DETAILED ACTION
Status of the Application
In view of the Appeal Brief filed on March 26, 2026, PROSECUTION IS HEREBY REOPENED. New grounds of rejection (edited versions of the previously presented 35 USC §§ 101 and 103 rejections) are set forth below.
To avoid abandonment of the application, appellant must exercise one of the following two options:
(1) file a reply under 37 CFR § 1.111 (if this Office action is non-final) or a reply under 37 CFR § 1.113 (if this Office action is final); or,
(2) initiate a new appeal by filing a notice of appeal under 37 CFR § 41.31 followed by an appeal brief under 37 CFR § 41.37. The previously paid notice of appeal fee and appeal brief fee can be applied to the new appeal. If, however, the appeal fees set forth in 37 CFR § 41.20 have been increased since they were previously paid, then appellant must pay the difference between the increased fees and the amount previously paid.
A Supervisory Patent Examiner (SPE) has approved of reopening prosecution by signing below:
/CHRISTINE M Tran/ Supervisory Patent Examiner, Art Unit 3695
This Second Action Non-Final Office Action is in response to Applicant’s Appeal Brief filed on March 26, 2026.
Claims 20-37 are pending, of which claims 20 and 29 are independent.
All pending claims have been examined on the merits.
Notice of Pre-AIA or AIA Status
The present application, 17/983,309 filed on 11/08/2022, is a Continuation of 16/510,384 filed on 07/12/2019, which claims priority to US Provisional application 62/697,038 filed on 07/12/2018.
The effective filing date is after the AIA date of March 16, 2013, and so the application is being examined under the “first inventor to file” provisions of the AIA .
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 20-37 are rejected under 35 U.S.C. §101 because the claimed invention is directed to non-statutory subject matter. The claimed invention is directed to an abstract idea, without “significantly more”.
The abstract idea elements in independent claim 29 are shown in regular font. The “additional elements” are shown in underlined font:
29. (Currently Amended) A patient communications system comprising:
software executing on a computer, the software receives patient data indicative of contact, insurance and demographic information for a plurality of patients;
said software comprising a neural network which uses an adaptable and self-adjusting set of heuristics and responsiveness information from the plurality of patients based on the patient data to tune said neural network utilizing the patient data to identify at least a first communications methodology likely to result in successful collection of a bill and generate a patient communication for a patient user with content based on contextualized information determined based on the patient data, the patient communication containing information related to a medical visit with a medical practice;
said software transmitting the patient communication to the patient user via the first communication methodology, said patient communication indicative of a bill associated with the medical visit, wherein said first communication methodology is determined by said neural network and said patient communication is transmitted via said first communication methodology in a manner determined by said neural network to be likely to result in successful collection of the bill;
said software receiving an indication of success or failure associated with bill collection of the bill for the patient communication and based on the success or failure said software updating a storage containing the patient data in order to train said neural network.
More specifically, claims 20-37 recite an abstract idea: “Certain Methods of Organizing Human Activity", specifically “Commercial or Legal Interactions (Including Agreements in the form of Contracts; Legal Obligations; Advertising, Marketing, or Sales Activities or Behaviors; Business Relations)”, as discussed in MPEP §2106(a)(2) Parts (I) and (II), and in the 2019 Revised Patent Subject Matter Eligibility Guidance.
The abstract idea elements include:
“identify at least a first communications methodology likely to result in successful collection of a bill”, and
“generate a patient communication for a patient user with content based on contextualized information determined based on the patient data”.
The “apply it” elements include:
“wherein said first communication methodology is determined by said neural network”.
The “additional” structural elements are: “software”, “computer”, and “said software comprising a neural network”.
The “additional” extra-solution elements are: “receives patient data indicative of contact, insurance and demographic information for a plurality of patients”, “transmitting the patient communication to the patient user via the first communication methodology”, “transmitted via said first communication methodology in a manner determined by said neural network to be likely to result in successful collection of the bill”, “receiving an indication of success or failure associated with bill collection for the patient communication”, and “updating a storage containing the patient data”.
This abstract idea is not integrated into a practical application, because:
The claim recites an abstract idea with additional generic computer elements. The generically recited computer elements (“software”, “computer”, and “said software comprising a neural network”) do not add a meaningful limitation to the abstract idea, because they amount to simply implementing the abstract idea on a computer.
