Why is child mortality so different across countries and regions with similar income levels ? Peter Boone May 2005 Questions: • Why is child mortality so different across countries and regions with similar income levels ? • Is it possible to replicate the low levels observed in some countries quickly in a RCT ? (at low cost ?) Child mortality and income Child Mortality Rate GNB MWI 200 150 MRT GIN CMR SWZ 100 NPL UZB 50 MDA ZAF VNM LKA CUB 0 6 6.5 7 7.5 8 8.5 Ln PPP GDP per capita 9 9.5 10 Case Study: Kerala • State created in 1956 • Communist government elected 1957 • 1993 and 1998/9 FHS provide large database • Literature reports improvements in almost every area of health as reason for Kerala’s relative success Child mortality in Indian States 160 Child Mortality Rate 140 Mad Utt Meg Raj Bih Ori Aru India Ass Guj Jam And Har Pun Tri ar Tam KNag Ben Mah Man Miz Him 120 100 80 60 40 20 Del Goa Ker 0 0.0 5.0 10.0 15.0 Per Capita Net State Domestic Product ('000 Rupees 1996) 20.0 25.0 Focus on ARI and Diarrhoea Cause of non-neonatal child death by major disease/syndrome 33% 13% 32% Source: Black et. Al. 2005 Diarrhea Pneumonia Malaria Other AIDS Measles Unknown Pathogens causing diarrhoea Presentation Organism Acute watery diarrhoea Salmonella, Staphylococci, B. cereus, C. perfringens, V. parahaemolyticus, Ecoli, Rotavirus + other enterovirsues, Cryptosporidium, V. cholerae Acute bloody diarrhoea Shigella sp., C. jejuni Chronic diarrhoea G. intestinalis Chronic diarrhoea with blood E. Histolytica, B. coli, S. mansoni Source: Webber (2005) Pathogens causing ALRI Presentation Organism Pneumonia Streptococcus pneumoniae, Haemophilius influenzae, Influenza A or B, RSV, Adenoviruses, Morbillivirus, Metapneumorvirus, and many more….. Meningitis Neisseria meningitis, Haemophilius influenzae, Streptococcus pneumoniae, Mycobacterium tuberculosis, Staphlycocci, Escherichia coli, Group B streptococci, Mumps virus, Rubella virus, and many more…. Source: Webber (2005) Steps to child mortality and possible interventions Exposure Improve sanitation, water, hygiene Clinical Manifestation Raise immune response: Better nutrition and vaccines Treatment Improve health seeking decisions: education, proximity of services, cost of services Outcome Raise healthcare worker training and incentives, medical supplies, new treatments Which stage is best targeted ? • No consistent answer in the literature – Low CMR countries intervened in all areas • Cuba, Sri Lanka, Kerala, Former communist block, Costa Rica – Consensus that vaccines (Measles, DPT, BCG) have been highly successful but Diarrhoea and ALRI not easy targets – Controlled trials demonstrate potential large impact of interventions at every stage – Cost effectiveness work (e.g. World Bank 1993) helps allocate resources across interventions but can’t capture synergies, social and institutional issues that could lead to a different package. Two alternative hypotheses For groups with similar income levels: A. Healthier populations have reduced exposure to disease, and hence less morbidity and less mortality. B. Healthier populations have similar morbidity but lower case fatality rates. Exposures & Morbidity • DHS surveys compare disease incidence across low income countries. • Sanitation, water and hygiene are the main interventions aimed to reduce pathogen exposures. International data is available. ARI and Diarrhoea Incidence (31 low income countries) ARI: average cases per year per child 10 NER 9 8 SDN 7 MDG 6 TCD ETH LBR 5 GAB 4 KAZ TZA 3 PHL 2 MDG KHM MWI VNM BWA BGD 1 0 0 2 4 6 Diarrhoea: average cases per year per child 8 10 Mortality and Disease Incidence Ln (Child Mortality Rate) Coef SE Ln (PPP GDP per capita) -3.073** 1.906 Ln (PPP GDP per capita)2 