A comparison of two models for breast cancer mortality for women with ductal carcinoma in situ: an SEER-based analysis

BACKGROUND

Approximately 1% of patients with ductal carcinoma in situ (DCIS) will die of breast cancer within 10 years. Women who develop an invasive breast cancer after DCIS have a much greater risk of dying than those who do not and it is often stated that these deaths are a consequence of metastases from the invasive in-breast recurrence. This progression is the result of a two-step process: first local invasive recurrence and then spread beyond the breast. A large proportion of women who die of DCIS have no record of invasive recurrence. We used SEER data and a simulation approach to test whether the actual mortality data are consistent with the two-step model.

METHODS

First, we constructed Kaplan-Meier mortality curves for all patients with pure DCIS and with small node-negative invasive breast cancers in the Surveillance, Epidemiology and End Results (SEER) registries database (1998-2014). We then constructed, through simulation, theoretical breast cancer mortality curves. To model the two-step scenario, we applied the annual incidence rates of incident invasive cancer following DCIS and of death from invasive cancer after DCIS to a theoretical cohort of 100,000 women.

RESULTS

The observed 15-year breast cancer-specific mortality rate for patients with pure DCIS in the SEER database was 2.0%. The expected mortality for DCIS patients (assuming a two-step process) was only 1.1% at 15 years. Assuming the mortality rates following DCIS were one-half of those observed for patients with small invasive breast cancers, the expected mortality at 15 years post-DCIS was 2.1%.

CONCLUSIONS

In the SEER database, we observed far more deaths from DCIS than would be expected under a model where all deaths from breast cancer occur amongst women who experience an invasive local recurrence. This lends support to the hypothesis that DCIS mortality is not restricted to those women who experience an in-breast invasive cancer and that DCIS has properties similar to small invasive breast cancers.

Impact of Physician Follow-Up Care on Psychiatric Readmission Rates in a Population-Based Sample of Patients With Schizophrenia

OBJECTIVE

The study evaluated the association between physician follow-up within 30 days after hospital discharge and psychiatric readmission within the subsequent 180 days.

METHODS

Among inpatients with schizophrenia who were discharged between 2007 and 2012 in Ontario (N=19,132), those who had a 30-day follow-up visit with a primary care physician (PCP) only, a psychiatrist only, or both were compared with a no-follow-up group. The primary outcome was psychiatric readmission in the subsequent 180 days. Secondary analyses stratified the sample on the basis of readmission risk at discharge.

RESULTS

About 65% of patients had follow-up care within 30 days postdischarge. Psychiatric readmission rates were similar among patients with any physician follow-up and significantly lower than among those with no follow-up (26%): PCP only: 22%; adjusted hazard ratio [aHR]=.88, 95% confidence interval [CI]=.81-.96; psychiatrist only, 22%; aHR=.84, CI=.77-.90; both, 21%, aHR=.82, CI=.75-.90). In stratified analyses, 66% of patients were in the category at highest risk of psychiatric readmission, and the effect of follow-up with any physician was significant for these patients, compared with high-readmission risk patients with no follow-up, who had a 29% readmission rate (PCP only, 20% readmission rate, aHR=.85, CI=.77-.94; psychiatrist only, 29%, aHR=.84, CI=.77-.92; both, 17%, aHR=.81, CI=.73-.90).

DISCUSSION

Timely physician follow-up was associated with reduced risk of psychiatric readmissions, with the greatest reduction among patients at high risk of readmission. Because more than one-third of patients had no physician visit within 30 days postdischarge, improving physician follow-up may help reduce psychiatric readmission rates.

Dysphagia screening after intracerebral hemorrhage

Background

Dysphagia screening is recommended after acute stroke to identify patients at risk of aspiration and implement appropriate care. However, little is known about the frequency and outcomes of patients undergoing dysphagia screening after intracerebral hemorrhage (ICH).

Methods

We used the Ontario Stroke Registry from 1 April 2010 to 31 March 2013 to identify patients hospitalized with acute stroke and to compare dysphagia screening rates in those with ICH and ischemic stroke. In patients with ICH we assessed predictors of receiving dysphagia screening, predictors of failing screening, and outcomes after failing screening.

