Labeling self-tracked menstrual health records with hidden semi-Markov models

Abstract

Globally, millions of women track their menstrual cycle and fertility via smartphone-based health apps, generating multivariate time series with frequent missing data. To leverage data from self-tracking tools in epidemiological studies on fertility or the menstrual cycle’s effects on diseases and symptoms, it is critical to have methods for identifying reproductive events, e.g. ovulation, pregnancy losses or births. We present two coupled hidden semi-Markov models that adapt to changes in tracking behavior, explicitly capture variable– and state– dependent missingness, allow for variables of different type, and quantify uncertainty. The accuracy on synthetic data reaches 98% with no missing data, 90% with realistic missingness, and 94% accuracy on our partially labeled real-world time series. Our method also accurately predicts cycle length by learning user characteristics. It is publicly available (HiddenSemiMarkov R package) and transferable to any health time series, including self-reported symptoms and occasional tests.

Tracking Sleep, Temperature, Heart Rate, and Daily Symptoms Across the Menstrual Cycle with the Oura Ring in Healthy Women

“…There is an emerging body of literature on the use of technology to investigate the menstrual cycle. 39–42 Some studies have used research-grade actigraphy, but to our knowledge, no studies published in peer-reviewed journals have used a wearable sleep tracker, to study sleep across the menstrual cycle. Only one research work, presented as conference proceedings, studied the relation between sleep and the menstrual cycle using a wrist-wearable device.…”

Sub-communities of the vaginal microbiota in pregnant and non-pregnant women

“…Menstrual cycles were identified from bleeding flows reported by participants on a scale from 0 (none) to 3 (heavy). Specifically, a hidden semi-Markov model was specified to account for empirically observed distribution of cycle length and bleeding patterns across the menstrual cycle (54). Data of participants who reported too few days with bleeding ( i .…”

“…Menstrual cycles were identified from bleeding flows reported daily by participants on a scale from 0 (none) to 3 (heavy). A hidden semi-Markov model was specified to account for empirically observed distributions of cycle length and bleeding patterns across the menstrual cycle, including spotting between menses (55). Data of participants who reported too few days with bleeding (i.e., less than 3/70 study days) or too many (i.e., more than 30/70 study days) were excluded from the menstrual cycle analyses.…”