A text-mining approach to obtain detailed treatment information from free-text fields in population-based cancer registries: A study of non-small cell lung cancer in California.

BACKGROUND: Population-based cancer registries have treatment information for all patients making them an excellent resource for population-level monitoring. However, specific treatment details, such as drug names, are contained in a free-text format that is difficult to process and summarize. We assessed the accuracy and efficiency of a text-mining algorithm to identify systemic treatments for lung cancer from free-text fields in the California Cancer Registry. METHODS: The algorithm used Perl regular expressions in SAS 9.4 to search for treatments in 24,845 free-text records associated with 17,310 patients in California diagnosed with stage IV non-small cell lung cancer between 2012 and 2014. Our algorithm categorized treatments into six groups that align with National Comprehensive Cancer Network guidelines. We compared results to a manual review (gold standard) of the same records. RESULTS: Percent agreement ranged from 91.1% to 99.4%. Ranges for other measures were 0.71-0.92 (Kappa), 74.3%-97.3% (sensitivity), 92.4%-99.8% (specificity), 60.4%-96.4% (positive predictive value), and 92.9%-99.9% (negative predictive value). The text-mining algorithm used one-sixth of the time required for manual review. CONCLUSION: SAS-based text mining of free-text data can accurately detect systemic treatments administered to patients and save considerable time compared to manual review, maximizing the utility of the extant information in population-based cancer registries for comparative effectiveness research.

Prevalence and characteristics of cancer patients receiving care from single vs. multiple institutions.

INTRODUCTION: Patients may receive cancer care from multiple institutions. However, at the population level, such patterns of cancer care are poorly described, complicating clinical research. To determine the population-based prevalence and characteristics of patients seen by multiple institutions, we used operations data from a state-mandated cancer registry. METHODS AND MATERIALS: 59,672 invasive cancers diagnosed in 1/1/2010-12/31/2011 in the Greater Bay Area of northern California were categorized as having been reported to the cancer registry within 365days of diagnosis by: 1) ≥1 institution within an integrated health system (IHS); 2) IHS institution(s) and ≥1 non-IHS institution (e.g., private hospital); 3) 1 non-IHS institution; or 4) ≥2 non-IHS institutions. Multivariable logistic regression was used to characterize patients reported by multiple vs. single institutions. RESULTS: Overall in this region, 17% of cancers were reported by multiple institutions. Of the 33% reported by an IHS, 8% were also reported by a non-IHS. Of non-IHS patients, 21% were reported by multiple institutions, with 28% for breast and 27% for pancreatic cancer, but 19%% for lung and 18% for prostate cancer. Generally, patients more likely to be seen by multiple institutions were younger or had more severe disease at diagnosis. CONCLUSIONS: Population-based data show that one in six newly diagnosed cancer patients received care from multiple institutions, and differed from patients seen only at a single institution. Cancer care data from single institutions may be incomplete and possibly biased.

Representativeness of breast cancer cases in an integrated health care delivery system.

BACKGROUND: Integrated health care delivery systems, with their comprehensive and integrated electronic medical records (EMR), are well-poised to conduct research that leverages the detailed clinical data within the EMRs. However, information regarding the representativeness of these clinical populations is limited, and thus the generalizability of research findings is uncertain. METHODS: Using data from the population-based California Cancer Registry, we compared age-adjusted distributions of patient and neighborhood characteristics for three groups of breast cancer patients: 1) those diagnosed within Kaiser Permanente Northern California (KPNC), 2) non-KPNC patients from NCI-designated cancer centers, and 3) those from all other hospitals. RESULTS: KPNC patients represented 32 % (N = 36,109); cancer center patients represented 7 % (N = 7805); and all other hospitals represented 61 % (N = 68,330) of the total breast cancer patients from this geographic area during 1996-2009. Compared with cases from all other hospitals, KPNC had slightly fewer non-Hispanic Whites (70.6 % versus 74.4 %) but more Blacks (8.1 % versus 5.0 %), slightly more patients in the 50-69 age range and fewer in the younger and older age groups, a slightly lower proportion of in situ but higher proportion of stage I disease (41.6 % versus 38.9 %), were slightly less likely to reside in the lowest (4.2 % versus 6.5 %) and highest (36.2 % versus 39.0 %) socioeconomic status neighborhoods, and more likely to live in suburban metropolitan areas and neighborhoods with more racial/ethnic minorities. Cancer center patients differed substantially from patients from KPNC and all other hospitals on all characteristics assessed. All differences were statistically significant (p < .001). CONCLUSIONS: Although much of clinical research discoveries are based in academic medical centers, patients from large, integrated medical centers are likely more representative of the underlying population, providing support for the generalizability of cancer research based on electronic data from these centers.

Racial/ethnic and socioeconomic differences in short-term breast cancer survival among women in an integrated health system.

