m-TERG Modeling mHealth Impact on Neonatal Survival using the Lives Saved Tool (LiST) Youngji Jo (Alain B. Labrique) December 11, 2014 PhD Candidate Health Systems Program International Health Department Johns Hopkins Bloomberg School of Public Health | JHU Global mHealth Initiative Outline • Why Lives Saved Tool (LiST)? • What is LiST? • Modeling LiST for mHealth Impact on Neonatal Survival in Bangladesh and Uganda • Caveats • Contributions • Acknowledgements Motivation Challenges of mHealth “pilotitis” VS. Interest and needs for mHealth scaling up Increasing mHealth impact evaluation/evidence • • • • • • Free et al (2013)—Systematic review / Meta Analysis Zuvorac et al (2009)—RCT study on ART adherence Lund et al (2013;2014)—RCT study on ANC visits Higgs et al (2014) – Recommendations for Research and Programs Anglada-Martinez (2014) – SMS for Adherence WHO mTERG Review and Recommendations Process Why LiST? mHealth as catalyst to existing interventions • Preventive and curative public health interventions of known efficacy exist and are well described. “mHealth helps overcome barriers to reaching effective coverage of interventions of known efficacy” • Coverage as a potential primary measure of mHealth impact • By promoting demand-side—eg. SMS text for IEC • By promoting supply-side—eg. Supply chain management, workflow management • By promoting access between the demand and supply Lives Saved Tool (LiST, www.livessavedtool.org) an evidence-based modeling tool What is LiST? • Initiated by the “Bellagio” modeling exercise and the Lancet Child Survival Series (2003) • Developed by the Child Health Epidemiology Reference Group (CHERG) for the WHO and UNICEF; • Managed by the JHU Bloomberg School of Public Health’s Institute for International Programs Objective: Estimate lives saved when introducing or scaling up key interventions Intervention types • • • • • • Maternal, fetal, neonatal, child Periconceptional, antenatal, birth, immediate postnatal, child Preventive, curative Nutritional, vaccination, water/ sanitation, treatment etc. Risk factors: Cause-of-death specific External (family planning, AIDS), internal (all others) Data needs (default data) • • • • Population data and trends (UN Population Division 1950-2050) (DemProj) Mortality rates/ratios (most recent) Cause of death structure (WHO/UNICEF/CHERG (2010) Intervention coverage (0-100%) (DHS/MICS/JMP/WHO-UNICEF (close to 2010) How it works? country specific health status coverage of intervention 1) Select countries 2) Select key interventions • Individual intervention • Combined interventions Efficacy of intervention Number of deaths averted 3) Determine target year • Based on policy goal • Based on achievable target 4) Determine target coverage • Absolute target • Relative target (by cause; by intervention) Impact can be categorized by: (i) Year of implementation; (ii) Cause of death; (iii) Population sub-group (e.g. mothers, newborns, children under 5 years); and (iv) Intervention Modeling analysis Select countries • Bangladesh and Uganda (NMR as 27 and 26 per 1,000 live births as of 2010) Select key interventions • Four key interventions—Antenatal care (ANC), Skilled birth attendance and/or Faci lity delivery (SBA/FD), Breastfeeding promotion (BF), and Postnatal care (PNC) • Bundled packages of interventions based on common mHealth strategies (“BF & P NC” and “ANC, SBA/FD, BF & PNC”, called All-combined). • The optimal mix of services and tradeoffs in coverage by comparing four individual and two bundled interventions scenarios, as described above. Determine target year--2015 Determine target coverage • Multiplying the baseline coverage of each intervention in 2011 by 110%, 130% and 150% in a relative manner, assuming linear trends of coverage increase over time. LiST Interventions and Coverage Increase Scenarios in Bangladesh and Uganda (input parameters of baseline year in 2011 and target year in 2015) LiST Interventions (selected) Pregnancy Childbirth Antenatal care (ANC 4+) Skilled birth attendance* Facility delivery* (Clinic and Hospital) Unassisted deliveries** Home deliveries** Assisted (% of all deliveries at deliveries) home** Facility deliveries** (% of all deliveries) Essential care ** BEmOC** CEmOC** Promotion of breastfeeding Breastfeeding promotion an Exclusive breastfeeding** d prevalence Predominant breastfeeding* * (<1 month) Partial breastfeeding** Preventive postnatal care (Thermal care, Clean Preventive postnatal practice) Baseline (2011) 25.5 31.7 28.8 Bangladesh Projected coverage increase (2015) 10% 30% 50% 28.1 33.2 38.3 34.8 41.2 47.6 31.7 37.4 43.2 Baseline (2011) 47.6 58.0 57.4 Uganda Projected coverage increase (2015) 10% 30% 50% 52.4 61.9 71.4 63.8 75.4 87.0 63.1 74.6 86.1 68.3 65.2 58.8 52.4 42.0 36.2 24.6 13.0 2.9 3.1 3.8 4.4 0.6 0.7 0.8 0.9 25.9 15.8 18.7 21.6 14.3 15.8 18.6 21.5 0.0 2.9 36.3 9.5 6.3 39.9 11.2 7.5 47.2 13 8.6 54.5 8.6 34.4 34.8 9.5 37.9 38.3 11.2 44.8 45.2 12.9 51.7 52.2 84.5 5.9 84.9 5.7 85.6 5.5 86.3 5.2 89.9 5.0 90.1 4.9 90.6 4.7 91.0 4.4 9.6 29.6 9.4 32.6 9.0 38.5 8.5 44.4 5.1 2.8 5.0 3.1 4.8 3.6 4.5 4.2 Neonatal Mortality Impacts by Various MNH Interventions and Coverage Scenarios in Bangla desh and Uganda in 2015 Interventions Illustrative mHealth Strategies ANC SBA/FD Data collection and management (e.g. Risk assessment and classific ation, Vital events tracking, adherence reminder); SMS texting for health promotion and scheduled visits reminder Emergency medical referral (e.g. r eferral calling) BF SMS texting for health promotion PNC SMS texting for health promotion BF & PNC SMS texting for health promotion All-combined Data collection and management ( : ANC,SBA/F e.g. Risk assessment and classifica D, BF &PNC tion, Vital events tracking, adhere nce reminder); SMS texting for he alth promotion and scheduled visi ts reminder; Emergency medical r eferral (e.g. referral calling) Coverage increase by 2015 Projected number of neonatal lives saved Bangladesh Uganda 2013 2014 2015 2012 2013 2014 2012 10% 30% 50% 0 0 0 0 1 1 0 1 2 0 (1)*** 1 (1) 2 (1) 1 3 5 2 7 11 3 10 17 5 (0.8) 14 (0.79) 23 (0.78) 10% 30% 50% 10% 30% 50% 10% 30% 50% 10% 30% 50% 10% 30% 50% 1038 1530 2021 4 12 20 98 290 482 102 302 502 1141 1820 2512 2055 3021 3984 8 24 41 194 576 958 202 600 999 2258 3587 4934 3048 4470 5882 12 36 61 289 858 1427 301 894 1487 3346 5298 7262 4016 (0.74) 5877 (0.74) 7717 (0.74) 16 (0.75) 48 (0.75) 80 (0.75) 383 (0.74) 1135 (0.74) 1888 (0.74) 399 (0.74) 1183 (0.74) 1968 (0.74) 4405 (0.74) 6951 (0.74) 9496 (0.74) 381 1141 1892 2 5 8 0 0 0 5 16 27 388 1160 1924 776 2312 3811 3 8 16 6 15 26 11 32 56 790 2349 3874 1187 3512 5753 5 14 26 12 30 54 18 50 86 1208 3569 5847 1611 (0.76) 