Retraining transplant pharmacy staff to reduce Medicare Part B prescription billing errors post-transplant.

BACKGROUND: Medicare Part B (MedB) imposes penalties for certain errors in prescription billing of post-transplant medications, which can greatly impact pharmacy revenue. To prevent MedB billing fines, pharmacy staff must be cognizant of specific MedB requirements. OBJECTIVE: The aim of this quality improvement project was to retrain certified pharmacy technicians (CPhTs) on common billing errors and evaluate changes in error rates and potential fines after retraining. We aimed to determine if retraining CPhTs minimizes MedB prescription billing errors and reduces potential fines owed by the Vanderbilt Transplant Pharmacy (VTP) to the Center for Medicare and Medicaid Services (CMS). METHODS: This was a single center, quality improvement study including post-transplant patients with at least one MedB prescription billing error who filled prescriptions through VTP. All CPhTs involved in MedB prescription billing received retraining focused on the top three errors in MedB billing identified at VTP: early refills, missing relationship of caller to patient and/or residence of patient on order documentation, or no day supply remaining recorded on the order file. Retraining consisted of developing a training checklist, testing current knowledge levels, individualized non-punitive coaching based on technician specific errors, and retesting for knowledge retention. Outcomes included the number of prescriptions with at least one MedB prescription billing error and the projected amount of dollars fined due to errors recorded during the three months before and three months after retraining. RESULTS: Fourteen CPhTs received retraining. Average refill too soon errors decreased by 37.5% (10.7% vs. 6.7%), average missing relationship by 21.7% (7.7% vs. 6%), and day supply errors by 39.7% (1.7% vs. 1%). Error reductions equaled a 28.2% decrease (approximately $12,700) in potential fines. CONCLUSION: Retraining focused on MedB billing error successfully reduced error frequency and fines from CMS. MedB billing error fines can be costly for pharmacies dispensing high- cost medications, therefore, identifying common errors and training staff can be useful and financially prudent.

Epidemiological Analysis of Traumatic Compartment Syndromes in Germany.

Background: Traumatic compartment syndrome is a critical condition that can lead to severe, lifelong disability. Methods: This retrospective study analyzed hospital billing data from 2015 to 2022, provided by the Federal Statistical Office of Germany, to examine the demographics and trends of traumatic compartment syndrome in Germany. The analysis included cases coded with ICD-10 codes T79.60 to T79.69 and any therapeutic OPS code starting with 5-79, focusing on diagnosis year, gender, ICD-10 code, and patient age. Results: The results showed that out of 13,305 cases, the majority were in the lower leg (44.4%), with males having a significantly higher incidence than females (2.3:1 ratio). A bimodal age distribution was observed, with peaks at 22-23 and 55 years. A notable annual decline of 43.87 cases in compartment syndrome was observed, with significant decreases across different genders and age groups, particularly in males under 40 (23.68 cases per year) and in the "foot" and "lower leg" categories (16.67 and 32.87 cases per year, respectively). Conclusions: The study highlights a declining trend in traumatic CS cases in Germany, with distinct demographic patterns. Through these findings, hospitals can adjust their therapeutic regimens, and it could increase awareness among healthcare professionals about this disease.

Treated versus self-reported prevalence of chronic pain and costs of patients' health services utilization: a population-based study of health administrative databases.

