IMPROVING MORTALITY SURVEILLANCE DATA IN HIGH MORTALITY SETTINGS

IMPROVING MORTALITY SURVEILLANCE DATA IN HIGH MORTALITY SETTINGS: CHAMPS AND COMSA Kathy Banke, Ph. D. IDM Symposium April 15, 2019 © Bill & Melinda Gates Foundation | 1

CHAMPS & COMSA CHAMPS Child Health and Mortality Prevention Surveillance High-precision cause of death identification Fine-grained data on causes of death among children under five, supported by analyzing samples collected using the Minimally Invasive Tissue Sampling (MITS) procedure. COMSA Precise & Scalable Mortality Surveillance Countrywide Mortality Surveillance for Action Representative sample to provide high quality information on births and deaths A robust sample registration system that captures the data that allows for calculation of mortality rates, birth rates, and cause-specific mortality fractions at the national and sub-national levels. 2

THE CHILD HEALTH AND MORTALITY PREVENTION SURVEILLANCE (CHAMPS) NETWORK Provides accurate, timely and reliable data on the causes of death for children under five Collects Minimally Invasive Tissue Samples (MITS), verbal autopsies, and medical records within a CHAMPS site Panel of specialists reviews MITS and all available information for each CHAMPS case and determines the most likely cause of death MITS data provide a more accurate and specific cause of death than would otherwise be available CHAMPS data are made available rapidly and shared openly with a range of stakeholders 3 Confidential © Bill & Melinda Gates Foundation | 3

How does Minimally Invasive Tissue Sampling (MITS) work? Brain Lung Blood CSF Heart Liver Stool Bone Marrow NP/OP swab • Abdominal approach - spleen / kidney • Placenta, umbilical cord if stillbirth or death immediately following birth • Skin lesion if present and lymph node if palpable High MITS consent 4

CHAMPS DETERMINATION OF CAUSE OF DEATH (DECODE) PANEL Data from site • Linked maternal data • Demographic and clinical data • Verbal autopsy cause of death and raw data • Microbiology, molecular biology (TAC), HIV, TB, malaria findings • Site histopathology findings Central Pathology Lab (CPL) • CPL histopathology • Special stains • Immunohistochemistry • PCR findings Data packet produced centrally Panel Members • Clinician • Pathologist • Microbiologist • Epidemiologist Central certifier review Panel Causes of Death Recorded in CHAMPS database, shared with clinicians and family (aggregate results shared with community) 5

What is COMSA? ü Mortality surveillance gaps remain but CHAMPS’ technical and financial investments limit its widespread delivery or why ü To complement CHAMPS, Countrywide Mortality Surveillance for Action (COMSA) was developed ü COMSA continuously records data on pregnancy outcomes, deaths, and cause of death from a population in a nationally representative sample of geographic areas ü Verbal autopsies conducted on all deaths ü MITS is conducted in a subsample of under five deaths to calibrate the cause of death information nationally from verbal autopsies 6

COMSA DEMONSTRATION PROJECTS IN MOZAMBIQUE AND SIERRA LEONE Status of Mozambique Status of Sierra Leone 700 clusters (~300 households each): • Phase I: 5 provinces, 422 clusters • Phase II: remaining 6 provinces, 278 clusters COMSA cluster Phase II Parks and reserves • Pilot in Bo district, May 2019 • Expand to all districts by end of 2019 © Bill & Melinda Gates Foundation | 7

COMSA approach Led by national statistics organizations with close collaboration from ministries of health, national civil registration authorities, national public health institutes § Four years of BMGF support for external technical assistance § Low running costs Select representative enumeration areas across a country § Cover ~ 3 -8% of entire population § Identify and report pregnancies outcomes and deaths (including stillbirths) § Conduct verbal autopsy (VA) on all deaths § Conduct MITS on some U 5 deaths from COMSA (outside of CHAMPS site) Assemble all data across the country and calculate statistics at the national and subnational levels § National and subnational crude birth and death rates § Age-group specific mortality rates and cause-specific mortality fractions and rates § Use MITS-VA pairs from CHAMPS and COMSA to calibrate national VA-based COD Integrate with existing data systems and share data promptly and continuously with local, national, and international stakeholders 8