Results

Among 1091 eligible patients with ICH, 354 (32.4%) patients did not have documented dysphagia screening. Patients with mild ICH were less likely to receive screening (40.4% of patients were omitted, adjusted odds ratio (aOR) 0.40, 95% confidence interval (CI) 0.26–0.63). Older age, greater stroke severity, speech deficits, lower initial level of consciousness, and admission to intensive care unit were predictive of failing the screening test. Failing screening was associated with poor outcomes, including pneumonia (aOR 5.3, 95% CI 2.36–11.88), severe disability (aOR 4.78, 95% CI 3.08–7.41), and 1-year mortality (adjusted hazard ratio 2.1, 95% CI 1.38–3.17). When compared to patients with ischemic stroke, patients with ICH were less likely to receive dysphagia screening (aOR 0.64, 95% CI 0.54–0.76) and more likely to fail screening (aOR 1.98, 95% 1.62–2.42).

Conclusion

One-third of patients with ICH did not have documented dysphagia screening, increasing to 40% in patients with mild clinical severity. Failing screening was associated with poor outcomes. Patients with ICH were less like to receive screening and twice as likely to fail compared to patients with ischemic stroke, and thus efforts should be made to include ICH patients in dysphagia screening protocols whenever possible.

The Variation of Statin Use Among Nursing Home Residents and Physicians: A Cross-Sectional Analysis

OBJECTIVES

To examine the variability of statin use among nursing home residents and prescribing physicians, and to assess statin use by resident frailty.

DESIGN

Population-based, cross-sectional analysis.

SETTING

All nursing home facilities (N = 631) in Ontario, Canada between April 1, 2013 and March 31, 2014.

PARTICIPANTS

All adults aged 66 years and older who received a full clinical assessment while residing in a nursing home facility and their assigned, most responsible, physician.

MEASUREMENTS

Statin use on date of clinical assessment. Resident- and physician-level characteristics ascertained through clinical assessment and health administrative data. Resident frailty was derived using a previously validated index.

RESULTS

Among 76,226 nursing home residents assigned to 1,919 physicians, 25,648 (33.6%) were statin users. There were 13,331 (30.1%) statin users among the 44,290 residents categorized as frail. In an adjusted mixed-effects logistic regression model, frail residents (adjusted odds ratio = 0.62, 95% confidence interval 0.58-0.65) were significantly less likely to be statin users compared with non-frail residents. After adjustment for resident characteristics, the intraclass correlation coefficient indicated that between-physician variability accounted for 9.1% of the residual unexplained variation in statin use (P < .001). Among the 894 physicians assigned 20 or more residents, funnel plots confirmed there were more low-outlying (17.4%) and high-outlying (12.0%) prescribers of statins than expected by chance. Physicians who were high-outlying prescribers had higher historical rates of statin prescribing.

CONCLUSIONS AND RELEVANCE

Statin prescribing was substantial within nursing homes, even among frail residents. After controlling for resident characteristics, the likelihood of statin prescribing varied significantly across physicians. Further studies are required to evaluate the risks and benefits of statin use, and discontinuation, among nursing home residents to better inform clinical practice in this setting.

Association of the Timing of Pregnancy With Survival in Women With Breast Cancer

Importance

Increasing numbers of women experience pregnancy around the time of, or after, a diagnosis of breast cancer. Understanding the effect of pregnancy on survival in women with breast cancer will help in the counseling and treatment of these women.

Objective

To compare the overall survival of women diagnosed with breast cancer during pregnancy or in the postpartum period with that of women who had breast cancer but did not become pregnant.

Design, Setting, and Participants

This population-based, retrospective cohort study linked health administrative databases in Ontario, Canada, comprising 7553 women aged 20 to 45 years at the time of diagnosis with invasive breast cancer, from January 1, 2003, to December 31, 2014.

Exposures

Any pregnancy in the period from 5 years before, until 5 years after, the index date of the diagnosis of breast cancer. Women were classified into the following 4 exposure groups: no pregnancy (the referent), pregnancy before breast cancer, pregnancy-associated breast cancer, and pregnancy following breast cancer.