Objectives: To compare treated to self-reported prevalence of chronic pain (CP) and to estimate health services utilization (HSU) costs of patients treated for CP in Alberta, Canada. Methods: Patients treated for CP were identified by the physician billing codes of health services for CP from the practitioner claims database in fiscal year 2021/22. The treated prevalence of CP (number of these patients divided by the population) was compared to the self-reported prevalence of CP previously estimated (doi:10.1371/journal.pone.0272638). Costs of patients' HSU included costs for general practitioner (GP), specialist, inpatient, emergency department, outpatient clinic services, and prescription drugs. Results: The treated prevalence of CP was 6.0% (4.4% among males and 7.8% among females) which was 30% to 41% of the self-reported prevalence. The highest treated prevalence (7.2%) was found in the age group of 18-64 years, followed by age groups of >64 years (7.0%) and <18 years (2.1%). The average cost per patient per year was $5096 ($5878 for males and $4652 for females), of which hospitalizations accounted for 65.0%, outpatient clinic visits 16.4%, ED visits 9.5%, prescription drugs 4.7%, GP visits 3.9%, and specialist visits 0.4%. The total cost of patients with CP for the health system was $1.37 billion (∼7% of total health expenditure), of which males accounted for 41.7% and females for 58.3%. Discussion: Our findings suggest that the economic burden of CP is considerable and that many people with self-reported CP do not use the public healthcare services. This can be multifactorial, including lack of availability and accessibility of publicly funded services, people's lack of awareness of available services, lower utilization due to COVID-19 pandemic, and reliance on self-management, private services, and alternative treatments. Further studies are warranted to inform future policies and health system initiatives aiming to reduce the burden of CP and improve lives of people living with it.

Geographical Differences in Surgeon Reimbursement, Volume, and Patient Characteristics in Primary Total Hip Arthroplasty.

BACKGROUND: The purpose of this study was to evaluate changes in regional and national variations in reimbursement to arthroplasty surgeons, procedural volumes, and patient populations for total hip arthroplasty (THA) from 2013 to 2021. METHODS: The Medicare Physician and Other Practitioners database was queried for all billing episodes of primary THA for each year between 2013 and 2021. Inflation-adjusted surgeon reimbursement, procedural volume, physician address, and patient characteristics were extracted for each year. Data were stratified geographically based on the United States Census regions and rural-urban commuting codes. Kruskal-Wallis and multivariable regressions were utilized. RESULTS: Between 2013 and 2021, the overall THA volume and THAs per surgeon increased at the highest rate in the West (+48.2%, +20.2%). A decline in surgeon reimbursement was seen in all regions, most notably in the Midwest (-20.3%). Between 2013 and 2021, the average number of Medicare beneficiaries per surgeon declined by 12.6%, while the average number of services performed per beneficiary increased by 18.2%. In 2021, average surgeon reimbursement was the highest in the Northeast ($1,081.15) and the lowest in the Midwest ($988.03) (P < .001). Metropolitan and rural areas had greater reimbursement than micropolitan and small towns (P < .001). Patient age, race, sex, Medicaid eligibility, and comorbidity profiles differ between regions. Increased patient comorbidities, when controlling for patient characteristics, were associated with lower reimbursement in the Northeast and West (P < .01). CONCLUSIONS: Total hip arthroplasty (THA) volume and reimbursement differ between US regions, with the Midwest exhibiting the lowest increase in volume and greatest decline in reimbursement throughout the study period. Alternatively, the West had the greatest increase in THAs per surgeon. Patient comorbidity profiles differ between regions, and increased patient comorbidity is associated with decreased reimbursement in the Northeast and the West. This information is important for surgeons and policymakers as payment models regarding reimbursement for arthroplasty continue to evolve.

"Electronic Phenotyping" Antimicrobials to Facilitate Outpatient Stewardship for Asymptomatic Bacteriuria and Urinary Tract Infection in Renal Transplant.

Asymptomatic bacteriuria and urinary tract infection in renal transplant are important antimicrobial stewardship targets but are difficult to identify within electronic medical records. We validated an "electronic phenotype" of antibacterials prescribed for these indications. This may be more useful than billing data in assessing antibiotic indication in this outpatient setting.

Trends in Medicare Submitted Charges to Allowed Payment Ratios for Ophthalmology Services.

