Patent Application
20160117465
The present disclosure relates generally to quantifying a health status of an individual and, more particularly, to a method and apparatus for calculating a health quality score and providing a health upside optimizing recommendation.
Health assessments and related indices have in the past been computed based on evidence based guidelines and the results of clinical tests. These are either principally clinical only or rule-driven approaches, which are not universally applicable in different sub-populations of people. Further, these approaches neglect other features that could go into assessing health risk.
Health risk assessments can benefit by taking into account the patients' or care-receivers' psyche and related factors. Coupled with clinical data, these non-clinical aspects can provide a more holistic picture of an individual's health.
According to aspects illustrated herein, there are provided a method, a non-transitory computer readable medium, and an apparatus for calculating an overall health quality index (HQI) and providing a health upside optimizing recommendation. One disclosed feature of the embodiments is a method that collects data associated with an individual from an external data source, filters the data to identify a plurality of features, divides each one of the plurality of features into one or more of six action classes, builds one or more models for each one of the six action classes, computes the overall HQI using the one or more models that are built for each one of the six action classes, identifies the health upside optimizing recommendation based on one or more important actionable features selected from the plurality of features and provides the overall HQI and the health upside optimizing recommendation to the individual.
Another disclosed feature of the embodiments is a non-transitory computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform an operation that collects data associated with an individual from an external data source, filters the data to identify a plurality of features, divides each one of the plurality of features into one or more of six action classes, builds one or more models for each one of the six action classes, computes the overall HQI using the one or more models that are built for each one of the six action classes, identifies the health upside optimizing recommendation based on one or more important actionable features selected from the plurality of features and provides the overall HQI and the health upside optimizing recommendation to the individual.
Another disclosed feature of the embodiments is an apparatus comprising a processor and a computer readable medium storing a plurality of instructions which, when executed by the processor, cause the processor to perform an operation that collects data associated with an individual from an external data source, filters the data to identify a plurality of features, divides each one of the plurality of features into one or more of six action classes, builds one or more models for each one of the six action classes, computes the overall HQI using the one or more models that are built for each one of the six action classes, identifies the health upside optimizing recommendation based on one or more important actionable features selected from the plurality of features and provides the overall HQI and the health upside optimizing recommendation to the individual.
The teaching of the present disclosure can be readily understood by considering the following detailed description in conjunction with the accompanying drawings, in which:
To facilitate understanding, identical reference numerals have been used, where possible, to designate identical elements that are common to the figures.
The present disclosure broadly discloses a method and non-transitory computer-readable medium for calculating an overall health quality index (HQI) and providing a health upside optimizing recommendation. As discussed above, health assessments and related indices have in the past been computed based on evidence based guidelines and the results of clinical tests. These are either principally clinical only or rule-driven approaches, which are not universally applicable in different sub-populations of people. Further, these approaches neglect other features that could go into assessing health risk.
Embodiments of the present disclosure provide a novel method for computing an objective health quality index (HQI) that is data driven. The present disclosure also provides recommendations for potential upside to improve the HQI over time. The embodiments of the present disclosure use a labeled test data set as its fundamental basis and add to it an individual's own actions and behaviors over time. The features in the combined data set include both clinical and non-clinical features. The embodiments of the present disclosure simultaneously factor the individual's own view of personal health and an expert body's view of the same person's health and train a suite of models.
In one embodiment, the AS 104 may be in communication with the communications network 102 over a wireless or wired connection. In one embodiment, the AS 104 may be deployed as a general purpose computer as described below with reference to
In one embodiment, the AS 104 may collect data from one or more external databases 108 and 110. Although only two external databases 108 and 110 are illustrated in
In one embodiment the external databases 108 and 110 may be clinical based data and non-clinical based data. For example, the external databases 108 and 110 may be data from the Center for Disease Control (CDC), a medical records database, a health insurance claims database, a prescription database, a lab results database, a clinical visits records database, a health plan enrollment application database, a survey database, and the like.
In one embodiment, the external data may be collected as input data and testing data for building one or more models that are used to compute the HQI. For example, a plurality of features may be extracted from the external data. The features may be anything that may affect a health of the individual that can be quantified from the external data. For example, features may include demographic information such as sex, address, age group, and the like. The features may include health related information such as weight, height, cholesterol levels, blood counts, blood sugar level, and the like. The features may include lifestyle data from surveys such as how often an individual gets a physical exam, how often the individual goes to a doctor's office, how often the individual exercises, a type of diet, and the like.
