ABSTRACT
BACKGROUND: Preoperative imaging before parathyroidectomy can localize adenomas and reduce unnecessary bilateral neck explorations. We hypothesized that (1) the utility of preoperative imaging varies substantially depending on the preoperative probability of having adenoma(s) and (2) that a selective imaging approach based on this probability could avoid unnecessary patient costs and radiation. METHODS: We analyzed 3,577 patients who underwent parathyroidectomy for primary hyperparathyroidism from 2001 to 2022. The predicted probability of patients having single or double adenoma versus hyperplasia was estimated using logistic regression. We then estimated the relationship between the predicted probability of single/double adenoma and the likelihood that sestamibi or 4-dimensional computed tomography was helpful for operative planning. Current Medicare costs and published data on radiation dosing were used to calculate costs and radiation exposure from non-helpful imaging. RESULTS: The mean age was 62 ± 13 years; 78% were women. Adenomas were associated with higher mean calcium (11.2 ± 0.74 mg/dL) and parathyroid hormone levels (140.6 ± 94 pg/mL) than hyperplasia (9.8 ± 0.52 mg/dL and 81.4 ± 66 pg/mL). The probability that imaging helped with operative planning increased from 12% to 65%, as the predicted probability of adenoma increased from 30% to 90%. For every 10,000 patients, a selective approach to imaging that considered the preoperative probability of having adenomas could save patients up to $3.4 million and >239,000 millisieverts of radiation. CONCLUSION: Rather than imaging all patients with primary hyperparathyroidism, a selective strategy that considers the probability of having adenomas could reduce costs and avoid excess radiation exposure.
Subject(s)
Adenoma , Hyperparathyroidism, Primary , Parathyroid Neoplasms , United States , Humans , Female , Aged , Middle Aged , Male , Parathyroidectomy/methods , Hyperparathyroidism, Primary/diagnostic imaging , Hyperparathyroidism, Primary/surgery , Parathyroid Neoplasms/diagnostic imaging , Parathyroid Neoplasms/surgery , Technetium Tc 99m Sestamibi , Hyperplasia/diagnostic imaging , Medicare , Radiopharmaceuticals , Parathyroid Hormone , Adenoma/diagnostic imaging , Adenoma/surgeryABSTRACT
Background: Patient-reported outcomes (PRO) allow clinicians to measure health-related quality of life (HRQOL) and understand patients' treatment priorities, but obtaining PRO requires surveys which are not part of routine care. We aimed to develop a preliminary natural language processing (NLP) pipeline to extract HRQOL trajectory based on deep learning models using patient language. Materials and methods: Our data consisted of transcribed interviews of 100 patients undergoing surgical intervention for low-risk thyroid cancer, paired with HRQOL assessments completed during the same visits. Our outcome measure was HRQOL trajectory measured by the SF-12 physical and mental component scores (PCS and MCS), and average THYCA-QoL score.We constructed an NLP pipeline based on BERT, a modern deep language model that captures context semantics, to predict HRQOL trajectory as measured by the above endpoints. We compared this to baseline models using logistic regression and support vector machines trained on bag-of-words representations of transcripts obtained using Linguistic Inquiry and Word Count (LIWC). Finally, given the modest dataset size, we implemented two data augmentation methods to improve performance: first by generating synthetic samples via GPT-2, and second by changing the representation of available data via sequence-by-sequence pairing, which is a novel approach. Results: A BERT-based deep learning model, with GPT-2 synthetic sample augmentation, demonstrated an area-under-curve of 76.3% in the classification of HRQOL accuracy as measured by PCS, compared to the baseline logistic regression and bag-of-words model, which had an AUC of 59.9%. The sequence-by-sequence pairing method for augmentation had an AUC of 71.2% when used with the BERT model. Conclusions: NLP methods show promise in extracting PRO from unstructured narrative data, and in the future may aid in assessing and forecasting patients' HRQOL in response to medical treatments. Our experiments with optimization methods suggest larger amounts of novel data would further improve performance of the classification model.