Purpose: Many physicians charge more than the Medicare insurance program pays. Current charge-to-payment ratios in ophthalmology and trends over the years are unknown. In this work, we examined physician charge-to-payment ratios in ophthalmology across procedures and consultations. Methods: We utilized data from 100% final-action physician/supplier Part B Medicare fee-for-service (FFS) population from 2015 to 2020. We retrieved data on ophthalmic procedures and consultations, both facility-based and non-facility-based, conducted by ≥ 50 ophthalmologists. We analyzed median charge-to-payment ratios, which were calculated as submitted charges divided by the Medicare-allowed payments, between ophthalmic procedures and consultations to assess for trends over the study period. Results: We find that the median charge-to-payment ratio for all current procedural terminology (CPT) codes in 2020 was 2.23 (Interquartile range (IQR): 1.54-3.27) as compared to 2.00 (IQR: 1.39-2.92) in 2015, an overall 2.76% average annual growth rate from 2015-2020. For ophthalmic procedures, the median charge-to-payment ratio in 2020 was 3.03 (IQR: 2.13-4.41) compared to 2.79 (IQR: 1.96-3.97) in 2015, corresponding to a 2.01% AAGR from 2015-2020. For consultations, those rates were 2.06 (IQR: 1.48-2.96), 1.85 (IQR: 1.33-2.59), and 2.71%, respectively. Conclusion: We found that the submitted charge-to-Medicare payment ratios among ophthalmic procedures and consultations have steadily increased since 2015. However, there was a relatively low rate of excess charges for ophthalmology services compared to other surgical-based specialties with minimal variation among providers.

Surprise billing in intensive care unit (ICU) hospitalizations.

Intensive care unit (ICU) care is expensive for patients and providers, and utilization and spending on ICU resources have increased. The No Surprises Act, passed in 2022, specifically prohibits balance billing by ICU specialists (intensivists) for emergency and most non-emergency care. The potential economic impact of this remains unclear, given few data exist on the magnitude of balance billing in the ICU. Using the MarketScan Commercial (IBM) database, we studied hospitalizations in which ICU care was provided ("ICU hospitalizations") between 2010 and 2019. Hospitalizations were characterized as fully in-network, fully out-of-network, or "mixed" (contained both in- and out-of-network services). The share of "mixed" hospitalizations among all ICU hospitalizations rose from 26% to 33% over the study period. Over half of these mixed hospitalizations contained out-of-network services specifically delivered within the ICU. Total hospitalization spending averaged $81 047, with ICU spending averaging $15 799. On average, 11% of ICU spending within these hospitalizations was out-of-network. Patients were plausibly balance-billed in approximately one-third of ICU hospitalizations, for thousands of dollars per hospitalization. Given that the No Surprises Act prevents this type of balance billing, the portended revenue loss may lead to changes in provider negotiations with insurers concerning network status and prices, which could affect the care patients receive.

Improving Anesthesia Start Time Documentation Through a Departmental Education Initiative at Yale New Haven Hospital, New Haven, United States.

Background Reimbursement for anesthetic services in the United States utilizes a formula that incorporates procedural and patient factors with total anesthesia time. According to the Centers for Medicare & Medicaid Services and the American Society of Anesthesiologists, the period of billable time starts when the anesthesia practitioner assumes care of the patient and may include transport to the operating room from the preoperative holding area. In this report on a quality improvement effort, we implemented a departmental education initiative aimed at improving the accuracy of anesthesia start-time documentation. Methods Utilizing de-identified, internal data on surgical procedures at Yale New Haven Hospital (YNHH), New Haven, United States, the difference between documented anesthesia start and patient in-room time was determined for all cases. Those with a difference between 0-1 minute were assumed "likely underbilled," and the total revenue lost for these cases was estimated using a weighted average of institutional reimbursement per unit of time. A monthly, department-wide educational email was then introduced to inform practitioners about the guidelines around start-time documentation, and the percentage of "likely underbilled" cases and lost revenue estimates trended over a one-year period. Results Baseline data in December 2020 showed that of the 6,877 total surgical cases requiring anesthesia at YNHH, 55.1% (N=3,790) had an anesthesia start to in-room time of 0-1 minute, which were considered "likely underbilled." The average start-to-in-room time for properly recorded cases (44.9%, N=3,087) was 4.42 minutes. The baseline revenue lost in December 2020 for underbilled cases was estimated at $52,302. Over the one-year quality improvement initiative, the proportion of underbilled cases showed a downward trend, decreasing to 29.2% of total cases by November 2021. The estimate of revenue lost due to underbilling also showed a downward trend, decreasing to $29,300 in November 2021. Conclusion This quality improvement study demonstrated that a relatively simple, department-wide educational email sent monthly correlated with an improvement in anesthesia start-time documentation accuracy and a reduction in estimated revenue lost to underbilling over a one-year period.

