Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
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 32-51 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea of a mathematical relationship and method of organizing human activity without significantly more. The claims recite an abstract idea including:
32. A computer-implemented method of deploying a claims resubmission predictive model, the computer-implemented method comprising, …
access a first set of payer remit data associated with a first set of healthcare claims, a first plurality of patients, a first plurality of provider identifiers, and a first payer entity;
process the first set of payer remit data to associate each of a plurality of remit data items with at least one of the first set of healthcare claims and an outcome status indicating either approval or denial for each respective healthcare claim to generate a set of denied claims whose outcome status indicates denial;
generate and send a request data package for submission to a server to apply a claims resubmission probability artificial intelligence/machine learning (AI/ML) model which is configured to, for each claim associated with the set of denied claims, predict a likelihood of being approved upon resubmission, wherein the claims resubmission probability AI/ML model has been trained using a first set of historical claim and remit data, and wherein training of the claims resubmission probability AI/ML model comprises:
accessing the first set of historical claim and remit data, wherein the first set of historical claim and remit data is associated with a training data set, and wherein the training data set is configured to be updated on a predetermined schedule,
accessing, from a data store, one or more model parameters, and
…
receive, from the server, a set of claims resubmission prediction data associated with the set of denied claims and associated with the first set of healthcare claims, the set of claims resubmission prediction data comprising: prediction indicators indicating a likelihood of being approved upon resubmission;
access first resubmission parameters associated with a first provider identifier of the first plurality of provider identifiers, the first provider identifier associated with a first subset of the set of denied claims;
process the first subset of denied claims using at least the first resubmission parameters and the respective prediction indicators indicating the likelihood of being approved upon resubmission, to generate a set of high-priority denied claims from the first subset of denied claims;
transmit the set of high-priority denied claims to a triage system for processing based on a high-priority flag;
access a second set of payer remit data associated with the set of high-priority denied claims and indicating a respective approval status or denial status of each claim; and
process the second set of payer remit data to associate the second set of payer remit data with the set of high-priority denied claims to determine approval or denial status for each claim in the set of high-priority denied claims, wherein the claims resubmission probability AI/ML model is additionally trained or updated using the determined approval or denial status for each claim in the set of high-priority denied claims.
33. The computer-implemented method of claim 32 further comprising specific executable instructions that:
process the first subset of denied claims using at least the first resubmission parameters, the respective prediction indicators indicating the likelihood of being approved upon resubmission, to generate a set of medium-priority denied claims from the first subset of denied claims; and
transmit the set of medium-priority denied claims to the triage system for processing based on a medium-priority flag.
34. The computer-implemented method of claim 33 further comprising specific executable instructions that:
process the first subset of denied claims using at least the first resubmission parameters, the respective prediction indicators indicating the likelihood of being approved upon resubmission, and to generate a set of low-priority denied claims from the first subset of denied claims; and
transmit the set of low-priority denied claims to the triage system for processing based on a low-priority flag.
35. The computer-implemented method of claim 34 further comprising specific executable instructions that:
process the first subset of denied claims using at least the first resubmission parameters, the respective prediction indicators indicating the likelihood of being approved upon resubmission, to automatically generate a set of write-off denied claims from the first subset of denied claims; and
flag the set of write-off denied claims with a write-off flag indicating that the set of write-off denied claims should be blocked from being sent to the triage system for processing.
36. The computer-implemented method of claim 32 further comprising specific executable instructions that:
access second resubmission parameters associated with a second provider identifier of the first plurality of provider identifiers, the second provider identifier associated with a second subset of set of denied claims.
37. The computer-implemented method of claim 32, wherein the set of claims resubmission prediction data includes claim denial reason data associated with at least one healthcare claim of the first set of healthcare claims indicating one or more reasons for denial of a respective claim.
38. The computer-implemented method of claim 32, wherein determining that the error threshold has been satisfied comprises determining that at least one of a false negative rate threshold or a false positive rate threshold has been satisfied.
39. The computer-implemented method of claim 32, wherein the one or more model parameters comprises at least one of a claim amount, procedure type, false positive rate, false negative rate, or payer entity.
This judicial exception is not integrated into a practical application because additional limitation of:
an AI/ML model…
based on the first set of historical claim and remit data and the one or more model parameters, training the claims resubmission probability AI/ML model, wherein the training includes validating the claims resubmission probability AI/ML model by determining that an error threshold has been satisfied;…
[receiving and transmitting data]
a server…
merely link the abstract idea to the field of computers and AI/ML models. The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the step of training does not improve the functioning of a computer or technical field (MPEP 2106.05(a)) and the generic computer parts such as a server, memory, and processors are instructions to apply an abstract idea to a generic computer. MPEP 2106.05(f).
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.
Claims 32-51 are rejected under 35 U.S.C. 103 as being unpatentable over US20230084146A1 to Singh et al and US20190244302A1 to Levy.
