Premature mortality is a major contributor to disease burden. According to the GBD estimates (Murray and Lopez, 1996) the YLL component of DALYs was about 50% in established market economies, where the epidemiological transition had already taken place. In former socialist economies, China, Latin America and the Caribbean (LAC), 57%-58% of disease burden was due to premature mortality. In India, Middle Eastern Crescent (MEC), the rest of Asia and islands (OAI), YLLs contributed 65-69% of total DALYs lost in 1990. In sub - Saharan Africa, more than 75% of DALYs was contributed by YLLs. In case of India 84% of the disease burden among children was estimated to be on account of premature mortality. A little more than half of DALYs, lost among older children and young adults, was due to premature mortality. For older people, the contribution of premature mortality to disease burden was 66% or higher. Hence the causes of these deaths is important for health policy.

Since the major share of disease burden would be from premature mortality, accurate estimate of causes of death would constrain the accuracy of burden of disease estimates. Here we describe, for rural and urban areas of AP, respectively, the existing statistical base for cause of death reporting. We present results from a study undertaken in Andhra Pradesh to improve the accuracy of cause of death data from the rural areas. Our study on causes of death in urban areas of AP is currently underway. Mean while, we have used medically certified cause of death statistics from the neighbouring state of Maharashtra, where coverage of the medical certification of cause of death is better.

At the national level, the Registrar General of India (RGI) is responsible for collection, collation and publication of cause of death statistics. At the state level, the Vital Statistics Division of the Directorate of Health deals with cause of death statistics. Cause of death reports originate from lay reporters in rural areas and medical attendants in urban areas. The reports reach the State Vital Statistics office through the primary health centre, in case of rural areas, and the municipal health office for urban areas. Tabulation is usually done at the state level but the statistics are published by the RGI. Until December 1998, cause of death data for the rural areas used to be collected under the Survey of Cause of Death Rural (SCD-Rural) scheme, from a sample of villages by a lay diagnosis and reporting system. A paramedical person from the PHC is designated as the field agent who undertakes the primary survey. (S)he identifies key informants and maintains liaison with them. A household register is drawn up and updated on a half yearly basis. For each death occurring in the village, the field agent identifies one or more persons having knowledge of the circumstances of death, interviews them and records the symptoms and circumstances of death in Form-7. A structured questionnaire is used to investigate cause of death using the symptoms and circumstances of death. The structured questionnaire is supplemented by a check list. The field agent arrives at a probable cause of death by applying the structured questionnaire to symptoms and circumstances recorded in Form-7. The check list entry against the probable cause of death is tallied with the symptoms and circumstances of death. The cause of death thus arrived is reported in Form-3. The PHC statistician is designated as the recorder of events reported by the field agent. Half-yearly verification of the household list is done by the recorder. Medical officer of the PHC is expected to check and certify the correctness of cause of death assignment by the field agent. Assignment of cause of death is done by the field agent based on a structured interview with a member of concerned household. The structured questionnaire currently in use was adopted after taking into account five years of field experience with a provisional questionnaire. The non medical list (NML) of causes of death was last revised in 1983 to correspond to ICD ninth revision (RGI, 1991). SCD-Rural used verbal autopsy (VA) to arrive at cause of deaths using paramedical personnel.

From January 1999 a cause of death component has been added to the SRS (RGI, 1999). We call this the SRS-COD component. Two more columns have been added to SRS Form-5 (Columns 16-17) and Form-10 (columns 12-13). The SRS part-time enumerator (PTE) records cause of death in column 16 and the code in column 17 of the revised Form-5. The SRS supervisor records similar information in columns 12 and 13 of the revised Form-10. A major departure from the SCD-Rural design is doing away with the symptom record (SCD-Rural Form-7). Another departure from the SCD-Rural is the elimination of the structured questionnaire. Instead the instructions contain a list of causes, related symptoms for some, and the corresponding ICD-10 code.

   Top 

In case of the urban areas, a medical certification of cause of death (MCCD) scheme is operational. This scheme has been acceded legal sanction under the Registration of Births and Deaths (RBD) Act. All medically attended deaths are expected to be registered (Form-2) along with cause of death reports in a format (Form-4) which is similar to what is prescribed by the WHO for International Classification of Cause of Death (ICD). The responsibility for reporting cause of death rests with the doctor / health care provider who last attended on the deceased. Reports are sent to the municipal health authorities, who forward them to the concerned state vital statistics office. The medical attendant is required to follow guidelines contained in the Physician's manual on medical certification of cause of death (RGI, 1992). This manual prescribed the WHO form for reporting cause of death according to the current version of ICD. Coding and tabulation is done according to the National List which is an adaptation of the ICD basic tabulation list. Since the MCCD essentially implements ICD coding and guidelines, the design of the system is considered satisfactory.

Ruzicka and Lopez (1990) have listed five criteria used by the World Health Organisation to assess fitness of country-level cause of death data for inclusion in its compilations. Firstly, the proportion of all deaths attributed to residual categories such as "Symptoms, signs and ill defined conditions" is within limits, say less than 10%. Secondly, the proportionate distribution of deaths by cause is consistent with the estimated mortality level for that country. Thirdly, no cause of death with a clear age-sex dependency has been incorrectly assigned. Fourthly, the age-sex distribution for major causes is consistent with what one may expect for each cause. Finally, data generated by the system are consistent with previous years. Note that these are basically plausibility checks. A data set failing these criteria is more likely to be biased. A data set satisfying these criteria may still not be usable, on account of poor statistical power of the generated estimates, and biases that are not readily noticeable. Building upon the criteria suggested by Ruzicka and Lopez (1990), we have identified the following nine criteria to assess the usability of any cause of death statistics:

Top 

We examine, below, the usability of the cause of death statistics in India from the rural and urban areas respectively. We take up each usability criteria, discuss its implications briefly and then examine how India's cause of death statistics fares, using national statistics and state level statistics from Andhra Pradesh. Where required, we supplement the published statistics with information about Andhra Pradesh, available to us from our study on cause of death in Andhra Pradesh. We have called this the Andhra Pradesh Rural Cause of Death (APRCD) study, 1998.

Design characteristics of the reporting system have a bearing on the usability of cause of death statistics. For example, changes in the guidelines of the international classification of causes of death (ICD) have been seen to cause reduction or increase in assignment of deaths to certain causes, depending on the specific changes brought by the particular version of the ICD. Analogously, guidelines for verbal autopsy can have some effect on the cause of death structure produced by the concerned cause of death reporting system.

  Top  

Next