4738 (0.76) 7714 (0.75) 7 (0.71) 18 (0.72) 35 (0.77) 17 (1) 47 (1) 83 (1) 24 (0.79) 68 (0.76) 118 (0.77) 1639 (0.76) 4814 (0.76) 7839 (0.75) 2015 Skilled birth attendance and increased facility delivery provide the biggest mortality impact Bangladesh Uganda Neonatal Mortality Impacts by Various MNH Interventions and Coverage Scenarios in Bangladesh and Uganda in 2015—individual intervention Breast Feeding Promotion & Postnatal Care provide relatively greater mortality impact in Bangladesh compared to Uganda Bangladesh Uganda Neonatal Mortality Impacts by Various MNH Interventions and Coverage Scenarios in Bangladesh an d Uganda in 2015—bundled packages Causes of Neonatal Deaths at Baseline in 2011 and Estimated Causes of Neonatal Deaths Averted (with 50% coverage scenario) in 2015 in Bangladesh and Uganda Causes of Neonatal Deaths Avered (2015) Coverage increase ANC (50%) SBA/FD (50%) by 2015 Neonatal-Diarrhea Neonatal-Sepsis Neonatal-Pneumonia Neonatal-Asphyxia Neonatal-Prematurity Neonatal-Tetanus Neonatal-Cogenital a nomalies Neonatal-Other BF (50%) PNC (50%) All-combined: AN BF & PNC (50%) C,SBA/FD, BF &P NC (50%) B U B U 17 13 17 13 787 40 1372 723 63 22 63 22 0 0 2382 3693 1082 41 5,627 3370 19 1 35 19 B 0 2 0 0 0 0 U 0 23 0 0 0 0 B 0 625 0 2382 4692 17 U 0 665 0 3693 3338 18 B 17 0 63 0 0 0 U 13 0 22 0 0 0 B 0 787 0 0 1082 19 U 0 40 0 0 41 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Caveats • LiST modeling assumptions—ex. Single cause of death, delivery care (BEmOC; CEmOC) • The level of mortality impact is influenced by the reported initial baseline coverage level and evidence from standard care practice. • Like many other modeling tools, the analysis (and LiST) does not systematically consider health systems constraints in achieving the target c overage. • This analysis does not consider effective coverage—quality of sub-components of a given intervention package. • It is important to ensure that the findings serve as added guidance for informed discussion / prioritization and not drive the selection of strategies in a vacuum. Contributions • Impact estimation: Estimates the neonatal mortality impact by improving coverage of specific MNH interventions around different target coverage levels, derived from rigorous, efficacy-trial driven expectations. • Prioritization of strategy: Determine mHealth programs and strategies in obtaining the likely highest-impact interventions for prioritization, considering the unique potential for synergy across m ultiple areas that mHealth solutions typically allow • Planning & Evaluation: Enable projects to set benchmarks and monitor progress against modeled targets, potential endpoints, and guidance for cost-assessment Validating the models – next steps… Empirical test with prospective impact evaluation research • To validate these models and provide more informed projections on potential coverage change associated with various mHealth solutions across different contexts over time. Effective coverage • Develop metrics for how mHealth can improve *quality* (completeness) of a given intervention? Scaling up/implementation science • How can mHealth *rapidly* expand a particular intervention? Barriers to coverage—Heath systems constraints • Geographical access • Medical supplies availability • Provider’s compliance • Patient’s adherence Mehl G, Labrique AB, Science 2014 Acknowledgements • • • • • • Alain B. Labrique, JHSPH Amnesty E. Lefevre, JHSPH Garrett Mehl, WHO Teresa Pfaff, JHSN Neff Walker, JHSPH Ingrid K. Friberg, JHSPH • JHU Global mHealth Initiative • WHO mHealth Technical Evidence Review Group References • USAID. LiST Manual. Available:http://www.jhsph.edu/departments/international-h ealth/centers-and-institutes/institute-for-international-programs/_documents/ma nuals/list_manual.pdf. • Labrique AB, Vasudevan L, Kochi E, Fabricant R, Mehl G (2013) mHealth innovation s as health system strengthening tools: 12 common applications and a visual frame work. Global Health: Science and Practice 1: 160-171 • Fischer Walker CL, Friberg IK, Binkin N, Young M, Walker N, et al. (2011) Scaling up diarrhea prevention and treatment interventions: a Lives Saved Tool analysis. PLoS Med 8: e1000428 • Walker N, Tam Y, Friberg IK (2013) Overview of the Lives Saved Tool (LiST). BMC Pu blic Health 13: S1 • Fox MJ, Martorell R, van den Broek N, Walker N (2011) Assumptions and methods i n the Lives Saved Tool (LiST). Introduction. BMC Public Health 11 Suppl 3: I1 • Bryce J, Friberg IK, Kraushaar D, Nsona H, Afenyadu GY, et al. (2010) LiST as a cataly st in program planning: experiences from Burkina Faso, Ghana and Malawi. Int J Ep idemiol 39 Suppl 1: i40-47. Project Country Organization Interventions Wired Mo Tanzania thers Maternal Ethiopia and Newb orn Health in Ethiopia Partnershi p (MaNHE P) E-IMCI Tanzania mHealth strategies mHealth benefit/impact evidence on service provision Danish Internati (i) Family planning (i) Data collection and m “The mobile phone intervention was associated with an increase in antena onal Developme (ii) Behavior chang anagement (e.g. Risk ass tal care attendance. In the intervention group 44% of the women received nt Cooperation, es through Inform essment and classificatio four or more antenatal care visits versus 31% in the control group (odds ra ation, Education a University of Co n, Vital events tracking, tio (OR), 2.39; 95% confidence interval (CI), 1.03-5.55). There was a trend nd Communicatio penhagen n (IEC) (iii) Antenat adherence reminder) (ii) towards improved timing and quality of antenatal care services across all s al care(ANC)/Expa SMS texting for health pr econdary outcome measures although not statistically significant.” [22] nded Program on I omotion and scheduled “The mobile phone intervention was associated with an increase in skilled mmunization (EPI) visits reminder (with mo delivery attendance: 60% of the women in the intervention group versus 4 /Postnatal care(PN bile phone voucher com 7% in the control group delivered with skilled attendance. The interventio C) (iv) Skilled birth ponents) n produced a significant increase in skilled delivery attendance amongst ur attendance(SBA)/F ban women (OR, 5.73; 95% CI, 1.51–21.81), but did not reach rural women acility delivery (FD .” [34] ) “The perinatal mortality rate was lower in the intervention clusters, 19 per 1000 births, than in the control clusters, 36 per 1000 births. The interventi on was associated with a significant reduction in perinatal mortality with a n OR of 0.50 (95% CI 0.27-0.93). Other secondary outcomes showed an ins ignificant reduction in stillbirth (OR 0.65, 95% CI 0.34-1.24) and an insignifi cant reduction in death within the first 42 