In one embodiment, each one of the plurality of features may be divided into one or more of six action classes. For example, a single feature may be placed into two or more of the action classes. In other words, each one of the plurality of features is not necessarily limited to a single action class. In one embodiment, the six action classes may include a clinical action class, a compliance action class, a demographics action class, an efficiency action class, a lifestyle action class and a readiness-to-change action class.
In one embodiment, the clinical action class may represent an assessment of health based on the individual's clinical attributes. These attributes may include any clinical test either performed explicitly or diseases that are diagnosable using a clinical test.
In one embodiment, the compliance action class may represent how an individual complies with medical guidelines and adhere to a treatment. For example, complying with guidelines and following a prescribed treatment has an important role to play in determining an individual's health.
In one embodiment, the demographics action class may contain information about a user's demography. For example, the demographics action class may contain information about where the person lives, raises his or her family, an occupation, and the like. Information about environmental and familial conditions may also affect the health of the individual and those factors are taken into account for computing the HQI.
In one embodiment, the efficiency action class may relate to whether the individual will utilize cost-effective, regular and preventive mechanisms as opposed to “last minute” efforts. One example of the efficiency action class may contain information relating to an individual going to a primary care physician versus using the emergency room. Another example may be taking flu shots on time as a preventive measure.
In one embodiment, the lifestyle action class may relate to a variety of features directly in the control of the individual. Examples of lifestyle features may include exercise, smoking, drinking, resting, eating habits or diet, and the like.
In one embodiment, the readiness-to-change action class may relate to whether the individual will progress through stages of change. For example, the readiness-to-change action class may track if the individual exercises or readily adopts defined measures in a set of regular visits to a doctor's office or health clinic.
The plurality of features that are divided into the six action classes may be stored in the DB 106. From the plurality of features, one or more actionable features may also be identified based on domain knowledge accumulated from the training data and testing data that are compiled from the data collected from the various external databases 108 and 110. For example, over time certain features may be known to be actionable features.
In one embodiment, actionable features may be defined as features which an individual may control. For example, actionable features may include weight (e.g., the individual may diet to regulate his or her weight), smoking, drinking, blood sugar level (e.g., the individual may control his or her sugar intake), regular physicals, and the like. Some features may not be actionable, such as for example, genetic or hereditary dispositions such as having a fatty liver, high blood pressure, allergies to certain foods that may affect the individual's diet, a handicap that prevents the individual from taking regular visits to a doctor's office, and the like. In one embodiment, the plurality of features that are identified as actionable features may be stored in the DB 106.
As noted above, once the features are divided into one or more of the action classes, one or more models may be built for each one of the six action classes based on the features that are included in each respective action class. In one embodiment, each action class may have the same one or more models. In another embodiment, each action class may have different ones of the one or more models.
The one or more models may then be used to calculate an HQI for each action class.
Any of the models may be combined with rules that a physician or subject matter expert may outline. For example, physicians may consider certain values or ranges of health factors to be outside the limits of what they would consider healthy. If a multitude of rules hold true, the effect of the rules may be combined using additive methods, multiplicative methods, aggregation methods or other suitable means. In this way, computation of scores for certain variables may get rerouted to a rules bank that either combine machine-learning methods and physician-determined rules or use standalone rules. This may add human factors to variables that may be considered exceptions.
In one embodiment, the HQI computation table 200 may include a column 202 for action classes and include rows 210, 212, 214, 216, 218 and 220 for each one of the six action classes. In one embodiment, a column 204 may be used for HQIs computed using a logistic regression model and a column 206 may be used for HQIs computed using a random forest model. In one embodiment, an aggregated column 208 may be used to aggregate the HQI for each one of the models in columns 204 and 206 for each one of the six classes. In one embodiment, an overall row 222 may be used to compute the overall HQI for each model based on all six classes. The overall HQIs in the row 222 and the aggregated HQIs in column 208 may then be aggregated to compute an overall HQI 224.
In one embodiment, the HQIs may be aggregated using a variety of computational methods. For example, the HQIs may be aggregated using a weighted average, a mean, and the like. In the example in
In one embodiment, during the computation of the HQIs, one or more important features may be identified from the features in each one of the six classes. In one embodiment, the important features may be defined as those features that create a largest delta in the HQI when the feature is included or excluded from the HQI computation. In other words, the important features may be the features that have the largest effect on the computation of the HQI.
In one embodiment, a top k number of important features may be selected from each one of the six action classes. In other words, dithering is performed on the features from each one of the six action classes to obtain the top k number of important features.
