Comparison of a Kidney Replacement Therapy Risk Score Developed in Kaiser Permanente Northwest vs Estimated Glomerular Filtration Rate in Advanced Chronic Kidney Disease Using Decision Curve Analysis.

INTRODUCTION: Use of kidney replacement therapy (KRT) prediction models for guiding arteriovenous fistula (AVF) referrals in advanced chronic kidney disease (CKD) is unknown. We aimed to compare a hypothetical approach using a KRT prediction model developed in Kaiser Permanente Northwest to estimated glomerular filtration rate (eGFR) for AVF referrals. METHODS: Our retrospective cohort consisted of patients with stage G4 CKD in Kaiser Permanente Northwest followed by nephrology. Two-year KRT risk was calculated at each nephrology visit up to 2 years from entrance into cohort based on a previously published model. We calculated sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) based on several 2-year KRT risk and eGFR cutoffs for outcome of hemodialysis at 18 months. We compared an approach of AVF referral using 2-year KRT risk and eGFR cutoffs using decision curve analysis. RESULTS: Two-year KRT risk better discriminated progression to hemodialysis compared to eGFR < 15 mL/min (AUC 0.60 vs 0.69 at 2-year KRT risk > 20% and 0.69 at 2-year KRT risk > 40%, p = 0.003 and 0.006, respectively) but not to eGFR of 20 mL/min (AUC 0.64, p = 0.16 and 0.19, respectively). Decision curve analysis showed that AVF referral guided by 2-year KRT risk score resulted in higher net benefit compared to eGFR at low thresholds for referral. CONCLUSION: In stage G4 CKD, a 2-year KRT risk model better predicted progression to KRT at 18 months compared to an eGFR of 15 mL/min but not to 20 mL/min and may improve timely referral for AVF placement in patients at lower thresholds for referral.

Predicted risk of renal replacement therapy at arteriovenous fistula referral in chronic kidney disease.

BACKGROUND AND OBJECTIVES: Optimal timing of arteriovenous fistula placement in chronic kidney disease remains difficult and contributes to high central venous catheter use at initial hemodialysis. We tested whether a prediction model for progression to renal replacement therapy developed at Kaiser Permanente Northwest may help guide decisions about timing of referral for arteriovenous fistula placement. DESIGN, SETTING, PARTICIPANTS, AND MEASUREMENTS: A total of 205 chronic kidney disease stage 4 patients followed by nephrology referred for arteriovenous fistula placement were followed for up to 2 years. Patients were censored if they died or discontinued Kaiser Permanente Northwest coverage. Survival analyses were performed for overall progression to renal replacement therapy divided by quartiles based on 2-year risk for renal replacement therapy and estimated glomerular filtrate rate at time of referral. RESULTS: By 2 years, 60% progressed to renal replacement therapy and 11% had died. 80% in the highest risk versus 36% in the lowest risk quartile progressed to renal replacement therapy (predicted risk 84% vs 17%). 75% in the lowest estimated glomerular filtrate rate versus 56% in the highest estimated glomerular filtrate rate quartile progressed to renal replacement therapy (mean estimated glomerular filtrate rate 13 mL/min vs 21 mL/min). The hazard ratio was significantly higher for each consecutive higher renal replacement therapy quartile risk while for estimated glomerular filtrate rate, the hazard ratio was only significantly higher for the lowest compared to the highest quartile. The extreme quartile risk ratio was higher for 2-year risk for renal replacement therapy compared to estimated glomerular filtrate rate (4.0 vs 2.4). CONCLUSION: In patients with chronic kidney disease stage 4 referred for arteriovenous fistula placement, 2-year renal replacement therapy risk better discriminated progression to renal replacement therapy compared to estimated glomerular filtrate rate at time of referral.

Association Between Social and Economic Needs With Future Healthcare Utilization.

INTRODUCTION: Unmet social and economic needs are associated with poor health outcomes, but little is known about how these needs are predictive of future healthcare utilization. This study examined the association of social and economic needs identified during medical visits with future hospitalizations and emergency department visits. METHODS: Individuals with electronic health record-coded social and economic needs during a primary care, emergency department, or urgent care visit at Kaiser Permanente Northwest from October 1, 2016 to November 31, 2017 (case patients) were identified, as well as individuals who had visits during that time period but had no electronic health record-coded needs (control patients). The 2 groups were compared on sociodemographic characteristics, comorbidities, and healthcare utilization in the prior year. Finally, logistic regression assessed the relationship between documented needs and hospitalizations and emergency department visits in the 12 months following the index visit, controlling for sociodemographic characteristics, comorbidities, and prior healthcare utilization. Statistical analysis was completed in April 2019. RESULTS: Case patients differed significantly from control patients on sociodemographic characteristics and had higher rates of comorbidities and prior healthcare utilization. Social and economic needs documented during the index visit were associated with significantly higher rates of hospitalization and emergency department visits in the 12 months following the visit, controlling for sociodemographic characteristics, comorbidities, and prior utilization. CONCLUSIONS: These results demonstrate that documented social and economic needs are a powerful predictor of future hospitalization and emergency department use and suggest the need for research into whether interventions to address these needs can influence healthcare utilization.