Can Publicly Available Artificial Intelligence Successfully Identify Current Procedural Terminology Codes for Common Procedures in Neurosurgery?

OBJECTIVE: Coding for neurosurgical procedures is a complex process that is dynamically changing year to year, through the annual introduction and removal of codes and modifiers. The authors hoped to elucidate if publicly available artificial intelligence (AI) could offer solutions for neurosurgeons with regard to coding. METHODS: Multiple publicly available AI platforms were asked to provide Current Procedural Terminology (CPT) codes and Revenue Value Units (RVU) values for common neurosurgical procedures of the brain and spine with a given indication for the procedure. The responses of platforms were recorded and compared to the currently valid CPT codes used for the procedure and the amount of RVUs that would be gained. RESULTS: Six platforms and Google were asked for the appropriate CPT codes for 10 endovascular, spinal, and cranial procedures each. The highest performing platforms were as follows: Perplexity.AI identified 70% of endovascular, BingAI identified 55% of spinal, and ChatGPT 4.0 with Bing identified 75% of cranial CPT codes. With regard to RVUs, the top performer gained 78% of endovascular, 42% of spinal, and 70% of cranial possible RVUs. With regard to accuracy, AI platforms on average outperformed Google (45% vs. 25%, P = 0.04236). CONCLUSIONS: The ability of publicly available AIs to successfully code for neurosurgical procedures holds great promise in the future. Future development of AI should focus on improving accuracy with regard to CPT codes and providing supporting documentation for its decisions. Improvement on the existing capabilities of AI platforms can allow for increased operational efficiency and cost savings for practices.

Financial Viability of the No Surprises Act Independent Dispute Resolution Process: Radiology and Other Hospital-Based Specialties.

BACKGROUND. The federal No Surprises Act (NSA), designed to eliminate surprise medical billing for out-of-network (OON) care for circumstances beyond patients' control, established the independent dispute resolution (IDR) process to settle clinician-payer payment disputes for OON care. OBJECTIVE. The purpose of our study was to assess the fraction of OON claims for which radiologists and other hospital-based specialists can expect to at least break even when challenging payer-determined payments through the NSA IDR process, as a measure of the process's financial viability. METHODS. This retrospective study extracted claims from a national commercial database (Optum's deidentified Clinformatics Data Mart) for hospital-based specialties occurring on the same day as in-network emergency department (ED) visits or inpatient stays from January 2017 to December 2021. OON claims were identified. OON claims batching was simulated using IDR rules. Maximum potential recovered payments from the IDR process were estimated as the difference between the charges and the allowed amount. The percentages of claims for which the maximum potential payment and one-quarter of this amount (a more realistic payment recovery estimate) would exceed IDR fees were determined, using US$150 and US$450 fee thresholds to approximate the range of final 2024 IDR fees. These values represented the percentage of OON claims that would be financially viable candidates for IDR submission. RESULTS. Among 76,221,264 claims for hospital-based specialties associated with in-network ED visits or inpatient stays, 1,482,973 (1.9%) were OON. The maximum potential payment exceeded fee thresholds of US$150 and US$450 for 55.0% and 32.1%, respectively, of batched OON claims for radiologists and 76.8% and 61.3% of batched OON claims for all other hospital-based specialties combined. At payment of one-quarter of that amount, these values were 26.9% and 10.6%, respectively, for radiologists and 56.6% and 38.4% for all other hospital-based specialties combined. CONCLUSION. The IDR process would be financially unviable for a substantial fraction of OON claims for hospital-based specialists (more so for radiology than for other such specialties). CLINICAL IMPACT. Although the NSA enacted important patient protections, IDR fees limit clinicians' opportunities to dispute payer-determined payments and potentially undermine their bargaining power in contract negotiations. Therefore, IDR rulemaking may negatively impact patient access to in-network care.