ABSTRACT
BACKGROUND: The COVID-19 (coronavirus disease 2019) pandemic rapidly expanded telemedicine scale and scope. As telemedicine becomes routine, understanding how specialty and diagnosis combine with demographics to impact telemedicine use will aid in addressing its current limitations. OBJECTIVES: To analyze the relationship between medical specialty, diagnosis, and telemedicine use, and their interplay with patient demographics in determining telemedicine usage patterns. METHODS: We extracted encounter and patient data of all adults who scheduled outpatient visits from June 1, 2020 to June 30, 2021 from the electronic health record of an integrated academic health system encompassing a broad range of subspecialties. Extracted variables included medical specialty, primary visit diagnosis, visit modality (video, audio, or in-person), and patient age, sex, self-reported race/ethnicity and 2013 rural-urban continuum code. Six specialties (General Surgery, Family Medicine, Gastroenterology, Oncology, General Internal Medicine, and Psychiatry) ranging from the lowest to the highest quartile of telemedicine use (video and audio) were chosen for analysis. Relative proportions of video, audio, and in-person modalities were compared. We examined diagnoses associated with the most and least frequent telemedicine use within each specialty. Finally, we analyzed associations between patient characteristics and telemedicine modality (video vs. audio/in-person, and video/audio vs. in-person) using a mixed-effects logistic regression model. RESULTS: A total of 2,494,296 encounters occurred during the study period, representing 420,876 unique patients (mean age: 44 years, standard deviation: 24 years, 54% female). Medical diagnoses requiring physical examination or minor procedures were more likely to be conducted in-person. Rural patients were more likely than urban patients to use video telemedicine in General Surgery and Gastroenterology and less likely to use video for all other specialties. Within most specialties, male patients and patients of nonwhite race were overall less likely to use video modality and video/audio telemedicine. In Psychiatry, members of several demographic groups used video telemedicine more commonly than expected, while in other specialties, members of these groups tended to use less telemedicine overall. CONCLUSION: Medical diagnoses requiring physical examination or minor procedures are more likely to be conducted in-person. Patient characteristics (age, sex, rural vs. urban, race/ethnicity) affect video and video/audio telemedicine use differently depending on medical specialty. These factors contribute to a unique clinical scenario which impacts perceived usefulness and accessibility of telemedicine to providers and patients, and are likely to impact rates of telemedicine adoption.
Subject(s)
COVID-19 , Gastroenterology , Telemedicine , Adult , Humans , Female , Male , Internal Medicine , Electronic Health RecordsABSTRACT
BACKGROUND: The Collaborative Endocrine Surgery Quality Improvement Program tracks thyroidectomy outcomes with self-reported data, whereas the National Surgical Quality Improvement Program uses professional abstractors. We compare completeness and predictive ability of these databases at a single-center and national level. METHOD: Data consistency in the Collaborative Endocrine Surgery Quality Improvement Program and the National Surgical Quality Improvement Program at a single institution (2013-2020) was evaluated using McNemar's test. At the national level, data from the Collaborative Endocrine Surgery Quality Improvement Program and the National Surgical Quality Improvement Program (2016-2019) were used to compare predictive capability for 4 outcomes within each data source: thyroidectomy-specific complication, systemic complication, readmission, and reoperation, as measured by area under curve. RESULTS: In the single-center analysis, 66 cases were recorded in both the Collaborative Endocrine Surgery Quality Improvement Program and the National Surgical Quality Improvement Program. The reoperation variable had the most discrepancies (2 vs 0 in the National Surgical Quality Improvement Program versus the Collaborative Endocrine Surgery Quality Improvement Program, respectively; χ2 = 2.00, P = .16). At the national level, there were 24,942 cases in the National Surgical Quality Improvement Program and 17,666 cases in the Collaborative Endocrine Surgery Quality Improvement Program. In the National Surgical Quality Improvement Program, 30-day thyroidectomy-specific complication, systemic complication, readmission, and reoperation were 13.25%, 2.13%, 1.74%, and 1.39%, respectively, and in the Collaborative Endocrine Surgery Quality Improvement Program 7.27%, 1.95%, 1.64%, and 0.81%. The area under curve of the National Surgical Quality Improvement Program was higher for predicting readmission (0.721 [95% confidence interval 0.703-0.737] vs 0.613 [0.581-0.649]); the area under curve of the Collaborative Endocrine Surgery Quality Improvement Program was higher for thyroidectomy-specific complication (0.724 [0.708-0.737] vs 0.677 [0.667-0.687]) and reoperation (0.735 [0.692-0.775] vs 0.643 [0.611-0.673]). Overall, 3.44% vs 27.22% of values were missing for the National Surgical Quality Improvement Program and the Collaborative Endocrine Surgery Quality Improvement Program, respectively. CONCLUSION: The Collaborative Endocrine Surgery Quality Improvement Program was more accurate in predicting thyroidectomy-specific complication and reoperation, underscoring its role in collecting granular, disease-specific variables. However, a higher proportion of data are missing. The National Surgical Quality Improvement Program infrastructure leads to more rigorous data capture, but the Collaborative Endocrine Surgery Quality Improvement Program is better at predicting thyroid-specific outcomes.
