Frequency of Parathyroid Hormone Assessment in the Evaluation of Hypercalcemia. A Comparison Between Patients With and Without a History of Malignancy in a 20-year Dataset of 20,954 Patients

Background: The purpose of this study was to evaluate whether a prior diagnosis of malignancy affected the assessment of parathyroid hormone (PTH) in hypercalcemic patients and whether the rate of this assessment changed over time. Methods: A retrospective cohort study was designed that included adult patients with hypercalcemia with and without a history of malignancy between January 1, 2000 and December 31, 2019 in the Marshfield Clinic Health System (MCHS). The overall and annual rates of PTH assessment in each group was determined. In patients with a PTH assessment, duration of time and number of elevated serum calcium levels between the first documentation of hypercalcemia and the assessment of PTH were recorded, as was the degree of hypercalcemia. Results: Approximately a quarter (23%) of the patients in each group had a PTH assessment. The rate of PTH assessment initially increased over time but later declined significantly. Although a more severe degree of hypercalcemia predicted a greater probability of PTH assessment, the rate of assessment declined with all degrees of hypercalcemia in the last 5 years. While most patients who had a PTH assessed did so within a few months of the first documentation of hypercalcemia, less than half (40%) had a delay of more than 2 years before a PTH level was drawn. Conclusion: This lack of appropriate and timely assessment may have significant health consequences in both groups of patients. Better education of providers about the appropriate and timely assessment of PTH in the evaluation of hypercalcemia is urgently needed.

Racial Differences in Clinical Phenotype and Hospitalization of Blastomycosis Patients

Background Dimorphic fungal infections, such as blastomycosis, cause significant morbidity and mortality. Historically, blastomycosis studies have focused on non-Hispanic whites, which limits our understanding of the clinical presentation and outcomes for patients of other races and ethnicities. We evaluated whether clinical presentation and disease severity varied across racial and ethnic groups. Methods Blastomycosis patients were identified from Marshfield Clinic Health System and data were abstracted from electronic medical records. Blastomyces genotyping was performed for cases with available isolates. Bivariate analyses (χ 2 tests/analysis of variance) assessed associations of race and/or ethnicity, Blastomyces spp, and hospitalization status with demographics and clinical presentation. Multivariable logistic regression was used to evaluate the association of race and/or ethnicity and hospitalization. Results In total, 477 patients were included. Age differences were observed across race and ethnicity categories (P < .0001). Non-Hispanic whites were oldest (median, 48 years; interquartile range [IQR], 31–62) and Asians were youngest (26 years; IQR, 19–41). Non-Hispanic whites (55%) and African Americans (52%) had underlying medical conditions more frequently than Hispanic whites (27%) and Asians (29%). Odds of hospitalization were 2 to 3 times higher for Hispanic whites (adjusted odds ratio [aOR], 2.9; 95% confidence interval [CI], 1.2–1.7), American Indian or Alaska Native (AIAN) (aOR, 2.4; 95% CI, 1.0–5.5), and Asian (aOR, 1.9; 95% CI, 1.0–3.6) patients compared with non-Hispanic white patients. Ninety percent of Blastomyces dermatitidis infections occurred in non-Hispanic whites, whereas blastomycosis in Hispanic whites, AIAN, and Asian patients was frequently caused by Blastomyces gilchristii (P < .0001). Conclusions Hispanic whites, AIAN, and Asian blastomycosis patients were younger and healthier but more frequently hospitalized. Patients in these racial and ethnic groups may need more aggressive treatment and closer therapeutic monitoring.

Evaluating electronic health record data sources and algorithmic approaches to identify hypertensive individuals

Objective: Phenotyping algorithms applied to electronic health record (EHR) data enable investigators to identify large cohorts for clinical and genomic research. Algorithm development is often iterative, depends on fallible investigator intuition, and is time- and labor-intensive. We developed and evaluated 4 types of phenotyping algorithms and categories of EHR information to identify hypertensive individuals and controls and provide a portable module for implementation at other sites. Materials and Methods: We reviewed the EHRs of 631 individuals followed at Vanderbilt for hypertension status. We developed features and phenotyping algorithms of increasing complexity. Input categories included International Classification of Diseases, Ninth Revision (ICD9) codes, medications, vital signs, narrative-text search results, and Unified Medical Language System (UMLS) concepts extracted using natural language processing (NLP). We developed a module and tested portability by replicating 10 of the best-performing algorithms at the Marshfield Clinic. Results: Random forests using billing codes, medications, vitals, and concepts had the best performance with a median area under the receiver operator characteristic curve (AUC) of 0.976. Normalized sums of all 4 categories also performed well (0.959 AUC). The best non-NLP algorithm combined normalized ICD9 codes, medications, and blood pressure readings with a median AUC of 0.948. Blood pressure cutoffs or ICD9 code counts alone had AUCs of 0.854 and 0.908, respectively. Marshfield Clinic results were similar. Conclusion: This work shows that billing codes or blood pressure readings alone yield good hypertension classification performance. However, even simple combinations of input categories improve performance. The most complex algorithms classified hypertension with excellent recall and precision.

A Rigorous Algorithm To Detect And Clean Inaccurate Adult Height Records Within EHR Systems

SummaryBackground: Height is a critical variable for many biomedical analyses because it is an important component of Body Mass Index (BMI). Transforming EHR height measures into meaningful research-ready values is challenging and there is limited information available on methods for “cleaning” these data.Objectives: We sought to develop an algorithm to clean adult height data extracted from EHR using only height values and associated ages.Results: The algorithm we developed is sensitive to normal decreases in adult height associated with aging, is implemented using an open-source software tool and is thus easily modifiable, and is freely available. We checked the performance of our algorithm using data from the Northwestern biobank and a replication sample from the Marshfield Clinic biobank obtained through our participation in the eMERGE consortium. The algorithm identified 1262 erroneous values from a total of 33937 records in the Northwestern sample. Replacing erroneous height values with those identified as correct by the algorithm resulted in meaningful changes in height and BMI records; median change in recorded height after cleaning was 7.6 cm and median change in BMI was 2.9 kg/m2. Comparison of cleaned EHR height values to observer measured values showed that 94.5% (95% C.I 93.8-% – 95.2%) of cleaned values were within 3.5 cm of observer measured values.Conclusions: Our freely available height algorithm cleans EHR height data with only height and age inputs. Use of this algorithm will benefit groups trying to perform research with height and BMI data extracted from EHR.Citation: Muthalagu A, Pacheco JA, Aufox S, Peissig PL, Fuehrer JT, Tromp G, Kho AN, Rasmussen-Torvik LJ. A rigorous algorithm to detect and clean inaccurate adult height records within EHR systems. Appl Clin Inf 2014; 5: 118–126 http://dx.doi.org/10.4338/ACI-2013-09-RA-0074