Evaluation of suicide rates in rural India using verbal autopsies, 1994-9

BMJ 2003; 326 doi: http://dx.doi.org/10.1136/bmj.326.7399.1121 (Published 22 May 2003)
Cite this as: BMJ 2003;326:1121
  1. A Joseph, professor1 (chad{at}cmcvellore.ac.in),
  2. S Abraham, professor1,
  3. J P Muliyil, professor1,
  4. K George, professor1,
  5. J Prasad, lecturer1,
  6. S Minz, lecturer1,
  7. V J Abraham, lecturer1,
  8. K S Jacob, professor2
  1. 1 Department of Community Health, Christian Medical College, Vellore 632002, India
  2. 2 Department of Psychiatry, Christian Medical College, Vellore
  1. Correspondence to: A Joseph
  • Accepted 5 March 2003


Suicide rates have increased in many developing countries.1 But the reported rates are misleading because population counts are unreliable, and identifying suicides is problematic because of inefficient civil registration systems, non-reporting of deaths, variable standards in certifying death, and suicide's legal and social consequences.

Suicide rates were between 8.1 and 58.3/100 000 population for different parts of India.2 Police records, which under-report, were used to calculate these rates.

We used verbal autopsies in the 85 villages of the Kaniyambadi region of southern India (area 127 km2; population 108 873 in 1999) to calculate mean age and sex specific suicide rates for the period 1994-9.

Methods and results

A community health worker (a resident of the village), health aide, community nurse, and doctor reached a consensus on the cause of death. The community health worker, health aide, and nurse independently visited the home of the deceased and collected information from relatives and neighbours of the deceased, traditional healers, and village leaders. These health professionals discussed the circumstances of the death with the doctor. The doctor independently collected information from the different sources in the village in case of any doubts about cause of death.3 The system was evaluated by independent interviewers in 1994 and 1996-7; they concluded that the method does not over-report the number of suicides. We estimated population using census data (from 1994), which are updated regularly.

The mean suicide rate for the 6 year period was 95.2/100 000 (range 83.7-106.3/100 000) and did not change significantly over time (96.7, 106.3, 83.7, 103.6, 89.8, and 90.9 in each 100 000 population for each year 1994-9). Suicides accounted for between 8% (89/1057) and 12% (112/940) of total deaths. Suicides in women were 0.84 times as likely as suicides among men; this ratio did not change significantly over time.

Older men were more likely to commit suicide than younger men (table). Most women who committed suicide were aged 15–24 or older than 65. We found more suicides among women (102/278) than among men (58/331) in the 15–24 years age group (χ2 for linear trend = 15.5; P < 0.001).

View this table:

Age and sex specific suicide rates in the Kaniyambadi region of southern India, 1994-9

Poisoning (275/609; 45%) and hanging (248/609; 41%) were the commonest methods overall. A greater proportion of women chose drowning or burning (χ2 = 52.2; df = 1; P < 0.0001) than men, who more often chose poisoning or hanging. People younger than 44 years tended to use poison; older people tended to choose hanging (χ2 = 44.1; df = 18; P < 0.001). From 1994 until 1999, the proportion of suicide by poisoning, hanging, and drowning did not change significantly. The number of suicides by burning increased from 4 in 1994 to 11 in 1999 (χ2 for linear trend = 7.25; P = 0.007). No suti—a widow committing suicide by burning in her husband's funeral pyre—was recorded.


Verbal autopsies can give a good idea of the cause of death from suicide in the developing world, where coroners' verdicts are not available. A community health programme in the Kaniyambadi region of India found that recent studies in India have under-reported suicide rates by two to three times.2 The independently verified method used verbal autopsies and found the rate in 1994–9 was 95.2/100 000 population—nine times the national average. The high rates are not likely to be peculiar to Kaniyambadi; they reflect more accurate data collection. Sentinel centres that accurately monitor suicide are needed in the developing world.


We thank the staff of the Department of Community Health for the census, data collection, computerisation, and verbal autopsies.

Contributors: AJ, SA, JPM, and KG designed and monitored the project and contributed to writing the paper. JP, SM, and VJA monitored the programme, reviewed verbal autopsies, and helped write the paper. KSJ analysed the data and wrote the paper. AJ is guarantor.


  • Funding No additional funding.

  • Competing interests None declared.


  1. 1.
  2. 2.
  3. 3.