Embed Size (px)
Citation preview
January 2012
Born, lived and died - but counted or not?
highlightsStudies of HIV in
African communities
Much of the developing world,especially sub-Saharan Africa, ischaracterised by incomplete or no
vital registration systems. With the majority ofdeaths occurring at home in such poor settings,they are usually not registered and causes ofdeath are hardly ascertained. The unavailabilityof cause-of-death information in these areaslimits efforts to establish a solid evidence basefor health policy formulation. It also presentssignificant challenges for effective planning,monitoring and evaluation of health caredelivery systems.
Over the past two decades, a number of fieldresearch sites have adopted, tested andcontinue to refine a technique called “verbalautopsy” (VA) to derive information on causesof death in remote rural settings in sub-SaharanAfrica and South-East Asia. Together withrecently developed computer-basedprogrammes to facilitate a consistentassignment of cause(s) of death, the method hasbecome an attractive alternative to potentiallyfill a sizeable chunk of the health informationgaps that exist in the world’s poorest nations.
VA is an indirect technique of ascertainingbiomedical causes of death from informationon signs and symptoms, as well ascircumstances preceding death, obtained froma lay interview of the deceased’s main carerprior to death. It involves the administration ofspecially designed questionnaires by lay, buttrained, enumerators to solicit the necessaryinformation. Each case is reviewed by at leasttwo qualified physicians to assign direct andunderlying causes of death. Only concordantcauses are adopted, and discordant reviewsare returned with an undetermined cause ofdeath.
This method, therefore, has great potential forgenerating the much needed cause-of-deathdata that is vital for health planning,intervention design and implementation, aswell as for ensuring equitable resourceallocation in settings with incomplete or no vitalregistration. This has been demonstrated inmany, albeit small, demographic surveillancesites in sub-Saharan Africa, and sampleregistration systems in India and China.
The VA procedure demands considerablehuman resources. Its use of physicians toreview cases and assign cause of death hasrestricted its applicability to the research
setting where it has evolved over the pasttwenty years. With the alarming levels ofdoctor-population ratios in the areas wherethe VA method is most suited, the fewavailable physicians will be better utilisedtreating the sick than reviewing VAs. As aresult, scaling up of the use of VA beyond thecurrent research setting can only be achievedwith the adoption of an efficient alternative tophysicians for assigning causes of death.Computer-based algorithms have beendesigned to rapidly process symptom-levelinformation from VAs and produce consistentoutcomes of probable causes of death. TheInterVA model is one such programme that hasbeen widely tested and validated in Africaand Asia. Designed to support the mostwidely used VA questionnaire with little humanresources requirement, it has demonstratedmuch promise as the main avenue for thescaling up of VA through sample registrationwith a view to making cause-of-deathinformation available at district, provincial andnational level within a short period of time.
With the need for population-based cause-of-death information becoming increasinglyurgent for effective health care planning andresource allocation, the time has come forthe relevant policy decisions to be taken andbacked by the appropriate legal frameworkto adopt the available standard VAquestionnaire and the InterVA model to scaleup and integrate VA into national vitalregistration systems.
Momodou JassehMedical Research CouncilPO Box 273, Banjul, The GambiaT +220 [email protected]
Jeannette QuarcoopomeINDEPTH NetworkPO Box KD 213, Kanda, Accra, Ghana T +233-302-519394/[email protected]
Francis LeviraIfakara Health Institute PO Box 78 373, Dar es Salaam, TanzaniaT +255 784 [email protected]
Vital informationfor vital statistics
A fieldworker conducting a verbal autopsy interview in rural Gambia. Main carers ofdeceased residents of a demographic surveillance area are visited at least 40 days after deathto answer questions relating to circumstances in the few days or hours leading to the death.The information gathered forms the basis for determining the probable direct and underlyingcauses of death; and is collectively used to inform local health authorities about the cause-of-death distribution in the area for effective health intervention planning and evaluation.© Mamadi Sidibeh, MRC The Gambia, 2011
T he fifth Millennium Development Goal(MDG5) aims to reduce maternal mortality by three quarters between
1990 and 2015. However, measuring progresstowards this target is proving challenging. Inpractice, MDG5 is being measured country bycountry, assessing at national level whethereach one is achieving the target.
Maternal mortality is commonly assessedusing the maternal mortality ratio (MMR)which is the number of women dying duringpregnancy or immediately afterwards,whose deaths are somehow connected totheir pregnancies, divided by the number ofbabies born alive within the samepopulation. Unfortunately, very little isknown about the number of such deaths inmany countries and consequently, MMR isoften estimated using mathematical modelsdeveloped by experts at the global level.