days of life (OR 0.79, 95% CI 0.3 6-1.74).” [40] (i) Family planning University Resea (i) SMS texting for healt “Women who had additionally attended 2 or more CMNH meetings with f rch Co., LLC, Qu (ii) Behavior chang h promotion and schedul amily members and had access to a health extension worker’s mobile pho ality Improveme es through IEC (iii) ed visits reminder (e.g. p ne number were 4.9 times more likely to have received postnatal care (OR ANC/EPI/PNC nt Advisor for th romotion of community , 4.86; 95% CI, 2.67-8.86; P _.001).” [41] “Notification of health extension e Maternal and maternal and newborn h workers for labor and birth within 48 hours was closely linked with receipt Newborn Health ealth family meetings an of postnatal care. Women with any antenatal care were 1.7 times more lik in Ethiopia Partn d labor and birth notifica ely to have had a postnatal care visit (OR, 1.67; 95%; 95% CI 1.10-2.54; P _ ership tion) .001).” [41] D-Tree (i) ANC/EPI/PNC (ii (i) Point of care decision “For all ten critical IMCI items included in both systems, adherence to the ) Behavior changes support through complia protocol was greater for eIMCI than for pIMCI. The proportion assessed un through IEC nce to IMCI protocols der pIMCI ranged from 61% to 98% compared to 92% to 100% under eIMC I (p < 0.05 for each of the ten assessment items).” [25] Project M Zambia, UNICEF wana Malawi (i) HIV-antiret (i) Data collectio “ SMS delivery of results can increase turnaround times by 50% on averag roviral therap n and managem e, with a greater positive impact in rural facilities” [42] y (ART) surveil ent (ii) SMS text lance and trea ing for health pr tment omotion and sch eduled visits rem inder Better Borde Thailand- Mahidol Universit (i) Family plannin (i) Data collection an “ANC/EPI coverage in the study area along the country border improved; numbers of AN r Healthcare Burma y, Thailand g d management (ii) C and EPI visits on-time as per schedule significantly increased; there was less delay of a Program (ii) ANC/EPI/PNC SMS texting for heal ntenatal visits and immunizations” [43] th promotion and sc heduled visits remin der RapidSMS-M Uganda Ministry of Healt (i) Family plannin (i) Data collection an Study reported “a 27% increase in facility based delivery from 72% twelve months befor CH h Uganda, UNICE g (ii) Behavior ch d management (ii) e to 92% at the end of the twelve months pilot phase.” [44] anges through IE SMS texting for heal F C (iii) ANC/EPI/P th promotion and sc NC (iv) SBA/FD heduled visits remin der Rural Extend Uganda Ministry of Healt (i) Behavior chan (i) Emergency medic “improved communication and transportation links between the Traditional Birth Atten ed Services a h, UN Population ges through IEC ( al referral (e.g. refer dants (TBAs) and the health posts resulted in increased and more timely referrals as well nd Care for U Fund and the Uga ii) ANC/EPI/PNC ( ral calling) with tran as the improved delivery of healthcare to a large number of pregnant women”… “The in ltimate Emer nda Population S iii) -SBA/FD sportation services creased number of deliveries under trained personnel and increased referrals to health gency Relief ( ecretariat units led to a reduction of about 50 percent in the maternal mortality rate (MMR) in thr RESCUER) ee years” [45] M4RH Kenya Ta USAID, FHI 360’s (i) Family plannin (i) Data collection an User interviews reported various positive responses including “the text messaging servic nzania, PROGRESS (Progr g (ii) Behavior ch d management (ii) e was perceived as being private, convenient, and cost-effective.” [33] am Research for anges through IE SMS texting for heal Strengthening Se C th promotion and sc rvices) heduled visits remin der PREVEN Aceh Besar Midwives MAMA MOTECH Peru Cell-Preven (i) Sexual and re productive healt h surveillance a nd service delive ry (i) Data collection an Lessons include “Two-way information systems are more than just collecting data. They d management (ii) provide feedback and support to health care workers in the field. Many times, only man SMS texting for heal agers have information that allows them to monitor and evaluate data but these system th promotion and sc s do not prove any aggregate value to health care workers in the field. A well-designed i heduled visits remin nformation system has to support and enhance the performance of all user levels in a se der cure environment.” [30] “Prahalad (2005) has reported that health workers in some dev eloping countries spend as much as 40% of their time filling out forms, compiling and co pying data from different pro-grams (e.g., tuberculosis, malaria, HIV/AIDS, etc.). By choo sing the most appropriate information technology, we can avoid duplication and deploy different devices—i.e., cell phones, Internet—to report from each public health program .” [30] Indonesia UNICEF, UNFPA, (i) Behavior chan (i) Data collection an “Findings from the project indicate that the mobile phone has proven to be an effective and World Vision ges through IEC ( d management (ii) and efficient device for facilitating smoother communication, and allowing speedier eme ii) ANC/EPI/PNC ( SMS texting for heal rgency response. The system also aids in gathering and disseminating health-related info th promotion and sc iii) SBA/FD rmation to midwives, who in turn convey this knowledge to the patient community.” [13 heduled visits remin der (iii) Emergency ] medical referral (e.g . referral calling) Banglade mHealth Alliance (i) Family plannin (i) Data collection an MAMA Bangladesh Aponjon project represented “a 37% increase over a 2011 national sh, India, g (ii) Behavior ch d management (ii) baseline of 26% attending four ANC visits. It is also important to note that 45% of the Ap and Sout anges through IE SMS texting for heal onjon subscribers went to a facility for delivery and 32% chose safe delivery at home” [3 h Africa C th promotion and sc 2] heduled visits remin der Ghana Grameen Founda (i) Family plannin (i) Data collection an Comprehensive observational studies demonstrated lessons learned and key future impl tion g (ii) Behavior ch d management (ii) ications. [28] Evaluation is on-going with Grameen Foundation, Healthcare Innovation T anges through IE SMS texting for heal echnology LAB (HITLAB), and Ghana’s School of Public Health.