Predicting clinically meaningful changes in behavioral health symptomatology using patient voice inputs
One example method for predicting clinically meaningful changes in behavioral health symptomatology using patient voice inputs includes receiving a baseline behavioral sample indicating an initial behavioral health symptomatology of an individual; receiving a subsequent behavioral sample including a supplemental voice recording from the individual; computing a difference score from a trained machine-learning (“ML”) model by providing the baseline behavioral sample and the subsequent behavioral sample as input to the trained ML model wherein the difference score indicates a predicted change in behavioral health symptomatology of the individual, the trained ML model being configured to output a score indicating a predicted change in behavioral health symptomatology based on input data; and outputting the predicted change in behavioral health symptomatology to a client device executing a graphical user interface indicating the change for an entity treating the individual.
1 . A method comprising:
receiving a baseline behavioral sample indicating an initial behavioral health symptomatology of an individual, the baseline behavioral sample comprising (i) an audio recording, (ii) a video recording, or (iii) a multimedia recording, and the initial behavioral health symptomatology indicative of a behavioral health disorder;
receiving a subsequent behavioral sample including a supplemental voice recording from the individual, the supplemental voice recording received after the baseline behavioral sample and comprising the same or different content as the baseline behavioral sample;
computing a difference score from a trained machine-learning (“ML”) model by providing the baseline behavioral sample and the subsequent behavioral sample as input to the trained ML model, wherein the difference score indicates a predicted change in behavioral health symptomatology of the individual based on determined changes between the baseline behavioral sample and the subsequent behavioral sample, the trained ML model being configured to output a score indicating a predicted change in behavioral health symptomatology based on input data;
outputting the predicted change in behavioral health symptomatology to a client device executing a graphical user interface indicating the predicted change for an entity treating the individual;
determining whether the predicted change in behavioral health symptomatology satisfies a first threshold of a plurality of thresholds; and
in response to determining that the predicted change in behavioral health symptomatology satisfies the first threshold, provide a notification to the individual to perform a first intervention.
2 . The method of claim 1 , wherein:
the baseline behavioral sample comprises a baseline voice recording and a baseline assessment score, wherein the baseline assessment score is derived from a clinically validated assessment in behavioral health; and
the predicted change in behavioral health symptomatology indicates a predicted change in behavioral health symptomatology from the baseline assessment score in accordance with the clinically validated assessment.
3 . The method of claim 2 , wherein the clinically validated assessment includes multiple questions, each question capable of generating an independent score based on a self-reported answer wherein the trained ML model comprises a plurality of trained ML models, each trained ML model is trained to assess a predicted change in a subset composite score of a subset of questions on a clinical assessment test used to generate the baseline assessment score, and further comprising:
providing the baseline behavioral sample and the subsequent behavioral sample as input to each trained ML model of the plurality of trained ML models; and
outputting, for each trained ML model of the plurality of trained ML models, the predicted change in a subset composite score of the respective subset of questions on the assessment.
4 . The method of claim 1 , further comprising:
determining that the baseline behavioral sample requires recalibration based on an environmental condition or a completion of a corresponding biological cycle; and
in response to determining that the baseline behavioral sample requires recalibration, generating a request to update the baseline behavioral sample at a digital device of the individual.
5 . The method of claim 1 , wherein receiving the baseline behavioral sample comprises receiving a plurality of baseline behavioral samples, each baseline behavioral sample corresponding to a point in time along a biological cycle; and wherein computing the difference score includes:
selecting a target baseline behavioral sample among the plurality of baseline behavioral samples based on a timestamp of the subsequent behavioral sample relative to the respective point in time along the biological cycle of the target baseline behavioral sample, and
providing the target baseline behavioral sample and the subsequent behavioral sample as input to the trained ML model.
6 . The method of claim 1 , further comprising generating the trained ML model by training a model using pairs of behavioral samples at two different time points of a population of individuals, wherein such population does not include the individual of claim 1 , and wherein each pair of behavioral samples includes a pair of voice recordings and a pair of assessment scores of a clinically validated assessment respectively at the two different time points.
7 . The method of claim 1 , wherein the difference score and the predicted change in behavioral health symptomatology represent a personalized screening of the individual and wherein the trained ML model is trained without personal data of the individual.