Such models can provide predictions ofMMR for each country, based on therelationship between maternal mortality, thelevel of fertility, gross domestic product(GDP), HIV prevalence and other indicatorswhich are thought to be easier to measure.But how confident can we be with thesemodelled estimates? They may be useful inthe absence of anything better, but theresults do need to be interpreted withcaution. Mathematical models are only asgood as the data on which they are based,and in the case of maternal deaths, thesedata are often very poor.
According to the WHO, about 148countries out of a total of 172 haveconducted surveys in which women were asked questions about their sister’sexperience of maternal mortality, or haveother direct data on maternal deaths – therest are totally dependent on modelledestimates. Of the 148, 110 have data fromthe last 10 years and the rest rely on trendextrapolation. Furthermore, the complexmodelling methods are hard for most peopleto understand.
A big uncertainty in these models is therelationship between HIV and maternalmortality. If an HIV-positive woman diesduring pregnancy, it is hard to judgewhether her pregnancy might haveaggravated her HIV-related illness, orwhether she would have probably diedirrespective of being pregnant. Thisinformation can only come from community-based data detailing mortality among arange of women whose HIV status andpregnancy status is well known at the time of their death.
Research sites in the ALPHA and INDEPTHnetworks have rich data on people living inwell defined areas in Africa, together withtheir causes of death. Each site cancontribute to overall understanding ofmaternal mortality by calculating MMRs andcomparing these to modelled estimates tohelp ascertain whether the MMRs producedby the models are reasonable. Furthermore,using HIV status data from the ALPHA sites,provides a unique opportunity to gain a
better understanding of the relationshipsbetween pregnancy and HIV, and thereforeimprove the design of models.
We could only manage without models ifreliable national data for every country wereavailable, and this is still a major challengefor many developing countries. Meanwhile,data from areas within countries lackingnational data can be used to validatecurrent models and provide better estimatesof the parameters driving the models.
Maquins Sewe and Paul OgindoKEMRI/CDC Study SiteP.O Box 1578-40100Kisumu, KenyaT+254 [email protected]
Peter ByassUmeå Centre for Global Health ResearchEpidemiology and Global HealthUmeå University90187 Umeå, [email protected]
Clara CalvertDepartment of Infectious Disease EpidemiologyLondon School of Hygiene & Tropical MedicineKeppel Street, London WC1E 7HT, [email protected]
Who is counting?
Page 2 www.ids.ac.uk/go/knowledge-services
See also‘The Imperfect World of Global HealthEstimates’, Byass, P., PLoS Med, 7(11):e1001006, 2010www.plosmedicine.org/article/info:doi/10.1371/journal.pmed.1001006
The ALPHA Network aims to maximise theusefulness of data generated in community-based longitudinal HIV studies in sub-Saharan Africa for national andinternational agencies involved indesigning or monitoring interventions andepidemiological forecasting.
The project links several existing HIV cohortstudies and runs training workshops tofacilitate replication of analyses ofdemographic correlates and consequencesof HIV infection previously published in justone or two sites. Comparative studies and
meta-analyses are undertaken oncomparable data sets, imposing a commonformat on data collected and stored in avariety of ways.
The ALPHA Network
How do weknow who diesof what?Not all deaths in the world are handledin the same way. In many countries thereis a legal requirement for a doctor’scertificate stating cause of death. But inother places, deaths are handled lessformally and no cause is recorded. Forhealth planners, though, cause-of-deathinformation is very important. So whenthe cause of a death is not formallyrecorded, another approach is to laterinterview family and friends about thecircumstances of the death. Then theirinformation can either be considered bya doctor, or fed into a computer model,to decide the likely cause of the death.This process is called “verbal autopsy” –not perfect, but much better than nothing.
A mother carrying her child to the maternaland child health clinic at Kisesa healthcentre in Tanzania. The data from thevarious ALPHA Network sites provideunique opportunities to gain a betterunderstanding of the relationships betweenpregnancy and HIV, and therefore improvethe design of models. © David Sevuri, True Vision, 2009
Every year, half a million women die ofpregnancy related causes. These deathshave long term devastating effects for
surviving children and families. We all knowthat pregnant women are at risk of deathfrom direct obstetric causes such asobstructed labour, excessive bleeding orsevere infection after delivery. This has led tothe prevailing perception that pregnantwomen are more vulnerable than non-pregnant women. However, most of thesepregnancy-related deaths are preventable bywell-established health interventions andtherefore, such loss of young women’s lives isinexcusable. What about other causes ofdeath unrelated to pregnancy such astuberculosis, viral/ bacterial infections,pneumonia, etc.?
