High fidelity clinical documentation improvement (CDI) smart scoring systems and methods
A clinical documentation improvement (CDI) smart scoring method may include predicting, via per-condition diagnosis machine learning (ML) models and based on clinical evidence received by a system, a probability that a medical condition is under-documented and, via per-condition documentation ML models and based on documentation received by the system, a probability that a medical condition is over-documented. The under- and over-documentation scores are combined in view of special indicators and queryability factors, which can also be evaluated using ML query prediction models, to generate an initial CDI score. This CDI score can be further adjusted, if necessary or desired, to account for factors such as length of stay, payer, patient location, CDI review timing, etc. The final CDI score can be used to prioritize patient cases for review by CDI specialists to quickly and efficiently identify meaningful CDI opportunities.
1. A computer-implemented clinical documentation improvement (CDI) scoring method, comprising:
receiving, by a CDI system over a network, real-time medical data from a data source, the real-time medical data comprising clinical data and documentation for a patient case, the CDI system embodied on at least one server machine having at least one processor and non-transitory computer memory;
determining, by the CDI system, a medical condition from at least one of the clinical data or the documentation for the patient case;
determining a condition-specific diagnosis machine learning (ML) model and a condition-specific documentation ML model, the determining performed by the CDI system for the medical condition;
invoking a condition prediction process, the condition prediction process comprising applying the diagnosis ML model to a plurality of input items associated with the patient case, the condition prediction process producing an under-documentation score for the medical condition;
invoking a documentation analysis process, the documentation analysis process comprising applying the documentation ML model to the plurality of input items associated with the patient case, the documentation analysis process producing an over-documentation score for the medical condition;
generating a CDI score for the patient case, the generating performed by the CDI system based at least on the under-documentation score and the over-documentation score for the medical condition, the under-documentation score representing a probability that the medical condition is clinically valid and is not sufficiently documented, the over-documentation score representing a probability that the medical condition is not clinically valid but is documented;
performing, by the CDI system, the determining, invoking, and generating steps for each patient case found in the real-time medical data received from the data source to thereby generate CDI scores for a plurality of patient cases;
prioritizing, by the CDI system, the plurality of patient cases based on the CDI scores; and
communicating, by the CDI system to a computing device communicatively connected to the CDI system, the plurality of patient cases prioritized based on the CDI scores for display on the computing device.
2. The computer-implemented CDI scoring method according to claim 1 , further comprising:
parsing the real-time medical data received from the data source, the parsing comprising extracting any feature present in the clinical data and extracting any medical concept present in the documentation for the patient case, wherein the medical condition is determined from at least one of an extracted feature or an extracted medical concept.
3. The computer-implemented CDI scoring method according to claim 1 , wherein determination of the diagnosis ML model comprises evaluating a plurality of diagnosis ML models specifically configured for the medical condition, each condition-specific diagnosis ML model of the plurality of diagnosis ML models comprising a multitude of individual factors drawn from the patient case.
4. The computer-implemented CDI scoring method according to claim 1 , wherein application of the diagnosis ML model to the plurality of input items associated with the patient case generates a condition predictiveness factor, the condition predictiveness factor representing a probability that a patient identified by the patient case has the medical condition.
5. The computer-implemented CDI scoring method according to claim 4 , wherein application of the documentation ML model to the plurality of input items associated with the patient case generates a documentation confidence factor, the documentation confidence factor representing a probability that the documentation received thus far for the patient case positively documents the medical condition.
6. The computer-implemented CDI scoring method according to claim 5 , wherein the under-documentation score is generated by multiplying the condition predictiveness factor by an inverse of the documentation confidence factor and wherein the over-documentation score is generated by multiplying the documentation confidence factor by an inverse of the condition predictiveness factor.
7. The computer-implemented CDI scoring method according to claim 1 , wherein the generating further comprises evaluating individual under-documentation scores for all medical conditions indicated in the patient case and determining an overall under-documentation score representing a probability that the patient case has query opportunities for clarifying or augmenting insufficient documentation.
8. The computer-implemented CDI scoring method according to claim 7 , wherein the generating further comprises statistically aggregating individual over-documentation scores across all medical conditions indicated in the patient case and determining an overall over-documentation score representing a probability that the patient case has query opportunities for addressing over-documentation.