0.184** 0.127 Diarrhoea (under age 5 cases per yr) 0.057 0.050 ARI (under age 5 cases per yr) 0.017 0.041 Measles vaccine (proportion popn) -1.005++ 0.488 Doctors per 100,000 -0.279++ 0.133 R2 0.732 N 31 ** : Jointly significant at 5% level ++: Significant at 5% level Mortality and Sanitation/Water Ln (Child Mortality Rate) Coef SE Ln (PPP GDP per capita) 1.057** 0.546 Ln (PPP GDP per capita)2 -0.103** 0.032 Improved Sanitation (proportion popn) -0.163 0.218 Improved Water (proportion popn) -0.160 0.311 Measles vaccine (proportion popn) -0.103** 0.646 DPT Vaccine (proportion popn) -1.051** 0.577 Doctors per 100,000 -0.202++ 0.044 R2 0.869 N 103 **: Jointly significant at 1% level (vaccines; income measures grouped separately) ++: Significant at 1% level Mortality and Latrine Access -1 -.5 0 .5 1 (103 countries) 0 20 40 60 80 % of population with access to latrines 100 India: ARI and Diarrhoea Incidence (26 States, children aged <4) 25 ARI Incidence (%) Tri 20 15 Man Ass Ben 10 5 Ker Nag India Raj Goa Kar Pun 0 0 5 Tam Ori Aru Him Bih 20 15 10 Diarrhoea Incidence (% in previous two weeks, < 4 yr olds) Jam Miz 25 India: Mortality and Disease Incidence (26 states) Ln (Child Mortality Rate) Coef SE Ln (SDP per capita) -0.0379 0.2011 Diarrhoea (under age 5 cases per yr) 0.0357 0.0344 ARI (under age 5 cases per yr) -0.0482 0.0482 All rec’d vaccines (proportion popn aged<4) -1.4000++ 0.3412 R2 0.54 N 26 ** : Jointly significant at 5% level ++: Significant at 5% level ris M ad ya sa Bi h Pr ar U a H ttar des im h ac P ra ha de An l P sh dh ra ra de Pr sh ad R esh aj a T a st h a m n il N ad u In Ka d rn ia at a H ka W ar y es an tB a en g G al u M ah jara Ja ara t m s m htr u a R eg io Pu n n M eg jab ha la ya G oa As Ar sa un m Tr ac ha ipu lP r ra a de s N ag h al an d Ke ra M la an ip ur D e M lhi iz or am O 100 % of households with access to latrines India: Household access to latrines 120 80 60 40 20 0 M M oa ad iz ya ora m Pr ad es h O r R issa a Ja ja s m m th a u n R eg H io im n ac ha Bih l P ar ra de sh An dh In Ar r un a P dia ra ac ha de l P sh r M ad ah es h ar as ht r G a uj ar a Tr t i T a pu ra m il U N tta r P adu ra d Ka es rn h at a H ka a r W es yan tB a en ga l D el h Pu i nj ab G a N ag la M alan eg d ha la ya M an ip u As r sa m Ke r % of households with piped or pumped water India: Access to improved water supply 120 100 80 60 40 20 0 Latrines and Diarrhoea Incidence Diarrhea incidence (children <4 yrs %) 35 Jam Him 30 Ori 25 20 Bih 15 Tri Mah Mad Utt Meg Raj India Guj Goa And Aru Nag Ben Pun Miz Man Tam Har Kar 10 Del Ker Ass 5 0 0 20 40 60 Households with latrines (%) 80 100 Diarrhea and Water/Sanitation: Trial results Interventions vs Diarrhoea 50 45 40 35 30 % risk reduction 25 20 15 10 5 0 water quality Source: Val Curtis water quantity sanitation hygiene prom hand wash Implication: Disease incidence and mortality • Disease incidence is weakly correlated with mortality • Sanitation and water improvements are weakly correlated with mortality • No evidence from survey data that Kerala’s success is due to lower morbidity B. Treatment regimes • Treatments in theory should be able to stop most mortality – Treatment trials wr to ARI and Diarrhea at healthcare centers typically have miniscule case fatality rates – Population-based RCT with frequent participant monitoring have low mortality rates • Regressions across countries and within India suggest treatment variables (vaccines, physicians) are more important than incidence measures Treatment Protocol (WHO) Presentation Treatment Suspected ALRI: Cough and rapid breathing Antibiotics at home, at 0.0% hospital if severe $0.50-$2.00 per course at home Diarrhea -Watery - ORT -Bloody - Antibiotics Suspected Malaria - Fever, parasites