Start Your Search

V. Gupta



Author of

  • +

    EP06 - e-Poster Session 6 (ID 57)

    • Event: e-Health 2018 Virtual Meeting
    • Type: e-Poster Session
    • Track: Executive
    • Presentations: 1
    • +

      EP06.01 - Re-Architecting Interoperability: A Creative Use of Constraints (ID 360)

      V. Gupta, InfoClin Analytics; Toronto/CA

      • Abstract
      • Slides

      Purpose/Objectives: To propose a new approach to interoperability in healthcare. Current technological approaches to interoperability generate increased effort for end-users, are costly to implement and do not provide sufficient value to stakeholders that they would pay for it out of their own pockets or contribute to its sustainability. We suggest that by introducing an economic value-based approach to data exchange, we can create a more compelling opportunity for adoption across multiple stakeholders. Using diabetes as a use case; we determined the economic value that accrues to the healthcare system that is associated with exchanging a single data element.

      Methodology/Approach: We used the British Design Council’s Double Diamond design method to identify and define key high value use cases. We identified screening for diabetic retinopathy as a use case due to the increasing burden of diabetes on the Canadian healthcare system. We developed and used an economic framework that assesses the Net Present Value of a stream of current costs (and savings by not screening 30% of patients) and compared them to the savings generated by increased screening (and associated costs).

      Finding/Results: Approximately 59% of the 3.4 million patients with diabetes in Canada have some form of diabetic retinopathy (DR), and 1.5% of those patients have vision loss due to DR (30,000/year). We identified that exchanging information on whether a diabetic patient completed their annual eye exam allowed for prevention, early detection of DR and timely treatment. Using our economic framework, we calculated the value of exchanging a single data element. (By definition, constraining exchenge to a single data element minimizes costs and minimizes effort.) This allowed us to find that the overall cost savings associated with exchanging one data element between optometrists (who provide eye exams) and family doctors (who provide care to patients with diabetes) would equal approximately $140-210 million per year. re-architecting interoperability -dr.png

      Conclusion/Implications/Recommendations: The implication of this method is to highlight that current interoperability approaches do not sufficiently incentivize stakeholders to exchange information. Interoperability is perceived as an end in itself; but our value-based approach suggests that it should be used only as a means to an end, in order to better support care planning, program design, and clinical decision support. As we have found, it is not necessary to exchange hundreds of data elements amongst stakeholders. Rather, sharing just one high-value data element for diabetes can benefit thousands of patients and provide cost-savings to government, with minimal data collection efforts from vendors and physicians. We believe we have found an alternative approach to solving the interoperability challenge.

      140 Character Summary: We undertook to find out why interoperability in healthcare is so elusive. We believe we have found an alternative approach to solving the interoperability problem.

      Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.

  • +

    OS14 - Leveraging Successes and Lessons Learned (ID 21)

    • Event: e-Health 2018 Virtual Meeting
    • Type: Oral Session
    • Track: Clinical Delivery
    • Presentations: 1
    • +

      OS14.03 - Pandemic Preparedness in Canada: Who Has Been Vaccinated? (ID 363)

      V. Gupta, InfoClin Analytics; Toronto/CA

      • Abstract
      • Slides

      Purpose/Objectives: Physicians need to identify patients within their practice who have been vaccinated against influenza and those who have not been vaccinated in case of a potential influenza pandemic. In most EMR systems in Canada, a physician would initiate this search through a simple text search. The effectiveness of the search would depend on the skill of the physician and whether the EMR would allow the physician to easily separate patients who have had their vaccinations from those who have not. We have developed a Smart Algorithm that can rapidly and reliably identify records of true vaccinations.

      Methodology/Approach: Our Smart Algorithm identifies those patients who received a vaccination, those who declined a vaccination and those who were not vaccinated in the clinic but received their vaccination elsewhere. We compared our Smart Algorithm, to a search that might be done by a physician in a traditional EMR system using a simple text search of “flu”. In the vaccination record with the ability to use the ‘not given’ flag to identify those who did not receive the flu shot. The ‘Reason for Not Given’ is a text field and we assume that the doctors would not be able to query it for information about why it was not given (patient refused, patient is allergic, vaccination received elsewhere).

      Finding/Results: The Smart Algorithm accurately identifies a greater number of patient records which have either have or have not been administered vaccinations for influenza from a file with a total number of 20,878 vaccine records. A Simple Search misses as many as 1/3 of patients who did get an immunization and mis-identifies about 2.5% of patients as having had one when in fact they had not. Algorithm TOTAL + - Simple Search + 4,184 112 4,296 - 2,091 14,491 16,582 TOTAL 6,275 14,603 20,878 Sensitivity (Recall): 66.7% Specificity: 99.2% PPV (Precision): 97.4% NPV: 87.4% F-score: 79.1

      Conclusion/Implications/Recommendations: When the simple search in the EMR finds a patient that has been vaccinated, the chances are high that the patient was indeed vaccinated (PPV=97.4%). However, a simple search performed by a physician would miss approximately 33.3% of vaccinated patients (Sensitivity of 66.7%), requiring them to spend time reviewing each record manually or contacting patients directly to confirm the patient’s vaccination status. When compared to a simple search in the EMR, the Smart Algorithm performs better and allows for better time management for physicians and their staff.

