A Markov chain model to assess the efficacy of screening for non-insulin dependent diabetes mellitus (NIDDM)

Background The high prevalence and severe consequences of non-insulin depen dent diabetes mellitus (NIDDM) in Taiwan calls for urgent measures to detec t this disease in the asymptomatic phase. However, the efficacy of early de tection of NIDDM is highly dependent on its natural history from the diseas e-free state, through the asymptomatic to the symptomatic phase and death f rom NIDDM or other causes. Methods In order to project the above progression, a five-state illness-and -death Markov chain model was proposed to estimate these transition paramet ers using data from two rounds of a blood sugar screening programme for NID DM in Puli, in central Taiwan. Results showed that the annual incidence for asymptomatic NIDDM was 10.67 p er 1000 (95% CI : 8.26-13.79) and the average duration between the asymptom atic and symptomatic phases (the sojourn time) was 8 years (95% CI: 5.74-11 .29). The 10-year survival rate for asymptomatic NIDDM (79.35%) was better than that for symptomatic NIDDM (69.45%). Prediction of deaths from NIDDM w as performed to assess how the efficacy of screening for NIDDM varied by di fferent screening frequencies (annual, biennial, 4-yearly and the control g roup). Results indicated there is no substantial difference in mortality re duction from NIDDM among the annual, biennial and 4-yearly screening regime ns. However, a 4-yearly screening regimen significantly reduced deaths from NIDDM by 40% (95% CI: 26-51%). Conclusions A long sojourn time and the substantial reduction in mortality suggest that a 4-yearly screening regime for NIDDM would be most effective and feasible in Taiwan. The proposed five-state Markov chain model can be a pplied to other similar NIDDM screening projects.