uganda.
Sustainability is a fraught question – but simply finding a good partner is not enough – many aid projects are dependent on donors and grants with 3-5 year grant cycles. Until we escape that funding model, we will never be “sustainable”.
And lastly – “mobile” health is a myth implying that our applications and the answer to health solutions resides inside aphone or inside little sensors. None of these solutions truly exist without a good server to back it up.n While admittedly the call of the mobile is lots more sexy than the giant energy sucking server room, we still need to think about what it means to develop and maintain the backends that work with the mobiles. They don’t exist in a vacuum.
Networks are unreliable and oversubscribed
Presented by Melissa R. Densmore,

14. Multiple Modes of Communication  Reliability and Consumer Confidence
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Multiple Modes of Communication  Reliability and Consumer Confidence
Courier
PO Box
Bulk SMS
In-Person
Phone (Voice)
Radio
The key to overcoming these constraints is not one singular technology, but the use of many information and communications technologies together. The stakeholders are linked together by their collaboration in the RHVP, and innovative uses of ICTs can help them overcome information constraints, improving how these stakeholders communicate over the course of the project.
The most critical component of this financial system – as it turns out, was not transfer of the money, or figuring out how to push the paperwork around. But how to effectively manage the communication channels with the stakeholders, and to keep them engaged, even through bottlenecks and delays.
Technology is not a silver bullet, and will not necessarily magically fix delays any better than effective program management will (and is probably more expensive in the long run).
Presented by Melissa R. Densmore,

15. Melissa R. Densmore mho@ischool.berkeley.edu http://www.ictdchick.com
Case Studies at the Intersection of Mobile Money and Mobile Health mHealth 2010, Tuesday, November 9, 2010, Washington, DC
Melissa R. Densmore
Acknowledge the funders … This work is sponsored primarily by the Blum Center for Developing Economies in Berkeley, California, USA, with additional funding from CITRIS as part of the Bears Breaking Boundaries competition.
Technology and Infrastructure for
Emerging Regions (TIER)
Presented by Melissa R. Densmore,