Using Verbal Autopsies to Improve Child Survival Interventions_Perry_Davis_5.3.12

Verbal Autopsies:Data Analysis

Tom Davis, MPHSenior Director of Program Quality Improvement

May 2012, CORE Group Spring Meeting

Introduction

What is a Verbal Autopsy?

Interview with a parent of a child that has died (or family of a woman who died in pregnancy or childbirth) concerning the circumstances surrounding the death.

An investigation to learn how we can prevent deaths.

When are Verbal Autopsies Done and with Whom?

• Allow for period of mourning: Conduct at 2-4 weeks after death

• Done with mother of the child that died or another person who was around child during the illness / death (if mother was not).

• Supervisor (preferable, but not always feasible) or CHW / Promoter

Verbal Autopsy ≠ Mortality Tracking Form

IMPORTANT:

• Not all programs use the VA forms to track mortality rates (which is often done with a separate form, or not at all).

• Verbal Autopsies can be collected in only a sample of programs areas (e.g., one Care Group or Community per district).

• Data from only a few areas can still be instructive in making system changes that lead to improvement, but better to have representative data.

What’s the burden on the Supervisor or CHW?

• FH/Moz: 148,444 U5 children

• 7 districts w/65 Promoters and 5 Supervisors

• U5MR: 150/1,000

• 148,444*0.15/5/12 = 371 deaths/month

• 5.7 VAs / Promoter / month

• If only doing VAs in 10% of Promoter's coverage area: 0.6 VAs / Promoter / month, or 7.5 VAs / Supervisor / month

• More typical project (SAWSO/Zambia): 25,000 U5 children

• 21 Facilitators and 5 Supervisors

• U5MR: ~150/1,000

• 25,000*0.15/5/12 = 63 deaths/month

• 3 VAs / Facilitator / month

• If only doing VAs in 10% of Facilitator's area: 0.3 VAs / Facilitator / month, or 1.3 VAs / Supervisor / month

What is your Experience?

What has been your experience with Verbal Autopsies?

Has your organization used them? Did you get useful data?

What challenges did you have?

Benefits of Verbal Autopsies • To understand ages when children are dying

• To understand the causes of death (e.g., diseases)

• To understand delays that may lead to some deaths.

• delays in recognizing signs/symptoms of disease,

• delays in seeking medical attention,

• delays in reaching a health provider,

• delays in receiving care once they reached a health provider,

• delays in starting treatment.

• To understand home care and follow-up care provided

• To offer condolences to the family

Overview of Rationale and Use of Verbal Autopsies

Henry PerryDepartment of International Health

Johns Hopkins Bloomberg School of Public Health

CORE Group Spring Meeting, 4 April 2012

Outline

• Origins of verbal autopsies• Examples of the practical value of verbal

autopsy information for strengthening program implementation

• Examples of the practical values of knowing the age at death

• A few methodological points• Conclusions

Origins of Verbal Autopsy:John Gordon and John Wyon

Early Use of Verbal Autopsies 3 Decades Ago by Andean Rural Health Care

• Home visitation to deliver services and to register vital events

• Verbal autopsy by higher-level field staff to identify cause of death

Early Findings on Cause of Death in Two Areas of Bolivia, 1990-1993

Practical Implications

• Led to focus on diarrhea prevention and control and on nutrition in Montero (tropical lowlands) and to pneumonia in Carabuco (in the mountains)

• It should have led us to more investigation of what “asphyxia” was – likely to have been hypothermia

Percentage of Deaths by Age Group in World Relief/Cambodia Child Survival Project,

March 2004-February 2005

N=174

Cause of Death, 1-11 Months of Age, March 2004-February 2005, World

Relief/Cambodia Child Survival Project

Pneumonia, other resp. conditions

36.7%

Diarrhea/dehydra-tion1.0%

Other infections17.8%

Accident/injury1.0%

Dengue5.2%

Malnutrition5.2%

Other27.3%

Neonatal tetanus3.9%

Umbilical cord hemorrhage

1.6%

Practical Implications

• Pneumonia among children 1-11 months of age is a target for reducing mortality

• Neonates are also a target for reducing under-5 mortality

• Children over year of age have a low risk of death and are therefore not a target

