A Standard Verbal Autopsy Method for Investigating Causes of Death in Infants and Children

Name: A Standard Verbal Autopsy Method for Investigating Causes of Death in Infants and Children

Purpose: Verbal Autopsies are designed to improve the measurement of cause-specific mortality in areas where medical certification of cause of death is rare.

Type of Design: Descriptive, cross sectional community-based survey.

Where Used: Validation studies were undertaken in Bangladesh, Nicaragua and Uganda.

Mortality Survey in Bolivia: The Final Report
Investigating and Identifying the Causes of Death for Children Under Five


BASIC INFORMATION

Name: A Standard Verbal Autopsy Method for Investigating Causes of Death in Infants and Children

Origin/Source: World Health Organization (WHO) [WHO/CDS/CSR/ISR/99.4]

Johns Hopkins School of Hygiene and Public Health

London School Hygiene and Tropical Health

Technical Area: Newborn Health

Basic Description: A standardized method of determining causes of deaths among infants and children based on semi structured interviews with the mother, next of kin or other caregivers. Specific algorithms for coding common causes of death in infants and children have been developed and tested for validity in Bangladesh, Nicaragua and Uganda. An underlying assumption of the method is that each cause of death investigated has a set of observable features that can be recalled during a verbal autopsy.

Country Applications: Validation studies have been undertaken in Bangladesh, Nicaragua and Uganda.

Language available: English.

The standard questionnaire should be translated into the local language by someone who is fluent both in English and in the local language, and who is familiar with the local lay medical terminology. A local anthropologist or a demographer should check the translation for accuracy of intended meaning. Back translation should also be done and compared to the original questionnaire to make sure that the intent of each question reflects the original medical meaning.

Alternative local terms for medical conditions, symptoms and signs of illness should be considered both during the translation phase and also during the interviewing training. Local terms can be printed directly on the questionnaire or inserted by the interviewer.

Purpose: Evaluation

Technical Scope / Purpose: Management and Assessment of Childhood Illness (infants and children); Measurement of cause-specific mortality.

Design: Descriptive, settings in which this method can be used are varried, is flexible ,ranging from cross sectional community-based surveys .to demographic surveillance sites-, to other settings in which deaths are identified and there is a need to know the cause of death.

Method: Qualitative and quantitative

Key Users of Information: Researchers, policy-makers and programme managers. for assessing impact of programme interventions.

Objectives and Scope of the Tool:

This tool is designed to improve measurement of cause-specific mortality in areas where medical certification of cause of death is rare.

The specific objectives are:

  1. To provide data on mortality by cause.
  2. To evaluate health interventions aimed at reducing mortality from specific causes of death, when these interventions are introduce into a small geographic on a trial basis.
  3. To identify ways to reduce unnecessary deaths. This can be achieve combining a verbal autopsy questionnaire with a household questionnaire asking about steps taken by the family and by the health services during the illness preceding death can make it possible to identify problems relating both to health seeking behavior and health service provision
  4. To facilitate research into factors associated with mortality from specific causes of death.

The standard verbal autopsy questionnaire uses a combination of open history questions followed by a series of closed ended questions that ask about specific signs and symptoms prior to death. The questionnaire is coded according to predetermined algorithms so that standardized causes of deaths can be derived.

Key Indicators: Many verbal autopsy questionnaires allow for only one cause of death. However, the standard verbal autopsy questionnaire for children (in the present document) allows for multiple causes of death. This is important since many children who die from a combination of causes would have survived if any of the causes had been prevented or treated early enough.

Specifically, information is collected on the major causes of childhood death.

These are:

  1. Acute Respiratory Infection (ARI); Pneumonia 8. Meningitis/Encephalitis
  2. Malaria
  3. LAccidents
  4. Sepsis
  5. Neonatal tetanus
  6. Measles;
  7. Diarrhea and dysentery.
  8. Meningitis/Encephalitis
  9. Low birth weight
  10. Birth asphyxia
  11. Premature births
  12. Congenital abnormality
  13. Birth trauma
  14. Local bacterial infection

Separate Algorithms are presented for causes of deaths in neonates and post neonates.

Tool Design:

The verbal autopsy is designed to be used in a variety of settings where medical certification of deaths is rare. The questionnaire needs to be adapted carefully to each setting. It is important to note that.

  1. Verbal autopsies work best at identifying causes of deaths with distinctive features not found in other causes of deaths. Injuries fit this requirement well.
  2. The sensitivity and specificity of the algorithms will vary from setting to setting and by cause of death. In previous studies the sensitivity and specificity for neonatal tetanus, measles and malnutrition and accidents are generally higher than for diarrhea and ALRI . These findings are also true of the standardized verbal autopsy validation studies.
  3. Verbal autopsy may be subject to relatively high misclassification errors and this may have a very large effect on the proportion of deaths due to a specific cause. Whilst investigators should be aware of these errors it is not possible to correct for these errors without doing a local validation study.
  4. Tables (as shown in the tool on page 10) that show the differences between the verbal autopsy estimate and the true cause-specific mortality fraction, for different levels of sensitivity and specificity, are useful when evaluating the feasibility of using a verbal autopsy is a particular setting.

 

Implementation and Training:

When adapting a verbal autopsy questionnaire to a local setting, it is important to have prior knowledge about the community, the health care system, common diseases and causes of death. Some false-positives can result from the verbal autopsy algorithms even if the condition is not present in a population.

Translation and back translation is an important part in adapting a questionnaire into the local language (please, see above, "translation issues").

Interviewers, supervisors and trainers should agree on the final draft of the questionnaire. Then, the interviewers should pre-test it by interviewing several caregivers of severely ill or recently deceased children, in a hospital or nearby homes. Problems must be resolved before taking the final questionnaire to the final to the field.

Lessons from Experience:

  1. Especially for diarrhea and pneumonia, symptoms are sometimes described in slightly different ways in the local languages. For example, local terms describing the stools and the local term for diarrhea are not necessarily identical, as their exact meanings vary with the local language.
  2. For pneumonia, the medical cause of death/severe illness is communicated to the caregiver at the time of death or discharge from hospital. This influences caregivers subsequent responses and consequently, it is quite likely that the current validation studies described in the tool may have over-estimated the sensitivities for the local term for pneumonia.
  3. Finally, sometimes, questions are not properly understood and mothers might be loosing concentration at a certain point of the interview. In future questionnaires it would be useful to add an introductory remark before a certain key question (such as the one about the size of the baby, while trying to assess malnutrition and premature birth).

Final Remarks

The current document contains a questionnaire, algorithms for coding specific causes of death, and the results of the validation studies in terms of the sensitivity and specificity of each algorithm for each validation study.

The underlying mix of causes of death in the population affects the accuracy of the verbal autopsy. This implies that accuracy of the estimates will vary according to the population under study. For example, it has been suggested that verbal autopsies may be better able to identify acute lower respiratory infections (ALRI) in malaria free areas than in areas in which malaria is a common cause of childhood deaths.

Cultural aspects may also influence the accuracy of verbal autopsy. The willingness to accept a verbal autopsy interview, the ability of the respondent to describe the final illness and the way the diseases are understood and described in the community will vary from culture to culture.

Contact Person: Martha ANKER, E-Mail: ankerm@who.ch
World Health Organization
Division of Emerging and other Communicable Diseases Surveillance & Control
20, Avenue Appia, 1211 Geneva 27
Switzerland
Tel: 41 22 791 2380/4551
Fax: 41 22 791 4198