Main Outcomes and Measures

Five-year actuarial survival rates for all exposure groups, age-adjusted and multivariable hazard ratios [HRs] of pregnancy for overall survival for all exposure groups, and time-dependent hazard ratios for women with pregnancy following breast cancer.

Results

Among the 7553 women in the study (mean age at diagnosis, 39.1 years; median, 40 years; range, 20-44 years) the 5-year actuarial survival rate was 87.5% (95% CI, 86.5%-88.4%) for women with no pregnancy, 85.3% (95% CI, 82.8%-87.8%) for women with pregnancy before breast cancer (age-adjusted hazard ratio, 1.03; 95% CI, 0.85-1.27; P = .73), and 82.1% (95% CI, 78.3%-85.9%) for women with pregnancy-associated breast cancer (age-adjusted hazard ratio, 1.18; 95% CI, 0.91-1.53; P = .20). The 5-year actuarial survival rate was 96.7% (95% CI, 94.1%-99.3%) for women who had pregnancy 6 months or more after diagnosis of breast cancer, vs 87.5% (95% CI, 86.5%-88.4%) for women with no pregnancy) (age-adjusted HR, 0.22; 95% CI, 0.10-0.49; P < .001).

Conclusions and Relevance

Pregnancy did not adversely affect survival in women with breast cancer. For breast cancer survivors who wish to conceive, the risk of death is lowest if pregnancy occurs 6 months or more after diagnosis.

Influences on the start, selection and duration of treatment with antibiotics in long-term care facilities

BACKGROUND

Understanding the extent to which current antibiotic prescribing behaviour is influenced by clinicians' historical patterns of practice will help target interventions to optimize antibiotic use in long-term care. Our objective was to evaluate whether clinicians' historical prescribing behaviours influence the start, prolongation and class selection for treatment with antibiotics in residents of long-term care facilities.

METHODS

We conducted a retrospective cohort study of all physicians who prescribed to residents in long-term care facilities in Ontario between Jan. 1 and Dec. 31, 2014. We examined variability in antibiotic prescribing among physicians for 3 measures: start of treatment with antibiotics, use of prolonged durations exceeding 7 days and selection of fluoroquinolones. Funnel plots with control limits were used to determine the extent of variation and characterize physicians as extreme low, low, average, high and extreme high prescribers for each tendency. Multivariable logistic regression was used to assess whether a clinician's prescribing tendency in the previous year predicted current prescribing patterns, after accounting for residents' demographics, comorbidity, functional status and indwelling devices.

RESULTS

Among 1695 long-term care physicians, who prescribed for 93 132 residents, there was wide variability in the start of antibiotic treatment (median 45% of patients, interquartile range [IQR] 32%-55%), use of prolonged treatment durations (median 30% of antibiotic prescriptions, IQR 19%-46%) and selection of fluoroquinolones (median 27% of antibiotic prescriptions, IQR 18%-37%). Prescribing tendencies for antibiotics by physicians in 2014 correlated strongly with tendencies in the previous year. After controlling for individual resident characteristics, prior prescribing tendency was a significant predictor of current practice.

INTERPRETATION

Physicians prescribing antibiotics exhibited individual, measurable and historical tendencies toward start of antibiotic treatment, use of prolonged treatment duration and class selection. Prescriber audit and feedback may be a promising tool to optimize antibiotic use in long-term care facilities.