In one embodiment, a health upside optimizing recommendation may then be based on one or more actionable important features selected from the actionable features and the important features. For example, the one or more actionable important features are features that are identified as being important and also identified as being actionable. In other words, even if a feature is important, but not actionable, the feature may not be helpful in being part of the health upside optimizing recommendation as the individual may not have any control over the feature and cannot improve his or her overall HQI based on the feature.
In one embodiment, once the overall HQI is computed and the health upside optimizing recommendation are selected, the overall HQI and the health upside optimizing recommendation may be provided to the individual. In one embodiment, the overall HQI may be presented via a graphical user interface (GUI) 300 illustrated in
In one embodiment, the GUI 300 may display the overall HQI 302 and the HQI for each one of the six action classes 304, 306, 308, 310, 312 and 314. In one embodiment, the individual's score may be displayed alongside a peer score so that the individual can compare his or her score to his or her peers.
In one embodiment, the peer score may be computed based on an average of a particular group that the individual belongs to. For example, the particular group may be a demographic group (e.g., age, region, and the like). In another embodiment, the particular group may be a group with a common ailment (e.g., diabetics, high blood pressure, and the like).
In one embodiment, the peer score may be computed using the HQIs and the overall HQIs of other individuals that are computed by the AS 104 and stored in the DB 106. For example, each time a new individual computes his or her HQIs and overall HQI, the data may be stored in the DB 106 and used for the peer scores displayed in the GUI 300.
In one embodiment, the health upside optimizing recommendation may be provided to the individual via a GUI 400 illustrated in
In addition, the health upside optimizing recommendation 416 may be provided. For example, the health upside optimizing recommendation 416 may be to increase exercise and eat fewer calories (e.g., dieting). In other words, if weight was selected as an important actionable feature, then weight may have a significant impact on the individual's overall HQI. Thus, the health upside optimizing recommendation 416 may help to optimize or improve the individual's overall HQI for future calculations.
In one embodiment, the overall HQI may be computed for each individual periodically over a period of time. The overall HQI may be tracked for each individual and graphically presented (e.g., via the GUI 300 or 400). As a result, individuals may be provided with an objective data driven view of his or her overall healthiness and be provided with recommendations on how to improve his or her overall HQI.
At step 502 the method 500 begins. At step 504, the method 500 collects data associated with an individual from an external data source. For example, the external data sources may be data from external databases such as the Center for Disease Control (CDC), a medical records database, a health insurance claims database, a prescription database, a lab results database, a clinical visits records database, a health plan enrollment application database, a survey database, and the like.
At step 506, the method 500 filters the data to identify a plurality of features. In one embodiment, the features may be anything that may affect a health of the individual that can be quantified from the external data. For example, features may include demographic information such as sex, address, age group, and the like. The features may include health related information such as weight, height, cholesterol levels, blood counts, blood sugar level, and the like. The features may include lifestyle data from surveys such as how often an individual gets a physical exam, how often the individual goes to a doctor's office, how often the individual exercises, a type of diet, and the like.
At step 508, the method 500 divides each one of the plurality of features into one of six action classes. In one embodiment, a single feature may be placed into two or more of the action classes. In other words, each one of the plurality of features is not necessarily limited to a single action class. In one embodiment, the six action classes may include a clinical action class, a compliance action class, a demographics action class, an efficiency action class, a lifestyle action class and a readiness-to-change action class.
In one embodiment, from the plurality of features, one or more actionable features may also be identified based on domain knowledge accumulated from the training data and testing data that are compiled from the data collected from the external data sources. For example, over time certain features may be known to be actionable features.
In one embodiment, actionable features may be defined as features which an individual may control. For example, actionable features may include weight (e.g., the individual may diet to regulate his or her weight), smoking, drinking, blood sugar level (e.g., the individual may control his or her sugar intake), regular physicals, and the like. Some features may not be actionable, such as for example, genetic or hereditary dispositions such as having a fatty liver, high blood pressure, allergies to certain foods that may affect the individual's diet, a handicap that prevents the individual from taking regular visits to a doctor's office, and the like.
At step 510, the method 500 builds one or more models for each one of the six action classes. In one embodiment, each action class may have the same one or more models. In another embodiment, each action class may have different ones of the one or more models. In one embodiment, the one or more models may include a logistic regression model, a random forest model, and the like.
At step 512, the method 500 computes an overall HQI using the one or more models. For example, a HQI for each one of the six classes may be computed for each one of the models that are built. An aggregate HQI may be computed for each one of the six classes and each model. An overall HQI may be computed for each model using each one of the six classes. The overall HQI may then be computed by aggregating all of the aggregate HQIs and the overall HQIs for each model, for example as illustrated in
In one embodiment, during the computation of the HQIs, one or more important features may be identified from the features in each one of the six classes. In one embodiment, the important features may be defined as those features that create a largest delta in the HQI when the feature is included or excluded from the HQI computation. In other words, the important features may be the features that have the largest effect on the computation of the HQI.