[29] C (iii) ANC/EPI/P th promotion and sc NC (iv) SBA/FD heduled visits remin der (iii) Emergency medical referral (e.g . referral calling) mHealth Health Service Coverage Increase Impact Model Mobile phone Health Information Systems Users Processes Outcomes Impact Information registration through mobile phone : ID, demographic information, geographic location etc. Systems: Calculate due dates for certain care events: child birth, ANC/PNC/EPI visit scheduling Identify clients with upcoming delivery dates, those who recently delivered and those who estimated due dates have passed without delivery Send alerts to CHWs when event is overdue Send BCC messages/reminders to mothers Mothers: receive SMS messages Promoting Information, Education and Communication (IEC) Behavior change: promoting care seeking behaviors CHWs: client management scheduling visits send alerts/reminder etc. Decision making supportin g tool Facilitatin g referrals Community empowerment: efficient/effective service delivery Physicians: case identification Improving health systems readiness and quality of care Improved responsiveness: providing timely and quality care Increasing uptake (coverage) of health services in ANC, SBA/FD, BF, or PNC Effectiveness of Interventions , Affected Fractions, and Assumptions to Neonatal Mortality in Bangladesh and Uganda Intervention Effectiveness (<1month) Affected fraction Neonatal-Diarrhea Curative after birth ORS-oral rehydration solution 0.93 1 Neonatal-Sepsis Pregnancy 0.97 0.006 0.4 0.23 0.28 0.65 0.8 1 1 1 1 1 0.42 1 Injectable antibiotics Full supportive care Curative after birth Full supportive care 0.75 0.9 0.05 1 1 1 Preventive Thermal care 0.2 1 KMC-Kangaroo mother care 0.51 1 Full supportive care 0.28 1 Pregnancy TT-Tetanus toxiod vaccination 0.94 1 Preventive Clean postnatal practice 0.4 1 Folic acid supplementation/fortification 0.35 1 0.1 1 Syphillis detection and treatment Clean postnatal practice Chlorhexidine Oral antibiotics Curative after birth Injectable antibiotics Full supportive care Preventive Oral antibiotics Neonatal-Pneumonia Curative after birth Neonatal-Asphyxia Neonatal-Prematurity Curative after birth Neonatal-Tetanus Neonatal-Cogenital anomalies Periconceptual Neonatal-Other Curative after birth Full supportive care Developing an evidence base for mHealth solutions at scale: Monitoring and Evaluation Framework Amnesty LeFevre PhD MHS Smisha Agarwal DDS MPH MBA Alain Labrique PhD MHS MS Garrett Mehl PhD MHS Presentation Overview • Global context for evaluating mHealth soluions • mHealth in South Africa: 1 million pregnant women registration initiative • Monitoring and Evaluation Framework • Implications and next steps © 2014, Johns Hopkins University. All rights reserved. Global Context • 2011 the Bellagio Call to Action on Global eHealth Evaluation highlights the need for rigorous evaluation in mHealth • US NIH, Global Donors and Academic calls for improved rigor in mHealth evaluation and reporting • Significant investment in evidence generation (eg. Innovations Working Group) • 2013 WHO mHealth Technical Evidence Review Group: Working Papers on mHealth Taxonomy, Evaluation, Indicators and Evidence Grading © 2014, Johns Hopkins University. All rights reserved. Considerations for evaluating mHealth solutions Pre-prototype Prototype Feasibility/ Usability Pilot Efficacy Scaled Demonstration Integration Effectiveness Implementation Science • Where the technology is in the stage of development? • What corresponding stage of evaluation is appropriate for that strategy? • What are the evidence claims the project wants to make? • What is the time point for evaluation initiation? • Available resources © 2014, Johns Hopkins University. All rights reserved. The evaluation process for mHealth solution • Outcome, impact assessments Is my solution effective? Is my solution good value for money? • Cost effectiveness / Utility analysis • Cost benefit analysis • Economic and financial evaluations of a single program Is my solution affordable? © 2014, Johns Hopkins University. All rights reserved. Is my solution scalable? • Policy analysis • Economic and financial costing of a single program • Sector wide planning for integration mHealth in South Africa • National program to enable approximately one million pregnant women to register in health facilities using interoperable mobile health services. • Three major features of the program 1. 2. 3. Improve early identification of pregnant women, increase e arly access to antenatal care, and repeated antenatal visits thr oughout the pregnancy. Register all pregnant women in the public health system usi ng the mobile health system early in their pregnancy; Subscribe pregnant women in the public health system to re ceive pregnancy-related information messages. © 2014, Johns Hopkins University. All rights reserved. How will this work in practice? Identification/ registration / subscription End-u ser Facility provider CHW Source: Debbie Rodgers/ Praekelt © 2014, Johns Hopkins University. All rights reserved. Monitoring and Evaluation Framework • Objectives: Develop a monitoring and evaluation framework that will • facilitate the measurement of a common set of indicators across all implementing partners, and • yet accommodate differentiation across services providers, impl ementing strategies, etc. • Challenges • Accommodating variation in integration o Programs at various stages of development o Health systems variability (CHW presence or not?) • Program growth o o Potential for linkages with other NDOH programs mHealth field is dynamic, always evolving © 2014, Johns Hopkins University. All rights reserved. Logframe Overview Inputs/ Proce sses Outputs Outcomes Partnerships 1. Increased service utilization Increased MNC H services utiliza tion Health Promotio n Messages 2. Strengthening of human resources an d facility readiness t o provide care Health facility a nd community in puts Technology 3. Improved service delivery (facility an d community) 4. Improved technol ogy use 5. Funding Funding 6. Supply side • Pregnancy • Delivery • Postnatal/ postp artum • Child health • HIV • Efficiency in ca reseeking • Improved conti nuity Improved knowl edge Supply side Funding Impact Reduction in nu mber of • Stillbirths • Neonatal mort ality • Maternal mort ality Monitoring and Evaluation Framework • Available on GSMA website http://www.gsma.com/mobilefordevelopment/mhealth-formnch-impact-model © 2014, Johns Hopkins University. All rights reserved. How will this work in practice? • Framework includes ~ 100 indicators across an expansive range of domains; anticipated that a sub-sample of these applied • Can be used by organizations that are involved in one or more components (identification/ subscription/ registration) o Example 1: Messaging component alone, no capacity to link w ith individual patient records o Example 2: Programs with CHW identification/ referral, facilit y based confirmation/ registration, and subscription © 2014, Johns Hopkins University. All rights reserved. How will this work in