The claim amounts to adding the words "apply it" (or an equivalent) with the abstract idea, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea, such as in the following feature: “wherein said first communication methodology is determined by said neural network”.
In regards to “apply it” (applying the abstract idea on a general purpose computer), the 35 USC § 101 rejections are based on the CAFC decision in Recentive Analytics, Inc. v. Fox Corp. April 18, 2025 (https://www.cafc.uscourts.gov/opinions-orders/23-2437.OPINION.4-18-2025_2500790.pdf).
The Recentive Analytics decision states (see page 10): “This case presents a question of first impression: whether claims that do no more than apply established methods of machine learning to a new data environment are patent eligible. We hold that they are not.”
The Examiner holds that Applicant’s description of the neural network merely describes “apply it” uses of a generic neural network. (See the description in para. [0106] and [0122] of the specification).
The extra-solution activities (“receives patient data indicative of contact, insurance and demographic information for a plurality of patients”, “transmitting the patient communication to the patient user via the first communication methodology”, “transmitted via said first communication methodology in a manner determined by said neural network to be likely to result in successful collection of the bill”, “receiving an indication of success or failure associated with bill collection for the patient communication”, and “updating a storage containing the patient data”) do not add a meaningful limitation to the method, as they are insignificant extra-solution activity;
The combination of the abstract idea with the additional elements (generically recited computer elements), and/or with the extra-solution activities, does not integrate the abstract idea into a practical application.
The claims do not include additional elements that are sufficient to amount to significantly more than the abstract idea, because:
When considering the elements "alone and in combination" (“software”, “computer”, and “said software comprising a neural network”), they do not add “significantly more” to the exception, because they amount to simply implementing the abstract idea on a computer.
Instead, they merely apply established methods of machine learning to a new data environment, as held to be unpatentable in the Recentive Analytics case.
In regards to the extra solution activities (“receives patient data indicative of contact, insurance and demographic information for a plurality of patients”, “transmitting the patient communication to the patient user via the first communication methodology”, “transmitted via said first communication methodology in a manner determined by said neural network to be likely to result in successful collection of the bill”, “receiving an indication of success or failure associated with bill collection for the patient communication”, and “updating a storage containing the patient data”), these are well-understood, routine, conventional computer functions recognized by the court decisions listed in MPEP § 2106.05(d).
More specifically, in regards to the “updating a storage containing the patient data” step, see the court cases Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015) (storing and retrieving information in memory); and OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1092-93 (Fed. Cir. 2015) (storing and retrieving information in memory).
More specifically, in regards to the “transmitting” and “receiving” steps, see the court cases OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network) and (presenting offers and gathering statistics), OIP Techs., 788 F.3d at 1362-63, 115 USPQ2d at 1092-93; buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network).
Independent claim 20 is rejected on the same grounds as independent claim 29.
All dependent claims are also rejected, because they merely further define the abstract idea.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim 20-37 are rejected under 35 U.S.C. 103 as being unpatentable over WO 2001/057756 A1 to Shao et al. (“Shao”. Eff. Filed on Feb 1, 2000. Published on Aug. 9, 2001) in view of AU 2003/254073 A1 to O’neill (“O’neill”. Eff. Filed on July 22, 2003. Published on February 9, 2004), and further in view of US 2017/0060930 to Elkherj et al. (“Elkherj”. Eff. Filed on Aug. 24, 2015. Published on Mar. 2, 2017).
In regards to claims 1-19, they are cancelled.
In regards to claim 20,
20. (Currently Amended) A method of patient communications comprising:
providing software executing on a computer,
said software comprising a neural network which uses an adaptable and self- adjusting set of heuristics and responsiveness information from the plurality of patients based on the patient data to tune said neural network utilizing the patient data to identify at least a first communications methodology likely to result in successful collection of a bill …,
(See Shao, page 4, lines 1-6: “The present invention provides an automated system and method for predicting the likelihood of collecting on a delinquent debt of an account. The system uses one or more predictive models, for example, a neural network, to evaluate individual debt holder accounts and predict the amount that will be collected on each account based on learned relationships among known variables.”)
(See Shao, Claim 62: “The method of claim 52, wherein the predictive model predicts the optimal method of communicating with a delinquent debtor.”)