Radiologists' Out-of-Network Billing Trends, 2007 to 2021.

PURPOSE: Given the financial hardships of surprise billing for patients, the aim of this study was to assess the degree to which radiologists effectively participate in commercial insurance networks by examining the trend in the share of radiologists' imaging claims that are out of network (OON). METHODS: A retrospective study over a 15-year period (2007-2021) was conducted using claims from Optum's deidentified Clinformatics Data Mart Database to assess the share of radiologists' imaging claims that are OON. Radiologists' annual OON rate was assessed overall as well as for claims associated with inpatient stays and emergency department (ED) visits. Rates were assessed for all imaging studies as well as by modality. Linear regression was conducted to assess OON rate time trends. RESULTS: From 2007 to 2021, 5,039,142 of radiologists' imaging claims (6.3%) were OON. This rate declined from 12.6% in 2007 to 1.1% in 2021. Over the study period, the OON rate was 5.0% during an inpatient stay and 2.1% on the same day as an ED visit that did not lead to an inpatient admission. The linear trend in the overall OON rate declined 0.74 percentage points annually (95% confidence interval [CI], -0.90 to -0.58 percentage points) over the study period. Likewise, the annual declines were 0.54 percentage points (95% CI, -0.71 to -0.36) and 0.26 percentage points (95% CI, -0.33 to -0.20 percentage points) for imaging claims associated with inpatient stays and ED visits, respectively. CONCLUSIONS: Radiologists' imaging claims that are OON has significantly declined from 2007 to a minimal level in 2021. This may indicate effective negotiations between radiologists and commercial payers and new state-level surprise billing laws.

Development and multicentre validation of the FLEX score: personalised preoperative surgical risk prediction using attention-based ICD-10 and Current Procedural Terminology set embeddings.

BACKGROUND: Preoperative knowledge of surgical risks can improve perioperative care and patient outcomes. However, assessments requiring clinician examination of patients or manual chart review can be too burdensome for routine use. METHODS: We conducted a multicentre retrospective study of 243 479 adult noncardiac surgical patients at four hospitals within the Mass General Brigham (MGB) system in the USA. We developed a machine learning method using routinely collected coding and patient characteristics data from the electronic health record which predicts 30-day mortality, 30-day readmission, discharge to long-term care, and hospital length of stay. RESULTS: Our method, the Flexible Surgical Set Embedding (FLEX) score, achieved state-of-the-art performance to identify comorbidities that significantly contribute to the risk of each adverse outcome. The contributions of comorbidities are weighted based on patient-specific context, yielding personalised risk predictions. Understanding the significant drivers of risk of adverse outcomes for each patient can inform clinicians of potential targets for intervention. CONCLUSIONS: FLEX utilises information from a wider range of medical diagnostic and procedural codes than previously possible and can adapt to different coding practices to accurately predict adverse postoperative outcomes.

Federal Advance Care Planning Policy Primer - Key Aspects, Barriers, and Opportunities.

Advance care planning (ACP) is a nuanced process where patients identify their goals and consider their preferences for medical care over time. Recent systematic reviews have shown mixed findings about the association of ACP with the provision of goal-concordant care, completion of advance directives, and health care utilization. Despite a lack of consistent benefit, patients and clinicians value ACP and policy makers at the state and federal level have been moving ACP policies forward. All fifty states have policies regarding advance directives, and federal policy has had important implications on promoting awareness of ACP and its corresponding legal documents such as advance directives. However, challenges to effectively incentivize and facilitate the delivery of high-quality ACP exist. This paper provides an overview of key federal policy aspects and barriers that affect ACP use including: limitations of Medicare ACP billing codes, disparities in telemedicine access, difficulties with interoperability of advance directives, and underutilization of ACP as a mandatory measure in federal programs. This paper highlights key opportunities to improve federal ACP policy. Because ACP is an essential part of high-quality care and is deeply embedded in state and federal policies, it is imperative that clinicians are knowledgeable about these issues so they may more effectively engage in ACP policy.