World Relief/Mozambique Child Survival Project, 2004-2009

• Level of effort:– Malaria, 20%– Nutrition, 20%– Diarrhea, 20%– STDs/HIV, 15%– Pneumonia, 10%– Immunization, 10%– Exclusive BF, 5%

Practical Implications

• Malaria should be given top priority

Curamericas/Guatemala Child Survival Project, 2003-2007

Age at Death

Methodological Points

The Current “Gold Standard” for Verbal Autopsies for All Age Groups

http://www.who.int/healthinfo/statistics/verbalautopsystandards/en/index.html

• Shortened version in preparation now

Computerized Diagnosis

• Helps to standardize the process• Lowers the level of person required

• Simplifies it – in terms of time involved• InterVA – computerized algorithm

– Peter Byass

Malaria

• Difficult to diagnose precisely with verbal autopsy and can often be confused with pneumonia

• Many illnesses produce fever• Fever produces rapid respirations• Malaria is a difficult diagnosis to make even in

hospitalize patients– Having circulating parasites does not prove that a

patient’s symptoms are from malaria

• In malaria-endemic areas, it might be best to consider malaria and pneumonia and joint diagnoses and target both

Conclusions

• The goal of child survival programs is to reduce mortality in under-5 mortality

• Obtaining information about cause of death and the conditions under which that death occurred is an important step in empowering program staffs to more effectively target interventions

• By obtaining information on cause of death, you are also obtaining information about age at death, and this is important for programming as well

Conclusions (cont.)

• National-level data on cause of death are better than no data, but local-level data better than national data since the situation can vary from place to place (but attention to quality is important!)– There is variation in cause of death (and age at

death) from one socio-ecological setting to another• Verbal autopsy data not perfect, but it is the

best we can get – and it must be interpreted appropriately

Using the Verbal Autopsy Questionnaire

• (Go through questionnaire.)

• (See Tom for maternal form.)

If doing a Sample:Avoiding Selection Bias

NOTE: It does not matter if the child died from a birth defect after two days or from diarrhea when she was 8 months old – if a child dies in the area chosen for Vas, fill out the form. AVOID SELECTION BIAS.

But do not fill out a form if the child was born dead – only for children who were born alive and later died.

Analysis of Verbal Autopsy Data:

Examining Patterns & Exploring Hypotheses

Look for patterns of deaths in….• Home care

• Preventive care

• Care seeking

• Treatment

• Delays

• Gender and age of the child

• Co-morbidity, clusters of symptoms (e.g., malnutrition)

• Etc.

How to Analyze Verbal Autopsy Data – The Mortality Review Meeting

a. If tracking mortality rates, review the total number of deaths of children (0-23m, 0-59m) in the project area compared to the previous period (using tracking form) and rates.

b. Determine the principal causes of death, the age pattern for deaths for each cause of death, and other patterns. Write out statements on patterns, for example:

• "42% of deaths were in the first month of life;"

• "52% of deaths of children 0-11m were due to malaria";

• “71% of children were seen by a CHW and referred";

• “23% of mothers knew 2+ danger signs.”

How to Analyze Verbal Autopsy Data – The Mortality Review Meeting

c. If your numbers are large enough and representative enough, you may be able to look for possible associations. For example, association between malnutrition and deaths (using 2x2 tables).

How to Analyze Verbal Autopsy Data – The Mortality Review Meeting

Look for possible associations and problem areas: • Nutritional status (#12)

• Knowledge of Danger Signs (#14).