Improving Health Care with Advanced Analytics: Practical Considerations.

Artificial intelligence (AI) is becoming ubiquitous in health care, largely through machine learning and predictive analytics applications. Recent applications of AI to common health care scenarios, such as screening and diagnosing, have fueled optimism about the use of advanced analytics to improve care. Careful and objective considerations need to be made before implementing an advanced analytics solution. Critical evaluation before, during, and after its implementation will ensure safe care, good outcomes, and the elimination of waste. In this commentary we offer basic practical considerations for developing, implementing, and evaluating such solutions based on many years of experience.

Exploring Different Approaches in Measuring EHR-based Adherence to Best Practice - A Case Study with Order Sets and Associated Outcomes.

In connection with a recent enterprise-wide rollout of a new electronic health record, Intermountain Healthcare is investing significant effort in building a central library of best-practice order sets. These order sets represent best practice guidelines for specific clinical scenarios and are deployed with the intent of standardizing care, reducing variation, and consistently delivering good clinical outcomes to the populations we serve. The importance of measuring their use and the level to which caregivers adhere to these standards becomes an important factor in understanding and characterizing the impact that they deliver. Notwithstanding the importance of these metrics, well- defined methods for measuring adherence to a given clinical guideline as delivered through an order set are not fully characterized in the medical literature. In this paper, we describe initial efforts at measuring compliance to a defined 'best practice' standard by means of content utilization analysis, a calculated adherence model, and relevant clinical key performance indicators. The degree to which specified clinical outcomes vary across these measurement models are compared for a group of order sets tied to treating coronary artery bypass graft patients and heart failure patients. While the patterns derived from this analysis show some uncertainty, more granular methods that look at line-item, or 'order level' detail reveal more significant differences in the corresponding set of outcomes than higher-level adherence surrogates.

Predicting readmission risk shortly after admission for CABG surgery.

BACKGROUND: Reducing preventable hospital readmissions after coronary artery bypass graft (CABG) surgery has become a national priority. Predictive models can be used to identify patients at high risk for readmission. However, the majority of the existing models are based on data available at discharge. We sought to develop a model to predict hospital readmission using data available soon after admission for isolated CABG surgery. METHODS: Fifty risk factors were included in a bivariate analysis, 16 of which were significantly associated (P < 0.05) with readmissions and were entered into a multivariate logistic regression and removed stepwise, using backward elimination procedures. The derived model was then validated on 896 prospective isolated CABG cases. RESULTS: Of 2589 isolated CABG patients identified between December 1, 2010, and June 30, 2014, 237(9.15%) were readmitted within 30 days. Five risk factors were predictive of 30-day all-cause readmission: age (odds ratio [OR] = 1.03; 95% confidence interval [CI]: 1.01-1.05; P = 0.004), prior heart failure (OR = 1.55; 95%CI: 1.07-2.24; P = 0.020), total albumin prior to surgery (OR = 0.68; 95%CI: 0.05-0.94; P = 0.021), previous myocardial infarction (OR = 1.44; 95%CI: 1.00-2.08; P = 0.50), and history of diabetes (OR = 1.54; 95%CI: 1.09-2.19; P = 0.015). The area under the curve c-statistic was 0.63 in the derivation sample and 0.65 in the validation sample showing good discrimination. CONCLUSIONS: A 30-day all-cause readmission among isolated CABG patients can be predicted soon after admission with a small number of risk factors.

It takes a village: Exploring the impact of social determinants on delivery system outcomes for heart failure patients.