Singh teaches claim 32. A computer-implemented method of deploying a claims resubmission predictive model, the computer-implemented method comprising, as implemented by one or more computing devices configured with specific executable instructions to: (Singh abs “A method of predicting an outcome of a prior-authorization, claim, or appeal…” and fig. 5)
access a first set of payer remit data associated with a first set of healthcare claims, (“claim forms”) a first plurality of patients, (“patient name”) a first plurality of provider identifiers, (“provider name”) and a first payer entity; (“payer name”) (Singh para 50 “The first step 110 of the method 100 includes receiving raw data from a set of documents or other sources described above. The data may be from the sources described above, or other documents that can include, for example, insurance claim forms prepared by one or more providers, insurance claim appeal forms prepared by one or more providers, doctor notes associated with one or more patients, explanation of benefit forms prepared by one or more payers, claim denial letters prepared by one or more payers, claim acceptance letters prepared by one or more payers, or any combination thereof.” Singh para 56 “structuring of the data may refer to identifying and categorizing each piece of data in each of the labeled sets. For example, the data in each labeled set can be categorized into one or more of a plurality of fields, including, for example, a provider name, a provider address, a provider phone number, a payer name, a payer address, a payer phone number, a patient name, a patient address, a patient phone number, a patient social security number or other identifier, a patient date of birth, doctor notes associated with the patient and/or procedure, a procedure code, a diagnosis code, a services provided date, a billed amount, a claim number, an insurance group number, a plan number, a claim denial date, an allowed amount, a deductible amount, a co-pay amount, an adjustment code, a payment amount, an appeal deadline date, or any combination thereof.”)
process the first set of payer remit data to associate each of a plurality of remit data items with at least one of the first set of healthcare claims and an outcome status indicating either approval or denial for each respective healthcare claim to generate a set of denied claims whose outcome status indicates denial; (Singh para 53 “ labeling sets of the training data selected during the second step 120. More specifically, the selected training data is labeled by grouping the selected training data into sets, for example, each set being designated as (1) being associated with an insurance prior-authorization, claim, or appeal that has been accepted by the payer or (2) being associated with an insurance prior-authorization, claim, or appeal that has been denied by the payer. For example, the selected data can be labeled with designations such as “appeal accepted,” “appeal denied,” “claim paid,” “claim not paid,” “claim paid in full,” “claim paid in part,” etc.”)
generate and send a request data package for submission to a server to apply a claims resubmission probability artificial intelligence/machine learning (AI/ML) model which is configured to, for each claim associated with the set of denied claims, predict a likelihood of being approved upon resubmission, (Singh para 78 “The fourth step 240 includes inputting the structured data subsets from the second step 220 into the machine learning algorithm selected during the third step 230.” Singh fig. 5 and para 105 “Any of the trained or untrained machine learning algorithms described herein can be stored in the memory device of the provider system 510, the server 530, and the cloud system 540, or any combination thereof.”) wherein the claims resubmission probability AI/ML model has been trained using a first set of historical claim and remit data, (Singh para 61 “The fifth step 150 of the method 100 includes inputting the labeled, structured training data from step 140 into a machine learning algorithm…”) and wherein training of the claims resubmission probability AI/ML model comprises:
accessing the first set of historical claim and remit data, wherein the first set of historical claim and remit data is associated with a training data set, and wherein the training data set is configured to be updated on a predetermined schedule, (Singh para 61 “The fifth step 150 of the method 100 includes inputting the labeled, structured training data from step 140 into a machine learning algorithm…” Singh fig. 1 shows that inputting the training data is “to train machine learning algorithm” [150]. Singh para 52 “the training data will have a time window of one year, two years, three years, 6 months or other suitable time windows. In some examples, the machine learning algorithms associated with each payer in a database may be newly trained with a new time window of data. This will ensure that the machine learning algorithms are up to date with the practices of each payer, which may evolve over time and therefore not be biased by older data.”)
accessing, from a data store, one or more model parameters, and (Applicant’s model parameters are not model parameters like learning rate and number of layers. Applicant’s model parameters are what phosita would consider to be input data such as “payer entity”. Claim 39. Singh para 56 “ the data in each labeled set can be categorized into one or more of a plurality of fields, including, for example, a provider name, … payer name… a billed amount,… an allowed amount…”)
based on the first set of historical claim and remit data and the one or more model parameters, training the claims resubmission probability AI/ML model, (Singh para 61 “The fifth step 150 of the method 100 includes inputting the labeled, structured training data from step 140 into a machine learning algorithm, such as, for example, decision trees (“DT”), Bayesian networks (“BN”), artificial neural network (“ANN”), support vector machines (“SVM”), deep learning algorithms…” Singh fig 1 [160] “Input training data to train machine learning algorithm”) wherein the training includes validating the claims resubmission probability AI/ML model by determining that an error threshold has been satisfied; (Singh para 69 “validation data associated with granted appeals and/or denied appeals can be labeled as such, structured, and input into the machine learning algorithm to determine whether the machine learning algorithm is sufficiently trained to predict appeal outcome (e.g., the machine learning algorithm predicts the correct outcome more than 80% of the time, the machine learning algorithm predicts the correct outcome more than 90% of the time, predicts the correct outcome more than 99% of the time, etc.)”)
