An incidence estimation model for multi‐stage diseases with differential mortality

Abstract

Prevalence and incidence are two important measures of the impact of a disease. For many diseases, incidence is the most useful measure for response planning. However, the longitudinal studies needed to calculate incidence are resource-intensive, so prevalence estimates are often more readily available. In 1986, Podgor and Leske (Statistics in Medicine, 5:573-578, 1986) developed a model to estimate incidence of a single disease from one survey of age-specific prevalence, even where the presence of the disease increases the mortality rate of patients. Here, we extend their model to the case of progressive diseases, where the incidence of all disease stages is desired. As an example, we consider the case of cataract disease in Africa, where ophthalmologists wish to distinguish between unilateral and bilateral cataract incidence in order to plan the number of cataract surgeries needed to prevent the occurrence of blindness as a result of the disease. Our method has successfully provided cataract incidence estimates on the basis of prevalence data from new Rapid Assessment of Avoidable Blindness surveys in Africa (Lewallen et al., Archives of Ophthalmology, 128(12):1584-1589, 2010). In this paper, we provide a more general form of the model in order to promote its applicability to other diseases.

Incidence estimation using a single cross-sectional age-specific prevalence survey with differential mortality

“…Therefore, they are subject to higher mortality compared with people who have never had a cataract of the same age, even if they have had the cataract removed by surgery . (Such an assumption was also made by Dray and Williams .) Work in progress by us (E. L. T. and R. L.) suggests that elevation in those with cataract or treated with surgery is approximately 20%.…”

“…Their work was extended by Podgor and Leske to account for differential mortality between diseased and non‐diseased, this time in the case of cataract, where the authors did not explicitly include a cured state. Dray and Williams further developed that work to allow for multi‐stages of disease with the example of two stages in the case of cataract (i.e. cataract in one eye vs. cataract in both eyes).…”

“…We demonstrate why the distinction between the diseased and cured can be ignored for incidence estimation but that it must be accounted for, together with the competing risk of death, when predicting future disease burden. As in , our method exploits important characteristics of many diseases such as cataract: incidence/prevalence varies with age (typically increases), and differential mortality (often elevated in the presence of disease). (See for a statistical perspective on age‐specific incidence and prevalence.)…”

Task shifting for cataract surgery in eastern Africa: productivity and attrition of non-physician cataract surgeons in Kenya, Malawi and Tanzania

“…Based on recent estimates of cataract prevalence, District level hospitals may not serve a large enough population to have adequate cases, if surgery is limited to eyes with a vision <3/60 or <6/60 [2,3]. If Ministries of Health place NPCSs in these settings, productivity will likely remain low and national targets for cataract surgery will not be achieved.…”

“…For individual countries, the known prevalence of vision loss due to cataract can be used to estimate the number of cataract operations (incident cases) needed to eliminate blindness or visual impairment due to cataract [2,3]. Achieving this level of productivity and improved quality of outcomes [4] would eliminate blindness due to cataract [5] and result in achievement of the International Agency for the Prevention of Blindness, VISION 2020 goals.…”