Statin therapy and mortality among new long-term care residents in Ontario, Canada: the contribution of clinical assessment data to a population-based cohort study

ABSTRACTObjectiveThere is limited evidence from randomized trials and observational studies to guide clinical practice regarding the use of statins in long-term care (LTC); the effectiveness of statins among those with limited life expectancy is not clear and there is concern that the risk of drug-related adverse events might outweigh any benefit. We examined the impact of initiating statin therapy on mortality for patients newly admitted to LTC. ApproachPopulation-based health administrative data from Ontario, Canada were used to conduct a retrospective cohort study of newly admitted LTC residents, aged 66+ years and no statin use in the previous year, between January 1 2011 and December 31 2014. This cohort was linked to Resident Assessment Instrument (interRAI) data to capture clinical and functional characteristics (including frailty, activities of daily living, and cognitive function). The primary exposure was statin use within 30 days following LTC entry; residents who died or did not receive an interRAI assessment within 30 days were excluded. A propensity score for receiving statins was computed using resident demographic, clinical and functional characteristics. We matched exposed to unexposed patients on the basis of age (±1 year), sex, prior myocardial infarction(MI)/stroke hospitalization, frailty and propensity score (±0.2 standard deviations). Patients were followed in an intention-to-treat manner from the end of the exposure window until the earliest of death or March 31 2015. Cox regression was used to compare mortality between the study groups. ResultsWe identified 39,560 newly admitted LTC residents aged 66+ years with no statin use in the previous year, of which 1,953 (4.9%) were prescribed a statin within 30 days of LTC entry. Propensity score matching resulted in 1,710 pairs of exposed and unexposed patients. In the matched cohort, those receiving statins had a lower rate of mortality compared with those not receiving statins (Hazard Ratio 0.77; 95% Confidence Interval [CI] 0.70-0.85). In pre-specified subgroup analyses, the association between statin use and reduced mortality persisted among those with and without a prior MI/stroke hospitalization and among those categorized as frail and not frail. ConclusionOur data suggest initiating statins may be beneficial in reducing mortality risk among LTC residents, despite the complexity and advanced age of the patients. By linking rich resident-level health and functional assessment data with health administrative data we were able to characterize the association between demographic and clinical characteristics (including frailty) and exposure to statins more fully than with administrative data alone.

Predictors and Outcomes of Dysphagia Screening After Acute Ischemic Stroke

Background and Purpose—

Guidelines advocate screening all acute stroke patients for dysphagia. However, limited data are available regarding how many and which patients are screened and how failing a swallowing screen affects patient outcomes. We sought to evaluate predictors of receiving dysphagia screening after acute ischemic stroke and outcomes after failing a screening test.

Methods—

We used the Ontario Stroke Registry from April 1, 2010, to March 31, 2013, to identify patients hospitalized with acute ischemic stroke and determine predictors of documented dysphagia screening and outcomes after failing the screening test, including pneumonia, disability, and death.

Results—

Among 7171 patients, 6677 patients were eligible to receive dysphagia screening within 72 hours, yet 1280 (19.2%) patients did not undergo documented screening. Patients with mild strokes were significantly less likely than those with more severe strokes to have documented screening (adjusted odds ratio, 0.51; 95% confidence interval [CI], 0.41–0.64). Failing dysphagia screening was associated with poor outcomes, including pneumonia (adjusted odds ratio, 4.71; 95% CI, 3.43–6.47), severe disability (adjusted odds ratio, 5.19; 95% CI, 4.48–6.02), discharge to long-term care (adjusted odds ratio, 2.79; 95% CI, 2.11–3.79), and 1-year mortality (adjusted hazard ratio, 2.42; 95% CI, 2.09–2.80). Associations were maintained in patients with mild strokes.

Conclusions—

One in 5 patients with acute ischemic stroke did not have documented dysphagia screening, and patients with mild strokes were substantially less likely to have documented screening. Failing dysphagia screening was associated with poor outcomes, including in patients with mild strokes, highlighting the importance of dysphagia screening for all patients with acute ischemic stroke.

The impact of nodal micrometastasis on mortality among women with early-stage breast cancer

PURPOSE

The clinical significance of nodal micrometastasis is debated. Our primary objective was to determine whether, among women with early-stage breast cancer, regional lymph node micrometastasis is an independent risk factor for mortality. The secondary objective was to identify subgroups of women who have the highest risk of death from early-stage breast cancer with micrometastases.