receive, from the server, a set of claims resubmission prediction data associated with the set of denied claims and associated with the first set of healthcare claims, the set of claims resubmission prediction data comprising: prediction indicators indicating a likelihood of being approved upon resubmission; (Signh para 78 “The fourth step 240 includes inputting the structured data subsets from the second step 220 into the machine learning algorithm … the fourth step 240 includes a first substep 242 that predicts an outcome of an appeal of the denied claim to the payer. For example, the selected machine learning algorithm predicts whether an appeal will be successful…”)
access first resubmission parameters associated with a first provider identifier of the first plurality of provider identifiers, the first provider identifier associated with a first subset of the set of denied claims; (Signh para 79 “one of the subsets of data from the second step 220 contains a procedure code associated with knee surgery and another one of the subsets of data from the second step 220 contains words from doctor notes that do not contain the word “knee” or “surgery,” but instead include words associated with another procedure (e.g., the procedure code was mistakenly entered). In this example, the machine learning algorithm (which has been trained using the method 100) may predict that an appeal of the denied claim will be denied…” The parameter is the doctor note.)
process the first subset of denied claims using at least the first resubmission parameters and the respective prediction indicators indicating the likelihood of being approved upon resubmission, (Signh para 79 “one of the subsets of data from the second step 220 contains a procedure code associated with knee surgery and another one of the subsets of data from the second step 220 contains words from doctor notes that do not contain the word “knee” or “surgery,” but instead include words associated with another procedure (e.g., the procedure code was mistakenly entered). In this example, the machine learning algorithm (which has been trained using the method 100) may predict that an appeal of the denied claim will be denied…” The parameter is the doctor note.)
access a second set of payer remit data associated with the set of (Singh para 45 “After that, the payer will deny or accept the appeal (or partially) which will provide more output data that can be fed into machine learning algorithms….”)
process the second set of payer remit data to associate the second set of payer remit data with the set of of high-priority denied claims. (Singh para 47-48 ‘Payers generally utilize algorithms for accepting or denying prior-authorizations, claims, or appeals, and those algorithms are largely unknown outside the payers, …[0048] Therefore, machine learning algorithms have been developed based on the data sources available that can automatically train and input data into the algorithms to predict the outcomes of the payers' decisions, based on the documents and other data available.”)
Singh doesn’t teach high-priority claim triage.
However, Levy teaches how to generate a set of high-priority denied claims from the first subset of denied claims; (Levy para 81 “The worklist editing menu 1520 may allow a manager to create or edit a worklist according to desired criteria, which may include priority level, payer type priority, client priority…” Levy para 86 “By selecting a high, medium, or low priority, a multiplier may be applied to the claims which causes the claims to move higher or lower on the list accordingly. For example, by choosing claim age as the highest priority, a large multiple is applied to older claims, which makes these claims appear higher on the list…”)
transmit the set of high-priority denied claims to a triage system for processing based on a high-priority flag; (Levy para 86 “By selecting a high, medium, or low priority, a multiplier may be applied to the claims which causes the claims to move higher or lower on the list accordingly. For example, by choosing claim age as the highest priority, a large multiple is applied to older claims, which makes these claims appear higher on the list… In another example, by choosing previous follow-up as the highest priority, claims that have not previously been followed-up on will appear higher than other claims of the same value that have been followed-up on. This is discussed in greater detail in the priority examples below.”)
process the second set of payer remit data to associate the second set of payer remit data with the set of high-priority denied claims to determine approval or denial status for each claim… (Levy para 81 “As shown in FIG. 15D, once categories are selected for each of the priority levels, additional filter options may become available. By way of example, if the category “Denials” is designated as high priority, the user may be able to specify a narrower type of denial on which the application can focus…”)
Singh, Levy and the claims are all for denying insurance responsibility for paying for contracted services. It would have been obvious to a person having ordinary skill in the art, at the time of filing, to update data periodically and add priority status to claims to add the highest predictive value to Singh application, by including more input data.
Singh teaches claim 33. The computer-implemented method of claim 32 further comprising specific executable instructions that:
process the first subset of denied claims using at least the first resubmission parameters, the respective prediction indicators indicating the likelihood of being approved upon resubmission,
(Signh para 79 “one of the subsets of data from the second step 220 contains a procedure code associated with knee surgery and another one of the subsets of data from the second step 220 contains words from doctor notes that do not contain the word “knee” or “surgery,” but instead include words associated with another procedure (e.g., the procedure code was mistakenly entered). In this example, the machine learning algorithm (which has been trained using the method 100) may predict that an appeal of the denied claim will be denied…” The parameter is the doctor note.)
Singh doesn’t teach medium priority.
However, Levy teaches to generate a set of medium-priority denied claims from the first subset of denied claims; and (Levy para 81 “Categories may be selected for high priority, medium priority, and low priority levels.”)
transmit the set of medium-priority denied claims to the triage system for processing based on a medium-priority flag. (Levy para 81 “As shown in FIG. 15D, once categories are selected for each of the priority levels, additional filter options may become available. By way of example, if the category “Denials” is designated as high priority, the user may be able to specify a narrower type of denial on which the application can focus… Levy para 81 “Categories may be selected for high priority, medium priority, and low priority levels.”)
Singh teaches claim 34. The computer-implemented method of claim 33 further comprising specific executable instructions that:
process the first subset of denied claims using at least the first resubmission parameters, the respective prediction indicators indicating the likelihood of being approved upon resubmission…(Signh para 79 “one of the subsets of data from the second step 220 contains a procedure code associated with knee surgery and another one of the subsets of data from the second step 220 contains words from doctor notes that do not contain the word “knee” or “surgery,” but instead include words associated with another procedure (e.g., the procedure code was mistakenly entered). In this example, the machine learning algorithm (which has been trained using the method 100) may predict that an appeal of the denied claim will be denied…” The parameter is the doctor note.)