(See Shao, page 8, lines 23-29: “Fig. 2 illustrates an embodiment of the development and use of predictive models for delinquent debt collection. A set of historical data is selected for use in model development 230. A suitable set of data is selected wherein the data contains sufficient information to properly train the desired predictive model. Suitable criteria for inclusion in the historical dataset is developed, taking into account such factors as the type of account information historically available, and the type of information that will typically be available when making a prediction for a currently delinquent debt.”)
(See Shao, page 35, lines 4-16: “Modeling the success or failure of a particular collection action is complicated due to the fact that by making action recommendations, the underlying distribution on which the model was built is changed (i.e. a feedback loop is created, because each current action taken effects the likelihood of the consequences of future actions). In order to explicitly model collection action effects to obtain better recoveries, it is preferable to avoid creating too many distinct actions to be monitored, to prevent undesired feedback. Assume, therefore, that all possible actions have been aggregated into a small number of action groups (e.g., soft reminder letter, harsh reminder letter, soft reminder call, harsh reminder call, threat to shut off authorizations, threat to close account, offer of partial pay, offer to re- age, etc.), denoted as a1, a2 ..., aq. Furthermore, assume that building individual predictive models that estimate the probability to pay for each action or action sequence is practically undesirable. Two different embodiments of the modeling process may be used, either modeling the effect of a single action, or modeling the effect of action sequences.”)
the patient communication associated with a medical visit with a medical practice;
(See Shao, claim 3: “The method of claim 1, wherein the delinquent debt was incurred on a medical service.”)
(See Shao, claim 81: “The method of claim 65, wherein the profile includes events related to a medical condition of the account debtor or family members of the account debtor.”)
In addition, Shao also teaches the following:
(See Shao, page 5, last paragraph: “The term "debt" as used throughout this document is defined to encompass a wide variety of different types of debts or credit obligations, for example, credit card debt, medical debts, utility bills, bounced checks, electronic transaction (Internet) debt, personal loan debt, secured or unsecured loans, and other types of unpaid bills.”)
(See Shao, page 19, lines 5-10: “A co-occurrence matrix is constructed 414 for the words in the set of documents dj, d2, ... dm. The context vector software collects documents and determines co-occurrences (words that appear commonly together) between sets of words within the documents. Cooccurrences are determined within a window of size w, where w indicates the number of words from which to infer content. For example, "sick can't pay" or "hospital bills no money" may occur commonly together and contain predictive information.”)
said software transmitting the patient communication to the patient user via the first communication methodology identified by the neural network, said patient communication indicative of the bill associated with the medical visit, wherein said first communication methodology is determined by said neural network and said patient communication is transmitted via said first communication methodology in a manner determined by said neural network to be likely to result in successful collection of the bill;
(See Shao, page 4, lines 1-6: “The present invention provides an automated system and method for predicting the likelihood of collecting on a delinquent debt of an account. The system uses one or more predictive models, for example, a neural network, to evaluate individual debt holder accounts and predict the amount that will be collected on each account based on learned relationships among known variables.”)
(See Shao, page 8, lines 23-29: “Fig. 2 illustrates an embodiment of the development and use of predictive models for delinquent debt collection. A set of historical data is selected for use in model development 230. A suitable set of data is selected wherein the data contains sufficient information to properly train the desired predictive model. Suitable criteria for inclusion in the historical dataset is developed, taking into account such factors as the type of account information historically available, and the type of information that will typically be available when making a prediction for a currently delinquent debt.”)
(See Shao, page 6, first paragraph: “There are a large number of actions that may be taken when determining how to attempt to collect a delinquent debt. For example, a letter may be sent, a phone call may be made by a collection specialist, or no action at all may be taken. Letters and phone calls may be made at a variety of different times, and may target both the debtor's home and work locations. Electronic mail may also be used to contact a debtor.”)
(See Shao, page 6, second paragraph: “The present invention includes a debt collection optimization system, which uses a predictive model to estimate the amount of a particular debt that will be recovered based upon information about the debt account and the collection actions taken on the account. The system gathers information and uses a predictive model to determine the optimal actions to use in debt collection.”)
said software receiving an indication of success or failure associated with bill collection of the bill for the patient communication and based on the success or failure said software updating a storage containing the patient data in order to train said neural network.
