num="0065">The detailed description of the exemplary embodiments and examples herein are by way of illustration only and are not limiting on the scope of the invention. A person of skill in the art will understand that changes can be made without departing from the spirit and scope of this disclosure.

By way of example, the elements in the network of services that interact to provide ECM and/or TCM are illustrated in FIG. 1. There may be other elements not mentioned in this description and omission of those other elements is not intended to be limiting on the invention.

Patients 101 are the focus of the system. Patients interact with the Health Care System (“HCS”) 102, which may include hospitals, outpatient clinics, family care physicians, specialist physicians, pharmacies, urgent care facilities, aged care facilities, nursing homes, in-home visits from a health care professional (“HCP”), and the like. The HCS 102 generally uses an Electronic Medical Record (“EMR”) 110. The system is configured to access other databases 111 for information not generally available in the EMR and may also receive data entered manually.

Other elements of the community which are not normally considered as part of the health care system can impact the patient's outcomes and quality of care. These elements contribute to the SDOH which have been shown to have a major impact on a patient's health and well-being. See for example, Braverman, Paula, Susan Egerter, and David R. Williams. “The social determinants of health: coming of age.” Annual review of public health 32 (2011): 381-398. It has been suggested that provision of mobile health aligned with the SDOH would provide better outcomes (Rogers, C. C., Jang, S. S., Tidwell, W., Shaughnessy, S., Milburn, J., Hauck, F. R., Williams, I. C., & Valdez, R. S. (2023). Designing mobile health to align with the social determinants of health. Frontiers in Digital Health, 5, https://doi.org/10.3389/fdgth.2023.1193920).

These elements may include the family support network 103 (co-resident, local, or extended family), friends and companions 104 (which may include animal companions and pets), the community support network 105 (which may include churches, volunteer and charitable services such as meals on wheels), independent contractors 106 which provide paid services as requested (e.g.: transportation services (taxis, Uber), food delivery services, house cleaning and laundry services, physical therapists, broad band internet and mobile phone service), and available community services 107 such as social workers, translation services, employment assistance, and tax help.

Although most communities have at least some of the elements, few communities have all the elements. Many patients have limited knowledge of available services, often there is poor coordination of services and unclear mechanisms to provide (and pay) for services to those who need them. In this invention, a Central Service Provider (“CSP”) 108 is configured to coordinate delivery of services to provide ECM and TCM. The CSP may provide some services directly (e.g.: telephone follow-up of a patient) and may initiate and coordinate the provision of services from the community elements or the HCS

The organization of the CSP is illustrated in FIG. 3, described below in detail. In summary, the CSP consists of as organization of people who coordinate the provision of services.

The CSP interacts and communicates 109 with the HCS. For example, the HCS may request or contract with the CSP to provide ECM or TCM services to the patient(s). The CSP may provide information back to the HCS about patient outcomes, arrange scheduling, and provide information about the patient's health care outside the HCS e.g.: in another hospital while traveling. The CSP may interact with the HCS EMR 110 to take data about the patients, or to update the EMR (and/or other databases 111) with patient outcomes and history of services provided.

In this invention, the CSP 200 is organized to provide ECM and TCM as illustrated in the block diagram in FIG. 2.

A menu of services 201 is provided. In a preferred embodiment, the menu of services 201 includes at least some of the services described above and illustrated in FIG. 1, which generally fit into the categories of clinical services and social services. Clinical services might include in-home visits by an HCP; telephone consultation with an HCP; counseling; scheduling hospital or clinic visits; drug delivery to the home; and other services. Social services might include arranging transportation (e.g.: for a clinic visit); language and translation services; in-home food delivery; family education; referral to a social worker; and others. As time goes on, additional services may be added to the menu of services, or obsolete services can be removed from the menu. In this manner, the menu of services is continually updated to reflect the best possibilities.

Each element in the menu of services is tagged to reflect geographical and time availability. Thus the available services 215 may be different for each patient and for different times of day. For example, some patients may be in an area in which social workers are not available, so social workers would excluded in the services available for that patient.