Singh doesn’t teach priority levels.
However, Lvey teaches how to generate a set of low-priority denied claims from the first subset of denied claims; and
transmit the set of low-priority denied claims to the triage system for processing based on a low-priority flag. (Levy para 81 “As shown in FIG. 15D, once categories are selected for each of the priority levels, additional filter options may become available. By way of example, if the category “Denials” is designated as high priority, the user may be able to specify a narrower type of denial on which the application can focus… Levy para 81 “Categories may be selected for high priority, medium priority, and low priority levels.”)
Singh teaches claim 35. The computer-implemented method of claim 34 further comprising specific executable instructions that:
process the first subset of denied claims using at least the first resubmission parameters, the respective prediction indicators indicating the likelihood of being approved upon resubmission… (Singh para 85 “The seventh step 270 includes ordering the appeal candidate entries in the database such that appeal candidate entries that are either (1) entries predicted to result in paid claim on appeal or entries determined to have a high percentage likelihood of resulting in a paid claim on appeal…”)
Singh doesn’t teach a write-off.
However, Levy teaches how to automatically generate a set of write-off denied claims from the first subset of denied claims; and
flag the set of write-off denied claims with a write-off flag indicating that the set of write-off denied claims should be blocked from being sent to the triage system for processing. (Singh para 85 “Filtering by claim balance data may include excluding claims with claim balances outside of a desired range from a worklist. For example, claims with claim balances beneath a threshold may be filtered.”)
Levy teaches claim 36. The computer-implemented method of claim 32 further comprising specific executable instructions that:
access second resubmission parameters associated with a second provider identifier of the first plurality of provider identifiers, the second provider identifier associated with a second subset of set of denied claims. (Levy para 44 “Medical claim information may also include a unique identifier for the claim, a patient identifier, a client identifier, a claim status, CPT codes, and other metadata helpful in processing a medical claim. … A client identifier may be a name, client number, or other identifying information used to identify the medical service provider seeking payment and processing of the medical claim.”)
Singh teaches claim 37. The computer-implemented method of claim 32, wherein the set of claims resubmission prediction data includes claim denial reason data associated with at least one healthcare claim of the first set of healthcare claims indicating one or more reasons for denial of a respective claim. (Singh para 55 “data may take the form of unstructured, natural language text as in the case of an appeal letter, claim letter, claim denial letter, doctor's notes…” Singh para 93 “During the first step 310, the raw data in the denial communication is received by and stored…”)
Singh teaches claim 38. The computer-implemented method of claim 32, wherein determining that the error threshold has been satisfied comprises determining that at least one of a false negative rate threshold or a false positive rate threshold has been satisfied. (Singh para 69 “input into the machine learning algorithm to determine whether the machine learning algorithm is sufficiently trained to predict appeal outcome (e.g., the machine learning algorithm predicts the correct outcome more than 80% of the time, the machine learning algorithm predicts the correct outcome more than 90% of the time, predicts the correct outcome more than 99% of the time, etc.)”)
Singh teaches claim 39. The computer-implemented method of claim 32, wherein the one or more model parameters comprises at least one of a claim amount, procedure type, false positive rate, false negative rate, or payer entity. (Singh para 56 “ the data in each labeled set can be categorized into one or more of a plurality of fields, including, for example, a provider name, a provider address, a provider phone number, a payer name, a payer address, a payer phone number, a patient name, a patient address, a patient phone number, a patient social security number or other identifier, a patient date of birth, doctor notes associated with the patient and/or procedure, a procedure code, a diagnosis code, a services provided date, a billed amount, a claim number, an insurance group number, a plan number, a claim denial date, an allowed amount, a deductible amount, a co-pay amount, an adjustment code, a payment amount, an appeal deadline date, or any combination thereof.”)
Singh teaches claim 40. A system for tuning a claims resubmission predictive model, the system comprising:
one or more processors;
a network communications interface;
a memory; and (Singh fig. 4-6)
computer code stored in the memory, wherein the computer code, when retrieved from the memory and executed by the one or more processors causes the one or more processors to: (Singh abs “A method of predicting an outcome of a prior-authorization, claim, or appeal…” and fig. 5)
access a first set of payer remit data associated with a first set of healthcare claims, (“claim forms”) a first plurality of patients, (“patient name”) a first plurality of provider identifiers, (“provider name”) and a first payer entity; (“payer name”) (Singh para 50 “The first step 110 of the method 100 includes receiving raw data from a set of documents or other sources described above. The data may be from the sources described above, or other documents that can include, for example, insurance claim forms prepared by one or more providers, insurance claim appeal forms prepared by one or more providers, doctor notes associated with one or more patients, explanation of benefit forms prepared by one or more payers, claim denial letters prepared by one or more payers, claim acceptance letters prepared by one or more payers, or any combination thereof.” Singh para 56 “structuring of the data may refer to identifying and categorizing each piece of data in each of the labeled sets. For example, the data in each labeled set can be categorized into one or more of a plurality of fields, including, for example, a provider name, a provider address, a provider phone number, a payer name, a payer address, a payer phone number, a patient name, a patient address, a patient phone number, a patient social security number or other identifier, a patient date of birth, doctor notes associated with the patient and/or procedure, a procedure code, a diagnosis code, a services provided date, a billed amount, a claim number, an insurance group number, a plan number, a claim denial date, an allowed amount, a deductible amount, a co-pay amount, an adjustment code, a payment amount, an appeal deadline date, or any combination thereof.”)
