Reference Intervals Generated by Electronic Medical Record Data Mining with Clinical Exclusions: Age-Specific Intervals for Thyroid-Stimulating Hormone from 33038 Euthyroid Patients.

BACKGROUND: Serum thyroid-stimulating hormone (TSH) reference intervals are dependent on population characteristics, including prevalent thyroid disease and iodine status. Studies in the US have demonstrated increasing TSH levels with age, and the American Thyroid Association recommends higher TSH goals for older patients taking thyroid supplementation, but few laboratories offer age-specific reference intervals for TSH. Our objective was to establish TSH reference ranges in our racially diverse population in northern California. METHODS: Data mining of electronic medical records was used with the a posteriori approach to select a euthyroid reference population for TSH reference intervals. A report gathered all TSH results from 2 weeks from >1 year in the past, excluding results from patients with thyroid-related disease or medication use at any time before or after the TSH test. RESULTS: The reference population numbered 33038 and consisted of approximately 44% of the total TSH results reported in the selected time periods. The population identified as 46.5% white, 18.3% Asian, 17.0% Hispanic/Latino, 8.0% black/African American, and 10.3% other or unknown. These data demonstrate an increase in the median and 97.5 percentile of TSH levels with increasing age in adults. No clinically significant difference was seen between female and male individuals or between the self-identified races, except for lower TSH levels in the black/African American population. CONCLUSIONS: The a posteriori approach using data mining for disease-specific criteria proved to be an efficient method for obtaining a large healthy reference population. Age-specific TSH reference ranges could prevent inappropriate diagnoses of subclinical hypothyroidism in older patients.

Age, sex, and racial influences on the Beckman Coulter AccuTnI+3 99th percentile.

BACKGROUND: Beckman Coulter recently released a new cardiac troponin I (cTnI) assay, AccuTnI+3, for the Access 2 and DxI platforms. We validated the stated 99th percentile (20 ng/l) using a large population of healthy adults representative of the Northern California population. METHODS: Within a large sample of healthy adult members receiving care at Kaiser Permanente, cTnI was quantified in residual specimens using the AccuTnI+3 assay. Patients were selected based on pre-defined criteria extracted from a comprehensive electronic medical record. All specimens with a cTnI concentration >30ng/l were repeated; specimens that had a reproducible result >30 ng/l were subject to heterophile blocking procedure. 99th percentiles were calculated based on age, sex, race and body mass index categories. RESULTS: Among 1764 tested subjects, the 99th percentile for all samples was 25 ng/l. Sex differences were observed; the male and female 99th percentiles were 31 and 21 ng/l, respectively (p=0.001). Age (range evaluated 18-89y, median 47y) also had a significant influence on the value (p=0.003), but there were no significant differences by race. False positive results were detected in 0.9% of specimens (0.6% "fliers" and 0.3% heterophile antibodies), corresponding to 52% of all results >30 ng/l. CONCLUSIONS: Among a large, representative cohort of healthy adults, we found a 99th percentile value consistent with prior studies based on highly selected small patient samples. Sex and age-specific upper reference limits for cTnI should be considered. In this cohort, about half the findings above the 99th percentile were false positives. Avoiding reporting erroneous results requires implementation of quality indicators.

Preliminary Development and Evaluation of an Algorithm to Identify Breast Cancer Chemotherapy Toxicities Using Electronic Medical Records and Administrative Data.

