Use of natural language processing in electronic medical records to identify pregnant women with suicidal behavior: towards a solution to the complex classification problem.

We developed algorithms to identify pregnant women with suicidal behavior using information extracted from clinical notes by natural language processing (NLP) in electronic medical records. Using both codified data and NLP applied to unstructured clinical notes, we first screened pregnant women in Partners HealthCare for suicidal behavior. Psychiatrists manually reviewed clinical charts to identify relevant features for suicidal behavior and to obtain gold-standard labels. Using the adaptive elastic net, we developed algorithms to classify suicidal behavior. We then validated algorithms in an independent validation dataset. From 275,843 women with codes related to pregnancy or delivery, 9331 women screened positive for suicidal behavior by either codified data (N = 196) or NLP (N = 9,145). Using expert-curated features, our algorithm achieved an area under the curve of 0.83. By setting a positive predictive value comparable to that of diagnostic codes related to suicidal behavior (0.71), we obtained a sensitivity of 0.34, specificity of 0.96, and negative predictive value of 0.83. The algorithm identified 1423 pregnant women with suicidal behavior among 9331 women screened positive. Mining unstructured clinical notes using NLP resulted in a 11-fold increase in the number of pregnant women identified with suicidal behavior, as compared to solely reliance on diagnostic codes.

The effect of experimentally induced sleep fragmentation and estradiol suppression on neurobehavioral performance and subjective sleepiness in premenopausal women.

STUDY OBJECTIVES: Menopause is associated with nighttime sleep fragmentation, declining estradiol, and impaired cognition. In a model of pharmacologically induced estradiol suppression mimicking menopause, we examined the impact of menopause-pattern sleep fragmentation on daytime neurobehavioral performance and sleepiness in premenopausal women. METHODS: Twenty premenopausal women completed two five-night inpatient studies in the mid-to-late follicular phase (estrogenized) and after pharmacological estradiol suppression (hypo-estrogenized). During each study, participants had an uninterrupted 8-hour sleep opportunity for two nights, followed by three nights where sleep was experimentally fragmented to mimic menopause-pattern sleep disturbance, and during which the sleep opportunity was extended to prevent shortening of the sleep duration. Neurobehavioral performance and subjective sleepiness were measured using the Psychomotor Vigilance Task and Karolinska Sleepiness Scale (KSS). RESULTS: Compared to unfragmented sleep, sleep fragmentation increased attentional lapses (+ 0.6 lapses, p < .05), slowed reaction time (+ 9.4 milliseconds, p < .01), and increased daytime sleepiness (+ 0.5 KSS score, p < .001). Estradiol suppression increased attentional lapses (+ 0.8; p < .001) and reaction time (+ 12.3, p < .01) but did not significantly affect daytime sleepiness. The effect of sleep fragmentation on neurobehavioral performance differed by estradiol state, such that the adverse effects of sleep fragmentation on attentional lapses (+ 0.9, trend p = .06) and reaction time (+ 15, p < .05) were observed only when estrogenized. CONCLUSIONS: Menopause-pattern sleep fragmentation and estradiol suppression worsened neurobehavioral performance and daytime sleepiness, even while sleep duration was not reduced. The adverse effects of sleep fragmentation in the context of an adequate sleep duration highlight the importance of sleep continuity as a vital aspect of good sleep health.

Effects of Sleep Fragmentation and Estradiol Decline on Cortisol in a Human Experimental Model of Menopause.

CONTEXT: Perturbations to the hypothalamic-pituitary-adrenal (HPA) axis have been hypothesized to increase postmenopausal cardiometabolic risk. Although sleep disturbance, a known risk factor for cardiometabolic disease, is prevalent during the menopause transition, it is unknown whether menopause-related sleep disturbance and estradiol decline disturb the HPA axis. OBJECTIVE: We examined the effect of experimental fragmentation of sleep and suppression of estradiol as a model of menopause on cortisol levels in healthy young women. METHODS: Twenty-two women completed a 5-night inpatient study during the mid-to-late follicular phase (estrogenized). A subset (n = 14) repeated the protocol after gonadotropin-releasing hormone agonist-induced estradiol suppression. Each inpatient study included 2 unfragmented sleep nights followed by 3 experimental sleep fragmentation nights. This study took place with premenopausal women at an academic medical center. Interventions included sleep fragmentation and pharmacological hypoestrogenism, and main outcome measures were serum bedtime cortisol levels and cortisol awakening response (CAR). RESULTS: Bedtime cortisol increased 27% (P = .03) and CAR decreased 57% (P = .01) following sleep fragmentation compared to unfragmented sleep. Polysomnographic-derived wake after sleep-onset (WASO) was positively associated with bedtime cortisol levels (P = .047) and negatively associated with CAR (P < .01). Bedtime cortisol levels were 22% lower in the hypoestrogenized state compared to the estrogenized state (P = .02), while CAR was similar in both estradiol conditions (P = .38). CONCLUSION: Estradiol suppression and modifiable menopause-related sleep fragmentation both independently perturb HPA axis activity. Sleep fragmentation, commonly seen in menopausal women, may disrupt the HPA axis, which in turn may lead to adverse health effects as women age.

A double-blind, randomized, placebo-controlled trial of suvorexant for the treatment of vasomotor symptom-associated insomnia disorder in midlife women.

