Background: Years of life lost (YLL) is a measure that summarizes the global impact of the disease in population in terms of premature deaths. The YLL is classically estimated using age of death as a starting point. In the electronic health records, as the measure of mortality is only available on short periods, alternative estimations using the age of diagnosis as a starting point and extrapolation on survival on longer follow-up than the one observed can be computed.
Objectives: To estimate the number of years of life lost among incident patients with multiple myeloma (MM).
Methods: Data from a retrospective observational cohort study of MM patients identified through the French NHI databases (SNDS) from 2014 to 2019 was used. The YLL was estimated using the classical method i.e., the weighted sum of life expectancy (LE) in the general population at the age of death among dead MM patients (method 1). Two other methods were implemented: the weighted sum of the difference of LE at the age of diagnosis in the general population obtained from life tables and the observed LE with a parametric gamma model adjusted on age and sex (method 2) or with linear extrapolation of a logit-transformed curve of the survival ratio between simulated general control population from life tables and MM study population (method 3).
Results: A total of 26,357 incident patients were diagnosed with MM between 2014 and 2019 (mean age:69.7 years) including 10,563 (40.1%) who died during follow-up (mean age:76.3 years). The mean follow-up duration was 2.2 years with a high censoring rate (46.9% of patients left after 2 years). The overall mean number of YLL was estimated at 13.3 years using method 1, 10.9 years using method 2 and, 12.3 years using method 3.
Conclusions: The mean number of YLL for MM patients varies between 10.9 and 13.3 years according to the method of estimation. The method 1 is widely used in the literature due to its simplicity and allows comparisons with other reference studies however it applies only to deceased patients and tends to overestimate YLL. In electronic health records and in general with heavily censored data, alternative estimations relying on hypotheses on survival among diseased patients will provide more accurate estimations by considering the distribution of ages of diagnosis for a given disease.