BACKGROUND: Local social determinants may act as effect modifiers for the impact of neighborhood material deprivation on patient-level healthcare outcomes. The objective of this study was to understand the mediating effect of local social determinants on neighborhood material deprivation and delivery outcomes in heart failure (HF) patients. MATERIAL AND METHODS: A retrospective cohort study was conducted using 4737 HF patients receiving inpatient care (n=6065 encounters) from an integrated healthcare delivery system from 2010 to 2014. Outcomes included post-discharge mortality, readmission risk and length of stay. Deprivation was measured using an area deprivation index by address of residence. Effect modifications measured included urban-rural residency and faith identification using generalized linear regression models. Patient-level data was drawn from the delivery system data warehouse. RESULTS: Faith identification had a significant protective effect on HF patients from deprived areas, lowering 30-day mortality odds by one-third over patients who did not identify with a faith (OR 0.35 95%CI:0.12-0.98;p=0.05). Significant effects persisted at the 90 and 180-day timeframes. In rural areas, lack of faith identification had a multiplicative effect on 30-day mortality for deprived patients (OR 14.0 95%CI:1.47-132.7;p=0.02). No significant effects were noted for other healthcare outcomes. CONCLUSIONS: The lack of expected association between area deprivation and healthcare outcomes in some communities may be explained by the presence of effect modifiers. IMPLICATIONS: Understanding existing effect modifiers for area deprivation in local communities that delivery systems serve can inform targeted quality improvement. These factors should also be considered when comparing delivery system performance for reimbursement and in population health management.

Real world MRI experience with nonconditional and conditional cardiac rhythm devices after MagnaSafe.

INTRODUCTION: The recent MagnaSafe Registry demonstrated safety of nonthoracic magnetic resonance imaging (MRI) with nonconditional cardiac implantable electronic devices (CIEDs). However, independent validation and extension to thoracic MRIs are needed. METHODS AND RESULTS: We prospectively examined 178 consecutive patients with CIEDs who underwent 212 MRI scans (62 with implantable cardioverter/defibrillators) for clinical reasons between February 2014 and August 2016. Scans were done in standard modes with a limit of 2 W/kg. Pacing modes were ODO or OVO for intrinsic rates > 40 and DOO or VOO for rates ≤ 40. Patients were cleared prescan by both radiology and cardiology, and pre- and postscan CIED interrogations were performed. Primary outcome events were death and generator or lead failure. Secondary events included battery voltage loss > 0.04 V, decrease in P wave voltage > 50% or R wave voltage > 25%, threshold increase of > 0.5 V, and impedance change > 50 Ω. Scan sites were 87 (41%) C-spine/head/neck, 28 (13%) T-spine/cardiac/shoulder (thoracic), 69 (33%) L-spine/abdomen/pelvis, and 28 (13%) lower extremity. No primary or secondary outcome events occurred, and no periscan disruption of pacing was noted. CONCLUSION: In a real-world MRI experience in patients with CIEDs representing a broad range of models, types, and scan sites (including thoracic scans), no adverse safety signals were noted. This experience validates and extends that of the large but inclusion-restricted MagnaSafe Registry, profiles MRI scanning in CIED patients in general clinical practice, and argues against replacing nonconditional with conditional devices when MRI is performed in a carefully controlled environment.

Clinical Decision Support to Efficiently Identify Patients Eligible for Advanced Heart Failure Therapies.

BACKGROUND: Patients who need and receive timely advanced heart failure (HF) therapies have better long-term survival. However, many of these patients are not identified and referred as soon as they should be. METHODS: A clinical decision support (CDS) application sent secure email notifications to HF patients' providers when they transitioned to advanced disease. Patients identified with CDS in 2015 were compared with control patients from 2013 to 2014. Kaplan-Meier methods and Cox regression were used in this intention-to-treat analysis to compare differences between visits to specialized and survival. RESULTS: Intervention patients were referred to specialized heart facilities significantly more often within 30 days (57% vs 34%; P < .001), 60 days (69% vs 44%; P < .0001), 90 days (73% vs 49%; P < .0001), and 180 days (79% vs 58%; P < .0001). Age and sex did not predict heart facility visits, but renal disease did and patients of nonwhite race were less likely to visit specialized heart facilities. Significantly more intervention patients were found to be alive at 30 (95% vs 92%; P = .036), 60 (95% vs 90%; P = .0013), 90 (94% vs 87%; P = .0002), and 180 days (92% vs 84%; P = .0001). Age, sex, and some comorbid diseases were also predictors of mortality, but race was not. CONCLUSIONS: We found that CDS can facilitate the early identification of patients needing advanced HF therapy and that its use was associated with significantly more patients visiting specialized heart facilities and longer survival.

Early inpatient calculation of laboratory-based 30-day readmission risk scores empowers clinical risk modification during index hospitalization.

