Objective: To explore the usefulness of incidence-prevalence-mortality (IPM) models in improving estimates of disease epidemiology.
Methods: Two artificial and four empirical data sets (for breast, prostate, colorectal, and stomach cancer) were employed in IPM models.
Findings: The internally consistent artificial data sets could be reproduced virtually identically by the models. Our estimates often differed considerably from the empirical data sets, especially for breast and prostate cancer and for older ages. Only for stomach cancer did the estimates approximate to the data, except at older ages.
Conclusion: There is evidence that the discrepancies between model estimates and observations are caused both by data inaccuracies and past trends in incidence or mortality. Because IPM models cannot distinguish these effects, their use in improving disease estimates becomes complicated. Expert opinion is indispensable in assessing whether the use of these models improves data quality or, inappropriately, removes the effect of trends.