(See Shao, page 14, lines 16-25: “Fig. 3 is a block diagram illustrating the creation of a profile 300. The profile 300 represents a delinquent debt account as a dynamic entity. A set of data 310A-I is collected regarding the account, for example, from the financial data facility 110 and the collection efforts data facility 130 shown in Fig. 1. These data inputs 310 are then used to create a set of derived variables 320A-F, which make up the profile 300. In delinquent debt account profiles, the profile 300 is initialized by pre-collection activities, such as the cardholder masterfile, authorizations, and historical payment information. The profile 300 is dynamically updated by each transaction or other interaction with the account holder, such as a phone call, a letter, or a debt payment. The profile 300, in addition to other static data sources, becomes the base data from which predictive statistical models 250 are built.”)
(See Shao, page 14, line 26 to page 15, line 2: “Predictive models 250 each combine the predictive information from a profile of an account to create a score that exploits the meanings in the interactions between pieces of information. In one embodiment, a statistical pattern recognition technology is used to develop a statistical predictive model that calculates an estimate of how likely a delinquent debt account is to pay, and a correlation of likely payment to estimated payment amount.”)
(See Shao, page 9, lines 15-20: “The neural network is trained by properly adjusting the weights of each connection until the connections between each element are optimized to match historical outcomes based upon the set of historical inputs. The training and use of neural networks is described further in U.S. Patent No. 5,819,226, the subject matter of which is herein incorporated by reference in its entirety.”
See Shao, Abstract “The predictive model is generated using profiles of delinquent debt accounts (132) summarizing patterns of events in the accounts and the success of the collection effort in each account.”)
However, under a conservative interpretation of Shao, it could be argued that Shao does not explicitly teach the italicized portions below (in underlined font), which are taught by O’neill:
said software comprising a neural network which uses an adaptable and self- adjusting set of heuristics and responsiveness information from the plurality of patients based on the patient data to tune said neural network utilizing the patient data to identify at least a first communications methodology likely to result in successful collection of a bill and generate a patient communication for a patient user with content based on contextualized information determined based on the patient data,
said software transmitting the patient communication to the patient user via the first communication methodology identified by the neural network, said patient communication indicative of the bill associated with the medical visit, wherein said first communication methodology is determined by said neural network and said patient communication is transmitted via said first communication methodology in a manner determined by said neural network to be likely to result in successful collection of the bill;
(See O’neill, page 14, line 26 to page 15, line 2: “A debt collection selection process disclosed on pages 4 - 17 of commonly-owned and co-pending application serial number 10/011,523 for "Improved Debt Collection Practices" is also incorporated herein by reference. The 5 debtor analysis engine 312 identifies a particular debt collection strategy to use by employing either rules-based software, neural network evaluative technology, applied analytic models, scrolling and segmentation, or any other decision-making tools, which can determine the most effective collection strategy to employ of potentially several different collection strategies. In some instances, the debtor analysis engine 10 312 will not identify any change to a collection strategy being used with a debtor. In other words, newly-acquired information will not always trigger selection of another strategy or adaptation of a strategy already being used. The debtor analysis engine 312 considers factors that include a creditor's credit policies 304, external data of the debtor 306 (identified in Figure 1 by reference 15 No. 102), debtor internal data 308 (identified by reference No. 104 in Figure 1), a collector profile 310 (which includes the skill set and experience level of the collector charged with the responsibility for collecting a particular debt from a debtor), and information collected from the debtor at step 204 (shown in Figure 2). Among other things, the debtor analysis engine 312 will consider past attempts or effectiveness of 20 various strategies used against debtors with similar or identical internal data, external data, creditor policies and collector profiles. Using such historical data, it is possible to predict with reasonable certainty the effectiveness of a particular strategy of particular types of debtors. When one or more pieces of internal data, external data, or creditor policies change, or if the collector ability profile changes, the collection 25 strategy that was previously used may become ineffective. Thus, the debtor analysis engine is able to consider these changes and modify the collection strategy accordingly. Upon the identification of a debt collection strategy, the debtor analysis engine 312 determines the best interaction treatment for the particular debtor and preferably 30 provides a script or set of scripts to a debt collection agent. Text of a script can be readily stored in random access memory or on disk for retrieval and printing or 7 WO 2004/008830 PCT/US2003/022810 display on a suitable display device for use by a collection agent. By following a scripted text, questions or statements can be made to the debtor to elicit information pertinent to the debt collection effort. The script can also provide suggestions or verbatim passages to deliver to the debtor either by mail, telephone or during a 5 personal contact. Debtor responses to the script comprise the debtor input at step 204, which is re-input to the debtor analysis engine 312. At step 314, the output of the debtor analysis engine 312 will identify the appropriate strategy to be used with the debtor and if the strategy has changed since the last execution of the method shown in Figure 2, the re-classification of the debtor 10 is performed at step 314.”)