Improving 30-day readmission continues to be problematic for most hospitals. This study reports the creation and validation of sex-specific inpatient (i) heart failure (HF) risk scores using electronic data from the beginning of inpatient care for effective and efficient prediction of 30-day readmission risk. METHODS: HF patients hospitalized at Intermountain Healthcare from 2005 to 2012 (derivation: n=6079; validation: n=2663) and Baylor Scott & White Health (North Region) from 2005 to 2013 (validation: n=5162) were studied. Sex-specific iHF scores were derived to predict post-hospitalization 30-day readmission using common HF laboratory measures and age. Risk scores adding social, morbidity, and treatment factors were also evaluated. RESULTS: The iHF model for females utilized potassium, bicarbonate, blood urea nitrogen, red blood cell count, white blood cell count, and mean corpuscular hemoglobin concentration; for males, components were B-type natriuretic peptide, sodium, creatinine, hematocrit, red cell distribution width, and mean platelet volume. Among females, odds ratios (OR) were OR=1.99 for iHF tertile 3 vs. 1 (95% confidence interval [CI]=1.28, 3.08) for Intermountain validation (P-trend across tertiles=0.002) and OR=1.29 (CI=1.01, 1.66) for Baylor patients (P-trend=0.049). Among males, iHF had OR=1.95 (CI=1.33, 2.85) for tertile 3 vs. 1 in Intermountain (P-trend <0.001) and OR=2.03 (CI=1.52, 2.71) in Baylor (P-trend < 0.001). Expanded models using 182-183 variables had predictive abilities similar to iHF. CONCLUSIONS: Sex-specific laboratory-based electronic health record-delivered iHF risk scores effectively predicted 30-day readmission among HF patients. Efficient to calculate and deliver to clinicians, recent clinical implementation of iHF scores suggest they are useful and useable for more precise clinical HF treatment.

Automated identification and predictive tools to help identify high-risk heart failure patients: pilot evaluation.

OBJECTIVE: Develop and evaluate an automated identification and predictive risk report for hospitalized heart failure (HF) patients. METHODS: Dictated free-text reports from the previous 24 h were analyzed each day with natural language processing (NLP), to help improve the early identification of hospitalized patients with HF. A second application that uses an Intermountain Healthcare-developed predictive score to determine each HF patient's risk for 30-day hospital readmission and 30-day mortality was also developed. That information was included in an identification and predictive risk report, which was evaluated at a 354-bed hospital that treats high-risk HF patients. RESULTS: The addition of NLP-identified HF patients increased the identification score's sensitivity from 82.6% to 95.3% and its specificity from 82.7% to 97.5%, and the model's positive predictive value is 97.45%. Daily multidisciplinary discharge planning meetings are now based on the information provided by the HF identification and predictive report, and clinician's review of potential HF admissions takes less time compared to the previously used manual methodology (10 vs 40 min). An evaluation of the use of the HF predictive report identified a significant reduction in 30-day mortality and a significant increase in patient discharges to home care instead of to a specialized nursing facility. CONCLUSIONS: Using clinical decision support to help identify HF patients and automatically calculating their 30-day all-cause readmission and 30-day mortality risks, coupled with a multidisciplinary care process pathway, was found to be an effective process to improve HF patient identification, significantly reduce 30-day mortality, and significantly increase patient discharges to home care.

Epinome - a novel workbench for epidemic investigation and analysis of search strategies in public health practice.

We present a novel user-centric visual analytics system that supports investigation of simulated disease outbreak and the study of decision-making. We developed Epinome as part of our research on decision making in public health and in particular, on the evaluation of information search strategies in public health practice. Epinome is a highly dynamic web-based system that provides a platform to track and study subjects' decision making and information search strategies, under controlled and repeatable conditions using simulated disease outbreaks. In this paper we focus on the design and implementation of Epinome and present relevant results from field tests we conducted in Utah and Colorado.

Rural-urban differences in colorectal cancer screening capacity in Arizona.

Colorectal cancer can be prevented via screening by the detection and removal of colorectal adenomas. Few data exist on screening capacity by rural/urban areas. Therefore, the aims of this work were to evaluate current colorectal cancer endoscopy screening capacity and to estimate potential volume for rural and urban regions in Arizona. Gastroenterologists and colorectal surgeons practicing in Arizona completed a survey (n = 105) that assessed current colonoscopy and sigmoidoscopy screening and estimated future capacity. Resources needed to increase capacity were identified, and differences between rural and urban regions were examined. Responders were more likely to practice in an urban region (89.5%). Physicians reported performing 8,717 endoscopic procedures weekly (8,312 in urban and 405 in rural regions) and the vast majority were colonoscopies (91% in urban and 97% in rural regions). Urban physicians estimated being able to increase their capacity by 35.7% (95% confidence interval 34.7-35.7) whereas rural physicians estimated an increase of 53.1% (95% confidence interval 48.1-58.0). The most commonly cited resource needed to increase capacity was a greater number of physicians in urban regions (52.1%); while the top response in rural areas was appropriate compensation (54.6%). Lastly, 27.3% of rural physicians noted they did not need additional resources to increase their capacity. In conclusion, Arizona has the ability to expand colorectal cancer screening endoscopic capacity; this potential increase was more pronounced in rural as compared to urban regions.