In one embodiment, a top k number of important features may be selected from each one of the six action classes. In other words, dithering is performed on the features from each one of the six action classes to obtain the top k number of important features.
At step 514, the method 500 identifies a health upside optimizing recommendation. In one embodiment, a health upside optimizing recommendation may then be based on one or more actionable important features selected from the actionable features and the important features. For example, the one or more actionable important features are features that are identified as being important and also identified as being actionable. In other words, even if a feature is important, but not actionable, the feature may not be helpful in being part of the health upside optimizing recommendation as the individual may not have any control over the feature and cannot improve his or her overall HQI based on the feature.
At step 516, the method 500 provides the overall HQI and the health upside optimizing recommendation. In one embodiment, the overall HQI and the health upside optimizing recommendation may be provided to a device of the individual via a GUI. In one embodiment, the overall HQI may be provided alongside an overall peer HQI to allow the individual to compare his or her overall health to her peer group.
In one embodiment, the peer score may be computed based on an average of a particular group that the individual belongs to. For example, the particular group may be a demographic group (e.g., age, region, and the like). In another embodiment, the particular group may be a group with a common ailment (e.g., diabetics, high blood pressure, and the like). The method 500 then proceeds to 518 where the method 500 ends.
As a result, the embodiments of the present disclosure transform general clinical data and non-clinical data to produce an overall health quality index that provides a holistic overview of an individual's health. The data is transformed from providing general disconnected data (e.g., clinical data and no-clinical data that are otherwise unrelated) into an overall HQI that provides a data driven objective overview of the individual's health based on a variety of different types of data. In other words, a new method is also provided that allows the transformation all of the different types of data that are collected into an objective score, i.e., the overall HQI.
Furthermore, the embodiments of the present disclosure improve the functioning of an application server or computer used for health analysis. For example, more accurate estimations of an individual's health may be performed by the computer (e.g., for pricing health insurance premiums, and the like). In other words, the technological art of estimating an individual's health is improved by providing a more accurate overview that is depicted by the overall HQI of the present disclosure.
It should be noted that although not explicitly specified, one or more steps, functions, or operations of the method 500 described above may include a storing, displaying and/or outputting step as required for a particular application. In other words, any data, records, fields, and/or intermediate results discussed in the methods can be stored, displayed, and/or outputted to another device as required for a particular application. Furthermore, steps, functions, or operations in
It should be noted that the present disclosure can be implemented in software and/or in a combination of software and hardware, e.g., using application specific integrated circuits (ASIC), a programmable logic array (PLA), including a field-programmable gate array (FPGA), or a state machine deployed on a hardware device, a general purpose computer or any other hardware equivalents, e.g., computer readable instructions pertaining to the method(s) discussed above can be used to configure a hardware processor to perform the steps, functions and/or operations of the above disclosed methods. In one embodiment, instructions and data for the present module or process 605 for calculating an overall health quality index (HQI) and providing a health upside optimizing recommendation (e.g., a software program comprising computer-executable instructions) can be loaded into memory 604 and executed by hardware processor element 602 to implement the steps, functions or operations as discussed above in connection with the exemplary method 500. Furthermore, when a hardware processor executes instructions to perform “operations”, this could include the hardware processor performing the operations directly and/or facilitating, directing, or cooperating with another hardware device or component (e.g., a co-processor and the like) to perform the operations.
The processor executing the computer readable or software instructions relating to the above described method(s) can be perceived as a programmed processor or a specialized processor. As such, the present module 605 for calculating an overall health quality index (HQI) and providing a health upside optimizing recommendation (including associated data structures) of the present disclosure can be stored on a tangible or physical (broadly non-transitory) computer-readable storage device or medium, e.g., volatile memory, non-volatile memory, ROM memory, RAM memory, magnetic or optical drive, device or diskette and the like. More specifically, the computer-readable storage device may comprise any physical devices that provide the ability to store information such as data and/or instructions to be accessed by a processor or a computing device such as a computer or an application server.
It will be appreciated that variants of the above-disclosed and other features and functions, or alternatives thereof, may be combined into many other different systems or applications. Various presently unforeseen or unanticipated alternatives, modifications, variations, or improvements therein may be subsequently made by those skilled in the art which are also intended to be encompassed by the following claims.