OBJECTIVES: We examined the combined influence of race/ethnicity and neighborhood socioeconomic status (SES) on short-term survival among women with uniform access to health care and treatment. METHODS: Using electronic medical records data from Kaiser Permanente Northern California linked to data from the California Cancer Registry, we included 6262 women newly diagnosed with invasive breast cancer. We analyzed survival using multivariable Cox proportional hazards regression with follow-up through 2010. RESULTS: After consideration of tumor stage, subtype, comorbidity, and type of treatment received, non-Hispanic White women living in low-SES neighborhoods (hazard ratio [HR] = 1.28; 95% confidence interval [CI] = 1.07, 1.52) and African Americans regardless of neighborhood SES (high SES: HR = 1.44; 95% CI = 1.01, 2.07; low SES: HR = 1.88; 95% CI = 1.42, 2.50) had worse overall survival than did non-Hispanic White women living in high-SES neighborhoods. Results were similar for breast cancer-specific survival, except that African Americans and non-Hispanic Whites living in high-SES neighborhoods had similar survival. CONCLUSIONS: Strategies to address the underlying factors that may influence treatment intensity and adherence, such as comorbidities and logistical barriers, should be targeted at low-SES non-Hispanic White and all African American patients.

Patterns and predictors of breast cancer chemotherapy use in Kaiser Permanente Northern California, 2004-2007.

Chemotherapy regimens for early stage breast cancer have been tested by randomized clinical trials, and specified by evidence-based practice guidelines. However, little is known about the translation of trial results and guidelines to clinical practice. We extracted individual-level data on chemotherapy administration from the electronic medical records of Kaiser Permanente Northern California (KPNC), a pre-paid integrated healthcare system serving 29 % of the local population. We linked data to the California Cancer Registry, incorporating socio-demographic and tumor factors, and performed multivariable logistic regression analyses on the receipt of specific chemotherapy regimens. We identified 6,004 women diagnosed with Stage I-III breast cancer at KPNC during 2004-2007; 2,669 (44.5 %) received at least one chemotherapy infusion at KPNC within 12 months of diagnosis. Factors associated with receiving chemotherapy included <50 years of age [odds ratio (OR) 2.27, 95 % confidence interval (CI) 1.81-2.86], tumor >2 cm (OR 2.14, 95 % CI 1.75-2.61), involved lymph nodes (OR 11.3, 95 % CI 9.29-13.6), hormone receptor-negative (OR 6.94, 95 % CI 4.89-9.86), Her2/neu-positive (OR 2.71, 95 % CI 2.10-3.51), or high grade (OR 3.53, 95 % CI 2.77-4.49) tumors; comorbidities associated inversely with chemotherapy use [heart disease for anthracyclines (OR 0.24, 95 % CI 0.14-0.41), neuropathy for taxanes (OR 0.45, 95 % CI 0.22-0.89)]. Relative to high-socioeconomic status (SES) non-Hispanic Whites, we observed less anthracycline and taxane use by SES non-Hispanic Whites (OR 0.63, 95 % CI 0.49-0.82) and American Indians (OR 0.23, 95 % CI 0.06-0.93), and more anthracycline use by high-SES Asians/Pacific Islanders (OR 1.72, 95 % CI 1.02-2.90). In this equal-access healthcare system, chemotherapy use followed practice guidelines, but varied by race and socio-demographic factors. These findings may inform efforts to optimize quality in breast cancer care.

Reducing the risk for distal forearm fracture: preserve bone mass, slow down, and don't fall!

A case-control study of 1,150 female and male distal forearm cases and 2,331 controls of age 45 years and older was undertaken from 1996-2001 in five Northern California Kaiser Permanente Medical Centers. Most information on possible risk factors was obtained by an interviewer-administered questionnaire, supplemented by a few tests of lower extremity neurological function. Previous fractures since 45 years of age, a rough marker of osteoporosis, were associated with an increased risk (adjusted odds ratio [OR] [95% confidence interval] = 1.48 [1.20-1.84 ] per previous fracture). Several factors thought to protect against low bone mass were associated with a reduced risk, including current use of menopausal hormone therapy (adjusted OR = 0.60 [0.49-0.74]), ever used thiazide diuretics or water pills for at least 1 year (adjusted OR = 0.79 [0.64-0.97]), high body mass index (weight in kg/height in m2) (adjusted OR = 0.96 [0.89-1.04] per 5 unit increase), and high dietary calcium intake (adjusted OR = 0.88 [0.75-1.03] per 500 mg/day). Falls in the past year and conditions associated with falling, such as epilepsy and/or use of seizure medication (adjusted OR = 2.07 [1.35-3.17]) and a history of practitioner-diagnosed depression (adjusted OR = 1.40 [1.13-1.73]), were associated with increased risks. Having difficulty performing physical functions and all lower-extremity problems measured in this study were associated with reduced risks. The results from this and other studies indicate that distal forearm fractures tend to occur in people with low bone mass who are otherwise in relatively good health and are physically active, but who are somewhat prone to falling (particularly on an outstretched hand), and whose movements are not slowed by lower extremity problems and other debilities. Thus, measures to decrease fall frequency and to slow down the pace of relatively healthy people with low bone mass should lead to a lower frequency of distal forearm fracture.