COMPARED TO OTHER SOURCES OF CAUSE OF DEATH, COMSA CAN PROVIDE COMPREHENSIVE DATA AT LOW Comprehensiveness of information COST 5 attributes of mortality surveillance HDSS Site Post-census VA surveys COMSA-Like SRS (like INCAM in Mozambique) (using VA for community deaths and integrating facility data) CRVS Nationally Representative N Y Y Adequately Powered N Y Y Frequent Data Collection Y N (every ~5 years) N (every ~10 years) Y Y Cause of Death Included Y (but often just mothers and children) Y (but only when followon VA survey done) Y Y N N N Y Y $250 k annually (population of 60 k in rural setting) ~$2 -3 M for VA survey, ~$2 -4 M for original DHS survey ~$2 M for VA survey, $1 -8 M for original census ~$1 -2 M annually ~$4 -7 M annually or more Country-Run Cost Post-DHS or post. MICS VA survey Traditional DHS, MICS, and censuses only include fact of death, not cause of death. Cause of death must be gathered through an additional survey. Only SRS and CRVS contain all 5 attributes © Bill & Melinda Gates Foundation | 9

SRS CAN BE IMPLEMENTED WHILE BUILDING UNIVERSAL CRVS SRS can help countries build capacity for future surveillance efforts… Establish government buy-in • Help ensure government agencies are willing to invest resources in improving country data • Produce statistics that demonstrate proof-of-concept for nation-wide surveillance Build capacity • Provide in-country groups with skills and experience for accurate data collection • Build a network of collaboration and data sharing across groups within the country • Sites can serve as lighthouse areas for business process innovation in surveillance …and serve as an important catalyst for a full CRVS system CRVS System Facility-based deaths and cause of death recording Community deaths and cause of death recording SRS can help build capacity for all components of a CRVS system, but particularly help fill a data gap in the area of community deaths Universal birth registration Source: Based on interview with Dr. Philip Setel, Vice President and Director of Civil Registration / Vital Statistics Program at Vital Strategies 10

COMSA MOZAMBIQUE RESULTS (AS OF MARCH 2019) © Bill & Melinda Gates Foundation | 11

POPULATION BY AGE AND SEX Percent Distribution of the COMSA Population by Age and Sex Ratio 18 1. 2 16 12 0. 8 10 0. 6 8 6 0. 4 4 0. 2 2 0 0 0 -4 10 -14 20 -24 Male 30 -34 40 -44 Age group Female 50 -54 All 60 -64 Sex ratio 70 -74 Sex Ratio 1 14 Percentage Population pyramid

CURRENT LISTING* OF COMSA POPULATION BY PROVINCE 140, 000 Total Population: 726, 315 Males: 354, 358 Females: 371, 957 124, 642 120, 000 104, 256 100, 000 84, 369 76, 643 75, 098 80, 000 55, 067 60, 000 41, 278 35, 946 40, 000 41, 659 45, 160 42, 197 Maputo Provincia Maputo Cidade 20, 000 0 Zambezia Tete Cabo Delgado Nampula Sofala Phase 1 (Since April 2018) *Listing still underway in some COMSA clusters Manica Inhambane Niassa Gaza Phase 2 (Since October 2018)

BIRTHS BY PROVINCE 1800 1648 1600 1400 Total Births: 6, 935 1371 1216 1200 1000 800 625 600 491 460 400 270 200 181 187 Niassa Gaza Maputo Provincia 200 107 0 Zambezia Tete Cabo Delgado Nampula Sofala Phase 1 (Since April 2018) Manica Inhambane Maputo Cidade Phase 2 (Since October 2018)