STUDY OBJECTIVES: The neuropeptide orexin promotes wakefulness, modulates thermoregulation, increases after menopause, and is normalized in women receiving estrogen therapy, suggesting a role for orexin antagonism as a treatment for the vasomotor symptom (VMS)-associated insomnia disorder. We tested the efficacy of the dual orexin receptor antagonist suvorexant for chronic insomnia related to nighttime VMS. METHODS: In a double-blind, placebo-controlled trial, 56 women with chronic insomnia associated with nighttime VMS, Insomnia Severity Index (ISI) scores ≥15, and >30 min of diary-rated wake after sleep-onset (WASO) were randomized to receive oral suvorexant 10-20 mg (n = 27) or placebo (n = 29) nightly for 4 weeks. Analysis of within-person change in ISI was adjusted for baseline ISI and race. RESULTS: Mean baseline ISI scores were 18.1 (95% CI, 16.8 to 19.4) and 18.3 (95% CI, 17.2 to 19.5) in the suvorexant and placebo groups, respectively (p = .81). The average 4-week ISI within-person decrease from baseline was greater on suvorexant (-8.1 [95% CI, -10.2 to -6.0]) compared to placebo (-5.6 [95% CI, -7.4 to -3.9], p = .04). Compared to placebo, nighttime diary-rated VMS frequency was significantly reduced with suvorexant (p < .01). While diary-rated WASO and total sleep time trended toward improvement on suvorexant, findings were not significant after adjustment for multiple comparisons. Daytime VMS and other sleep-related outcomes did not differ between groups. Suvorexant was well tolerated. CONCLUSION: These results suggest that suvorexant is likely a well-tolerated and efficacious treatment for VMS-associated insomnia disorder and reduces nighttime VMS. Antagonism of orexin receptors could provide a novel therapeutic option for midlife women with VMS-associated chronic insomnia. CLINICAL TRIAL INFORMATION: Efficacy of Suvorexant in the Treatment of Hot Flash-associated Insomnia, https://clinicaltrials.gov/ct2/show/NCT03034018, ClinicalTrials.gov Identifier: NCT03034018.

Sleep Fragmentation and Estradiol Suppression Decrease Fat Oxidation in Premenopausal Women.

CONTEXT: Body fat gain associated with menopause has been attributed to estradiol (E2) withdrawal. Hypoestrogenism is unlikely to be the only contributing factor, however. OBJECTIVE: Given the links between sleep and metabolic health, we examined the effects of an experimental menopausal model of sleep fragmentation on energy metabolism. METHODS: Twenty premenopausal women (age 21-45 years) underwent a 5-night inpatient study during the mid-to-late follicular phase (estrogenized; n = 20) and the same protocol was repeated in a subset of the participants (n = 9) following leuprolide-induced E2 suppression (hypo-estrogenized). During each 5-night study, there were 2 nights of unfragmented sleep followed by 3 nights of fragmented sleep. Indirect calorimetry was used to assess fasted resting energy expenditure (REE) and substrate oxidation. RESULTS: Sleep fragmentation in the estrogenized state increased the respiratory exchange ratio (RER) and carbohydrate oxidation while decreasing fat oxidation (all P < 0.01). Similarly, in the hypo-estrogenized state without sleep fragmentation, RER and carbohydrate oxidation increased and fat oxidation decreased (all P < 0.01); addition of sleep fragmentation to the hypo-estrogenized state did not produce further effects beyond that observed for either intervention alone (P < 0.05). There were no effects of either sleep fragmentation or E2 state on REE. CONCLUSION: Sleep fragmentation and hypoestrogenism each independently alter fasting substrate oxidation in a manner that may contribute to body fat gain. These findings are important for understanding mechanisms underlying propensity to body fat gain in women across the menopause transition.

Hypothalamic-pituitary-adrenal axis, subjective, and thermal stress responses in midlife women with vasomotor symptoms.

OBJECTIVE: Dysregulated responses to experimental stress paradigms may indicate exposure to chronic stress. Vasomotor symptoms (VMS) are linked with diminished quality of life and psychological stress, but induced stress responsivity has received limited investigation. We examined whether women with and without VMS differ in their evoked hypothalamic-pituitary-adrenal axis, subjective, hemodynamic, and thermal stress responses. METHODS: A total of 37 midlife women (27 VMS+; 10 VMS-) completed 2 experimental stress paradigms: (1) Montreal Imaging Stress Task (MIST; computerized social-evaluative stressor) and (2) Quantitative Sensory Testing (QST; thermal stress task). Responses on a five-domain (range 0-50) Visual Analog Scale, salivary cortisol (hypothalamic-pituitary-adrenal axis), and hemodynamic indices (blood pressure, heart rate) were measured before and after each task to compare within-person change between groups. Thermal sensitivity was assessed on the QST. RESULTS: On the MIST, the VMS+ group showed a smaller cortisol release (0.01 vs 0.07 µg/dL; P = 0.046; corresponding to 54% vs 83% increases), and subjective stress response (21.2- vs 31.1-point Visual Analog Scale increase, P = 0.05; corresponding to 2427% vs 2863% increases) but no hemodynamic difference, compared to the VMS- group. The QST did not provoke stress responses via cortisol release or subjective report, but the VMS+ group tended to perceive heat at a higher temperature (38.5°C vs 36.4°C, P = 0.08). CONCLUSIONS: Women with VMS exhibited both diminished cortisol and subjective stress responses to the MIST, and reduced thermal sensitivity on QST compared to women without VMS. Dysregulated stress responsivity provides preliminary evidence suggesting that VMS may represent a chronic stress condition.