Subject(s)
Data Accuracy , Postoperative Complications , Humans , Postoperative Complications/epidemiology , Postoperative Complications/etiology , Thyroidectomy/adverse effects , Quality Improvement , Reoperation/adverse effects , Retrospective StudiesABSTRACT
Objective: A personalized simulation tool, p-THYROSIM, was developed (1) to better optimize replacement LT4 and LT4+LT3 dosing for hypothyroid patients, based on individual hormone levels, BMIs, and gender; and (2) to better understand how gender and BMI impact thyroid dynamical regulation over time in these patients. Methods: p-THYROSIM was developed by (1) modifying and refining THYROSIM, an established physiologically based mechanistic model of the system regulating serum T3, T4, and TSH level dynamics; (2) incorporating sex and BMI of individual patients into the model; and (3) quantifying it with 3 experimental datasets and validating it with a fourth containing data from distinct male and female patients across a wide range of BMIs. For validation, we compared our optimized predictions with previously published results on optimized LT4 monotherapies. We also optimized combination T3+T4 dosing and computed unmeasured residual thyroid function (RTF) across a wide range of BMIs from male and female patient data. Results: Compared with 3 other dosing methods, the accuracy of p-THYROSIM optimized dosages for LT4 monotherapy was better overall (53% vs. 44%, 43%, and 38%) and for extreme BMI patients (63% vs. ~51% low BMI, 48% vs. ~36% and 22% for high BMI). Optimal dosing for combination LT4+LT3 therapy and unmeasured RTFs was predictively computed with p-THYROSIM for male and female patients in low, normal, and high BMI ranges, yielding daily T3 doses of 5 to 7.5 µg of LT3 combined with 62.5-100 µg of LT4 for women or 75-125 µg of LT4 for men. Also, graphs of steady-state serum T3, T4, and TSH concentrations vs. RTF (range 0%-50%) for untreated patients showed that neither BMI nor gender had any effect on RTF predictions for our patient cohort data. Notably, the graphs provide a means for estimating unmeasurable RTFs for individual patients from their hormone measurements before treatment. Conclusions: p-THYROSIM can provide accurate monotherapies for male and female hypothyroid patients, personalized with their BMIs. Where combination therapy is warranted, our results predict that not much LT3 is needed in addition to LT4 to restore euthyroid levels, suggesting opportunities for further research exploring combination therapy with lower T3 doses and slow-releasing T3 formulations.