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Enhancing delinquent debt collection using statistical models of debt historical information and account events”, as taught by Shao above, with “Method to improve debt collection practices”, as taught by O’neill above, because both references are in the same field of endeavor, and both references uses neural networks to determine the optimal method and optimal texts for debt collection.
However, under a conservative interpretation of Shao in view of O’neill, it could be argued that Shao in view of O’neill does not explicitly teach the italicized portions below (in underlined font), which are taught by Elkherj:
the software receives patient data indicative of contact, insurance and demographic information for a plurality of patients;
Elkherj teaches a “user churn engine” (see para. [0050]) that uses a machine learning model trained on historical data of users (see para. [0076]), including user feature information that includes demographic information, user preferences including communication techniques (see para. [0125]), and indicative of insurance (see para. [0076]: “connected services”) and (see para. [0009]: “insurance policies”).
(See Elkherj, para. [0009]: “Particular embodiments of the subject matter described in this specification can be implemented so as to realize one or more of the following advantages. A system can determine that one or more user records are each associated with a same person, even if the user records describe the person differently (e.g., different name, different address, and so on). The system therefore allows for insights into user records that might have errors associated with input (e.g., an employee entered a name incorrectly), or errors associated with out of date information (e.g., a person has a different address on two user records because he/she moved in the intervening time period). Additionally, the system can associate all user records with respective people, and/or combine all user records into a combined user record database, which can reduce the total storage space and reduce processing time when searching for user records or information about each person. After determining unique users from a large pool of electronic user records, the system can provide identifications of services (e.g., insurance policies, subscriptions to services) to recommend to each unique user, and also identifications of services at risk for cancellation by each unique user. The system generates user interfaces that enable a reviewing user (e.g., a sales analyst) to obtain a holistic view of each unique user, and also summary data of all similar unique users.”)
(See Elkherj, para. [0050]: “The user determination system 100 includes a user churn engine 114, which can access stored information (e.g., described above) describing services each user is connected with. The user churn engine 114 can determine a likelihood (e.g., assign a score) that each user will disconnect from respective services in a period of time (e.g., the next 3 months, 6 months, and so on).”)
(See Elkherj, para. [0076]: “Features utilized for clustering can include age, a type of domicile (e.g., apartment, house), a type of geographic area a domicile is located (e.g., a suburb, a rural area, a city), specific geographic area (e.g., particular city, neighborhood in city), salary, gender, connected services, familial relationships (e.g., a user has kids), married or single, owns a car, leases a car, credit score, method of communicating with a business that provides the services (e.g., email, phone, in person), and so on.”)
(See Elkherj, para. [0125]: “The summary data of users 810 includes user preferences of types of communication techniques 824. For instance, the user interface 800 identifies that users prefer to be contacted by “Mail” firstly, and “Email” secondly. The preferences can be based off user records, and specifically the system can determine that users in the selected cluster 802 respond to communications sent via mail instead of e-mail, and/or determine that users have connected with services in a period of time (e.g., a month) after receiving mail.“)
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Enhancing delinquent debt collection using statistical models of debt historical information and account events”, as taught by Shao above, with “Method to improve debt collection practices”, as taught by O’neill above, because both references are in the same field of endeavor of debt collection, and both references uses neural networks to determine the optimal method and optimal texts for debt collection, and to further include “Feature clustering of users, user correlation database access, and user interface generation system” as taught by Elkherj above, because all three references are in the similar field of endeavor of debt collection or , and both references uses neural networks to determine the optimal method and optimal texts for communicating with the customer/debtor.
