# US 12,602,598 B2 · Index for risk of non-adherence in geographic region with patient-level projection

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,602,598 B2
- Document ID: US12602598B2
- Status: Granted
- Kind: B2
- Application number: 17/824,351
- Filed: 2022-05-25
- Published: 2023-12-21
- Granted: 2026-04-14
- Earliest claimed: 2022-05-25
- Assignees: IQVIA Inc.
- Applicants: IQVIA Inc.
- Inventors: Arturo Silva
- Web page: https://patentgrep.com/patent/US12602598B2

## CPC

- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 20/20: Ensemble learning
- G16H 10/60: for patient-specific data, e.g. for electronic patient records
- G16H 50/30: for calculating health indices; for individual health risk assessment
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA

## Abstract

Methods and systems to train and use an ensemble of artificial intelligence/machine learning (AI/ML) models to extract information from social determinants of health (SDoH), including training each of multiple dimensionality reduction models to reduce dimensionality of socio-demographic variables associated with a respective one of multiple SDoH categories, training a predictive model to predict a patient behavior for a geographic region (e.g., risk of non-adherence to treatment regimens) based on dimensionally reduced SDoH (alone or in combination with selected socio-demographic variables and/or other data), training a patient classification model to classify patients based on prescription transactions, and/or training a regional similarity model to determine a measure of similarity between geographic regions based on SDoH and/or dimensionally reduced SDoH. Also disclosed are techniques to visually represent outputs of the models on a user-interactive display.

## Claims

### Claim 1

An apparatus, comprising:
a processor and memory configured to train and utilize an ensemble of artificial intelligence/machine learning (AI/ML) models, including to,
train each of multiple dimensionality reduction models to reduce a dimensionality of socio-demographic variables associated with a respective one of multiple social determinants of health (SDoH) categories,
process multiple sets of SDoH training data with the dimensionality reduction models to generate multiple respective sets of dimensionally reduced SDoH training data,
train a predictive model to correlate the sets of dimensionally reduced SDoH training data to respective training measures of a patient behavior,
process categories of SDoH of a first geographic region with respective ones of the dimensionality reduction models to provide dimensionally reduced SDoH of the first geographic region,
process the dimensionally reduced SDoH of the first geographic region with the predictive model to provide a predicted measure of the patient behavior for the first geographic region, and
present a visual indication of the predicted measure of the patient behavior on a user-interactive display.

### Claim 2 (depends on 1)

The apparatus of claim 1, wherein the processor and memory are further configured to:
train the predictive model based further on one or more socio-demographic variables of the respective sets of SDoH training data.

### Claim 3 (depends on 1)

The apparatus of claim 1, wherein the processor and memory are further configured to:
train the predictive model based further on patient-level data that identifies risk factors reported for patients.

### Claim 4 (depends on 1)

The apparatus of claim 1, wherein the processor and memory are further configured to:
determine a relative contribution of one or more of the categories of SDoH of the first geographic region and of one or more of the socio-demographic variables of the SDoH of the first geographic region, to the predicted measure of the patient behavior; and
present a visual indication of the relative contribution on the user-interactive display.

### Claim 5 (depends on 1)

The apparatus of claim 1, wherein the processor and memory are further configured to:
process SDoH of a second geographic region with the dimensionality reduction models to provide dimensionally reduced SDoH of the second geographic region;
determine a measure of similarity between the first and second geographic regions based at least in part on the dimensionally reduced SDoH of the first and second geographic regions; and
present a visual indication of the measure of similarity on the user-interactive identify the second geographic region as similar to the first geographic region, on the user-interactive display.

### Claim 6 (depends on 1)

The apparatus of claim 1, wherein the predicted measure of the patient behavior includes a risk of non-adherence to treatment regimens, and wherein the processor and memory are further configured to:
train a classification model to identify individual patients as at-risk or not at-risk of non-adherence to treatment regimens based on prescription transactions of the patients, wherein the prescription transactions of the patient are determined from insurance claims.

### Claim 7 (depends on 6)

The apparatus of claim 6, wherein the processor and memory are further configured to:
identify a second set of patients as not at risk of non-adherence based on dissimilarity of prescription transaction information of a first set of patients who are known to be at-risk of non-adherence to medical treatment regimens and demographic information of the second set of patients; and
train the classification model to distinguish between the first and second sets of patients based on the prescription transaction information of the respective patients.

### Claim 8 (depends on 7)

The apparatus of claim 7, wherein the processor and memory are further configured to:
use the classification model to identify other patients as at-risk of non-adherence to treatment regimens based on prescription transaction information of the respective other patients; and
present a visual indication of the classifications of the other patients on the user-interactive display.

### Claim 9

A non-transitory computer readable medium encoded with a computer program that comprises instructions to cause a processor to:
train each of multiple dimensionality reduction models to reduce a dimensionality of socio-demographic variables associated with a respective one of multiple social determinants of health (SDoH) categories;
process multiple sets of SDoH training data with the dimensionality reduction models to generate multiple respective sets of dimensionally reduced SDoH training data;
train a predictive model to correlate the sets of dimensionally reduced SDoH training data to respective training measures of a patient behavior;
process categories of SDoH of a first geographic region with respective ones of the dimensionality reduction models to provide dimensionally reduced SDoH of the first geographic region;
process the dimensionally reduced SDoH of the first geographic region with the predictive model to provide a predicted measure of the patient behavior for the first geographic region; and
present a visual indication of the predicted measure of the patient behavior on a user-interactive display.