practice? Example 1: Messaging component alone INPUTS OUTPUTS OUTCOMES Health promotion messag es developed Posters/ Fl iers in place; Accessible to users TechnologyNetwork co verage and power, provider / user mobile equipment Health workers reporte d use of mobile tools Platform functionality functional performanc e of the service, user sati sfaction Improved reported utilization of MCH servi ces • improved ANC attendance; improvements i n HIV identification / treatment Woman wit h a suspecte d pregnancy sees a poste r Improved knowledge Subscribers Women receives staged base messaging through out her pregnancy Dials in Rec eives Min message set Goes to health Registered to facility; pregna receive full s ncy confirmed et of messag es © 2014, Johns Hopkins University. All rights reserved. How will this work in practice? Example 2: Programs with multiple components INPUTS OUTPUTS OUTCOMES Health promotion messag es developed Posters/ Fl iers in place; Accessible to users TechnologyNetwork co verage and power, provider / user mobile equipment Health workers reporte d use of mobile tools Platform functionality functional performanc e of the service, user sati sfaction Improved utilization of MCH services • improved ANC attendance; improved early ANC 1 attendance; improvements in HIV id entification / treatment Improved knowledge Subscribers; provider s (Phase II) Supply side Utilization of HelpDesk; respon siveness of HelpDesk Women receives staged base messaging through out her pregnancy CHW identifies woman with sus pected pregnan cy Dials in Rec eives MID SIZE messa ge set Goes to health facility; pregnancy confirmed Registered to receive full s et of messag es © 2014, Johns Hopkins University. All rights reserved. What will this evidence tell us? Minimum evidence claims • Effect on critical outcomes MNCH service utilization, changes in knowledge (users and provider), and potential for modeled impact (LiST)) • Information to attract new investors, inform existing donors • Draw additional funding; identify avenues of commercial sustainability © 2014, Johns Hopkins University. All rights reserved. What will this evidence tell us? Optimal use of logframe • Prospective program monitoring to allow continuous improvement o Complementary to existing evaluations – avoids duplication wit h existing evaluation efforts-- allow for tracking at a scale o Allow improved planning at a national level: linkages with MN Os, forecasting of resources and health status (MDG targets) • Support to complementary streams of research (process documentation, efficiency, economic analyses) cutting across all programs to o allow for differentiation / improved understanding of implement ation (what works in some areas and not others?) o standardized methodology © 2014, Johns Hopkins University. All rights reserved. Implementation and next steps Is implementation feasible? • Wide array of indicators intended to accommodate a wide array of programs (established, forthcoming) • Need for additional modification to accommodate linkages with programs that have existing streams of data collection / evaluation partners • What’s the value to providing data on their programs? o Unique opportunity to contribute to macro-level generation of eviden ce claims on mHealth o Opportunity for differentiation Comparison of different models of implementation / integration Ability for continued learning given variability in where programs will li e along the continuum of efficacy implementation science © 2014, Johns Hopkins University. All rights reserved. Acknowledgements This document is an output from a project funded by the UK Department for International Development (DFID) for the benefit of developing countries, managed through HLSP Mott Macdonald. The views expressed are not necessarily those of DFID or HLSP. © 2014, Johns Hopkins University. All rights reserved. Further information • • • • • Full Logframe www.gsma.com Amnesty LeFevre PhD MHS, aelefevre@gmail.com Alain Labrique PhD MS, alabriqu@gmail.com Smisha Agarwal DDS MPH MBA, smishaa@gmail.com Garrett Mehl PhD, mehlg@who.int © 2014, Johns Hopkins University. All rights reserved. Supplementary slides © 2014, Johns Hopkins University. All rights reserved. Intervention Validation versus Delivery Strategy Evaluation • Two roles for mHealth strategies: 1. Intervention with known efficacy, 2. Interventions with an independent effect on Outcome C Intervention A Measles Vaccination Problem: Outcome C: Reduced measles transmissi on Measles outbreak mHealth strategy: Vaccine stockout n otification © 2014, Johns Hopkins University. All rights reserved. Data sources Data Source Description Purpose Facility HMIS Records Routinely collected facility-based records for clients served Data on facility registration, gestational age, demographics, receipt of ANC, SBA, PNC services CHW records Data on community-level client identification, Routinely collected CHW and referrals for registration (at the health records for clients facility) and subscription. Data on subscribed client’s satisfaction with Client phone Cross-sectional phone mHealth services, knowledge and practice of survey at survey addressed to a sub- key behaviors. Will include post-partum 3, 6, 12, 18 sample of subscribers surveys for data on continuity of care months parameters. Facility and community level provider surveys Healthcare Cross-sectional survey to assess satisfaction with mHealth platform, provider addressed to facility and frequency and regularity of use. May include a survey community providers component of direct observation, if feasible. © 2014, Johns Hopkins University. All rights reserved. How will this work in practice? Supply side Registered M om receives c are at a healt h facility* Dials public lin e 1. Baby/ preg nancy help 2. Complime nts and co mplaints 1. 