Follow-up Data

Annex F.2: Defining limits for Mortality comparison score Sangita Kulathinal1,2 and Kari Kuulasmaa1 for the MORGAM Project3

¹ MORGAM Data Centre, National Public Health Institute, Helsinki, Finland
² Since January 2007 at Indic Society for Education and Development (INSEED), Nashik, India
³ See Annex for the sites and key personnel of contributing MORGAM Centres

© National Institute for Health and Welfare and the MORGAM Project investigators Last updated: 4 July 2007 For more information, please contact Kari Kuulasmaa (firstname.lastname@thl.fi)

Contents

• Introduction
• Effect of randomness of cohort mortality and limited availability of official statistics
• Effect of survey non-response
• Limits for Mortality comparison score
• References
• Updates to this document

Introduction

Most of the MORGAM cohorts consist of the respondents of population surveys. Therefore, the comparison of mortality in the population, which is usually available from the official statistics, and the mortality in the cohorts provides an indicator of the coverage of the mortality follow-up. We aim here to define reasonable limits to the difference between the observed mortality in the cohorts and the mortality estimated from the population statistics.

Effect of randomness of cohort mortality and limited availability of official statistics

Because of approximations used in the estimation of the expected mortality (see Annex F.1) and random variation in the cohort mortality, we cannot expect that the cohort mortality is equal to the estimate from the population. Furthermore, the last years of the population mortality are often based on extrapolation. Therefore, it should be reasonable to assign a high Mortality comparison score if the cohort mortality is at least 90% of the population mortality. On the other hand, a cohort mortality less than 70% of the population mortality should indicate major concern.

Effect of survey non-response

An additional explanation to a difference between the two mortalities relates to survey non-response. The response rates in the surveys were generally between 65% and 80% (see Full descriptions of MORGAM cohorts). There is a fairly consistent finding that the the total mortality as well as cardiovascular mortality of the non-respondents of such surveys is about twice the mortality of the respondents (see e.g. [1]). Therefore, the number of deaths in the cohorts, estimated using population mortality data, is higher than the expected number of deaths in the cohorts, which consist of the survey respondents only. We estimate the effect of non-response on the comparison:

Let n, n_r, n_nr be the cohort size, number of the respondents and number of the non-respondents. Then n=n_r+n_nr. Let d, d_r, d_nr be the mortality rate for the whole cohort, for the respondents and for the non-respondents, respectively. Then nd=n_rd_r+n_nrd_nr. Assuming that d_nr=2d_r we get d_r=d/(2-n_r/n). For response rate 75% (=n_r/n), which is close to the average of the MORGAM cohorts, we get d_r=d/1.25.

Limits for Mortality comparison score

To correct the limits 90% and 70% above for the effect of non-response, we have to divide the limits by 1.25. This gives limits about 70% and 55%.

References

1. Jousilahti P, Salomaa V, Kuulasmaa K, Niemelä M, Vartiainen E. Total and cause specific mortality among participants and non-participants of population based health surveys: a comprehensive follow up of 54 372 Finnish men and women. J Epidemiol Community Health 2005;59:310-315.

Updates to this document