Patient characteristics include clinical characteristics such as diagnosis, duration of current clinical condition, prior treatments, medications, comorbidities (e.g.: diabetes, obesity, hypertension), history, physical characteristics (height, weight, age, BMI), and vital signs. Many of these clinical characteristics may be obtained from the hospital patient records either automatically (computer-to-computer interface for the EMR), manual entry, or some combination. Patient characteristics might also include social characteristics such as marital status, ethnicity, employment status, family support, languages spoken, access to transportation, analysis of family or community support systems, access to nutrition, and income level. Patient characteristics may also include annotations from providers such as clinicians, nurses, and medical assistants.

The EMR is not the sole source of information on the needs of patients. The invention described herein incorporates the ability to retrieve data from other databases 111 with which it interfaces. Examples of such databases include employment, social services,, home nursing care program, family members, or other external entity engaged in elements of care delivery for the patient.

Some of the patient's characteristics may need to be entered manually.

Patient characteristics can include categorical variables (e.g.: sex) and continuous variables (e.g.: height). Patient characteristics can also include derivative variables e.g.: whether or not the patient's blood glucose is within or outside the normal range for that patient.

Key elements of patient characteristics are location information and time. Location can be both a permanent location (i.e.: place of residence) or a current location (e.g.: when working away from the home). Current location information can be provided, for example by a mobile phone which interacts with the system. As described above, Location and Time Information 214 modifies the menu of services to reflect local availability so the Available Services 215 for that patient at that time are one input to the Service Selection Module 205. Location information is also used in other ways as will be described below.

Patient Characteristics 202 is one set of inputs to the Prioritization Module 203. The Prioritization Module generates numerical Service Priorities to the menu of services to prioritize which services should be provided. The Prioritization module uses a computer system (which may be local or a cloud based computer system, or a distributed system) and uses one or more algorithms or, in a preferred embodiment, an AI engine to generate the Priorities.

Prioritization Module 203 generates numerical Service Priorities 204 to the Service Selection Module 205 which uses the priorities and the Menu of Services 201 to select one or more services to be provided to the patient. The Service Priorities are used to prioritize services based on analysis of the patient's individual data from Patient Characteristics 202 and other inputs e.g.: historical data, location and time. The historical data could include data aggregated from multiple patients that have used the system, as well as more general population statistics, and any other source of data that could be used to train an AI engine. Historical data may include data from each individual patient. As will be appreciated, the more factors that are used in the Prioritization module to determine Priorities to prioritize services, the more robust the system will be to missing data.

The Service Selection Module uses the Service Priorities 204 and Available Services 215 to provide Recommended Services 206 which are communicated to a Sequencing Module 207. The sequencing module uses the recommended services and integrates with other information such as the availability of selected services (e.g.: a clinic visit must fit in with the clinic schedule) and information about if and how sequencing of services makes a difference to the outcome, among other factors. For example, prior information about a particular patient may show that the particular patient usually needs two reminder phone calls to attend a clinic visit, as well as arranging transportation, whereas another particular patient needs only one reminder via text message and has her own transportation, or another patient requires communication and integration of family members in assisting the patient with support. The Service Selection Module 205 and the Sequencing Module 207 thus work together to optimize the services to be provided for the best outcomes at minimal cost, tailored to each individual patient.

The recommended and sequenced services for a particular patient 208 are communicated 209 back to the Prioritization module 203 which maintains a record of services for each patient. The recommended and sequenced services are also communicated to the CSP 210 which coordinates the provision of services as illustrated in FIG. 1 and described in detail eow with reference to FIG. 3. As the services 211 are provided, a Scheduling and Follow-up Module 212 keeps track of services provided (or attempted to be provided), and patient outcomes.

From time to time, the patient is followed up coordinated by a Scheduling and Follow-up Module 213. In a preferred embodiment, at least some of the follow-up could be automatic, for example a text message with a reminder for the patient to take medications or schedule a clinic visit. In a further embodiment, follow-up could be in person e.g.: with a clinic visit at the discharging hospital or another healthcare provider, a telephone follow-up, or an in-home follow-up. In a further embodiment, follow-up could be automatic e.g.: with a home blood pressure monitor (systolic and diastolic blood pressure, and heart rate) or bathroom scales (patient weight) that connects (e.g.: via Wi-Fi or the cell phone system) to automatically provide follow-up data.