2. Receives sy stem gener ated messa ge within 2 4 hours Compliant logged/ c ase opened Thank you response Standard content SMS m essage sent *Separate stream for end-user s that are not registered Message sent indicating question cannot be answe red advocates referral © 2014, Johns Hopkins University. All rights reserved. Responds to m essage with 1. Baby/ preg help question Web Interfa ce 2. Compliment or complaints Employee r esponds Complaints/ compliments Question can be answered with st andard content Question cannot be answered wit h standard conten t How will this work in practice? Example 2: CHW identification and referral, facility registration / subscription to full messaging INPUTS OUTPUTS OUTCOMES Health promotion messag es developed Posters/ Fl iers in place; Accessible to users TechnologyNetwork co verage and power, provider / user mobile equipment Health workers reported use of mobile tools Platform functionality functi onal performance of the service, user satisfaction Client Profile Messages sent; Proportion of and profile of clien ts that opt-out; Improved utilization of MCH servi ces • improved ANC attendance; • improved early ANC 1 attendance; improvements in HIV identificatio n / treatment Improved knowledge Subscribers Women receives staged base messaging through out her pregnancy CHW identifies woman with sus pected pregnan cy Dials in Rec eives MID SIZE messa ge set Goes to health facility; pregnancy confirmed Registered to receive full s et of messag es © 2014, Johns Hopkins University. All rights reserved. Vocabulary • mHealth TECHNOLOGY • The hardware / software underlying an mHealth solu tion • mHealth PROJECT • The use of a particular technology(ies) to achieve a particular goal in a specific location(s) • mHealth STRATEGY • The generic approach that is being undertaken using mHealth, agnostic of project or technology. © 2014, Johns Hopkins University. All rights reserved. mHealth Summit 11th of December 2014 Hajo van Beijma hvanbeijma@ttcmobile.com @hajovanbeijma Healthy Pregnancy Healthy Ba by Partners • • • • • • Tanzanian Ministry of Health and Social Welfare mHealth Tanzania Partnership – led by the CDC Foundation Financial support from the US Government Airtel Tanzania Wazazi Nipendeni multi-media campaign Johns Hopkins Bloomberg School of Public Health Background • Healthy Pregnancy Healthy Baby as part of Tanzania’s efforts in the cam paign on Accelerated Reduction of M aternal Mortality in Africa. • Tanzania has one of the highest mat ernal mortality rates in the world. Healthy Pregnancy Healthy Ba by SMS Information Categories • Prevention of mother to child transmission of HIV/aids • Ante-natal clinic visit reminders • Malaria prevention • Individual birth preparedness plan • Nutrition • Danger signs • Family planning/Birth Spacing • Fun/Interesting (e.g. fetal development) • Post-partum care Healthy Pregnancy Healthy Ba by How does it work TO: 15001 mtoto One short c ode for all n etworks Healthy Pregnancy Healthy Ba by • • • • • • • HPHB the largest interactive mobile h ealth program in Africa – to date We offered 31 million maternal health and early childcare text messages over 600,000 registrants On average we count an extra 21,500 r egistrants every month Majority of users self-registered, with only 1.41% of users utilizing the health -facility assisted registration option High level of engagement Users on average opt-in weekly Goal: 1.000.000 registrants by 2016 Users 51.50% male 37.81% female 10.70% did not want to report gender Timing • Day of the Week Wednesday > Tuesday > Thursday > Friday > Monday > Satur day > Sunday • Time of Day Morning 8:00 – 11:00 Lunch time 11:05 – 14:00 After work: 17:05 – 20:00 Traffic Thank you! Any questions? hvanbeijma@tt cmobile.com www.ttcmobile. com Using Globa l South Data to Improve mHealth for All Global mHealth Forum D ecember 11, 2014 Meagan Demitz Hesperian Resources Mission: Provide informatio n & tools that help all peopl e take greater control over t heir health & work to elimin ate underlying causes of po or health. Resources: Primary health care; women’s health; repro ductive health and rights; mi dwifery; health worker traini ng; community dentistry; su pport for women and childre n with disabilities; occupatio nal health and safety; enviro nmental health; early childh ood development Collective development to creat e empowering materials 1. Start from people’s own experiences 2. Offer practical solutions 3. Design materials based on listening carefully, rep orted needs, data collection and evaluation 4. Encourage participation and action in ways that d o not depend on literacy, and that encourage criti cal thinking 5. Collect and incorporate end-user feedback when ever possible Adapting the Hesperian mod el for digital resources • Feedback from consumers via field testing, expert review, an d beta-testing (pre and post) • On-going collection of global u ser data (Google analytics) • Content dev and design, user fl ow and user experience • Impact evaluation - qualitative and quantitative feedback • On-going adaptation of conten t/user experience The HealthWiki • 13 languages • 16,000 visitors per day; 2 million over the last year; majority from Global Sout h • Top 10 countries: US, Mex ico, Brazil, Spain, Colombi a, India, Peru, Argentina, UK, Venezuela • 57% of traffic via mobile d evices; 243,000 per mont h • Most visited: Reproductiv e health; maternal & child health; abortion; belly pai n/worms/diarrhea Launched in October 2011! Hesperian’s model for mHealth Need: Access to information in th e field without Internet Focus: Women’s health based on HealthWiki traffic; partner feedba ck Approach: Applying Hesperian’s collective development process Lessons learned: Adapting cont ent for mobile is time intensive Editorial, design adaptation Internal and external review Field-testing Beta testing before and after laun ch Result: Safe Pregnancy and Birth app Mobile App: Safe Pregnancy and Birth Over 215,365 apps downloaded fro m 149 countries Translations underway in Hindi, Tam il, Kannada, Malayalam, Bengali, Gr eek and Nepali. Field Test: Compañeros en Salud Facilitated by CES - PIH health partn ers in Chiapas, Mexico: • Field-testing was conducted with the help of CHWs, midwives, and patients • App was used to educate patients about the i mportance of health, and warning signs duri ng pregnancy Feedback: • Health workers valued skill-building sections • Images allowed illiterate patients to better u nderstand information • CES plans to continue to use the app in pre-n atal monitoring • Request for future apps: Communicable and non-communicable diseases; birth control; i nfant malnutrition; other common health iss ues Applying lessons for future mHealth resources • Collaborative development and partners hips • Strengths-based, community-centric app roach • Time-intensive content development, inc luding design and user interface • Ongoing evaluation through field-testing and beta-testing Awards: mHIFA working group; Ashoka “She will innovate” HealthWiki: http://en.hesperian.org/hh g/Healthwiki Meagan Demitz Foundation Relations an d Grants Manager Hesperian Health Guides 1919 Addison Street Berkeley, CA 94704 www.hesperian.org (510) 845-1447 meagan@hesperian.org Safe Pregnancy & Birth: http://hesperia n.org/books-and-resources/safe-pregn ancy-and-birth-mobile-app/ Data Use and Indicators fo r the mHealth pragmatist Smisha Agarwal, Amnesty Lefevre, Lavanya Vasudevan, Alain Lab rique Mobile data collection • Faster, cheaper, easier………. But is