9. The computer-implemented CDI scoring method according to claim 8 , wherein the generating further comprises combining the overall under- and over-documentation scores to yield an initial CDI score.
10. The computer-implemented CDI scoring method according to claim 9 , wherein the initial CDI score is adjusted to account for at least one of a length of stay, a payer, a patient location, or a review timing indicated in the patient case to thereby generate the CDI score for the patient case.
11. A clinical documentation improvement (CDI) scoring system, comprising:
at least one processor;
at least one non-transitory computer readable medium; and
stored instructions translatable by the at least one processor to perform:
receiving, over a network, real-time medical data from a data source, the real-time medical data comprising clinical data and documentation for a patient case;
determining a medical condition from at least one of the clinical data or the documentation for the patient case;
determining a condition-specific diagnosis machine learning (ML) model and a condition-specific documentation ML model for the medical condition;
invoking a condition prediction process, the condition prediction process comprising applying the diagnosis ML model to a plurality of input items associated with the patient case, the condition prediction process producing an under-documentation score for the medical condition;
invoking a documentation analysis process, the documentation analysis process comprising applying the documentation ML model to the plurality of input items associated with the patient case, the documentation analysis process producing an over-documentation score for the medical condition;
generating a CDI score for the patient case based at least on the under-documentation score and the over-documentation score for the medical condition, the under-documentation score representing a probability that the medical condition is clinically valid and is not sufficiently documented, the over-documentation score representing a probability that the medical condition is not clinically valid but is documented;
performing the determining, invoking, and generating steps for each patient case found in the real-time medical data received from the data source to thereby generate CDI scores for a plurality of patient cases;
prioritizing the plurality of patient cases based on the CDI scores; and
communicating, to a computing device communicatively connected to the CDI system, the plurality of patient cases prioritized based on the CDI scores for display on the computing device.
12. The CDI scoring system of claim 11 , wherein the stored instructions are further translatable by the at least one processor to perform:
parsing the real-time medical data received from the data source, the parsing comprising extracting any feature present in the clinical data and extracting any medical concept present in the documentation for the patient case, wherein the medical condition is determined from at least one of an extracted feature or an extracted medical concept.
13. The CDI scoring system of claim 11 , wherein determination of the diagnosis ML model comprises evaluating a plurality of diagnosis ML models specifically configured for the medical condition, each condition-specific diagnosis ML model of the plurality of diagnosis ML models comprising a multitude of individual factors drawn from the patient case.
14. The CDI scoring system of claim 11 , wherein application of the diagnosis ML model to the plurality of input items associated with the patient case generates a condition predictiveness factor, the condition predictiveness factor representing a probability that a patient identified by the patient case has the medical condition.
15. The CDI scoring system of claim 14 , wherein application of the documentation ML model to the plurality of input items associated with the patient case generates a documentation confidence factor, the documentation confidence factor representing a probability that the documentation received thus far for the patient case positively documents the medical condition.
16. The CDI scoring system of claim 15 , wherein the under-documentation score is generated by multiplying the condition predictiveness factor by an inverse of the documentation confidence factor and wherein the over-documentation score is generated by multiplying the documentation confidence factor by an inverse of the condition predictiveness factor.
17. The CDI scoring system of claim 11 , wherein the generating further comprises evaluating individual under-documentation scores for all medical conditions indicated in the patient case and determining an overall under-documentation score representing a probability that the patient case has query opportunities for clarifying or augmenting insufficient documentation.
18. The CDI scoring system of claim 17 , wherein the generating further comprises statistically aggregating individual over-documentation scores across all medical conditions indicated in the patient case and determining an overall over-documentation score representing a probability that the patient case has query opportunities for addressing over-documentation.
19. The CDI scoring system of claim 18 , wherein the generating further comprises combining the overall under- and over-documentation scores to yield an initial CDI score.
20. The CDI scoring system of claim 19 , wherein the initial CDI score is adjusted to account for at least one of a length of stay, a payer, a patient location, or a review timing indicated in the patient case to thereby generate the CDI score for the patient case.