The outcomes from follow-up are input to a Patient Outcomes module 208 which accumulates and aggregates the follow-up data. The patient outcomes are an input to the Prioritization module 203 which uses those outcome data (among others) to adjust the Prioritization for the service selection module 205. Patient Outcomes may be communicated 109 back to the HCS 102 in real time or occasionally (batch mode).

Information about the recommended and sequenced services 208 is one set of inputs to the Prioritization Module 203. In some cases, the services recommended might not be provided e.g.: because a patient misses a follow-up visit, so it is important that the Prioritization Module 203 has inputs of the services actually provided, not just those recommended. For example, if a patient routinely misses clinic follow-up visits, the Prioritization module could take that into account to de-prioritize clinic follow-up visits and emphasize telephone follow-up. The information about services actually provided is provided by the Patient Outcomes Module 213.

In a preferred embodiment, the Prioritization module 203 incorporates AI such that the Service Selection Module is informed by the prior experience of a large set of patients, as well as the individual outcomes of each patient. The AI engine in the Prioritization module 203 can be any of the structures well-known to those skilled in the art. An AI system can incorporate multiple inputs and outputs with accuracy and time beyond the capabilities of any human or group of humans.

An example of one embodiment of an AI based Prioritization module is shown in FIG. 5. This is a two-layer artificial neural network (ANN) with an input layer 501, two hidden layers 502 and 503, and an output layer 504. As will be appreciated by one skilled in the art, an ANN may have any number of hidden layers, and number of inputs and outputs. The inputs to the input layer, {In₁, In₂, In₃, In₄, . . . In_n} include Patient Characteristics 202, Patient Outcomes 213, and recommended and sequenced services 209. The outputs of the ANN {Out₁, Out₂, Out₃, . . . Out_n} are the Priorities which are the input to the Service Selection Module 205. The connections between the layers are for illustration only, and a person skilled in the art will understand that there can be more or fewer connections between the nodes in each layer.

The intermediate outcomes in the hidden layer A 502 {HA₁, HA₂, HA₃, HA₄, . . . HA_n} depend on the inputs and the weights between the inputs and the intermediate outcomes that arise from the training of the ANN. Similarly, the intermediate outcomes in the hidden layer B 503 {HB₁, HB₂, HB₃, HB₄, HB₅, . . . HB_n} depend on the intermediate outcomes in Hidden Layer A and the weights between them that arise from the training of the ANN. Finally, the Outputs {Out₁, Out₂, Out₃, . . . Out_n} depend on the intermediate outcomes in Layer B and the weights that arise from the training of the ANN.

In one embodiment, the ANN uses uplift modelling, a technique to model the change in probability of an outcome resulting from an action. In an embodiment for TCM, the initial probability of hospital readmission is calculated based on the patient characteristics. Then one or more interventions (services) are applied, and the readmission probability is recalculated. In an iterative process, different interventions are added (or subtracted), and the ANN settles on the interventions which minimizes the probability of readmission.

In this embodiment, the inputs {In₁, In₂, In₃, In₄, . . . In_n} include patient characteristics patient outcomes, available services, cost of services, and location information. The outputs {Out₁, Out₂, Out₃, . . . Out_n} are the priorities of intervention—i.e.: weights to the service selection module with the service that has the highest impact on reducing risk of readmission having the highest weight.

The outputs can also include the relative timing of intervention—that is, straightforward weighting does not account for the sequence of interventions. In a simple example, ordering transportation cannot be done until a follow up visit is scheduled and confirmed.

The outputs can also include an estimate of the cost effectiveness of intervention which can be used as another input to prioritize services. For example, certain drugs to treat diabetes are relatively new and expensive, whereas there may be inexpensive generic drugs with similar clinical performance.

In one embodiment, the ANN is initially trained using medical records e.g.: obtained from an EMR. The records may be curated e.g.: to ensure complete data without obvious errors. For instance, the ANN may be initially trained on previous instances of input data (e.g., a data set associated with an individual patient) and may be adapted to learn when data has been corrected (which may require further reference to a curated input data set). As one example, an input data set might include the following inputs, which would be indicative of obvious errors that might have been corrected later: Sex=Male, Date of Last Menstruation=30 Days Prior, Age=9, Number of Children=2. Alternatively, or in addition, the input data set might include more subtle errors such as inconsistent biomarker data, which the ANN might learn to recognize. Accordingly, the AI system may be adapted to provide ongoing and substantially real-time error-checking of input data on a large scale in a manner that could not be manually performed by a human due to the enormous amount of data involved in executing the disclosed systems and methods at scale with respect to ECM, TCM, or other care management environments. Training methods for the ANN can include supervised learning in which the curated data from the EMR is used. The ANN can be trained using back propagation, in which the input data are derived from the EMR and other databases, the known interventions which produce known results are compared to the ANN's output, and the weightings for intermediate nodes are adjusted to minimize error between the observed results and the ANN's results.