it? Starting with the end in mind What is the stage of development of the program What are the evidence claims we want to make? Identifying specific set of indicators to address claims Stage of Development What is the stage of development of the program What are the evidence claims we want to make? Identifying specific set of indicators to address claims Stages of Development Pre-prototype Prototype Pilot Scaled Demonstration Integration Prioritizing relevant evidence claims What is the stage of development of the program What are the evidence claims we want to make? Identifying specific set of indicators to address claims Data Stake-hol ders Source- MEASURE Evaluation Decisions Key Considerations for Selection of Indicators What is the stage of development of the program What are the evidence claims we want to make? Process Indicators Identifying specific set of indicators to address claims Intervention of kn own efficacy Absence of evidence-b ase for underlying inte rvention Barometer of mHealth In dicators Outcome Indicators Categorization of mhealth Indicators Does the technolog y work? • Technical Facto rs •Organizational Factors How do people int eract with technol ogy? •User Coverage • User Respons e • User Adoption How does technolog y improve implemen tation process? • Availability • Cost •Efficiency • Quality •Utilization How do improvem ents in service deli very affect health? • Improved hea lth outcomes Performance of Routine Information Syst em Management (PRISM) Emphasis on Health Information Systems (HIS) performance Consideration of Or ganizational, Technic al and Behavioral co mponents Linking Indicators to Taxonomic Stages of Developm ent Stages of Development Pre-prototype Prototype Pilot Scaled Demonstration Integration Pre-prototype: This stage includes hypothesis building, needs/context assessment, and usability/feasibility testing. This is the first stage of the development of a project. Earliest Stage of Developmen t to Capture Indicators for “D oes the Technology Work?” Does The Technology Work? Metric Area Indicators Technical Factors Connectivity % of target population with mobile phone signal at time of interview Power % of target population who have current access to a power source for recharging a mobile phone device Skilled local staf % of mHealth programs with current access* to local technical support for troubl f eshooting % of users who report having access to local technical support systems for troub leshooting Maintenance % devices that are not currently operational (misplaced/broken/not working) Functionality % of mobile devices that are operational in the language of the users % target population who are literate in the language used by the mHealth strateg y % of target population who report ever use of Short Message Service (SMS) capa bilities % of data fields from original paper based system that technology captures Organizational Factors Training Total hours of initial training attended by program staff in use and deployment of technology Total hours of refresher training attended by program staff in use and deployme Linking Indicators to Taxonomic Stages of Developm ent Stages of Development Pre-prototype Prototype Pilot Limited Demonstration Integration Prototype: During this phase, user-focused de signs are created and tested, and functionality, stability and usability are tested in an iterative process. Ways to improve the project are exam ined to enhance relevance. Earliest Stage of Development to Capture Indicators for “How do People Interact with Techn ology?” How do People Interact with Technology? Metric Area Indicator User Cov % of users who demonstrate proficiency in use of intended mobile application erage % intended users observed using the tool in preceding reference period of time No of transmissions sent by intended users over reference period of time % of transmissions successfully sent* in ‘x’ period of time User Res % of users who rate technology as "easy to use" ponse % of users rating technology “transmits information as intended” % of users who report satisfaction with the content of health information received th rough mobile device % of users motivated/intend to use technology User Ado % of messages/amount of data transmission sent from server that are responded to a ption ppropriately** by end user within reference period of time Number of messages/forms/amount of data transmission sent by end-user within re ference period of time % of data fields/forms that are left missing/incomplete over specified period of time *Successful transmission is reflective of the network coverage in the user area ** ‘Appropriately’ could refer to completion of intended action to reflect that the message has b een read e.g. Acknowledgement of message Feedback Loop Technology inputs affects performance Does the Technology Wor k? How do People Interact wi th the Technology? User feedback informs technology development proc ess Spiral Model of Software Development and Enhancement Linking Indicators to Taxonomic Stages of Developm ent Stages of Development Pre-prototype Prototype Pilot Scaled Demonstration Integration Earliest Stage of Development to C apture Indicators for “How does Te chnology Improve Implementation Process?” Pilot: This stage examines whether or not th e project has the ability to produce the desir ed effect under controlled circumstances and is usually a single deployment. This correspo nds with the evaluation stage of Efficacy. How Does Technology Improve Implementation P rocess? I. Health syste ms level Registration and vital events tracking Real time indicator reporting Human Resource management, accountability Electronic health records Supply Chain Management II. Provider le vel Decision Support Scheduling and Reminders Provider training, service updates III. Patient le vel Client education and Self-Efficacy Behavior Change Communication Adherence to Care Emergency services information mHealth functions and strategies Improvements i n • Availability • Costs •Efficiency • Quality •Utilization Health Systems-level: How Does Technology Improve Imp lementation Process? mHealth metri c by taxonomic Indicators constraint Availability % of target population who have access intervention ‘X’ over reference period % of health facilities in a target geographical area that use mHealth platform services Total number of clients seeking health service “x” at health facility with mHealth platform s ervices Efficiency Total cumulative time in minutes over reference period for all health workers in a facility u sing mHealth platform to enter data about intervention ‘x’* Total time taken in minutes for all health workers over reference period to