identify a set of denied claims of the first set of healthcare claims based on the first set of payer remit data; (Singh para 53 “ labeling sets of the training data selected during the second step 120. More specifically, the selected training data is labeled by grouping the selected training data into sets, for example, each set being designated as (1) being associated with an insurance prior-authorization, claim, or appeal that has been accepted by the payer or (2) being associated with an insurance prior-authorization, claim, or appeal that has been denied by the payer. For example, the selected data can be labeled with designations such as “appeal accepted,” “appeal denied,” “claim paid,” “claim not paid,” “claim paid in full,” “claim paid in part,” etc.”)
generate and send a request data package for submission to a server to apply a claims resubmission probability model which is configured to, for each claim associated with the set of denied claims, predict a likelihood of being approved upon resubmission, (Singh para 78 “The fourth step 240 includes inputting the structured data subsets from the second step 220 into the machine learning algorithm selected during the third step 230.” Singh fig. 5 and para 105 “Any of the trained or untrained machine learning algorithms described herein can be stored in the memory device of the provider system 510, the server 530, and the cloud system 540, or any combination thereof.”) wherein the claims resubmission probability model has been trained using a first set of historical claim and remit data, (Singh para 61 “The fifth step 150 of the method 100 includes inputting the labeled, structured training data from step 140 into a machine learning algorithm…”) and wherein training of the claims resubmission probability model comprises:
accessing the first set of historical claim and remit data, wherein the first set of historical claim and remit data is associated with a training data set, and wherein the training data set is configured to be updated on a predetermined schedule, (Singh para 61 “The fifth step 150 of the method 100 includes inputting the labeled, structured training data from step 140 into a machine learning algorithm…” Singh fig. 1 shows that inputting the training data is “to train machine learning algorithm” [150]. Singh para 52 “the training data will have a time window of one year, two years, three years, 6 months or other suitable time windows. In some examples, the machine learning algorithms associated with each payer in a database may be newly trained with a new time window of data. This will ensure that the machine learning algorithms are up to date with the practices of each payer, which may evolve over time and therefore not be biased by older data.”)
accessing, from a data store, one or more model parameters, and (Applicant’s model parameters are not model parameters like learning rate and number of layers. Applicant’s model parameters are what phosita would consider to be input data such as “payer entity”. Claim 39. Singh para 56 “ the data in each labeled set can be categorized into one or more of a plurality of fields, including, for example, a provider name, … payer name… a billed amount,… an allowed amount…”)
based on the first set of historical claim and remit data and the one or more model parameters, training the claims resubmission probability model, wherein the training includes validating the claims resubmission probability model by determining that an error threshold has been satisfied; (Singh para 69 “validation data associated with granted appeals and/or denied appeals can be labeled as such, structured, and input into the machine learning algorithm to determine whether the machine learning algorithm is sufficiently trained to predict appeal outcome (e.g., the machine learning algorithm predicts the correct outcome more than 80% of the time, the machine learning algorithm predicts the correct outcome more than 90% of the time, predicts the correct outcome more than 99% of the time, etc.)”)
receive, from the server, a set of claims resubmission prediction data associated with the set of denied claims; (Signh para 78 “The fourth step 240 includes inputting the structured data subsets from the second step 220 into the machine learning algorithm … the fourth step 240 includes a first substep 242 that predicts an outcome of an appeal of the denied claim to the payer. For example, the selected machine learning algorithm predicts whether an appeal will be successful…”)
access a second set of payer remit data associated with the set (Singh para 45 “After that, the payer will deny or accept the appeal (or partially) which will provide more output data that can be fed into machine learning algorithms….”)
process the second set of payer remit data to determine approval or denial status for each claim in the set of (Singh para 47-48 ‘Payers generally utilize algorithms for accepting or denying prior-authorizations, claims, or appeals, and those algorithms are largely unknown outside the payers, …[0048] Therefore, machine learning algorithms have been developed based on the data sources available that can automatically train and input data into the algorithms to predict the outcomes of the payers' decisions, based on the documents and other data available.”)
Singh doesn’t teach high-priority claim triage.
However, Levy teaches how to generate a set of high-priority denied claims from the set of denied claims based on the claims resubmission prediction data; (Levy para 81 “The worklist editing menu 1520 may allow a manager to create or edit a worklist according to desired criteria, which may include priority level, payer type priority, client priority…” Levy para 86 “By selecting a high, medium, or low priority, a multiplier may be applied to the claims which causes the claims to move higher or lower on the list accordingly. For example, by choosing claim age as the highest priority, a large multiple is applied to older claims, which makes these claims appear higher on the list…”)
transmit the set of high-priority denied claims to a triage system for processing based on a high-priority flag…(Levy para 86 “By selecting a high, medium, or low priority, a multiplier may be applied to the claims which causes the claims to move higher or lower on the list accordingly. For example, by choosing claim age as the highest priority, a large multiple is applied to older claims, which makes these claims appear higher on the list… In another example, by choosing previous follow-up as the highest priority, claims that have not previously been followed-up on will appear higher than other claims of the same value that have been followed-up on. This is discussed in greater detail in the priority examples below.”)