Like all AI systems, the reliability and usefulness of the outputs depends on the training dataset. In this invention, the training dataset can be obtained from multiple examples of patient data in the EMR from one or more HCS. A curated training dataset will use patient characteristics (as described above), interventions, and patient outcomes (among others) to train the AI system. In a preferred embodiment, the AI system will be continuously self-improving using data derived from new patients. U.S. Pat. No. 11,100,373 describes one example of an Autonomous and Continuously Self-Improving learning system.

In a preferred embodiment, the Prioritization module 203 can operate with sparse data (i.e.: less than complete data in the Input Layer 501) such that the services selected are the “best estimate” based on the data available.

FIG. 6 shows an example of the operation of the Service Selection Module. The Menu of Services provided to the Service Selection Module is tagged with a modifier that reflects the availability of the service based, for example, on location or time (e.g.: a social worker may be available only on some days, whereas a telephone follow up is available 24/7). In this example, the modifier is one if the service is available, and zero if it is not. The Service Priorities 204 from the Prioritization Module 203 are modified by the availability of Services to generate a score of Recommended Services 206 to the Sequencing Module 207. In this example, the Recommended and Sequenced Services (ranked high to low) are Daily Reminder, Telephone follow up, clinic visit, medication delivery, transportation and food.

As shown in FIG. 2, this list is provided to the Central Service Provider 210 which coordinates the provision of services. For example, the CSP care coordinator may call the clinic to schedule a visit, arrange transportation for the client, initiate a daily reminder (e.g.: via mobile phone), call the pharmacy to order medication delivery, and then call the client to advise what action has been taken, notify of the date of the clinic visit and that transportation has been arranged and ask if food delivery is needed.

In one example, the patient suffers from advanced heart failure (HF). Following hospital discharge, the Service Selection Module recommends daily telephone follow-up by a medical assistant for the first week, followed by weekly follow-up thereafter. In addition, an automatic message is delivered to the patient's phone to remind the patient to take the prescribed medications at the same time every day. The message could be delivered via text message (e.g.: SMS, WhatsApp), by automated voice message, or personal contact with the Care Coordinator. In addition, based on the patient's social situation, in-home food delivery is scheduled automatically for every evening. In addition, the patient is provided with a “smart” bathroom scale that communicates with the Patient Outcomes module 213, and the patient is reminded to weigh himself at the same time every day e.g.: upon waking in the morning. In addition, the patient has a questionnaire pushed to his mobile phone to ask questions about his symptoms. After one week, the system records that the patient's weight is increasing, and he reports shortness of breath. Based on these new data points, the Prioritization module adjusts the services recommended and for example may recommend an in-home visit by a healthcare professional, medication adjustments, clinic visit, or alteration of exercise regime.

FIG. 3 is a representation of one embodiment of the organization of the CSP 300.

The CSP 300 coordinates the provision of services 308 to the patient 307 from the CSP. In some cases, the CSP may not have the resources to provide certain services, and they may be provided by an External Network of Services 309 to the patient 310. For example, if one of the prioritized services is telephone counseling in the patient's first language, then the CSP might employ telephone counselors fluent in that language from the External Network of Services 309. If one of the required services is transportation for clinical follow-up visits, then the CSP might arrange transportation services as needed.

In a preferred embodiment, the CSP 300 establishes contracts with the external network of services to predetermine availability, pricing, and quality.

In a preferred embodiment, the system is implemented on a cloud-based computing and data storage module shown as 301. Cloud-based services allow access to the system from virtually anywhere, at any time, e.g.: from a desktop computer or a smartphone app, offering flexibility for healthcare providers to monitor the progress of any patient regardless of the patient's location. The cloud database is updated in real time as new data become available. Of course, local computing and storage are also used as necessary for example personal computers, tablets, and smart phones.