In regards to claim 21,
21. (Previously Presented) The method of claim 20 wherein the indication is of failure and the system identifying a different communication methodology from the first communication methodology and repeating the transmitting and receiving steps but for the different communication methodology.
(See Shao, page 31, line 25 to page 32, line 10: “A probability to pay model incorporates historical information such as past delinquencies, broken promises, authorizations, credit limit, behavior scores, etc. A best- time-to-call predictive model has, as an output, whether successful telephone contact is made with the correct party, and as input various information about the delinquent debt account, as well as call-attempt-specific information such as the time and the date of the call attempt. The best time to call prediction will utilize information about past successful failed contacts, but must be tempered by the fact that there is a limited "collector bandwidth" (i.e., only a limited number of accounts can be contacted within a certain time frame). The collector bandwidth is a parameter that is determined by the operational situation of the collection organization. It may be dependent on the number of employees, the length of calls, and other site-specific parameters. These site-specific parameters are supplied as fixed parameters in the best time to call decision making process. It may not always be possible for a collection organization to call each account at the precise time suggested by the best time to call predictive model, as this may be inconsistent with the organization's available operational loads and legal restrictions.”)
In regards to claim 22,
22. (Previously Presented) The method of claim 20 wherein based on the indication being failure, the software determines one or more parameters associated with one or more different patient user(s) that have resulted in bill collection success, the one or more different patient user(s) being selected by the software from one or more similar cohorts of patients.
(See Shao, page 40, lines 11-21: “In one embodiment, individual optimized account-level value predictions are rolled- up at the portfolio level. Typically, late-delinquency accounts are sold as a group, or portfolio, to a secondary debt collection agency. The secondary collection agency will evaluate the expected collection return from the portfolio in order to determine a reasonable purchase price. Using a predictive model and the global optimization methods disclosed herein, a secondary collection agency can estimate the maximum expected collection rate on all of the accounts in a portfolio (assuming that properly optimized collection actions will be taken on each of the accounts). The secondary collection agency can also estimate the cost of the optimized collection actions that will be taken on the portfolio accounts. This produces a global value estimate for the entire portfolio, and aids in setting a proper price for the worth of the portfolio.”)
In regards to claim 23,
23. (Previously Presented) The method of claim 20 wherein based on the indication being failure, the software generates a second patient communication with different content compared to the patient communication based on one or more parameters associated with one or more different patient user(s) that have resulted in bill collection success.
(See Shao, page 35, lines 4-16: “Modeling the success or failure of a particular collection action is complicated due to the fact that by making action recommendations, the underlying distribution on which the model was built is changed (i.e. a feedback loop is created, because each current action taken effects the likelihood of the consequences of future actions). In order to explicitly model collection action effects to obtain better recoveries, it is preferable to avoid creating too many distinct actions to be monitored, to prevent undesired feedback. Assume, therefore, that all possible actions have been aggregated into a small number of action groups (e.g., soft reminder letter, harsh reminder letter, soft reminder call, harsh reminder call, threat to shut off authorizations, threat to close account, offer of partial pay, offer to re- age, etc.), denoted as a1, a2 ..., aq. Furthermore, assume that building individual predictive models that estimate the probability to pay for each action or action sequence is practically undesirable. Two different embodiments of the modeling process may be used, either modeling the effect of a single action, or modeling the effect of action sequences.”)
(See Shao, page 38, lines 12-18: “ “By scoring the entire population, the desired segmentation is obtained. The probability to pay given a complex action is computed either by using the predictive model estimate or by using the previously discussed prior probability. The prior probability for a population segment is computed as the probability to pay given all possible actions whose marginal probabilities exceed the corresponding thresholds 7}, Tjt...,Tκ. As sufficient data is gathered for sparse complex actions, the prior probability can be modified to reflect the success or failure of the complex action.”)
In regards to claim 24,
24. (Previously Presented) The method of claim 20 wherein based on the indication being failure, the software sends the patient communication the patient user using a second communication methodology based on one or more parameters associated with one or more different patient user(s) that have resulted in bill collection success.