PURPOSE: Breast cancer chemotherapy toxicity is not well documented outside of randomized trials. We developed and conducted preliminary evaluation of an algorithm to detect grade 3 and 4 toxicities using electronic data from a large integrated managed care organization. METHODS: The algorithm used administrative, pharmacy, and electronic data from outpatient, emergency room, and inpatient records of 99 women diagnosed with breast cancer from 2006 to 2009 who underwent chemotherapy. Data were abstracted for 12 months post-treatment initiation (24 months for trastuzumab recipients). An oncology nurse independently blindly reviewed records; these results were the "gold standard." Sensitivity and specificity were calculated for overall toxicity, categories of toxicities, and toxicity by age or regimen. The algorithm was applied to an independent sample of 1,575 patients with breast cancer diagnosed during the study period to estimate prevalence rates. RESULTS: The overall sensitivity for detecting chemotherapy-related toxicity was 89% (95% CI, 77% to 95%). The highest sensitivity was for identification of hematologic toxicities (97%; 95% CI, 84% to 99%). There were good sensitivities for infectious toxicity, but rates dropped for GI and neurological toxicities. Specificity was high within each category (89% to 99%), but when combined to measure any toxicity, it was lower (70%; 95% CI, 57% to 81%). When applied to an independent chemotherapy sample, the algorithm estimates a 26% rate of hematologic toxicity; rates were higher among patients age ≥ 65 years versus less than 65 years. CONCLUSIONS: If validated in other samples and health care settings, algorithms to capture toxicity could be useful in comparative and cost-effectiveness evaluations of community practice-delivered treatment.

Falsely undetectable TSH in a cohort of South Asian euthyroid patients.

CONTEXT: An index case of a clinically euthyroid woman of South Asian descent was identified with discordant TSH results: undetectable TSH on our routine assay and normal TSH on an alternate assay. Low TSH concentrations due to functionally compromising TSH mutations have been reported. Here we describe a new phenomenon of functional TSH that is undetectable by 4 widely used US Food and Drug Administration (FDA)-approved TSH immunoassays marketed by a single vendor. OBJECTIVE: The purpose of this study was to identify additional cases and investigate the cause of the falsely undetectable TSH. DESIGN: All samples with TSH results of <0.01 µIU/mL were retested with a second TSH assay. Discordant samples were evaluated on up to 8 FDA-approved TSH immunoassays and the TSHß gene was sequenced. Retrospectively, thyroid function tests, diagnoses, and medications from 1.6 million individuals were analyzed. RESULTS: Out of approximately 2 million individuals, we have identified a cohort of 20 hypothyroid and euthyroid patients of shared ethnicity with falsely undetectable TSH (<0.01 µIU/mL) in 4 of 8 commercially available TSH assays. Half of these individuals were initially treated based on repeated falsely undetectable TSH values (7 euthyroid patients were treated with methimazole and 2 hypothyroid patients had doses of levothyroxine decreased). In all cases, a retrospective chart review revealed that clinical assessments and free T4 and total T3 results were inconsistent with the undetectable TSH results. Specific antibodies failing to detect TSH in these cases were identified in the 4 affected assays. A novel TSHß point mutation was identified. CONCLUSIONS: Our data suggest that these individuals have a previously unrecognized, functionally normal, TSH variant to which some monoclonal antibodies fail to bind. To assure appropriate patient management, clinicians and laboratorians need to be aware that certain TSH variants may be undetectable in some hyperselective TSH assays.

Patterns and predictors of first-line chemotherapy use among adults with advanced non-small cell lung cancer in the cancer research network.

BACKGROUND: Relatively low rates of chemotherapy receipt have been observed in older patients diagnosed with advanced non-small cell lung cancer (NSCLC) in SEER-Medicare-based studies. However, little is known about variation in first-line NSCLC chemotherapy use in younger patients, health maintenance organization (HMO)-based settings, and for high-cost, novel agents, such as bevacizumab and erlotinib. METHODS: A cohort of 6614 stage IIIB/IV NSCLC patients aged ≥ 21 years diagnosed between 2000 and 2007 was identified at four HMOs that participate in the Cancer Research Network (CRN). Demographic, comorbidity, tumor characteristics, and chemotherapy treatment data were included in logistic regression models to identify factors associated with chemotherapy receipt and tests of association examined secular and age-specific variation in first-line chemotherapy regimens. RESULTS: Within 120 days of diagnosis, 3612 (55%) patients received chemotherapy; increasing from 52% of patients diagnosed in 2000 to 59% in 2007 (p<0.001). Receipt was significantly higher for patients aged <65 years (64% versus 46% in ≥ 65) and was inversely related to stage and comorbidites (all p<0.001). Carboplatin and paclitaxel were received most frequently. Erlotinib and bevacizumab use in the later years of the study was associated with a significant change in distributions of first-line chemotherapies (p<0.001). CONCLUSIONS: For patients alive 30 days post diagnosis, chemotherapy use was higher in the aged population (>65 years) than previously published estimates, and higher still among younger patients. Chemotherapy use increased over the observation period, and the mix of first-line therapies used changed substantially over time. Of note, novel, high cost treatments were used in first-line therapy prior to FDA approval, increasing significantly throughout the study period. These findings demonstrate the utility of HMO CRN data to augment SEER-Medicare to conduct comparative effectiveness research related to chemotherapy use and the use of specific agents, especially among younger patients.

