1 / 17

1 London School of Hygiene and Tropical Medicine 2 Centers for Disease Control and Prevention

A new method to estimate mortality based on community informants: Validity and feasibility Bayard Roberts 1 , Oliver Morgan 2 , Francesco Checchi 1 Presented by Keith Sabin, WHO. 1 London School of Hygiene and Tropical Medicine 2 Centers for Disease Control and Prevention

herb
Download Presentation

1 London School of Hygiene and Tropical Medicine 2 Centers for Disease Control and Prevention

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. A new method to estimate mortality based on community informants: Validity and feasibility Bayard Roberts1, Oliver Morgan2, Francesco Checchi1 Presented by Keith Sabin, WHO 1 London School of Hygiene and Tropical Medicine 2 Centers for Disease Control and Prevention Collaborators: Ministry of Public Health, Afghanistan; United Nations High Commissioner for Refugees; Médecins Sans Frontières France; Aide Médicale Internationale

  2. Rationale • Mortality surveillance is costly and deteriorates rapidly without supervision • Retrospective mortality surveys have a number of limitations • Various selection and response biases • No time to explore causes and circumstances of deaths • Can’t really get real-time estimates (unfeasibly high sample sizes) • Hypothesis: key community informants know about a very high proportion of recent deaths in their community • Often assumed by people doing rapid, rough assessments • Needed to test this formally against an acceptable gold standard measure of mortality • Meant mainly for crisis settings • Operational questions

  3. How the method works (1) • Divide community into sectors based on obvious administrative boundaries • Rapid qualitative work to identify two key informants • Focus group discussion with community members • Topic guide, rapid analysis • There must be one of each in each sector • Exhaustive search process • Key informants refer investigators to all the households they know of in the sector where someone died in the last two months • Households are also asked to refer team to other households (snowball) • Sector is exhausted when informants cannot refer to any more households EM = Exhaustive Method

  4. How the method works (2) • Household questionnaire • Multiple questions to clearly establish date • Actually, restrict analysis for the last month only (or a similar well-defined period) • Questionnaire about variables of interest (optional: Verbal Autopsy) • Denominator for mortality rate = person-time • Time: defined recall period • Person: population size estimated through rapid methods (e.g. structure count + mini-survey of structure occupancy), if not already available through registration/census activities

  5. Methods Specific to Malawi • August-September 2008 • Focus group discussions with community • identify key community informants • EM - capture deaths over a two month recall period • Key informants = village headmen & village sage-women (fumukazi) • Population estimations • Capture-recapture analysis used to measure sensitivity • WHO standard verbal autopsy & analyses

  6. Malawi results • Verbal autopsies feasible • Among 50 deaths for which a cause of death was attributed, • 48.8% were due to infectious causes • children under 5 years - 45.3% neonatal causes. • HIV/AIDS - (16.6%), • Mortality rate of 0.05 (95% CI 0.01-0.10) deaths per 10,000 person-days

  7. Field testing (July-October 2008) • Urban district of Kabul city, Afghanistan (MoPH) • Extremely chaotic layout • Lots of population movement, short-term renters • Mae La refugee camp, Thailand (AMI) • Burmese Karen refugees • Well-established but with recent movement due to new arrivals and third-country emigration • Chiradzulu district, Malawi (MSF-France) • Rural, very scattered (spatial sampling of a fraction of the villages) • High HIV/AIDS burden • Lugufu and Mtabila camps, Tanzania (UNHCR) • DRC, Rwanda, Burundi, Sudanese refugees • Very well-established, well-serviced

  8. Validation and feasibility Validation: • Capture-recapture analysis / Multiple Systems Estimation on three lists • Gold standard measure of the true number of deaths (denominator for sensitivity) • List 1: from the method itself • Lists 2 and 3 from other sources • Bayesian Model Averaging Feasibility: • Kept track of costs and time inputs • Compared to expected inputs for a SMART mortality survey • Considered ethical implications

  9. Results: attrition • Rapidly canvassed large populations exhaustively

  10. Results: yield of the method • Few referrals from households (no snowball process) • Contrary to preliminary observations in Darfur • Need sufficient number of events?

  11. Results: sensitivity

  12. Results: sensitivity for deaths <5y • District 1, Kabul: 52.6% (60d period), ≤20% (30d) • Mae La camp, Thailand: ≤12.5% (60d), 0% (30d) • Chiradzulu district, Malawi: 66.7% (60d), ≤80.0% (30d) • Tanzania camps: 53.7% (60d), 47.1% (30d) • Key informants inappropriate for children, except Malawi • But low numbers

  13. Results: feasibility (financial inputs)

  14. Results: feasibility (time inputs)

  15. Major problems noted • Uninformative qualitative work • Difficult to get away from community leaders • “We know about every death!” • “If a family goes to bury their child somewhere, how am I supposed to know?” • Sectors a bit too big • Wouldn’t expect key informants to be able to cover them • No household chain referral process • Events too rare?

  16. Appraisal of the method Advantages: • Cheap and rapid data collection • No sampling bias • Data entry and analysis can be done by non-epidemiologists without specialised software • Assuming population estimate is available • Facilitates in-depth exploration of circumstances (of death in our case) Disadvantages: • Low sensitivity, esp. for children (but may be improved by adding informants) • Especially for deaths of young children • Sensitivity is probably context-specific • Need good qualitative skills to identify key informants • May need to do population estimation (unlike surveys)

  17. Acknowledgements • FANTA/AED funding • Megan Deitchler Co-investigators: • Mark Myatt (Brixton Health) • Daniel Chandramohan, Egbert Sondorp, Mohammed Ghaus Sultani (LSHTM) • Sunday Rwebangila (UNHCR) • Peter Nyasulu (MSF-France) Report now available: http://www.fantaproject.org/publications/EM_method.shtml

More Related