Singh, Levy and the claims are all for denying insurance responsibility for paying for contracted services. It would have been obvious to a person having ordinary skill in the art, at the time of filing, to update data periodically and add priority status to claims to add the highest predictive value to Singh application, by including more input data.
Levy teaches claim 41. (New) The system of Claim 40,
wherein the computer code, when retrieved from the memory and executed by the one or more processors causes the one or more processors to:
access first resubmission parameters associated with a first provider identifier of the first plurality of provider identifiers, the first provider identifier associated with a first subset of the set of denied claims. (Levy para 44 “Medical claim information may also include a unique identifier for the claim, a patient identifier, a client identifier, a claim status, CPT codes, and other metadata helpful in processing a medical claim. … A client identifier may be a name, client number, or other identifying information used to identify the medical service provider seeking payment and processing of the medical claim.”)
Levy teaches claim 42. (New) The system of Claim 41,
wherein the computer code, when retrieved from the memory and executed by the one or more processors causes the one or more processors to:
access second resubmission parameters associated with a second provider identifier of the first plurality of provider identifiers, the second provider identifier associated with a second subset of set of denied claims. (Levy para 44 “Medical claim information may also include a unique identifier for the claim, a patient identifier, a client identifier, a claim status, CPT codes, and other metadata helpful in processing a medical claim. … A client identifier may be a name, client number, or other identifying information used to identify the medical service provider seeking payment and processing of the medical claim.” It’s a plurality of claims, so there is more than one provider.)
Singh teaches claim 43. (New) The system of Claim 41,
wherein the set of claims resubmission prediction data comprises prediction indicators for each of the set of denied claims indicating a likelihood of the respective denied claim being approved upon resubmission, wherein the computer code, when retrieved from the memory and executed by the one or more processors causes the one or more processors to:
process the set of denied claims using at least the first resubmission parameters and the prediction indicators… (Signh para 79 “one of the subsets of data from the second step 220 contains a procedure code associated with knee surgery and another one of the subsets of data from the second step 220 contains words from doctor notes that do not contain the word “knee” or “surgery,” but instead include words associated with another procedure (e.g., the procedure code was mistakenly entered). In this example, the machine learning algorithm (which has been trained using the method 100) may predict that an appeal of the denied claim will be denied…” The parameter is the doctor note.)
Levy teaches how to generate a set of medium-priority denied claims from the set of denied claims; and
transmit the set of medium-priority denied claims to the triage system for processing based on a medium-priority flag. (Levy para 86 “By selecting a high, medium, or low priority, a multiplier may be applied to the claims which causes the claims to move higher or lower on the list accordingly. For example, by choosing claim age as the highest priority, a large multiple is applied to older claims, which makes these claims appear higher on the list… In another example, by choosing previous follow-up as the highest priority, claims that have not previously been followed-up on will appear higher than other claims of the same value that have been followed-up on. This is discussed in greater detail in the priority examples below.”)
Singh teaches claim 44. (New) The system of Claim 41,
wherein the set of claims resubmission prediction data comprises prediction indicators for each of the set of denied claims indicating a likelihood of the respective denied claim being approved upon resubmission, wherein the computer code, when retrieved from the memory and executed by the one or more processors causes the one or more processors to:
process the set of denied claims using at least the first resubmission parameters and the prediction indicators… (Signh para 79 “one of the subsets of data from the second step 220 contains a procedure code associated with knee surgery and another one of the subsets of data from the second step 220 contains words from doctor notes that do not contain the word “knee” or “surgery,” but instead include words associated with another procedure (e.g., the procedure code was mistakenly entered). In this example, the machine learning algorithm (which has been trained using the method 100) may predict that an appeal of the denied claim will be denied…” The parameter is the doctor note.)
Levy teaches how to generate a set of low-priority denied claims from the set of denied claims; and
transmit the set of low-priority denied claims to the triage system for processing based on a low-priority flag. (Levy para 86 “By selecting a high, medium, or low priority, a multiplier may be applied to the claims which causes the claims to move higher or lower on the list accordingly. For example, by choosing claim age as the highest priority, a large multiple is applied to older claims, which makes these claims appear higher on the list… In another example, by choosing previous follow-up as the highest priority, claims that have not previously been followed-up on will appear higher than other claims of the same value that have been followed-up on. This is discussed in greater detail in the priority examples below.”)
Singh teaches claim 45. (New) The system of Claim 41,
wherein the set of claims resubmission prediction data comprises prediction indicators for each of the set of denied claims indicating a likelihood of the respective denied claim being approved upon resubmission, wherein the computer code, when retrieved from the memory and executed by the one or more processors causes the one or more processors to:
process the set of denied claims using at least the first resubmission parameters and the prediction indicators… (Singh para 85 “The seventh step 270 includes ordering the appeal candidate entries in the database such that appeal candidate entries that are either (1) entries predicted to result in paid claim on appeal or entries determined to have a high percentage likelihood of resulting in a paid claim on appeal…”)
Singh doesn’t teach a write-off.
However, Levy teaches how to automatically generate a set of write-off denied claims from the set of denied claims; and
flag the set of write-off denied claims with a write-off flag indicating that the set of write-off denied claims should be blocked from being sent to the triage system for processing. (Singh para 85 “Filtering by claim balance data may include excluding claims with claim balances outside of a desired range from a worklist. For example, claims with claim balances beneath a threshold may be filtered.”)