• Social status of family (#15)

• Signs / symptoms (#16)

• Who attended child (#19)

• Location of death during illness / death (#20)

• Care seeking outside of home (#21)

• Care providers sought (#22)

• Hydration (#23)

• Medications given (#24)

How to Analyze Verbal Autopsy Data – The Mortality Review Meeting

d. Identify the delays / barriers to identification and care and explore what can be changed concerning those delays / blocks.

• Delay in recognition that the child had a problem requiring medical attention (see questions #28/29)

• Delay in seeking medical care (after recognizing that care required medical attention). (see questions #30/31)

• Delay in reaching a health facility (after decision to seek care). (See questions #32)

• Delay in receiving care once at health facility & starting treatment (see questions #33/34)

• You should also see if there were problems with home care or follow-up care after receiving medical care/advice (see questions #35/36)

How to Analyze Verbal Autopsy Data – The Mortality Review Meeting

e. Determine areas where the family can improve their

management of child illness (including values and

attitudes).

f. Determine areas where health staff can improve their

management of child illness (including values and

attitudes).

g. Propose new methods and strategies to prevent

future child deaths.

How to Analyze Verbal Autopsy Data – The Mortality Review Meeting

h. Follow-up on previous methods and strategies proposed at earlier Mortality Review Meetings.

i. Possible actions to take using the VA results.• Improvements / changes in behavior change communication • Improvements / changes in protocols (e.g., referral, f/u, tx)• Improvements / changes in monitoring • Improvements / changes in access to care

• Improvements / changes in quality of care

j. Report findings from the Mortality Review Meeting to the MOH and Other Stakeholders who are not already in attendance.

Your Turn: Analysis of Verbal Autopsy Data: Example #1

Pneumonia deaths in FOCAS/Haiti project area: • 60% in urban slum, 40% rural areas (same as project distrib.).

• 30% of pneum. deaths were in the first month of life, 30% in 1-11m period, and 40% in the 12-59m period.

• About half of children who died from pneumonia were males.

• In general, children who died with pneumonia were not malnourished.

• Only 60% of mothers took the child to a health facility during the illness.

• High proportion of these mothers knew danger signs (gen'l and of pneumonia).

• Some Health Agents who worked with children were more highly trained than others, but children were no more likely to die if they were seen by a more-highly trained Agent than a lesser-trained Agent.

Analysis of Verbal Autopsy Data: Example #1

• Health Agents only saw 6 of the 10 children who died during the illness preceding the child’s death. (Mothers were supposed to bring their children to either the clinic or the Health Agent if they had signs of pneumonia.)

• Sometimes there were stock-outs of antibiotics because Health Agents would give away (to poor families) antibiotics that they were supposed to sell, and then not have money to buy enough antibiotics to replenish their boxes.

• An “ARI Case Management Form” was used to track CHW activities in terms of ARI. Looking at those forms for all children in the registers (not just those who died) we found:

Analysis of Verbal Autopsy Data: Example #1

• 19% of children with pneumonia, severe pneumonia, or severe disease were followed up within two days.

• 43% of children (3 of 7) who should have been referred during the initial contact were not referred.

• 33% of children (2 of 6) who were the same or worse at follow-up (according to what the agent marked on the form) were not referred.

• 40% of children whose respiration rate had increased were not referred.

• Over half (52%) either had follow-up at one month or later, or did not show follow-up documented on the form.

Analysis of Verbal Autopsy Data: Questions to Examine

Questions to Examine (small groups, 10-15 mins + report out):

• Given this information, what do you think the problems

are? On what improvements should the project focus?

• What specific things should be considered to prevent

similar deaths in the future there?

Haiti: Analysis of Verbal Autopsy Data: What were the Problems?

• Health Agents who saw the children were not following the

protocols.

• Many children who needed referral (initially and at follow-up)

were not referred, and

• very few (19%) received any follow-up visits in the home.

Important since sometimes the first-line antibiotic does not work

or the mother does not use it properly.

• The staff also need to explore why mothers are not taking

children to clinic even when the child has danger signs and they

know danger signs (e.g., through Barrier Analysis).

