BACKGROUND Shortening behaviour during peritoneal dialysis training can be easily measured, and likened to the skipping behaviour in haemodialysis subjects, although its effect on peritoneal dialysis outcomes is now well understood. We studied the clinical impact of failing to adhere to a peritoneal dialysis training programme among incident dialysis patients. METHODS This study included 159 consecutive inception peritoneal dialysis patients in a single centre from September 1999 through November 2002. We evaluated the effects of behavioural compliance quantified by the per cent time arriving late for scheduled peritoneal dialysis training. The patients were categorized by whether they arrived late in >20% of their peritoneal dialysis training sessions. RESULTS Of the 159 incident peritoneal dialysis patients (mean age 57 +/- 13 years) who attended peritoneal dialysis training, 70 subjects (44%) arrived late in >20% of the sessions. They were younger by 5 years than patients who arrived late < or =20%. Mean peritonitis-free time for subjects who arrived late for training in >20% the of sessions was 30.9 months, as compared with 41.8 months in subjects with < or =20% late attendance behaviour (log rank test, P = 0.038). Multivariable Cox proportional hazards analysis showed that late attendance behaviour and baseline serum albumin were the only independent risk factors for the time to a first peritonitis after adjustment for diabetes mellitus and relevant coexisting medical factors. Late arrival in >20% of the peritoneal dialysis training sessions was associated with >50% increased likelihood of subsequent peritonitis, with an adjusted risk ratio of 1.56 (95% confidence interval, 1.02-2.39; P = 0.04). CONCLUSIONS These findings show that the behavioural measure of late attendance for peritoneal dialysis training has a crucial role in predicting peritonitis. It may therefore represent a practical strategy for identifying poor adherence or predicting medical outcomes.