It may be surprising to learn that pregnantwomen die much less from these causes ascompared to women who are not pregnant.The reason why pregnant women are at alower risk of death from causes unrelated topregnancy is because in general, womenwho are able to get pregnant tend to behealthier than their counterparts of similarages who are not pregnant. Given that
pregnant women are young and healthier,why is it that we do not see pregnantwomen’s initial health advantage translateinto significantly lower mortality ascompared to non-pregnant women on thewhole? A simple answer is that healthypregnant women are subjected to significantlevels of additional risk from direct obstetriccauses that are preventable by medical andpublic health interventions.
Policymakers and the medical communityneed to focus on preventing mortality fromsuch pregnancy-related causes. The researchcommunity can play a key role in thisendeavour by undertaking a carefulexamination of the causes of death inpregnant women. A key limitation in thisregard is the lack of good cause-of-deathinformation that distinguishes direct obstetriccauses of death from those arising fromcauses unrelated to pregnancy. One way toaddress this gap is to make better use ofverbal autopsies collected by severaldemographic surveillance sites (such asALPHA Network and INDEPTH sites) toidentify the direct and indirect causes ofmaternal mortality. Quantifying excess
mortality due to preventable causes will help policymakers design prevention andtreatment programmes to decrease theexcess burden of maternal mortality.
Urgent action is needed to stop thisunnecessary loss of healthy women,especially if the Millennium DevelopmentGoal of lowering maternal mortality by 75 percent by 2015 is to be achieved.
Carine RonsmansDepartment of Infectious Disease EpidemiologyLondon School of Hygiene & Tropical Medicine Keppel Street, London WC1E 7HT, UKT +44 (0)20 7927 [email protected]
Jessica Nakiyingi-MiiroMedical Research Council (MRC)/UgandaVirus Research Institute (UVRI)Research Unit on AIDSP.O. Box 49, Entebbe, UgandaT +256 (0)312 262910/[email protected]
Gayatri SinghBrown University Providence, Rhode Island 02912, [email protected]
www.ids.ac.uk/go/knowledge-services Page 3
What is killing pregnant women?
Antenatal check-up in a maternity clinic in Kenya. To reduce and bring down the high maternalmortality, the government has to address several challenges including the need to ensure theavailability of adequate maternity services and skilled personnel to attend to complicationscaused by unsafe/induced abortion, malaria, and HIV/AIDS, among others. © Peter Barker/Panos
See also‘Deaths Attributable to Childbearing inMatlab, Bangladesh: Indirect Causes ofMaternal Mortality Questioned’, Khlat, M.and Ronsmans, C., American Journal ofEpidemiology, 151.3 (2000): 300-6http://aje.oxfordjournals.org/content/151/3/300.full.pdf
‘HIV and the Magnitude of Pregnancy-Related Mortality in Pointe Noire, Congo’,Le Coeur, S., Khlat, M., Halembokaka, G.,Augereau-Vacher, C., Batala-M'Pondo, G.,Baty, G., and Ronsmans, C., AIDS, 19.1(2005): 69-75http://journals.lww.com/aidsonline/Fulltext/2005/01030/HIV_and_the_magnitude_of_pregnancy_related.8.aspx
‘Evidence for a 'Healthy Pregnant WomanEffect' in Niakhar, Senegal?’ Ronsmans, C.,Khlat, M., Kodio, B., Ba, M., De Bernis, L.,Etard, JF., International Journal ofEpidemiology, 30.3 (2001): 467-73http://ije.oxfordjournals.org/content/30/3/467.full
The InternationalNetwork for theDemographicEvaluation ofPopulations and TheirHealth in DevelopingCountries (INDEPTH) is
a global network of members who conductlongitudinal health and demographicevaluation of populations
in low- and middle-income countries(LMICs). INDEPTH aims to strengthenglobal capacity for Health andDemographic Surveillance Systems (HDSS),and to mount multi-site research to guidehealth priorities and policies in LMICs,based on up-to-date scientific evidence.
INDEPTH
Ongoing investment in HIV preventionand treatment systems requiresevaluation of the success of these
systems and in particular, monitoring trendsin deaths due to AIDS. This in turn requiresaccurate data on the causes of death, andreliable methods for identifying deaths due toAIDS. Many countries where AIDS is a majorcause of death have poor or non-existentvital registration systems and cause-of-deathcertification. This makes it difficult toaccurately determine the numbers of AIDSdeaths in these countries.