      140 Character Summary: Pandemic Preparedness: We developed a Smart Algorithm that rapidly and reliably identifies records of vaccinations for influenza compared to a simple EMR search.

      Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.

  • +

    OS19 - Expanding EMR Use in Communities (ID 29)

    • Event: e-Health 2018 Virtual Meeting
    • Type: Oral Session
    • Track: Technical/Interoperability
    • Presentations: 1
    • +

      OS19.03 - Standardizing Free Text in EMRs: Automating Data Coding in Primarycare (ID 361)

      V. Gupta, InfoClin Analytics; Toronto/CA

      • Abstract
      • Slides

      Purpose/Objectives: To automate the standardization of clinical data in electronic medical record systems (EMRs). Data collected in EMRs do not always accurately and consistently map to medical concepts that physicians need to identify individuals at high risk of specific health conditions and their complications. Asking physicians to enter structured data puts too great a burden on already busy clinicians. Addressing data quality problems through new advanced artificial intelligence and machine learning techniques holds the promise of saving doctors hundreds of hours of coding time while increasing data quality in their EMRs.

      Methodology/Approach: We developed several advanced text mining Smart Algorithms to identify patients with a range of chronic diseases that can lead to severe complications such as blindness, cancer, stroke, kidney failure, heart attacks and death. We compared our Smart algorithms (SA) to a simple search (SS) that a reasonable physician might conduct in their EMR for the same disease.

      Finding/Results: Disease: GERD Breast Ca Prostate Ca Lung Ca Schizophrenia Depression Diabetes Polycystic Ovarian Disease % Identified by SS 67% 63% 58% 68% 88% 81% 90% 46% % Missed by SS 33% 37% 42% 32% 12% 19% 10% 54% % False Positive by SS 3% 13% 13% 24% 7% 3% 5% 7% % Accuracy of SS 80% 73% 68% 72% 90% 88% 92% 62% % Identified = patients identified by SS compared to SA.Table. Comparison of Physician Search in EMR against a Smart Algorithm % Missed = percent patients with disease but not identified by SS. % False positive = percent of patients that SS detects who actually don’t have the disease. % Accuracy = how well SS compares to SA in terms of not making any errors in detection. Simple searches in EMRs can find as many as 90% of patients with a disease and as few as 46% (Table). Although the false positive rate is quite low (3-13% for the most part), it can be as high as 24% in some cases (Lung ca). The accuracy of SS in EMR can be as high as 92% (e.g., diabetes), but is usually lower than that and can get as low as 62%.

      Conclusion/Implications/Recommendations: Smart data cleaning approaches are required to overcome problems raised by heath data inconsistency and to help physicians accurately identify high risk and targeted groups. These data standardization algorithms are also important for more advanced uses such as predictive analytics, patient engagement, health system management reports and machine learning applications.

      140 Character Summary: New advanced Smart data cleaning Algorithms standardize data in EMRs more consistently than doctors.. Smart Algorithms help doctors provide better patient care.

      Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.

  • +

    RF01 - Innovation Is No Longer an Option in Digital Health (ID 10)

    • Event: e-Health 2018 Virtual Meeting
    • Type: Rapid Fire Session
    • Track: Clinical Delivery
    • Presentations: 1
    • +

      RF01.03 - Addressing the Opioid Epidemic with Evidence (ID 376)

      V. Gupta, InfoClin Analytics; Toronto/CA

      • Abstract
      • Slides

      Purpose/Objectives: Canada is facing a severe opioid overdose crisis. At least 2,458 Canadians died from opioid?related overdose in 2016. Many patients want to control their use of opioids, however physicians find tapering opioids time-consuming and fraught with error. We aimed to develop a web-based Opioid Manager tool for physicians that can dramatically reduce the time needed to develop a customized, safe opioid tapering schedule for patients, allowing physicians to spend more time counselling patients and to see more patients that need guidance with opioid use.

      Methodology/Approach: We developed a web-based Opioid Manager, based on the latest opioid management guidelines. We incorporated an expert system Opioid tapering algorithm based on the heuristics used by expert clinical pharmacists. The tapering algorithm is used by the Opioid Manager to generate a prescription that can be given to a patient to assist them in safely lowering their dose of opioids slowly over time. We validated the tapering algorithm through review with a clinical pharmacist.

      Finding/Results: Creating a customized opioid taper schedule takes a highly experienced physician or pharmacist 30?45 minutes per patient; longer for less experienced providers. Opioid tapering is emotionally and physically demanding on patients, requiring support from the extended care team to ensure that they follow through on the taper and that problems are solved quickly and efficiently, before the patient loses confidence and reverts to prior high doses. The tapering algorithm can automatically generate an opioid tapering prescription in less than 30 seconds. The algorithm automatically calculates the patient’s total current opioid dose and proposes a tapering regime that effectively utilizes real-world tablet formulations during the taper.

      Conclusion/Implications/Recommendations: The Opioid tapering calculator is reliable and can potentially save a physician or clinical pharmacist at least 30 minutes per patient, allowing them more time to explain the tapering process to the patient, to support the patient more effectively in tapering their dose and to assist more patients in managing their opioid medications. Payors and policy makers may also be interested in this tool, as it provides them a way to reduce the burden of addictions and reduce the overall cost to provide opioid management services.

      140 Character Summary: New Opioid Manager saves doctors 30 minutes per patient while decreasing opioid overdose risk.

      Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.