Subject(s)
Hypothyroidism , Patient-Specific Modeling , Thyroxine , Triiodothyronine , Body Mass Index , Dose-Response Relationship, Drug , Female , Humans , Hypothyroidism/blood , Hypothyroidism/drug therapy , Male , Thyroid Hormones/administration & dosage , Thyroid Hormones/blood , Thyroid Hormones/pharmacology , Thyroid Hormones/therapeutic use , Thyrotropin/blood , Thyroxine/administration & dosage , Thyroxine/blood , Thyroxine/pharmacology , Thyroxine/therapeutic use , Triiodothyronine/administration & dosage , Triiodothyronine/blood , Triiodothyronine/pharmacology , Triiodothyronine/therapeutic useSubject(s)
Adenocarcinoma , Thyroid Neoplasms , Humans , Quality of Life , Surveys and Questionnaires , Thyroid Neoplasms/surgeryABSTRACT
OBJECTIVE: Natural language processing (NLP) systems convert unstructured text into analyzable data. Here, we describe the performance measures of NLP to capture granular details on nodules from thyroid ultrasound (US) reports and reveal critical issues with reporting language. METHODS: We iteratively developed NLP tools using clinical Text Analysis and Knowledge Extraction System (cTAKES) and thyroid US reports from 2007 to 2013. We incorporated nine nodule features for NLP extraction. Next, we evaluated the precision, recall, and accuracy of our NLP tools using a separate set of US reports from an academic medical center (A) and a regional health care system (B) during the same period. Two physicians manually annotated each test-set report. A third physician then adjudicated discrepancies. The adjudicated "gold standard" was then used to evaluate NLP performance on the test-set. RESULTS: A total of 243 thyroid US reports contained 6,405 data elements. Inter-annotator agreement for all elements was 91.3%. Compared with the gold standard, overall recall of the NLP tool was 90%. NLP recall for thyroid lobe or isthmus characteristics was: laterality 96% and size 95%. NLP accuracy for nodule characteristics was: laterality 92%, size 92%, calcifications 76%, vascularity 65%, echogenicity 62%, contents 76%, and borders 40%. NLP recall for presence or absence of lymphadenopathy was 61%. Reporting style accounted for 18% errors. For example, the word "heterogeneous" interchangeably referred to nodule contents or echogenicity. While nodule dimensions and laterality were often described, US reports only described contents, echogenicity, vascularity, calcifications, borders, and lymphadenopathy, 46, 41, 17, 15, 9, and 41% of the time, respectively. Most nodule characteristics were equally likely to be described at hospital A compared with hospital B. CONCLUSIONS: NLP can automate extraction of critical information from thyroid US reports. However, ambiguous and incomplete reporting language hinders performance of NLP systems regardless of institutional setting. Standardized or synoptic thyroid US reports could improve NLP performance.
Subject(s)
Lymphadenopathy , Natural Language Processing , Humans , Thyroid Gland/diagnostic imagingABSTRACT
BACKGROUND: The ACS-NSQIP surgical risk calculator (SRC) often guides preoperative counseling, but the rarity of complications in certain populations causes class imbalance, complicating risk prediction. We aimed to compare the performance of the ACS-NSQIP SRC to other classical machine learning algorithms trained on NSQIP data, and to demonstrate challenges and strategies in predicting such rare events. METHODS: Data from the NSQIP thyroidectomy module ys 2016 - 2018 were used to train logistic regression, Ridge regression and Random Forest classifiers for predicting 2 different composite outcomes of surgical risk (systemic and thyroidectomy-specific). We implemented techniques to address imbalanced class sizes and reported the area under the receiver operating characteristic (AUC) for each classifier including the ACS-NSQIP SRC, along with sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) at a 5% - 15% predicted risk threshold. RESULTS: Of 18,078 included patients, 405 (2.24%) patients suffered systemic complications and 1670 (9.24%) thyroidectomy-specific complications. Logistic regression performed best for predicting systemic complication risk (AUC 0.723 [0.658 - 0.778]); Random Forest with RUSBoost performed best for predicting thyroidectomy-specific complication risk (0.702; 0.674 - 0.726). The addition of optimizations for class imbalance improved performance for all classifiers. CONCLUSIONS: Complications are rare after thyroidectomy even when considered as composite outcomes, and class imbalance poses a challenge in surgical risk prediction. Using the SRC as a classifier where intervention occurs above a certain validated threshold, rather than citing the numeric estimates of complication risk, should be considered in low-risk patients.