in blood (if measured) - Antimalarial chemoprophylaxis Source: Work in progress Case fatality if treated early and cost 0.0%, 10 cents per packet 0.0%, $0.50-$2.00 per course at home Near 0%, $0.25-$5.00 per course Surprising longevity of control study participants (if you visit them)…. Study Location Luby et. al, 2005 Pakistan Participants Baseline (intervention, mortality control) 3,163 ; 1,528 CMR 140 90 nonneonatal Schellenber Tanzania 350 ; 351 IMR 100 g et. al. 2003 60 nonneonatal The studies were not powered to measure mortality change Study mortality rate (intervention; control) 24 ; 38 20 ; 23 Treatment proxies and mortality (31 countries) Ln (Child Mortality Rate) Coef SE Ln (PPP GDP per capita) -3.073** 1.906 Ln (PPP GDP per capita)2 0.184** 0.127 Diarrhoea (under age 5 cases per yr) 0.057 0.050 ARI (under age 5 cases per yr) 0.017 0.041 Measles vaccine (proportion popn) -1.005++ 0.488 Doctors per 100,000 -0.279++ 0.133 R2 0.732 N 31 ** : Jointly significant at 5% level ++: Significant at 5% level India: Treatment proxies and Mortality (26 states) Ln (Child Mortality Rate) Coef SE Ln (SDP per capita) -0.0379 0.2011 Diarrhoea (under age 5 cases per yr) 0.0357 0.0344 ARI (under age 5 cases per yr) -0.0482 0.0482 All rec’d vaccines (proportion popn aged<=3) -1.4000++ 0.3412 R2 0.54 N 26 ** : Jointly significant at 5% level ++: Significant at 5% level India: Vaccine coverage All Recommended Vacines (children <4 yrs) 100 90 80 70 60 50 40 30 20 10 0 Goa Guj Ben Utt Mad Aru Raj Nag Bih Meg Tri India Man Ker Pun Har Kar And Tam Ass 0 20 40 60 Measles Vaccine (%) 80 100 India: Proximity to health care and spending on healthcare Population with healthcare facilities in village (%) 120 100 Ori Ker 80 60 Bih 40 20 0 0.00 Goa Tam Pun Him Raj 20.00 Aru 40.00 60.00 State healthcare expenditures per capita (Rp) 80.00 Why differences in treatment ? • Poverty/Price ? – How do household’s value treatment without knowing its potential benefit ? • Coordination and regulation failure ? – In Kerala, 2/3 of treatments are done by the private sector • Knowledge (both within the health system and within the family) ? – Essential to make health decisions…. Two sorts of “health knowledge” • Selected behavior in species • Acquired knowledge Disgust: Selected mechanism to control disease • With child mortality rates at 40% over a long history, small deviations in behavior due to genetic factors that raised odds of survival would eventually dominate the population • Our sense of disgust and beauty is probably intrinsically related to the fight against microscopic enemines Two fluids x= 1.6 x= 2.6 x= 1.6 x= 3.9 Source: Val Curtis 3.6 vs. 4.6 Households that heard of AIDS (%) Knowledge of AIDS and ORT 100 Man Tam 80 Del Nag Ker Miz Goa Mah Aru Kar Pun And Tri Meg India Har Ori Ass Jam Guj Ben Mad Raj Utt 60 40 20 Bih 0 30 50 70 Households that know of ORT (%) 90 Him or am K er al O a ris M sa an ip ur D el T r hi H im N ipu ac ag ra ha a l l P an ra d de s P h un H j ab ar ya n A a A s ru K sa na ar m ch na t M al P aka ad r ya ad Ja P esh m rad m u esh R M eg eg i o ha n la ya B ih ar U tta In rP d ra ia de A sh nd hr a Go P ra a de G sh T a uj a m ra M il N t ah a ar du a R sht a r W j as a es th t B an en ga l M iz % of households with access to latrines Should you reduce fluids when a child has diarrhoea ? 