METHODS

206,625 women diagnosed with early-stage breast cancer (IA, IB, and IIA) from 2004 to 2012 were identified in the Surveillance, epidemiology, and end results database. Nodal status was classified as node-negative, isolated-tumor cells, micrometastases, and macrometastases. Women were classified into eight ethnic groups. Logistic regression was performed to estimate the odds ratio of being diagnosed with micrometastases. The Cox proportional hazard model was used to estimate the hazard ratio (HR) of breast cancer-specific death associated with micrometastases for each ethnic group.

RESULTS

The 8-year breast cancer-specific survival was 96.6 % for women with node-negative breast cancers and was 94.6 % for women with micrometastases (adjusted HR 1.49; 95 % CI 1.31-1.69; P < .001). Among women with micrometastases, the 8-year breast cancer-specific survival was 95.1 % for white women and was 90.6 % for black women (HR 1.80; 95 % CI 1.29-2.52; P = .0006).

CONCLUSION(S)

Nodal micrometastasis is an independent risk factor for breast cancer mortality among women with early-stage breast cancer. Black women are more likely to die from breast cancer with micrometastases than white women.

Naive Bayes classifiers for verbal autopsies: comparison to physician-based classification for 21,000 child and adult deaths

BACKGROUND

Verbal autopsies (VA) are increasingly used in low- and middle-income countries where most causes of death (COD) occur at home without medical attention, and home deaths differ substantially from hospital deaths. Hence, there is no plausible "standard" against which VAs for home deaths may be validated. Previous studies have shown contradictory performance of automated methods compared to physician-based classification of CODs. We sought to compare the performance of the classic naive Bayes classifier (NBC) versus existing automated classifiers, using physician-based classification as the reference.

METHODS

We compared the performance of NBC, an open-source Tariff Method (OTM), and InterVA-4 on three datasets covering about 21,000 child and adult deaths: the ongoing Million Death Study in India, and health and demographic surveillance sites in Agincourt, South Africa and Matlab, Bangladesh. We applied several training and testing splits of the data to quantify the sensitivity and specificity compared to physician coding for individual CODs and to test the cause-specific mortality fractions at the population level.

RESULTS

The NBC achieved comparable sensitivity (median 0.51, range 0.48-0.58) to OTM (median 0.50, range 0.41-0.51), with InterVA-4 having lower sensitivity (median 0.43, range 0.36-0.47) in all three datasets, across all CODs. Consistency of CODs was comparable for NBC and InterVA-4 but lower for OTM. NBC and OTM achieved better performance when using a local rather than a non-local training dataset. At the population level, NBC scored the highest cause-specific mortality fraction accuracy across the datasets (median 0.88, range 0.87-0.93), followed by InterVA-4 (median 0.66, range 0.62-0.73) and OTM (median 0.57, range 0.42-0.58).

CONCLUSIONS

NBC outperforms current similar COD classifiers at the population level. Nevertheless, no current automated classifier adequately replicates physician classification for individual CODs. There is a need for further research on automated classifiers using local training and test data in diverse settings prior to recommending any replacement of physician-based classification of verbal autopsies.

Why have ovarian cancer mortality rates declined? Part III. Prospects for the future

Over the last 40 years, the age-adjusted ovarian cancer mortality rate in the USA declined by 23%. The decline in mortality paralleled a decline in incidence, which was largely due to changes in reproductive risk factors. There was no reduction in ovarian cancer case-fatality at 12 years, indicating that improvements in early detection or in treatment did not contribute to the decline in mortality. Here, we discuss potential strategies to further reduce ovarian cancer mortality through prevention, early detection and treatment. The first approach is to increase genetic testing, in order to identify women who are at a high risk of developing ovarian cancer and offer them preventive bilateral salpingo-oophorectomy. At present, up to 17% of ovarian cancers are potentially preventable through population-based genetic testing of known ovarian cancer susceptibility genes. The second approach is to increase the proportion of ovarian cancer patients who achieve a status of no residual disease through primary debulking surgery and subsequently receive adjuvant intraperitoneal chemotherapy. We believe that through a combination of screening to better identify low-volume advanced stage ovarian cancer, aggressive surgery to leave no residual disease and adjuvant intraperitoneal chemotherapy, the cure rate of ovarian cancer might be improved significantly.