(See Shao, page 35, lines 4-16: “Modeling the success or failure of a particular collection action is complicated due to the fact that by making action recommendations, the underlying distribution on which the model was built is changed (i.e. a feedback loop is created, because each current action taken effects the likelihood of the consequences of future actions). In order to explicitly model collection action effects to obtain better recoveries, it is preferable to avoid creating too many distinct actions to be monitored, to prevent undesired feedback. Assume, therefore, that all possible actions have been aggregated into a small number of action groups (e.g., soft reminder letter, harsh reminder letter, soft reminder call, harsh reminder call, threat to shut off authorizations, threat to close account, offer of partial pay, offer to re- age, etc.), denoted as a1, a2 ..., aq. Furthermore, assume that building individual predictive models that estimate the probability to pay for each action or action sequence is practically undesirable. Two different embodiments of the modeling process may be used, either modeling the effect of a single action, or modeling the effect of action sequences.”)
(See Shao, page 38, lines 12-18: “ “By scoring the entire population, the desired segmentation is obtained. The probability to pay given a complex action is computed either by using the predictive model estimate or by using the previously discussed prior probability. The prior probability for a population segment is computed as the probability to pay given all possible actions whose marginal probabilities exceed the corresponding thresholds 7}, Tjt...,Tκ. As sufficient data is gathered for sparse complex actions, the prior probability can be modified to reflect the success or failure of the complex action.”)
In regards to claim 25,
25. (Previously Presented) The method of claim 20 wherein the patient data used to generate the patient communication for the patient user is at least in part based on patient data from one or more patient users who are different from the patient user.
(See Shao, page 38, lines 12-18: “ “By scoring the entire population, the desired segmentation is obtained. The probability to pay given a complex action is computed either by using the predictive model estimate or by using the previously discussed prior probability. The prior probability for a population segment is computed as the probability to pay given all possible actions whose marginal probabilities exceed the corresponding thresholds 7}, Tjt...,Tκ. As sufficient data is gathered for sparse complex actions, the prior probability can be modified to reflect the success or failure of the complex action.”)
(See Shao, page 40, lines 11-21: “In one embodiment, individual optimized account-level value predictions are rolled- up at the portfolio level. Typically, late-delinquency accounts are sold as a group, or portfolio, to a secondary debt collection agency. The secondary collection agency will evaluate the expected collection return from the portfolio in order to determine a reasonable purchase price. Using a predictive model and the global optimization methods disclosed herein, a secondary collection agency can estimate the maximum expected collection rate on all of the accounts in a portfolio (assuming that properly optimized collection actions will be taken on each of the accounts). The secondary collection agency can also estimate the cost of the optimized collection actions that will be taken on the portfolio accounts. This produces a global value estimate for the entire portfolio, and aids in setting a proper price for the worth of the portfolio.”)
In regards to claim 26,
26. (Previously Presented) The method of claim 20 wherein the patient data used to generate the patient communication for the patient user based on patient data is associated with the patient user.
(See Shao, page 40, line 22 to page 41, line 3: “In another embodiment, results are globally aggregated across a portfolio of accounts, but different statistical predictive models are constructed and used for different segments of the portfolio of accounts. This embodiment allows additional individual tailoring of predictive models to represent a particular account type. Such a set of predictive models may more precisely predict collection results for their particular account segment, resulting in improved overall global predictions of collection results.”)
In regards to claim 27,
27. (Previously Presented) The method of claim 20 wherein based on receiving an indication of failure associated with the patient user being unresponsive, the software uses a set of actions previously applied to a different patient user to the patient user to generate and send a new communication and/or send the communication with a different communication methodology.
(See Shao, page 35, lines 4-16: “Modeling the success or failure of a particular collection action is complicated due to the fact that by making action recommendations, the underlying distribution on which the model was built is changed (i.e. a feedback loop is created, because each current action taken effects the likelihood of the consequences of future actions). In order to explicitly model collection action effects to obtain better recoveries, it is preferable to avoid creating too many distinct actions to be monitored, to prevent undesired feedback. Assume, therefore, that all possible actions have been aggregated into a small number of action groups (e.g., soft reminder letter, harsh reminder letter, soft reminder call, harsh reminder call, threat to shut off authorizations, threat to close account, offer of partial pay, offer to re- age, etc.), denoted as a1, a2 ..., aq. Furthermore, assume that building individual predictive models that estimate the probability to pay for each action or action sequence is practically undesirable. Two different embodiments of the modeling process may be used, either modeling the effect of a single action, or modeling the effect of action sequences.”)