National patterns in the use of International Statistical Classification of Diseases and Health Related Problems, tenth revision Z codes in ambulatory surgery from 2016 to 2019.

BACKGROUND: In the tenth revision of the International Statistical Classification of Disease and Health Related Problems (ICD-10), Z codes were added to improve documentation and understanding of health-related social needs. We estimated national Z code use in the ambulatory surgery setting from 2016 to 2019. METHODS: Using the Nationwide Ambulatory Surgery Sample (NASS), we identified encounters for ambulatory surgery with an ICD-10 code between Z55.0 and Z65.9. Data were stratified by Z code domains from the Centers for Medicare and Medicaid Services (CMS). RESULTS: This analysis of 41,827 ambulatory surgery encounters with documented Z codes found that the most documented determinants of health related to multiparity or unwanted pregnancy, homelessness, and incarceration. There was a 16.1% increase in the use of Z codes from 2016 to 2019. CONCLUSION: Rates of Z code use in the ambulatory surgery setting are increasing with current documentation serving as a specific but not sensitive measure of socioeconomic need.

Maximizing the potential value of the nursing workforce.

This panel paper is the fifth installment in a six-part Nursing Outlook special edition based on the 2022 Emory Business Case for Nursing Summit. The 2022 summit convened national nursing, health care, and business leaders to explore possible solutions to nursing workforce crises, including the nursing shortage. Each of the summit's four panels authored a paper in this special edition on their respective topic, and this panel paper focuses on maximizing the potential value of the nursing workforce. It addresses topics including the need to create a nursing-inclusive federal health care billing system improve nursing salaries by designing/testing nurse-informed compensation models, and strengthen nursing's national professional infrastructure.

Tele-Neuropsychology: From Science to Policy to Practice.

OBJECTIVE: The primary aim of this paper is to accelerate the number of randomized experimental studies of the reliability and validity in-home tele-neuropsychological testing (tele-np-t). METHOD: We conducted a critical review of the tele-neuropsychology literature. We discuss this research in the context of the United States' public and private healthcare payer systems, including the Centers for Medicare & Medicaid Services (CMS) and Current Procedural Terminology (CPT) coding system's telehealth lists, and existing disparities in healthcare access. RESULTS: The number of tele-np publications has been stagnant since the onset of the COVID-19 pandemic. There are less published experimental studies of tele-neuropsychology (tele-np), and particularly in-home tele-np-t, than other tele-np publications. There is strong foundational evidence of the acceptability, feasibility, and reliability of tele-np-t, but relatively few studies of the reliability and validity of in-home tele-np-t using randomization methodology. CONCLUSIONS: More studies of the reliability and validity of in-home tele-np-t using randomization methodology are necessary to support inclusion of tele-np-t codes on the CMS and CPT telehealth lists, and subsequently, the integration and delivery of in-home tele-np-t services across providers and institutions. These actions are needed to maintain equitable reimbursement of in-home tele-np-t services and address the widespread disparities in healthcare access.

How to Exercise Integrity in Medical Billing: Don't Distort Prices, Don't Free-Ride on Other Physicians.

This paper proposes that billing gamesmanship occurs when physicians free-ride on the billing practices of other physicians. Gamesmanship is non-universalizable and does not exercise a competitive advantage; consequently, it distorts prices and allocates resources inefficiently. This explains why gamesmanship is wrong. This explanation differs from the recent proposal of Heath (2020. Ethical issues in physician billing under fee-for-service plans. J. Med. Philos. 45(1):86-104) that gamesmanship is wrong because of specific features of health care and of health insurance. These features are aggravating factors but do not explain gamesmanship's primary wrong-making feature, which is to cause diffuse harm not traceable to any particular patient or insurer. This conclusion has important consequences for how medical schools and professional organizations encourage integrity in billing. To avoid free-riding, physicians should ask themselves, "could all physicians bill this way?" and if not, "does the patient benefit from the distinctive service I am providing under this code?" If both answers are "no," physicians should refrain from the billing practice in question.

