Proposed is a multicomponent model for the estimation of light-emitting diode (LED) lumen maintenance using test data that were acquired in accordance with the test standards of the Illumination Engineering Society of North America, i.e., LM-80-08. Lumen maintenance data acquired with this test do not always follow exponential decay, particularly data collected in the first 1000 h or under low-stress (e.g., low temperature) conditions. This deviation from true exponential behavior makes it difficult to use the full data set in models for the estimation of lumen maintenance decay coefficient. As a result, critical information that is relevant to the early life or low-stress operation of LED light sources may be missed. We present an efficiency-decay model approach, where all lumen maintenance data can be used to provide an alternative estimate of the decay rate constant. The approach considers a combined model wherein one part describes an initial break-in period and another part describes the decay in lumen maintenance. During the break-in period, several mechanisms within the LED can act to produce a small (typically <; 10%) increase in luminous flux. The effect of the break-in period and its longevity is more likely to be present at low-ambient temperatures and currents, where the discrepancy between a standard TM-21 approach and our proposed model is the largest. For high temperatures and currents, the difference between the estimates becomes nonsubstantial. Our approach makes use of all the collected data and avoids producing unrealistic estimates of the decay coefficient.