Singh teaches claim 46. (New) The system of Claim 40,
wherein the set of claims resubmission prediction data includes claim denial reason data associated with at least one denied claim of the set of denied claims indicating one or more reasons for denial of a respective claim. (Singh para 55 “data may take the form of unstructured, natural language text as in the case of an appeal letter, claim letter, claim denial letter, doctor's notes…” Singh para 93 “During the first step 310, the raw data in the denial communication is received by and stored…”)
Singh teaches claim 47. (New) The system of Claim 40,
wherein to determine that the error threshold has been satisfied, the computer code, when retrieved from the memory and executed by the one or more processors causes the one or more processors to determine that at least one of a false negative rate threshold or a false positive rate threshold has been satisfied. (Singh para 69 “input into the machine learning algorithm to determine whether the machine learning algorithm is sufficiently trained to predict appeal outcome (e.g., the machine learning algorithm predicts the correct outcome more than 80% of the time, the machine learning algorithm predicts the correct outcome more than 90% of the time, predicts the correct outcome more than 99% of the time, etc.)”)
Singh teaches claim 48. (New) The system of Claim 40,
wherein the one or more model parameters comprises at least one of a claim amount, procedure type, false positive rate, false negative rate, or payer entity. (Singh para 56 “ the data in each labeled set can be categorized into one or more of a plurality of fields, including, for example, a provider name, a provider address, a provider phone number, a payer name, a payer address, a payer phone number, a patient name, a patient address, a patient phone number, a patient social security number or other identifier, a patient date of birth, doctor notes associated with the patient and/or procedure, a procedure code, a diagnosis code, a services provided date, a billed amount, a claim number, an insurance group number, a plan number, a claim denial date, an allowed amount, a deductible amount, a co-pay amount, an adjustment code, a payment amount, an appeal deadline date, or any combination thereof.”)
Singh teaches claim 49. (New) One or more non-transitory storage media storing instructions that, when
executed by at least one processor, cause the at least one processor to: (Singh abs “A method of predicting an outcome of a prior-authorization, claim, or appeal…” and figs. 4-6)
access a first set of payer remit data associated with a first set of healthcare claims, (“claim forms”) a first plurality of patients, (“patient name”) a first plurality of provider identifiers, (“provider name”) and a first payer entity; (“payer name”) (Singh para 50 “The first step 110 of the method 100 includes receiving raw data from a set of documents or other sources described above. The data may be from the sources described above, or other documents that can include, for example, insurance claim forms prepared by one or more providers, insurance claim appeal forms prepared by one or more providers, doctor notes associated with one or more patients, explanation of benefit forms prepared by one or more payers, claim denial letters prepared by one or more payers, claim acceptance letters prepared by one or more payers, or any combination thereof.” Singh para 56 “structuring of the data may refer to identifying and categorizing each piece of data in each of the labeled sets. For example, the data in each labeled set can be categorized into one or more of a plurality of fields, including, for example, a provider name, a provider address, a provider phone number, a payer name, a payer address, a payer phone number, a patient name, a patient address, a patient phone number, a patient social security number or other identifier, a patient date of birth, doctor notes associated with the patient and/or procedure, a procedure code, a diagnosis code, a services provided date, a billed amount, a claim number, an insurance group number, a plan number, a claim denial date, an allowed amount, a deductible amount, a co-pay amount, an adjustment code, a payment amount, an appeal deadline date, or any combination thereof.”)
identify a set of denied claims of the first set of healthcare claims based on the first set of payer remit data; (Singh para 53 “ labeling sets of the training data selected during the second step 120. More specifically, the selected training data is labeled by grouping the selected training data into sets, for example, each set being designated as (1) being associated with an insurance prior-authorization, claim, or appeal that has been accepted by the payer or (2) being associated with an insurance prior-authorization, claim, or appeal that has been denied by the payer. For example, the selected data can be labeled with designations such as “appeal accepted,” “appeal denied,” “claim paid,” “claim not paid,” “claim paid in full,” “claim paid in part,” etc.”)
generate and send a request data package for submission to a server to apply a claims resubmission probability model which is configured to, for each claim associated with the set of denied claims, predict a likelihood of being approved upon resubmission, (Singh para 78 “The fourth step 240 includes inputting the structured data subsets from the second step 220 into the machine learning algorithm selected during the third step 230.” Singh fig. 5 and para 105 “Any of the trained or untrained machine learning algorithms described herein can be stored in the memory device of the provider system 510, the server 530, and the cloud system 540, or any combination thereof.”) wherein the claims resubmission probability model has been trained using a first set of historical claim and remit data, (Singh para 61 “The fifth step 150 of the method 100 includes inputting the labeled, structured training data from step 140 into a machine learning algorithm…”)
and wherein training of the claims resubmission probability model comprises:
accessing the first set of historical claim and remit data, wherein the first set of historical claim and remit data is associated with a training data set, and wherein the training data set is configured to be updated on a predetermined schedule, (Singh para 61 “The fifth step 150 of the method 100 includes inputting the labeled, structured training data from step 140 into a machine learning algorithm…” Singh fig. 1 shows that inputting the training data is “to train machine learning algorithm” [150]. Singh para 52 “the training data will have a time window of one year, two years, three years, 6 months or other suitable time windows. In some examples, the machine learning algorithms associated with each payer in a database may be newly trained with a new time window of data. This will ensure that the machine learning algorithms are up to date with the practices of each payer, which may evolve over time and therefore not be biased by older data.”)