OBAYA (obesity and adverse health outcomes in young adults): feasibility of a population-based multiethnic cohort study using electronic medical records.

BACKGROUND: Although obesity is a risk factor for many chronic diseases, we have only limited knowledge of the magnitude of these associations in young adults. A multiethnic cohort of young adults was established to close current knowledge gaps; cohort demographics, cohort retention, and the potential influence of migration bias were investigated. METHODS: For this population-based cross-sectional study, demographics, and measured weight and height were extracted from electronic medical records of 1,929,470 patients aged 20 to 39 years enrolled in two integrated health plans in California from 2007 to 2009. RESULTS: The cohort included about 84.4% of Kaiser Permanente California members in this age group who had a medical encounter during the study period and represented about 18.2% of the underlying population in the same age group in California. The age distribution of the cohort was relatively comparable to the underlying population in California Census 2010 population, but the proportion of women and ethnic/racial minorities was slightly higher. The three-year retention rate was 68.4%. CONCLUSION: These data suggest the feasibility of our study for medium-term follow-up based on sufficient membership retention rates. While nationwide 6% of young adults are extremely obese, we know little to adequately quantify the health burden attributable to obesity, especially extreme obesity, in this age group. This cohort of young adults provides a unique opportunity to investigate associations of obesity-related factors and risk of cancer in a large multiethnic population.

The prevalence of obesity and obesity-related health conditions in a large, multiethnic cohort of young adults in California.

PURPOSE: To identify population groups that are most susceptible to obesity-related health conditions at young age. METHODS: For this population-based cross-sectional study, measured weight and height, diagnosis, laboratory, and drug prescription information were extracted from electronic medical records of 1,819,205 patients aged 20 to 39 years enrolled in two integrated health plans in California in 2007 through 2009. RESULTS: Overall, 29.9% of young adults were obese. Extreme obesity (body mass index [BMI] ≥ 40 kg/m(2)) was observed in 6.1% of women and 4.5% of men. The adjusted relative risk (RR) for diabetes, hypertension, dyslipidemia, and the metabolic syndrome increased sharply for those individuals with a BMI of 40 or greater, with the sharpest increase in the adjusted RR for hypertension and the metabolic syndrome. The association between weight class and dyslipidemia, hypertension, and the metabolic syndrome but not diabetes was stronger among 20.0- to 29.9-year-olds compared with 30.0- to 39.9-year-olds (P for interaction < .05). For example, compared with their normal weight counterparts of the same age group, young adults with a BMI of 40.0 to 49.9, 50.0 to 59.9, and 60 or greater kg/m(2) had a RR for hypertension of 11.73, 19.88, and 30.47 (95% confidence interval [CI], 26.39-35.17) at 20 to 29 years old, and 9.31, 12.41, and 15.43 (95% CI, 14.32-16.63) at 30 to 39 years old. CONCLUSIONS: Although older individuals were more likely to be extremely obese, the association between obesity-related health conditions was stronger in younger individuals. Hispanics and Blacks are also more likely to be obese, including extremely obese, putting them at an elevated risk for premature cardiovascular disease and some cancers relative to non-Hispanic Whites.