Subject(s)
Postoperative Complications , Thyroidectomy , Humans , Postoperative Complications/diagnosis , Postoperative Complications/epidemiology , Postoperative Complications/etiology , ROC Curve , Retrospective Studies , Risk Assessment/methods , Risk Factors , Thyroidectomy/adverse effectsABSTRACT
BACKGROUND: Up to 30% of thyroid nodules are classified as indeterminate after fine needle aspiration biopsy. These indeterminate thyroid nodules (ITNs) require surgical pathology for definitive diagnosis. Molecular testing provides additional pre-operative cancer risk stratification but adds expense and invasive testing. The purpose of this study is to utilize a machine learning (ML) algorithm to predict malignancy of ITNs using data available from less invasive tests. MATERIALS AND METHODS: We conducted a retrospective study using medical records from one academic and one community center. Thyroid nodules with an indeterminate diagnosis on fine needle aspiration biopsy and completed diagnostic pathology were included. Linear, non-linear, and non-linear-ensemble ML methods were tested for accuracy when predicting malignancy using 10-fold cross-validation. Classifiers were evaluated using area under the receiver operating characteristics curve (AUROC). RESULTS: A total of 355 nodules met inclusion criteria. Of these, 171 (48.2%) were diagnosed with cancer. A Random Forest classifier performed the best, producing an accuracy of 79.1%, a sensitivity of 75.5%, specificity of 82.4%, positive predicative value of 80.3%, negative predictive value of 79.0%, and an AUROC of 0.859. CONCLUSIONS: ML methods accurately risk stratify ITNs using data gathered from existing, non-invasive, and inexpensive diagnostic tests. Applying an ML model with existing data can become a cost-effective alternative to molecular testing. Future studies will prospectively evaluate the performance of this ML approach when combined with expert judgment.
Subject(s)
Thyroid Neoplasms , Thyroid Nodule , Biopsy, Fine-Needle , Humans , Machine Learning , Retrospective Studies , Sensitivity and Specificity , Thyroid Neoplasms/diagnosis , Thyroid Neoplasms/pathology , Thyroid Nodule/pathologyABSTRACT
BACKGROUND: Thyroid nodules are common; up to 67% of adults will show nodules on high-quality ultrasound, and 95% of these nodules are benign. FNA cytology is a crucial step in determining the risk of malignancy, and a false negative diagnosis at this stage delays cancer treatment. The purpose of this study is to develop a predictive model using machine learning which can identify false negative FNA results based on less-invasive clinical data. MATERIALS AND METHODS: We conducted a retrospective medical record review at one academic and one community center. Inclusion criteria were thyroid nodules evaluated by ultrasound and FNA with a Bethesda II (benign) result or malignancy detected on pathology or FNA. Linear, non-linear, and ensemble models were generated with scikit-learn using 10-fold cross validation with repetition and compared with AUROC. The classification task was the prediction of malignancy using information acquired from less-invasive ultrasound and FNA. RESULTS: A total of 604 subjects met inclusion criteria; 38 were diagnosed with malignancy. Of all algorithms tested, a Random Forest method achieved the best AUROC (0.64) in separating benign and malignant nodules, though the improvement over other tested algorithms was not statistically significant. CONCLUSIONS: A Random Forest model performed better than random chance using readily available data obtained via standard evaluation of thyroid nodules. The diagnostic probability threshold of this model can be varied to minimize false positives at the cost of increasing the number of false negatives. Future studies will prospectively evaluate the model's performance.
Subject(s)
Thyroid Neoplasms , Thyroid Nodule , Adult , Biopsy, Fine-Needle , Humans , Machine Learning , Retrospective Studies , Thyroid Neoplasms/diagnostic imaging , Thyroid Neoplasms/pathology , Thyroid Nodule/diagnostic imaging , Thyroid Nodule/pathologyABSTRACT
BACKGROUND: The COVID-19 pandemic led to dramatic increases in telemedicine use to provide outpatient care without in-person contact risks. Telemedicine increases options for health care access, but a "digital divide" of disparate access may prevent certain populations from realizing the benefits of telemedicine. OBJECTIVES: The study aimed to understand telemedicine utilization patterns after a widespread deployment to identify potential disparities exacerbated by expanded telemedicine usage. METHODS: We performed a cross-sectional retrospective analysis of adults who scheduled outpatient visits between June 1, 2020 and August 31, 2020 at a single-integrated academic health system encompassing a broad range of subspecialties and a large geographic region in the Upper Midwest, during a period of time after the initial surge of COVID-19 when most standard clinical services had resumed. At the beginning of this study period, approximately 72% of provider visits were telemedicine visits. The primary study outcome was whether a patient had one or more video-based visits, compared with audio-only (telephone) visits or in-person visits only. The secondary outcome was whether a patient had any telemedicine visits (video-based or audio-only), compared with in-person visits only. RESULTS: A total of 197,076 individuals were eligible (average age = 46 years, 56% females). Increasing age, rural status, Asian or Black/African American race, Hispanic ethnicity, and self-pay/uninsured status were significantly negatively associated with having a video visit. Digital literacy, measured by patient portal activation status, was significantly positively associated with having a video visit, as were Medicaid or Medicare as payer and American Indian/Alaskan Native race. CONCLUSION: Our findings reinforce previous evidence that older age, rural status, lower socioeconomic status, Asian race, Black/African American race, and Hispanic/Latino ethnicity are associated with lower rates of video-based telemedicine use. Health systems and policies should seek to mitigate such barriers to telemedicine when possible, with efforts such as digital literacy outreach and equitable distribution of telemedicine infrastructure.