Verbal autopsies (VAs) can be used in theabsence of vital registration to ascertaincauses of death. Interviews are conductedwith caregivers or other friends or relativesof the deceased who were living with orclose to them during their final illness.Questions can be asked about the perceivedcause of death and the health of thedeceased prior to death, including key signs and symptoms of common illnesses.The data can be collected by trainedinterviewers, and then reviewed by a doctoror analysed by a computer programme toascertain causes of death.
It can be difficult to identify deaths due toAIDS using VAs. This is because HIV affectsthe immune system and makes sufferersmore at risk of developing other illnesses. It is often these illnesses that ultimately leadto death and it is the symptoms of thesediseases that are usually reported bycaregivers or friends during the VAs. Inmany countries, large proportions of thepopulation have never had an HIV test, sothe person being interviewed and thedeceased themselves may not have beenaware of their HIV status. In such
circumstances, it is difficult to ascertaindeaths due to AIDS. Other factors, such asstigma associated with HIV and confusionover the meaning of medical terminologycan further exacerbate these problems.
Research is needed to investigate alternativemethods of collecting and interpreting datafrom VAs in order to increase the accuracyof identification of deaths due to AIDS. Inparticular, the development of simpler,shorter VA questionnaires, and standardisedmethods for their interpretation is important.The ALPHA and INDEPTH networks havecollected VA data from several study sitesfrom across the world and are currentlyworking collaboratively to analyse theirdata. Prior knowledge of HIV status of manydeceased persons in ALPHA sites can beused to improve diagnostic accuracy andaddress many of these issues.
Improved coverage of HIV testing and HIVand AIDS education services may also helpto increase the number of people living withHIV, and their families, who are aware oftheir status, which would improve reportingof this during VAs. In the longer term, furtherinvestment in establishing vital registration,health records and death certification is alsorequired.
In conclusion, obtaining accurate data onthe frequency of deaths due to AIDSpresents many challenges. This is becauseHIV affects the immune system so that thosedying from AIDS suffer a wide variety ofillnesses and symptoms. Furthermore, manypeople are not aware of their HIV status orthat of their close family members andfriends. Further work is required to improvemethods for collecting and analysing VA
data in the context of the HIV epidemic andto improve coverage of HIV testing and HIVand AIDS education services.
Tom LutaloRakai Health Sciences ProgramPO Box 49, Entebbe, UgandaT +256 [email protected]
Denna Michael MkwashapiNational Institute for Medical ResearchKisesa HIV cohort StudyPO BOX 1462, Mwanza, TanzaniaT +255 28 [email protected]
Laura RobertsonDepartment of Infectious Disease EpidemiologyImperial College London South Kensington Campus London SW7 2AZ, UKT+44 2075943288 [email protected]
Page 4 www.ids.ac.uk/go/knowledge-services
This publication was produced by IDS KnowledgeServices with partners in ALPHA and INDEPTH.
Analysis funded by the Health Metrics Network(HMN), publication funded by Wellcome Trust.The views expressed in this publication do notnecessarily reflect those of the HMN, WellcomeTrust, the London School of Hygiene & TropicalMedicine and IDS.
IDS is a Charitable Company No.877338 limitedby guarantee and registered in England. © Institute of Development Studies 2012
Highlights is licensed under aCreative Commons Attribution
3.0 Unported License.This means that you are free to share and copythe content, provided IDS and originatingauthor(s) are acknowledged. ISSN 1746-8655
ALPHA Network, Population Studies Department,London School of Hygiene & Tropical Medicine,London WC1E 7HT, UKT +44 207 299 [email protected] www.lshtm.ac.uk/eph/psd/alpha
IDS Knowledge ServicesInstitute of Development StudiesUniversity of Sussex, Brighton BN1 9RE, UKT +44(0)1273 915777 F +44 (0) 1273 [email protected] www.ids.ac.uk/go/knowledge-services
INDEPTH NetworkP. O. Box KD 213, Kanda, Accra, Ghana T +233 302 519394 F +233 302 519395 [email protected]
See also‘Verbal Autopsy can Consistently MeasureAIDS Mortality: a Validation Study inTanzania and Zimbabwe’, Lopman, B. etal., Journal of Epidemiology andCommunity Health, 64(4), 330-334, 2010 http://jech.bmj.com/content/64/4/330.full.pdf
‘Creating and Validating an Algorithm toMeasure AIDS Mortality in the AdultPopulation Using Verbal Autopsy’, Lopman,B., et al., PLoS Med, 3(8), e312, 2006 www.plosmedicine.org/article/info:doi/10.1371/journal.pmed.0030312
Keywords: HIV, AIDS, verbal autopsy, vital registration systems, InterVA model, maternal mortality, cause of death, death
How do we know if someone has died from AIDS?