Haiti: Analysis of Verbal Autopsy Data: What were the Solutions?

• Follow-up with Health Agents more often to assure that they are referring children properly and giving high priority to follow-up of pneumonia cases.

• The CSP Manager started monitoring these forms more closely.

• Made changes to their forms to make it easier for Health Agents to know what they should do with a child with pneumonia.

• Gave Health Agents extra stocks of antibiotics to use with poor families who could not afford it.

Your Turn: Haiti Case History #2

• FOCAS Child Survival project in Haiti using

verbal autopsies (1997 – 2004)

• Operated in two project areas in Haiti.

• From 1999-2002, 31% reduction in the under

five mortality rate in first project area; 65%

reduction in second project area.

Haiti Case History #2

• 47% of deaths with VA form completed in 11m period were due to diarrhea and pneumonia.

• Focused on these deaths associated with diarrhea and pneumonia.

The Facts• Ten month period (Jan – Nov 2002). (All facts pertain to children

who died.)

• OBDC project area: Two deaths in 1-11m age group, 5 in 12-59m age group.

• MEI project area: 6 deaths, all among infants.

• Children who died were not malnourished.

• No relation to gender.

• Not related to transportation

• Most had diarrhea + vomiting

• Only ½ had been to a HF during illness.

• Most who went to HF received medications.

• Most mothers received counseling on ORS (either at home or clinic)

• Most saw a C-IMCI-trained Health Agent

• Many died day after Health Agent contact (before 2-day rtn visit).

What key facts would you focus on? Why? What might you do to improve the situation?

• (10 mins in small groups:)

• OBDC project area: Two deaths in 1-11m age group, 5 in 12-59m age group.

• MEI project area: 6 deaths, all among infants.

• Children who died were not malnourished.

• No relation to gender.

• Not related to transportation

• Most had diarrhea + vomiting

• Only ½ had been to a HF during illness.

• Most who went to HF received medications.

• Most mothers received counseling on ORS (either at home or clinic)

• Most saw a C-IMCI-trained Health Agent

• Many died on day after Health Agent contact (before 2-day rtn visit).

Report Out

• What key facts would you focus on?

• Why?

• What might you do to try to change the situation?

What was done: Changes to SBC and Protocols

• More emphasis and changed protocol on follow-up visits: Next-day

follow-up (not two day) of children with (1) diarrhea + signs of

dehydration or severe disease, (2) all children referred for treatment,

and (3) children with diarrhea + vomiting.

• Refer all children who had diarrhea + vomiting if child is unable to

tolerate liquids.

• Explain to every mother what to do when child vomits, and how to

give ORS after vomiting.

• CHW observes ORT if child vomited recently.

Also:• Worked on purification of water.

What was done: Changes to Administrative / M&E Procedures

• Developed a form to document case management,

referral, and follow-up of diarrheal cases.

• Made changes to Supervisor Monthly Report form

to report up case management.

• Added a section to the Program Manager Monthly

Report Form to report on the quality of these forms.

Remember to Examine Delays

Delay Delay was a problem

Delay was not a

problemTotal

Recognition that the child had a problem requiring medical attention

68% 32% 208

Seeking medical care 18% 82% 206

Reaching a health facility in a timely way 26% 74% 207

Receiving care once at health facility 7% 93% 208

Problems with follow-up care or home care after receiving medical care/advice 86% 14% 208

• What’s the main problems that people are having here?

• What could be done to decrease these types of delays?

“Taking off your Health Glasses”:Renew your Eyes

• Look for problems that the numbers don't show. Read the narratives!

• Bolivia Verbal Autopsy Case History #1: Wasted child

• Bolivia Verbal Autopsy Case History #2: Smothered child

Verbal Autopsies Can Save Lives

• They help you and your staff to better understand patterns of deaths.

• They help you and your staff to better understand system failures.

• They help you to uncover causes of deaths "hidden in the numbers."

• They help your staff to connect with families who have lost children and show you care.