Subject(s)
COVID-19/epidemiology , Health Services Accessibility/statistics & numerical data , Healthcare Disparities/statistics & numerical data , Pandemics/statistics & numerical data , Telemedicine/statistics & numerical data , Adolescent , Adult , Aged , Cross-Sectional Studies , Female , Humans , Male , Middle Aged , Retrospective Studies , Young AdultABSTRACT
The Internet is a key source of health information, yet little is known about resources for low-risk thyroid cancer treatment. We examined the timeliness, content, quality, readability, and reference to the 2015 American Thyroid Association (ATA) guidelines in websites about thyroid cancer treatment. We identified the top 60 websites using Google, Bing, and Yahoo for "thyroid cancer." Timeliness and content analysis identified updates in the ATA guidelines (n = 6) and engaged a group of stakeholders to develop essential items (n = 29) for making treatment decisions. Website quality and readability analysis used 4 validated measures: DISCERN; Journal of the American Medical Association (JAMA) benchmark criteria; Health on the Net Foundation certification (HONcode); and the Suitability Assessment of Materials (SAM) method. Of the 60 websites, 22 were unique and investigated. Content analysis revealed zero websites contained all updates from the ATA guidelines and rarely (18.2%) referenced them. Only 31.8% discussed all 3 treatment options: total thyroidectomy, lobectomy, and active surveillance. Websites discussed 28.2% of the 29 essential items for making treatment decisions. Quality analysis with DISCERN showed "fair" scores overall. Only 29.9% of the JAMA benchmarks were satisfied, and 40.9% were HONcode certified. Readability analysis with the SAM method found adequate readability, yet 90.9% scored unsuitable in literacy demand. The overall timeliness, content, quality, and readability of websites about low-risk thyroid cancer treatment is fair and needs improvement. Most websites lack updates from the 2015 ATA guidelines and information about treatment options that are necessary to make informed decisions.
Subject(s)
Consumer Health Information , Thyroid Neoplasms , Benchmarking , Comprehension , Humans , Internet , Thyroid Neoplasms/therapyABSTRACT
Introduction: Little is known about the experiences and concerns of patients recently diagnosed with thyroid cancer or an indeterminate thyroid nodule. This study sought to explore patients' reactions to diagnosis with papillary thyroid cancer (PTC) or indeterminate cytology on fine needle aspiration. Methods: We conducted semistructured interviews with 85 patients with recently diagnosed PTC or an indeterminate thyroid nodule before undergoing thyroidectomy. We included adults with nodules ≥1 cm and Bethesda III, IV, V, and VI cytology. The analysis utilized grounded theory methodology to create a conceptual model of patient reactions. Results: After diagnosis, participants experienced shock, anxiety, fear, and a strong need to "get it out" because "it's cancer!" This response was frequently followed by a sense of urgency to "get it done," which made waiting for surgery difficult. These reactions occurred regardless of whether participants had confirmed PTC or indeterminate cytology. Participants described the wait between diagnosis and surgery as difficult, because the cancer or nodule was "still sitting there" and "could be spreading." Participants often viewed surgery and getting the cancer out as a "fix" that would resolve their fears and worries, returning them to normalcy. The need to "get it out" also led some participants to minimize the risk of complications or adverse outcomes. Education about the slow-growing nature of PTC reassured some, but not all patients. Conclusions: After diagnosis with PTC or an indeterminate thyroid nodule, many patients have strong emotional reactions and an impulse to "get it out" elicited by the word "cancer." This reaction can persist even after receiving education about the excellent prognosis. Understanding patients' response to diagnosis is critical, because their emotional reactions likely pose a barrier to implementing guidelines recommending less extensive management for PTC.