accessing, from a data store, one or more model parameters, and based on the first set of historical claim and remit data and the one or more model parameters, (Applicant’s model parameters are not model parameters like learning rate and number of layers. Applicant’s model parameters are what phosita would consider to be input data such as “payer entity”. Claim 39. Singh para 56 “ the data in each labeled set can be categorized into one or more of a plurality of fields, including, for example, a provider name, … payer name… a billed amount,… an allowed amount…”) training the claims resubmission probability model, wherein the training includes validating the claims resubmission probability model by determining that an error threshold has been satisfied; (Singh para 69 “validation data associated with granted appeals and/or denied appeals can be labeled as such, structured, and input into the machine learning algorithm to determine whether the machine learning algorithm is sufficiently trained to predict appeal outcome (e.g., the machine learning algorithm predicts the correct outcome more than 80% of the time, the machine learning algorithm predicts the correct outcome more than 90% of the time, predicts the correct outcome more than 99% of the time, etc.)”)
receive, from the server, a set of claims resubmission prediction data associated with the set of denied claims; (Signh para 79 “one of the subsets of data from the second step 220 contains a procedure code associated with knee surgery and another one of the subsets of data from the second step 220 contains words from doctor notes that do not contain the word “knee” or “surgery,” but instead include words associated with another procedure (e.g., the procedure code was mistakenly entered). In this example, the machine learning algorithm (which has been trained using the method 100) may predict that an appeal of the denied claim will be denied…” The parameter is the doctor note.)
access a second set of payer remit data associated with the set of high-priority denied claims and indicating a respective approval status or denial status of each claim; and (Singh para 45 “After that, the payer will deny or accept the appeal (or partially) which will provide more output data that can be fed into machine learning algorithms….”)
process the second set of payer remit data to determine approval or denial status for each claim in the set of high-priority denied claims, wherein the claims resubmission probability model is additionally trained or updated using the determined approval or denial status for each claim in the set of high-priority denied claims. (Singh para 47-48 ‘Payers generally utilize algorithms for accepting or denying prior-authorizations, claims, or appeals, and those algorithms are largely unknown outside the payers, …[0048] Therefore, machine learning algorithms have been developed based on the data sources available that can automatically train and input data into the algorithms to predict the outcomes of the payers' decisions, based on the documents and other data available.”)
Singh doesn’t teach high-priority claim triage.
However, Levy teaches how to generate a set of high-priority denied claims from the set of denied claims based on the claims resubmission prediction data; (Levy para 81 “The worklist editing menu 1520 may allow a manager to create or edit a worklist according to desired criteria, which may include priority level, payer type priority, client priority…” Levy para 86 “By selecting a high, medium, or low priority, a multiplier may be applied to the claims which causes the claims to move higher or lower on the list accordingly. For example, by choosing claim age as the highest priority, a large multiple is applied to older claims, which makes these claims appear higher on the list…”)
transmit the set of high-priority denied claims to a triage system for processing based on a high-priority flag; (Levy para 86 “By selecting a high, medium, or low priority, a multiplier may be applied to the claims which causes the claims to move higher or lower on the list accordingly. For example, by choosing claim age as the highest priority, a large multiple is applied to older claims, which makes these claims appear higher on the list… In another example, by choosing previous follow-up as the highest priority, claims that have not previously been followed-up on will appear higher than other claims of the same value that have been followed-up on. This is discussed in greater detail in the priority examples below.”)
Singh, Levy and the claims are all for denying insurance responsibility for paying for contracted services. It would have been obvious to a person having ordinary skill in the art, at the time of filing, to update data periodically and add priority status to claims to add the highest predictive value to Singh application, by including more input data.
Levy teaches claim 50. (New) The non-transitory storage media of Claim 49, wherein the instructions that,
when executed by at least one processor, cause the at least one processor to:
access first resubmission parameters associated with a first provider identifier of the first plurality of provider identifiers, the first provider identifier associated with a first subset of the set of denied claims. (Levy para 44 “Medical claim information may also include a unique identifier for the claim, a patient identifier, a client identifier, a claim status, CPT codes, and other metadata helpful in processing a medical claim. … A client identifier may be a name, client number, or other identifying information used to identify the medical service provider seeking payment and processing of the medical claim.”)
Levy teaches claim 51. (New) The non-transitory storage media of Claim 50, wherein the instructions that,
when executed by at least one processor, cause the at least one processor to:
access second resubmission parameters associated with a second provider identifier of the first plurality of provider identifiers, the second provider identifier associated with a second subset of set of denied claims. (Levy para 44 “Medical claim information may also include a unique identifier for the claim, a patient identifier, a client identifier, a claim status, CPT codes, and other metadata helpful in processing a medical claim. … A client identifier may be a name, client number, or other identifying information used to identify the medical service provider seeking payment and processing of the medical claim.” Levy has multiple claims with multiple providers.)
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Austin Hicks whose telephone number is (571)270-3377. The examiner can normally be reached Monday - Thursday 8-4 PST.
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.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Mariela Reyes can be reached at (571) 270-1006. The fax phone 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.
/AUSTIN HICKS/ Primary Examiner, Art Unit 2142
Prosecution Insights