8 . A system comprising:
one or more processors; and
a non-transitory computer-readable medium comprising program code that is executable by the one or more processors to:
receive a behavioral sample from an individual;
determine, using a trained machine-learning (“ML”) model, a predicted change in behavioral health symptomatology based on a baseline behavioral sample for the individual and the behavioral sample, the predicted change in behavioral health symptomatology based on a difference score computed by the trained ML model based on the baseline behavioral sample for the individual and the behavioral sample;
determine whether the predicted change in behavioral health symptomatology satisfies a first threshold of a plurality of thresholds; and
in response to determining that the predicted change in behavioral health symptomatology satisfies the first threshold, output the predicted change in behavioral health symptomatology to a client device executing a graphical user interface indicating the predicted change for an entity treating the individual.
9 . The system of claim 8 , wherein the baseline behavioral sample comprises a baseline voice recording and a baseline assessment score, wherein the baseline assessment score is derived from a clinically validated assessment in behavioral health, and wherein the predicted change in behavioral health symptomatology indicates a predicted change in behavioral health symptomatology from the baseline assessment score in accordance with a clinically validated assessment.
10 . The system of claim 8 , wherein the baseline behavioral sample comprises a baseline voice recording and a baseline assessment score, wherein the baseline assessment score is derived from a clinically validated assessment in behavioral health and the clinically validated assessment includes multiple questions, each question capable of generating an independent score based on a self-reported answer, wherein the trained ML model comprises a plurality of trained ML models, each trained ML model is trained to assess a predicted change in a subset composite score of a subset of questions on a clinical assessment test used to generate the baseline assessment score, and the non-transitory computer-readable medium further comprises program code that is executable by the one or more processors to:
provide the baseline behavioral sample and a subsequent behavioral sample as input to each trained ML model of the plurality of trained ML models; and
output, for each trained ML model of the plurality of trained ML models, the predicted change in a subset composite score of the respective subset of questions on the assessment.
11 . The system of claim 8 , wherein receiving the baseline behavioral sample comprises receiving a plurality of baseline behavioral samples, each baseline behavioral sample corresponding to an enumerated environmental condition; and wherein the non-transitory computer-readable medium further comprises program code that is executable by the one or more processors to:
determine a match between contextual information associated with a subsequent behavioral sample and the enumerated environmental condition of a target baseline behavioral sample, and
provide the target baseline behavioral sample and the subsequent behavioral sample as input to the trained ML model.
12 . The system of claim 8 , wherein the trained ML model is configured to determine the predicted change in behavioral health symptomatology by analyzing sentiment, emotion, paralinguistic parameters, or a combination thereof associated with speech in a voice recording.
13 . The system of claim 12 , wherein the non-transitory computer-readable medium further comprises program code that is executable by the one or more processors to perform speech recognition on the voice recording and providing the recognized speech to analyze sentiment or emotion.
14 . The system of claim 10 , wherein the non-transitory computer-readable medium further comprises program code that is executable by the one or more processors to:
determining a recommended intervention based on the predicted changes in the subset composite scores; and
providing the recommended intervention to the entity treating the individual.
15 . A non-transitory computer-readable medium comprising program code that is executable by one or more processors to:
receive a behavioral sample from an individual;
determine, using a trained machine-learning (“ML”) model, a predicted change in behavioral health symptomatology based on a baseline behavioral sample for the individual and the behavioral sample, the predicted change in behavioral health symptomatology based on a difference score computed by the trained ML model based the baseline behavioral sample for the individual and the behavioral sample;
determine a recommended intervention based on the predicted change in behavioral health symptomatology; and
output the recommended intervention to a client device executing a graphical user interface indicating the predicted change for an entity treating the individual.
16 . The non-transitory computer-readable medium of claim 15 , further comprising program code that is executable by the one or more processors to:
determine whether the predicted change in behavioral health symptomatology satisfies a first threshold of a plurality of thresholds; and
in response to determining that the predicted change in behavioral health symptomatology satisfies the first threshold, provide a notification to the individual to perform a first intervention.
17 . The non-transitory computer-readable medium of claim 16 , wherein the first intervention comprises retaking an assessment.
18 . The non-transitory computer-readable medium of claim 16 , further comprising program code that is executable by the one or more processors to:
determine whether the predicted change in behavioral health symptomatology satisfies a second threshold of a plurality of thresholds, the second threshold greater than the first threshold; and
in response to determining that the predicted change in behavioral health symptomatology satisfies the second threshold, provide a notification to the entity treating the individual of the predicted change in behavioral health symptomatology.
19 . The non-transitory computer-readable medium of claim 18 , further comprising program code that is executable by the one or more processors to, in response to determining that the predicted change in behavioral health symptomatology satisfies the second threshold, provide one or more recommended interventions to the entity treating the individual.
20 . The non-transitory computer-readable medium of claim 19 , further comprising program code that is executable by the one or more processors to transmit content to a digital device of the individual, the content comprising coaching regarding strategies to remediate the predicted change.