Subject(s)
Health Knowledge, Attitudes, Practice , Patients/psychology , Thyroid Cancer, Papillary/psychology , Thyroid Neoplasms/psychology , Thyroid Nodule/psychology , Adult , Anxiety/etiology , Anxiety/psychology , Biopsy, Fine-Needle , Fear , Female , Grounded Theory , Humans , Interviews as Topic , Male , Middle Aged , Neoplasm Staging , Patient Education as Topic , Predictive Value of Tests , Qualitative Research , Randomized Controlled Trials as Topic , Thyroid Cancer, Papillary/pathology , Thyroid Cancer, Papillary/surgery , Thyroid Neoplasms/pathology , Thyroid Neoplasms/surgery , Thyroid Nodule/pathology , Thyroid Nodule/surgery , Thyroidectomy , Tumor Burden , Waiting ListsABSTRACT
BACKGROUND: Critical thyroid nodule features are contained in unstructured ultrasound (US) reports. The Thyroid Imaging, Reporting, and Data System (TI-RADS) uses five key features to risk stratify nodules and recommend appropriate intervention. This study aims to analyze the quality of US reporting and the potential benefit of Natural Language Processing (NLP) systems in efficiently capturing TI-RADS features from text reports. MATERIALS AND METHOD: This retrospective study used free-text thyroid US reports from an academic center (A) and community hospital (B). Physicians created "gold standard" annotations by manually extracting TI-RADS features and clinical recommendations from reports to determine how often they were included. Similar annotations were created using an automated NLP system and compared with the gold standard. RESULTS: Two hundred eighty-two reports contained 409 nodules at least 1-cm in maximum diameter. The gold standard identified three nodules (0.7%) which contained enough information to calculate a complete TI-RADS score. Shape was described most often (92.7% of nodules), whereas margins were described least often (11%). A median number of two TI-RADS features are reported per nodule. The NLP system was significantly less accurate than the gold standard in capturing echogenicity (27.5%) and margins (58.9%). One hundred eight nodule reports (26.4%) included clinical management recommendations, which were included more often at site A than B (33.9 versus 17%, P < 0.05). CONCLUSIONS: These results suggest a gap between current US reporting styles and those needed to implement TI-RADS and achieve NLP accuracy. Synoptic reporting should prompt more complete thyroid US reporting, improved recommendations for intervention, and better NLP performance.
Subject(s)
Image Processing, Computer-Assisted/methods , Natural Language Processing , Thyroid Gland/diagnostic imaging , Thyroid Nodule/diagnosis , Academic Medical Centers/standards , Academic Medical Centers/statistics & numerical data , Data Systems , Hospitals, Community/standards , Hospitals, Community/statistics & numerical data , Humans , Practice Guidelines as Topic , Radiology/standards , Retrospective Studies , Societies, Medical/standards , Ultrasonography/standards , Ultrasonography/statistics & numerical dataABSTRACT
BACKGROUND: Current recommendations using Hounsfield units (HU) ≤ 10 to identify adrenal adenomas on unenhanced computed tomography (CT) miss 10-40% of benign adenomas. We sought to determine if changing HU threshold and adding absolute percent contrast washout (APW) criteria would identify adrenal adenomas better than current recommendations. METHODS: Imaging characteristics were compared between patients with adenomas (n = 128) and those with non-adenomas (n = 54) after unilateral adrenalectomy. Sensitivity, specificity, positive and negative predictive values (PPV, NPV) were calculated. RESULTS: Using HU ≤ 10 to identify adenomas had a sensitivity of 47.6%, specificity of 93.3% (AUC = 0.71, p < 0.001), PPV of 95.3%, and NPV of 58.1% for identifying adrenal adenomas. Applying HU ≤ 16 improved sensitivity (65.4%) without reducing specificity (93.3%) (AUC = 0.79, p < 0.001), PPV increased to 96.3%, and NPV decreased to 47.6%. Applying HU ≤ 16 as the initial criterion followed by APW > 60% for lesions exceeding 16 HU, sensitivity increased to 93.4%, specificity was 93.3% and PPV 96.6%, and NPV improved to 85.7% (AUC = 0.96, p < 0.001). CONCLUSIONS: Criteria of initial threshold of HU ≤ 16 followed by APW > 60% for lesions exceeding 16 HU yielded improved sensitivity and specificity in identification of adrenal adenomas.