(See Shao, page 38, lines 12-18: “ “By scoring the entire population, the desired segmentation is obtained. The probability to pay given a complex action is computed either by using the predictive model estimate or by using the previously discussed prior probability. The prior probability for a population segment is computed as the probability to pay given all possible actions whose marginal probabilities exceed the corresponding thresholds 7}, Tjt...,Tκ. As sufficient data is gathered for sparse complex actions, the prior probability can be modified to reflect the success or failure of the complex action.”)
(See Shao, page 40, line 22 to page 41, line 3: “In another embodiment, results are globally aggregated across a portfolio of accounts, but different statistical predictive models are constructed and used for different segments of the portfolio of accounts. This embodiment allows additional individual tailoring of predictive models to represent a particular account type. Such a set of predictive models may more precisely predict collection results for their particular account segment, resulting in improved overall global predictions of collection results.”)
In regards to claim 28,
28. (Previously Presented) The method of claim 20 wherein the patient communication includes information about discounts or a payment plan available to the patient user.
(See Shao, page 21, lines 3-7: “The set of keywords for each cluster provides contextual meaning for the cluster. For example, cluster 18 appears to deal with illness, cluster 7 appears to deal with criminal and legal issues, cluster 6 appears to deal with payment plans and settlements, and cluster 15 with foreclosure and job issues. Keywords such as "jail" appear in more than one cluster, which indicates that this word is an important component of several clusters.”)
(See Shao, page 27, lines 23-30: “The transition into state S3 is characterized by the fact that at approximately 90 days past due the cardholder's account will be closed, meaning that most cardholder accounts will not be re-opened for transactions (the exceptional cases of reopening past 90 days past due are not taken into consideration in the diagram). Therefore, there is no transition from this state to states S2, Si, or So. The account will typically be terminated, irrespective of whether the debt is paid or not. If the account holder pays his/her debt, the account will go to S5. Otherwise, if the bank wants to continue to try to collect what is owed, the account will go to the asset recovery state S4. In state the account is off the collections books (legally a debt must be written off after it is 180 days past due) and the account is worked by the asset recovery management group. Actions available to this group include arranging payment plans, taking legal actions, accepting some fraction of the owed amount, or selling the account to an external collection agency. In state S5 the account is taken off of the debt issuer's books and the account holder's relationship with the debt issuer is terminated.”)
In regards to claims 29-37, they are rejected on the same grounds as claims 20-28.
Response to Amendments
Re: Claim Rejections - 35 USC § 101
The 35 USC § 101 rejections have been amended to more clearly describe how the CAFC decision in Recentive Analytics, Inc. v. Fox Corp. April 18, 2025 (https://www.cafc.uscourts.gov/opinions-orders/23-2437.OPINION.4-18-2025_2500790.pdf) is applicable to the independent claims 20 and 29 in the present application.
More specifically, the holding in Recentive Analytics states (see page 10): “This case presents a question of first impression: whether claims that do no more than apply established methods of machine learning to a new data environment are patent eligible. We hold that they are not.”
In the edited 35 USC § 101 rejections, the holding in Recentive Analytics is directly applied to the claims in the present application.
The Examiner holds that Applicant’s description of the neural network merely describes “apply it” uses of a generic neural network. (See the description in para. [0106] and [0122] of the specification).
Re: Claim Rejections - 35 USC § 103
The 35 USC § 103 rejections have been amended to more clearly describe how the Shao reference teaches the feature “the patient communication associated with a medical visit with a medical practice”.
Conclusion
Applicants are invited to contact the Office to schedule an in-person interview to discuss and resolve the issues set forth in this Office Action. Although an interview is not required, the Office believes that an interview can be of use to resolve any issues related to a patent application in an efficient and prompt manner.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
Any inquiry concerning this communication or earlier communications should be directed to Examiner Ayal Sharon, whose telephone number is (571) 272-5614, and fax number is (571) 273-1794. The Examiner can normally be reached from Monday to Friday between 9 AM and 6 PM. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Christine M Behncke can be reached on (571) 272-8103. The fax number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
Sincerely,
/Ayal I. Sharon/
Examiner, Art Unit 3695
June 17, 2026
Prosecution Insights