transmit data ab out intervention ‘x’ from community logs to health facility information systems Total cumulative time taken in minutes over reference number of events from identification of an adverse event to care provision for intervention ‘x’ across levels of a health system Total days in reference period for which a health facility reports stock out of commodity ‘x’ Quality Total number of healthcare workers observed to be providing clinical services related to m Health strategy % change in reported events of “stock out” of commodity ‘x” over reference period ** % change in data entry errors over reference period ** % of target health workers who receive initial training on health intervention ‘x’ using mHe alth platform % of target health workers who receive refresher training on health intervention ‘x’ using m Health platform (initial and refresher training) Health Systems-level: How Does Technology Improve Imp lementation Process? mHealth metri c by taxonomic Indicators constraint Utilization Total number of clients seeking health service ‘x’ over a specified period of time % of clients in a specified area who are receiving health service ‘x’ through mHealth strat egy over reference period Costs Change in costs of transporting paper forms and manual data entry over reference** Change in costs of human resources for data entry** Change in costs associated with timely and appropriate management of illness** Changes in reported out of pocket expenditures over time ** Total population level savings in out-of-pocket payments as a function of timely and appr opriate care seeking** * Aggregated facility-level indicator ( Corresponding indicator at provider-level is disaggregated) ** Assumption of data collection at two-points before and after the implementation of the mHealth strateg y X: To be replaced by specific health intervention targeted by the mHealth platform Provider- level: How Does Technology Improve Implemen tation Process? mHealth metri c by taxonomic Indicator constraint Availability % of targeted health workers who utilize mHealth platform about intervention ‘X’ throug h their phones over reference period % of health workers observed to use mHealth platform during their last client contact % of health workers who use mHealth intervention to connecta with medical staff to receiv e real-time clinical information and decision support Total number clients attendedb by a health worker using mHealth platform over reference period Efficiency Total reported/observed time in minutes for last client counseling using mHealth platform about intervention ‘x’ Total reported/observed time spent on health record keeping about intervention ‘x’ over r eference period Total time taken in minutes per health worker over reference period to transmit data abou t intervention ‘x’ from community logs to health facility information systems Total individual health provider time taken in minutes over reference number of events fro m identification of an adverse event to care provision for intervention ‘x’ across levels of a health system Total time taken in minutes by health worker to report important adverse events (stock-o uts) Provider- level: How Does Technology Improve Implemen tation Process? mHealth metri c by taxonomic Indicator constraint Quality % of health workers who report adequatec knowledge of topic ‘x’ % of care standards* observed to be met using mHealth intervention about intervention “x” during a client-provider consultation % of providers observed to be using mHealth intervention during their patient consultati on Costs Estimated cost savings due to improvement in provider technical efficiencyd a: “Connect” could be via phone call. E.g. community health workers might call health supervisors for susp ected complication and received decision support via phone call or other mHealth supported means from a high level provider b: “Attended” could be via phone call or personal home-visit or other modes of communication using mHe alth strategy c: ‘Adequate’ could be defined by program intervention, eg: % of target health workers who know 3 pregn ancy danger signs d: Composite Indicator derived through monetizing time savings for administrative functions X: To be replaced by specific health intervention targeted by the mHealth platform Client-level: How Does Technology Improve Implementati on Process? mHealth m etric by tax onomic con straint Indicators Availability % of target clients who report adequatea knowledge about signs and symptoms requiring care-see king for health area “X” % of target clients who report adequatea knowledge about ‘x’ health area Technical E Incremental time in minutes between mHealth prompt received about intervention ‘x’ and care-see fficiency king with provider Total number of in-person consultation with qualified health provider about intervention ‘x’ by tar get clients as a result of accessing required services using mHealth strategy over reference period b Quality Duration of illness episode in days Total time in minutes spent with a health provider about health intervention ‘x’ in the last visit % of messages received through mHealth strategy that clients are able to recall about intervention ‘ x’ during client exit interviews % of target clients who report correctly adhering to prescribed care protocol about intervention ‘x ’ a: ‘Adequate’ could be defined by program intervention, eg: % of target clients who know 3 pregnancy danger sig ns X: To be replaced by specific health intervention targeted by the mHealth platform b: Required collection at multiple time points to yield estimates of “averted” incidences Client-level: How Does Technology Improve Implementati on Process? mHealth m etric by tax Indicators onomic con straint Utilization % of emergency events where mobile phones were used by patients to expedite treatment over re ference period % of target clients who report receiving a health information about intervention ‘x’ through their mobile phone within reference period % of target clients who report contactc with a qualified health care provider using mobile phone str ategy about intervention ‘x’ over reference period Costs % changes in reported client out-of-pocket payments in illness management over specified period of time d c: Contact: To be determined based on mHealth strategy medium of health service delivery. Could include teleph onic consultation, home visit by health worker, or clinic visit by patient where the use of the mHealth strategy ha s played a role in the receipt of services. d: Composite indicator- could be sub-categorized into individual components of interest where cost-savings are i ntended e.g. travel cost, days wages lost ,