Subject(s)
Adenoma/diagnostic imaging , Adrenal Gland Neoplasms/diagnostic imaging , Tomography, X-Ray Computed , Adult , Aged , Contrast Media , Female , Humans , Male , Middle Aged , Retrospective Studies , Sensitivity and Specificity , Tomography, X-Ray Computed/methods , Tomography, X-Ray Computed/statistics & numerical dataABSTRACT
BACKGROUND: Thyroid cancer diagnoses are often discovered after diagnostic thyroid lobectomy. Completion thyroidectomy (CT) may be indicated for intermediate or high-risk tumors to facilitate surveillance and/or adjuvant treatment. The completeness of thyroid resection and the safety of CT compared to total thyroidectomy (TT) is unclear. We assessed outcomes after TT or CT to determine completeness of resection and risk of complications. METHODS: Patients undergoing TT or CT between 2000 and 2018 were retrospectively reviewed. Pathology, unstimulated thyroglobulin (uTg), parathyroid hormone (PTH), rates of hematoma, and recurrent laryngeal nerve (RLN) injury were compared. RESULTS: Differentiated thyroid cancer (DTC) was identified in 954 patients undergoing TT and 142 patients undergoing CT. Postoperative uTg at 6 months was not different between TT and CT, 0.2 vs 0.2 ng/mL, P = .37. Transient hypoparathyroidism with immediate postoperative PTH less than 10 was more common after TT, 14.3 vs 6.0% (P = .009). No differences were noted regarding postoperative hematoma, transient RLN injury, permanent hypoparathyroidism, and permanent RLN injury. CONCLUSIONS: If CT is required for DTC, a complete resection, as assessed by postoperative uTg, can be achieved. Furthermore, CT is significantly less likely to result in transient hypoparathyroidism and poses no additional risk of RLN injury, hematoma, or permanent hypoparathyroidism.
ABSTRACT
Clinical informatics is an interdisciplinary specialty that leverages big data, health information technologies, and the science of biomedical informatics within clinical environments to improve quality and outcomes in the increasingly complex and often siloed health care systems. Core competencies of clinical informatics primarily focus on clinical decision making and care process improvement, health information systems, and leadership and change management. Although the broad relevance of clinical informatics is apparent, this review focuses on its application and pertinence to the discipline of surgery, which is less well defined. In doing so, we hope to highlight the importance of the surgeon informatician. Topics covered include electronic health records, clinical decision support systems, computerized order entry, data analytics, clinical documentation, information architectures, implementation science, quality improvement, simulation, education, and telemedicine. The formal pathway for surgeons to become clinical informaticians is also discussed.
Subject(s)
General Surgery/organization & administration , Medical Informatics/organization & administration , Professional Role , Surgeons/organization & administration , Decision Support Systems, Clinical/organization & administration , HumansABSTRACT
BACKGROUND: This study analyzed independent factors associated with post-thyroidectomy Emergency Room (ER) visits and Hospital Readmissions (HR). METHODS: This is a retrospective review from the CESQIP registry of 8381 thyroidectomy patients by 173 surgeons at 46 institutions. A total of 7142 ER visits and 7265 HR were analyzed. Multivariable logistic regression analysis was performed to determine the risk factors for an ER visit or HR. RESULTS: Within 30-days of surgery, rates of all ER visits were 3.4% (n = 250) and all HR were 2.3% (n = 170). Hypocalcemia was the reason for 21.9% of ER encounters and 36.4% of HR. BMI >40 kg/m2 was a risk factor for both ER visit (OR1.86) and HR (OR1.94). Surgical duration >3 h (OR2.63), and transection of recurrent laryngeal nerve (OR4.58) were risk factors for HR. CONCLUSIONS: Strategies to decrease hypocalcemia and improve perioperative care of patients with BMI >40 kg/m2 may improve post-thyroidectomy outcome.