70 60 50 40 30 20 10 0 Uganda: Perceived causes of Malaria What causes Malaria ? Mosquitoes % Comment: 95.3 Most believed transmitted by drinking eggs or larvae Bad air 30.8 Contagious from others 51.5 Drinking water 67.0 Witchcraft Convulsions cannot be cured by modern medicine: Source: Nuwaha (2002) 9.8 51.0 What causes a common cold ? (Massachusetts, 197 families) Virus Germs 93% 88% Bacteria Changes in the weather Not enough clothes 66% 60% 56% Wet hair in cold weather 41% Cold weather Teething 38% 28% Walking barefoot 26% Source: Lee et. al. (2005) Tanzanian RCT healthcare quality Assessment Indicators IMCI Non-IMCI (%) (%) P-Value Pneumonia correctly treated 75 40 <0.01 Malaria correctly treated 88 25 <0.001 Child needing antibiotic /antimalarial prescribed correctly 73 35 <0.001 Child not needing an antibiotic leaves with an antibiotic 14 43 <0.001 Child needing vaccinations gets all needed 12 0 >0.10 Caretaker of sick child advised to give extra liquids and continue feeding 90 4 <0.00` Conclusion Exposure Clinical Manifestation Treatment Outcome Treatment variations probably explain the differences in mortality at similar income levels across regions and countries. An underlying factor leading to differerences in treatment outcomes may be variations in the perceived (and actual) value of existing available treatments. Next steps • Conduct one or more RCT with the goal being to reduce mortality sharply in a high mortality region through changes in the treatment regime – Preliminary research to understand treatment seeking, quality of healthcare in selected region – Outreach program in RCT - Intervention arm: • provide information to mothers on disease recognition and treatment • monitor child’s health frequently to ensure treatment regime has changed • free provision of treatment and vaccination coverage • Antenatal and neonatal program – Follow-up studies to measure duration of impact and changes in treatment seeking behavior once free assistance is ended SOAP Trial Africa • Objectives: – Determine the impact of an outreach program, including treatment for children that develop disease, on child mortality – Determine the impact of hand-washing behavior on incidence of acute lower respiratory infections and diarrhoea, and on the need for treatment SOAP Trial Africa Neighborhoods: High mortality region (randomized to intervention arms or control) Households within intervention neighborhoods randomized to intervention arms Arm 1: Handwashing & treatment 800 children Note: Suggested structure Arm 3: Mortality tracked 2800 children Arm 2: Treatment 800 children SOAP Trial: Endpoints • Comparison of incidence of ALRI and Diarrhoea in Arm 1 vs Arm 2 – Powered at 80% to capture a 25% reduction in ALRI in Arm 1 • Comparison of mortality in Arm 1 + Arm 2 versus Arm 3 – Powered at 80% to capture a 60% reduction in child mortality (assuming control 100 CMR) • Comparison of the cost of treatment in Arm 1 versus Arm 2 SOAP Trial: Other components • Surveys to understand (before and/or after the study): – Baseline disease incidence, mortality, and household characteristics – Hygiene behavior and beliefs – Treatment seeking behavior and beliefs – Quality of care in existing healthcare centers SOAP Trial: Timeline • Agreement on study design and structure – Within one month • Location, local partner, detailed protocol – End 2005 • Randomization – 1Q 2006 Cause of neonatal deaths 16% 24% 7% 24% Source: Black et. al., 2005 29% Severe Infections Birth Asphyxia Prematurity Tetanus Other Naandi Neonatal deaths trial • Objectives – To determine the impact on neonatal mortality of a targeted outreach and education program Neonatal Mortality Trial Villages: (randomized to intervention arms or control) Intervention Arm: Monthly education and healthcare visits Screening for high risk families Targeted outreach program (2000 births) Note: Suggested structure Control Arm Delayed introduction (2000 births) Naandi Trial: Timeline • Decision whether in principle we wish to proceed (within 3 months) • Study team selected to work out detailed timeline and program (6 months) • Start surveys and trial in 2006