Social network analyses of patient-healthcare worker interactions: implications for disease transmission.

Patients and healthcare workers (HCW) in healthcare settings represent a unique social network in which the risk of transmission of an infection is considered to be higher for both HCW and patients. Using data from existing clinical informatics resources, we constructed social networks of patient-HCW interactions in the emergency department of a tertiary care pediatric hospital. The structural properties of these networks were analyzed and compared to other well known networks. Patient-HCW networks do not demonstrate the classical power-law distribution of scale-free networks, thus indicating that they are different from social networks of individuals in a community. The clustering coefficient is larger as compared to a random network, indicating small world properties. The eigenvector centrality, used to identify the most important nodes, reveals HCW to be more connected than patients. These properties imply differences that must be taken into account when analyzing patient-HCW networks and planning interventions and mitigation strategies to prevent the spread of infectious diseases in healthcare settings.

Interactive agent based modeling of public health decision-making.

Agent-based models have yielded important insights regarding the transmission dynamics of communicable diseases. To better understand how these models can be used to study decision making of public health officials, we developed a computer program that linked an agent-based model of pertussis with an agent-based model of public health management. The program, which we call the Public Health Interactive Model & simulation (PHIMs) encompassed the reporting of cases to public health, case investigation, and public health response. The user directly interacted with the model in the role of the public health decision-maker. In this paper we describe the design of our model, and present the results of a pilot study to assess its usability and potential for future development. Affinity for specific tools was demonstrated. Participants ranked the program high in usability and considered it useful for training. Our ultimate goal is to achieve better public health decisions and outcomes through use of public health decision support tools.

Serum 25(OH)D levels, dietary intake of vitamin D, and colorectal adenoma recurrence.

There is strong epidemiological and laboratory evidence that vitamin D may be protective against colorectal neoplasia. Therefore, we sought to assess the relationship between serum 25(OH)D levels, dietary intake of vitamin D, and colorectal adenoma recurrence in our ursodeoxycholic acid trial. A total of 568 participants were randomly selected for analysis of serum 25(OH)D levels. The range of total 25(OH)D was 5.5-66.1 ng/ml, with a median of 25.6 ng/ml. After categorizing 25(OH)D levels into tertiles based on the population distribution, the adjusted odds ratios (95% CI) for adenoma recurrence in the second and third tertiles were 0.88 (0.56-1.39) and 0.78 (0.49-1.24), respectively. The association between serum 25(OH)D and adenoma recurrence appeared to be stronger among women than men. As compared to those below the median value, women with serum 25(OH)D levels above the median had an OR (95% CI) of 0.59 (0.30-1.16); the corresponding OR (95% CI) for men was 0.95 (0.60-1.49). Analyses by dietary vitamin D intake revealed no statistically significant associations. In summary, the results of this study show a moderate, nonsignificant inverse association between serum 25(OH)D levels and reduced risk for colorectal adenoma recurrence, particularly among women.

SPSS and SAS programming for the testing of mediation models.

BACKGROUND: Mediation modeling can explain the nature of the relation among three or more variables. In addition, it can be used to show how a variable mediates the relation between levels of intervention and outcome. The Sobel test, developed in 1990, provides a statistical method for determining the influence of a mediator on an intervention or outcome. Although interactive Web-based and stand-alone methods exist for computing the Sobel test, SPSS and SAS programs that automatically run the required regression analyses and computations increase the accessibility of mediation modeling to nursing researchers. OBJECTIVES: To illustrate the utility of the Sobel test and to make this programming available to the Nursing Research audience in both SAS and SPSS. METHODS: The history, logic, and technical aspects of mediation testing are introduced. The syntax files sobel.sps and sobel.sas, created to automate the computation of the regression analysis and test statistic, are available from the corresponding author. RESULTS: The reported programming allows the user to complete mediation testing with the user's own data in a single-step fashion. A technical manual included with the programming provides instruction on program use and interpretation of the output. CONCLUSION: Mediation modeling is a useful tool for describing the relation between three or more variables. Programming and manuals for using this model are made available.