PROPORTION OF BIRTHS BY PLACE OF DELIVERY AND PROVINCE 100% 39 80% 60 31 51 21 33 13 10 8 30 42 53 60% 40% 61 20% 40 69 49 79 67 87 90 92 70 58 47 0% Zambezia Tete (n=1340) (n=1638) Cabo Nampula Delgado (n=485) (n=1408) Sofala (n=619) Manica Inhambane Niassa (n=490) (n=266) (n=197) Health facility Phase 1 (Since April 2018) Gaza (n=184) Maputo Provincia Cidade (n=197) (n=111) Total (n=6935) Non Health facility Phase 2 (Since October 2018)

NUMBER OF DEATHS BY AGE AND PROVINCE 500 450 400 350 300 250 200 150 100 50 0 2, 910 deaths 1, 941 VAs Phase 1 (Since April 2018) Data through March 29, 2019. Includes stillbirths. ad e ia id nc ut o C ap M to ap u Stillbirths ro vi P G az M In ha m an M a ne ba a 12+Y ic a ss ia N So fa la pu la am N o Te te 5 -11 Y ab o D 0 -4 Y C Za m el g ad be zi a (67%) Phase 2 (Since October 2018)

PROPORTION OF HEALTH FACILITY DEATHS BY PROVINCE 100% 80% 60% 44 83 77 66 81 67 66 73 83 52 71 77 40% 56 20% 17 23 19 17 Cabo Delgado (n=546) Nampula (n=154) 34 33 Sofala (n=121) Manica (n=95) 34 29 Niassa (n=56) Gaza (n=89) 27 48 23 0% Zambezia Tete (459) (n=722) Health facility Phase 1 (Since April 2018) Inhambane (n=135) Maputo Provincia (n=55) Maputo Cidade (n=42) Non Health facility Phase 2 (Since October 2018) Total (n=2474)

NEW APPROACH ENABLES USE OF MITS/VA PAIRS TO INCREMENTALLY ADJUST VAS AND IMPROVE CAUSE OF DEATH Current State: Verbal Autopsies (VAs) are the standard DATA approach to evaluate cause of death (COD) in 226 child 17% The challenge: How many gold standard MITS paired with VAs do we need to start adjusting our understanding of COD from the thousands of current and historical VAs? 26% 16% Calibrated COMSA VAs based on Bayesian hierarchical approach with MITS-VA error matrix* Other Malnutriti on Diarrhea 0% Malaria 30% 20% 10% 0% 27% Pneumon ia % of deaths from COMSA Mozambique VAs calibrated based on 40 child deaths from CHAMPS Mozambique Cause Density Impact of MITS: MITS and VA done blindly on the same 167 CHAMPS deaths (1 -59 -month-olds) confirms and quantifies VA misclassification Verbal Autopsy Malaria Diarrhea Other infections HIV Malnutrition Other Total Pneumonia 42% 8% 25% 8% 0% 0% 17% 24 Malaria 13% 33% 13% 20% 0% 0% 20% 15 Diarrhea 31% 0% 38% 6% 19% 6% 0% 16 Other infections 50% 0% 5% 15% 0% 0% 30% 20 HIV 30% 4% 26% 4% 22% 4% 9% 23 Malnutrition 15% 10% 15% 5% 20% 15% 20 Other 29% 4% 12% 0% 0% 41% Total Cause-Specific Mortality Fraction VAs calibrated based on 167 child deaths from CHAMPS network Density Decoded MITS Pneumonia 49 167 Cause-Specific Mortality Fraction *Datta A, Fiksel J, Amouzou A, and Zeger S. Local calibration of verbal autopsy algorithms. Submitted to the Journal of the American Statistical Association (Oct 2018). ar. Xiv: 1810. 10572 [stat. ME]. Note: The data on this © Bill & Melinda Gates Foundation | 18

THANK YOU! Special thanks to: Mozambique: National Institute of Statistics (INE), National Institute of Health (INS), Ministry of Health (MISAU), Ministry of Justice, CISM (Manhica) Sierra Leone: Statistics Sierra Leone, Ministry of Health and Sanitation, National Civil Registration Authority, CDC Sierra Leone Grantees: Institute for International Programs – Johns Hopkins University (COMSA Mozambique), Centre for Global Health Research (COMSA Sierra Leone) © Bill & Melinda Gates Foundation | 19