### Claim 10 (depends on 9)

The non-transitory computer readable medium of claim 9, further comprising instructions to cause the processor to:
train the predictive model based further on one or more socio-demographic variables of the respective sets of SDoH training data.

### Claim 11 (depends on 9)

The non-transitory computer readable medium of claim 9, further comprising instructions to cause the processor to:
train the predictive model based further on patient-level data that identifies risk factors reported for patients.

### Claim 12 (depends on 9)

The non-transitory computer readable medium of claim 9, further comprising instructions to cause the processor to:
determine a relative contribution of one or more of the categories of SDoH of the first geographic region and of one or more of the socio-demographic variables of the SDoH of the first geographic region, to the predicted measure of the patient behavior; and
present a visual indication of the relative contribution on the user-interactive display.

### Claim 13 (depends on 9)

The non-transitory computer readable medium of claim 9, further comprising instructions to cause the processor to:
process SDoH of a second geographic region with the dimensionality reduction models to provide dimensionally reduced SDoH of the second geographic region;
determine a measure of similarity between the first and second geographic regions based at least in part on the dimensionally reduced SDoH of the first and second geographic regions; and
present a visual indication of the measure of similarity on the user-interactive identify the second geographic region as similar to the first geographic region, on the user-interactive display.

### Claim 14 (depends on 9)

The non-transitory computer readable medium of claim 9, wherein the predicted measure of the patient behavior includes a risk of non-adherence to treatment regimens, further comprising instructions to cause the processor to:
train a classification model to identify individual patients as at-risk or not at-risk of non-adherence to treatment regimens based on prescription transactions of the patients, wherein the prescription transactions of the patient are determined from insurance claims.

### Claim 15 (depends on 14)

The non-transitory computer readable medium of claim 14, further comprising instructions to cause the processor to:
identify a second set of patients as not at risk of non-adherence based on dissimilarity of prescription transaction information of a first set of patients who are known to be at-risk of non-adherence to medical treatment regimens and demographic information of the second set of patients; and
train the classification model to distinguish between the first and second sets of patients based on the prescription transaction information of the respective patients.

### Claim 16 (depends on 15)

The non-transitory computer readable medium of claim 15, wherein the processor and memory are further configured to:
use the classification model to identify other patients as at-risk or not at-risk of non-adherence to treatment regimens based on prescription transaction information of the respective other patients; and
present a visual indication of the classifications of the other patients on the user-interactive display.

### Claim 17

A computing machine implemented method, comprising:
training each of multiple dimensionality reduction models to reduce a dimensionality of socio-demographic variables associated with a respective one of multiple social determinants of health (SDoH) categories;
processing multiple sets of SDoH training data with the dimensionality reduction models to generate multiple respective sets of dimensionally reduced SDoH training data;
training a predictive model to correlate the sets of dimensionally reduced SDoH training data to respective training measures of a patient behavior;
processing categories of SDoH of a first geographic region with respective ones of the dimensionality reduction models to provide dimensionally reduced SDoH of the first geographic region;
processing the dimensionally reduced SDoH of the first geographic region with the predictive model to provide a predicted measure of the patient behavior for the first geographic region; and
presenting a visual indication of the predicted measure of the patient behavior on a user-interactive display.

### Claim 18 (depends on 17)

The method of claim 17, wherein the training a predictive model comprises:
training the predictive model based further on one or more socio-demographic variables of the respective sets of SDoH training data.

### Claim 19 (depends on 17)

The method of claim 17, wherein the training a predictive model comprises:
training the predictive model based further on patient-level data that identifies risk factors reported for patients.

### Claim 20 (depends on 17)

The method of claim 17, further comprising:
determining a relative contribution of one or more of the categories of SDoH of the first geographic region and of one or more of the socio-demographic variables of the SDoH of the first geographic region, to the predicted measure of the patient behavior; and
presenting a visual indication of the relative contribution on the user-interactive display.

### Claim 21 (depends on 17)

The method of claim 17, further comprising:
processing SDoH of a second geographic region with the dimensionality reduction models to provide dimensionally reduced SDoH of the second geographic region;
determining a measure of similarity between the first and second geographic regions based at least in part on the dimensionally reduced SDoH of the first and second geographic regions; and
presenting a visual indication of the measure of similarity on the user-interactive identify the second geographic region as similar to the first geographic region, on the user-interactive display.

### Claim 22 (depends on 17)

The method of claim 17, wherein the predicted measure of the patient behavior includes a risk of non-adherence to treatment regimens, the method further comprising:
training a classification model to identify individual patients as at-risk of non-adherence to treatment regimens based on prescription transactions of the patients, wherein the prescription transactions of the patient are determined from insurance claims.

### Claim 23 (depends on 22)

The method of claim 22, wherein the training a classification model comprises:
identifying a second set of patients as not at risk of non-adherence based on dissimilarity of prescription transaction information of a first set of patients who are known to be at-risk of non-adherence to medical treatment regimens and demographic information of the second set of patients; and
training the classification model to distinguish between the first and second sets of patients based on the prescription transaction information of the respective patients.

### Claim 24 (depends on 23)

The method of claim 23, further comprising:
using the classification model to identify other patients as at-risk of non-adherence to treatment regimens based on prescription transaction information of the respective other patients; and
presenting a visual indication of the classifications of the other patients on the user-interactive display.