Does declaration of competing interests affect readers' perceptions? A randomised trial.

line of the death certificate, including underlying and contributory conditions. We calculated age group specific annual mortality using midyear population estimates from the Office for National Statistics. MRSA was mentioned on 1387/6723 (20.6%) death certificates that included an ICD-9 code for staphylococcal infection (table). The percentage of certificates mentioning MRSA increased from 7.5% in 1993 to 25.0% in 1998. The final underlying causes of death indicated by death certificates that also mentioned MRSA included infections, neoplasms, and disease of nearly every system of the body. The number of certificates mentioning staphylococ-cal infection and the number of deaths with staphyloco-ccal infection as the underlying cause increased each year. Each year, a similar proportion of certificates mentioned staphylococcal infection as the underlying cause of death; in these certificates, the proportion mentioning MRSA increased from 8% in 1993 to 44% in 1998 (13/ 156 v 114/258). MRSA accounted for all of the increase in deaths due to staphylococcal infection in this period: MRSA in staphylococcal septicaemia increased from 3% to 28% (3/87 v 37/134), staphylococcal pneumonia from 13% to 44% (6/47 v 24/54), and unspecified bacterial infection, staphylococcus from 19% to 83% (4/21 v 53/64). In certificates mentioning MRSA where staphylo-coccal infection was the final underlying cause of death, mortality was higher in men and in older people. For 86% of the certificates, the age of the person who died was over 64. In 1998 mortality ranged from 0.4 per 100 000 for women aged 45-64 to 14.8 per 100 000 for men over 84. Comment Infections due to MRSA seem to be an increasing cause of mortality in England and Wales. Improved reporting is unlikely to explain the increase. The greatest rise in MRSA occurred for deaths in which invasive staphylococcal infection was given as the final underlying cause, so antimicrobial resistance probably influenced the success of medical management. Our study highlights the limitations of using routine mortality data for monitoring the impact of MRSA. There is no code for this infection in either ICD-9 or ICD-10. The Office for National Statistics could introduce routine automated searches of computerised text or assign one of the unused ICD-10 " U " codes available for special studies to MRSA. Further improvements in surveillance and control of healthcare associated infection and mortality should be a priority if MRSA related deaths are to be prevented. 4 Recent initiatives, such as the requirement since April 2001 for …