Estimating the completeness of physician billing claims for diabetes case ascertainment: a multiprovince investigation. / Estimation de l'exhaustivité des demandes de remboursement des médecins dans une optique de détermination des cas de diabète : étude menée dans plusieurs provinces.

INTRODUCTION: Previous research has suggested that how physicians are paid may affect the completeness of billing claims for estimating chronic disease. The purpose of this study is to estimate the completeness of physician billings for diabetes case ascertainment. METHODS: We used administrative data from eight Canadian provinces covering the period 1 April 2014 to 31 March 2016. The patient cohort was stratified into two mutually exclusive groups based on their physician remuneration type: fee-for-service (FFS), for those paid only on that basis; and non-fee-for-service (NFFS). Using diabetes prescription drug data as our reference data source, we evaluated whether completeness of disease case ascertainment varied with payment type. Diabetes incidence rates were then adjusted for completeness of ascertainment. RESULTS: The cohort comprised 86 110 patients. Overall, equal proportions received their diabetes medications from FFS and NFFS physicians. Overall, physician payment method had little impact upon the percentage of missed diabetes cases (FFS, 14.8%; NFFS, 12.2%). However, the difference in missed cases between FFS and NFFS varied widely by province, ranging from -1.0% in Nova Scotia to 29.9% in Newfoundland and Labrador. The difference between the observed and adjusted disease incidence rates also varied by province, ranging from 22% in Prince Edward Island to 4% in Nova Scotia. CONCLUSION: The difference in the loss of cases by physician remuneration method varied across jurisdictions. This loss may contribute to an underestimation of disease incidence. The method we used could be applied to other chronic diseases for which drug therapy could serve as reference data source.

Contrastive diagnostic embedding (CDE) model for automated coding - A case study using emergency department encounters.

BACKGROUND: Billing codes are utilized for medical reimbursement, clinical quality metric valuation and for epidemiologic purposes to report and follow disease trends and outcomes. The current paradigm of manual coding can be expensive, time-consuming, and subject to human error. Though automation of the billing codes has been widely reported in the literature via rule-based and supervised approaches, existing strategies lack generalizability and robustness towards large and constantly changing ICD hierarchical structure. METHOD: We propose a weakly supervised training strategy by leveraging contrastive learning, contrastive diagnosis embedding (CDE) to capture the fine semantic variations between the diagnosis codes. The approach consists of a two-phase contrastive training for generating the semantic embedding space adapted to incorporate hierarchical information of ICD-10 vocabulary and a weakly supervised retrieval scheme. Core strength of the proposed method is that it puts no limit on the 70 K ICD-10 codes set and can handle all rare codes for coding the diagnosis. RESULTS: Our CDE model outperformed string-based partial matching and ClinicalBERT embedding on three test cases (a retrospective testset, a prospective testset, and external testset) and produced an accurate prediction of rare and newly introduced diagnosis codes. A detailed ablation study showed the importance of each phase of the proposed multi-phase training. Each successive phase of training - ICD-10 group sensitive training (phase 1.1), ICD-10 subgroup sensitive training (phase 1.2), free-text diagnosis description-based training (phase 2) - improved performance beyond the previous phase of training. The model also outperformed existing supervised models like CAML and PLM-ICD and produced satisfactory performance on the rare codes. CONCLUSION: Compared to the existing rule-based and supervised models, the proposed weakly supervised contrastive learning overcomes the limitations in terms of generalization capability and increases the robustness of the automated billing. Such a model will allow flexibility through accurate billing code automation for practice convergence and gains efficiencies in a value-based care payment environment.