Update 2021

​​​​​​​​​​​​​​​​​​​​​​​Welcome to the first overview of The Permanente Medical Group’s Delivery Science and Applied Research (DARE) program!

Founded in 2018, under the leadership of Rich Isaacs and the Associate Executive Directors, DARE provides infrastructure, connections, and analytic support to clinician-researchers for answering questions that will change care. It fosters collaborations between clinician-investigators and Division of Research scientists and provides support from initial idea development to when data informs implementation. 

The DARE program includes:

  • Multiple funding mechanisms  for investigations of varying duration/complexity;
  • The Physician Researcher Program, providing high-level support for selected clinicians across specialties; 
  • A one-stop idea/funding submission portal and efficient core administrative team;
  • Specialty-specific research networks that build communities of evidence-driven clinicians. 
  • Research tools, collaborations, and funding for dissemination.

We invite you to learn more within, including the “DARE Dozen”:  twelve projects drawn from almost ninety DARE investigations in 2019-20 alone.  These are poised to change care or have already done so.

We look forward, in 2021, to increased support for implementation - connecting research results with operational leaders to translate to clinical actions. 

It has been an exciting two years – TPMG’s clinician-investigators’ talent, productivity, and dedication to evaluating and improving care is inspiring.  We look forward to your new research ideas and welcome recommendations.

Warm regards on behalf of the DARE team,

Douglas Corley, MD, PhD
Director, Delivery Science and Applied Research, TPMG

Tracy Lieu, MD, MPH
Director, Kaiser Permanente Division of Research

Yi Fen Irene Chen, MD and Smita Rouillard, MD
Associate Executive Directors, TPMG


Featured Projects - "DARE Dozen"

AuthorsJulia L. Marcus, PhD, MPH; Leo B. Hurley, MPH; Scott Chamberland, PharmD; Jamila H. Champsi, MD; Laura C. Gittleman, RN, MBA; Daniel G. Korn, MD; Jennifer B. Lai, MSc, PharmD; Jennifer O. Lam, PhD, MPH; Mary Patricia Pauly, MD; Charles P. Quesenberry, Jr., PhD; Joanna Ready, MD; Varun Saxena, MD; Suk Seo, MD; David J. Witt, MD; and Michael J. Silverberg, PhD, MPH
ChallengesGuidelines for hepatitis C treatment indicate that most people can be treated with ledipasvir and sofosbuvir for shorter durations (i.e., 8 weeks instead of the originally approved 12 weeks) but Black people with HCV were not included in this shorter recommendation based on data from older HCV treatments.
Existing EvidencePrior observational studies suggested reduced response for black patients with hepatitis C receiving 8 weeks of therapy. However, because prior studies did not limit analyses to black patients otherwise eligible for 8 weeks (i.e., treatment-naive, no cirrhosis, HIV-uninfected, and HCV RNA <6 million IU/mL), black patients receiving 8 and 12 weeks may have differed with respect to key factors for treatment response.
Target PopulationKPNC patients with HCV genotype 1 infection eligible for 8 week direct-acting antiviral regimen of ledipasvir/sofosbuvir.
Intervention or Exposure8 or 12 week ledipasvir (LDV)/sofosbuvir (SOF).
Outcomes/Key FindingsOf 2653 patients eligible for 8 weeks of treatment with LDV/SOF, 1958 (73.8%) received 8 weeks of treatment and 695 (26.2%) received 12 weeks; the proportions of patients with sustained virologic response 12 weeks after the end of treatment (SVR12) were 96.3% for those given 8-weeks and 96.3% for those given 12 weeks of treatment (P = .94). Similarly, when stratified by race, there was no difference in SVR12 by regimen duration (see figure). Specifically, for Black people with HCV, the percentages with SVR12 for 8- and 12-week regimens were 95.6% vs 95.8%, respectively, with an adjusted relative risk of 1.0 (P = .88).
Resulting Action/ChangeThese findings changed national KP clinical practice to an 8-week course of direct-acting anti-hepatitis C treatment for eligible black patients (instead of 12-weeks). This decreased patient inconvenience, decreased cost by one third, and may decrease toxicity associated with longer durations of treatment. In 2019, new liver society guidelines cited this study and revised guidance to shorter regimens for black people.
Additional RecommendationsThese results recommend evaluations for treatment differences/similarities for other medication regimens with conflicting data in the literature.
Implementation ToolsNew KP and national liver society guidelines based on these results.
Implementation and
Follow-up Measures
Proportions of people with guideline concordant care, consistent with these findings: (e.g. 8 weeks for eligible patients); pharmacy utilization/cost; virologic response.

​SVR12 for HCV treatment of 8 vs. 12 weeks among 2653 HCV-infected individuals otherwise eligible for 8-week regimens

doi: 10.1016/j.cgh.2018.03.003. Epub 2018 Mar 11. 

AuthorsTheodore R. Levin, MD, Christopher D. Jensen, PhD, MPH, Neetu M. Chawla, PhD, Lori C. Sakoda, PhD, MPH, Jeffrey K. Lee, MD, MAS, Wei K. Zhao, MPH, Molly A. Landau, MPH, Ariel Herm, MPH, Eryn Eby, MPH, Charles P. Quesenberry, PhD, Douglas A. Corley, MD, PhD
ChallengesAfrican Americans have increased incidence of colorectal cancer (CRC) before age 50 years, lower CRC screening rates, later stage at diagnosis and poorer survival compared to other races, but no prospective data on screening younger populations exist.
Existing EvidenceSome guidelines recommend starting CRC screening before age 50 years for African Americans, but there are few data on screening uptake, yield and long-term benefits of different screening tests below age 50 in this population.
Target PopulationAfrican Americans age 45–50 years.
Intervention or ExposureA pilot study-directed mailed fecal immunochemical test (FIT) screening outreach program to the target populations.
Outcomes/Key FindingsAmong 10,232 African Americans ages 45–50 mailed a FIT, screening was successfully completed by 33.1% and abnormal results were comparable to those routinely screened ages >50. Among the 4% with positive test results, 85.3% completed a follow-up colonoscopy: 57.8% had any adenoma, 33.6% had an advanced adenoma (adenoma with advanced histology or polyp 10 mm), and 2.6% were diagnosed with CRC. African Americans in the early screening group were modestly more likely to have completed screening than previously unscreened African Americans, whites, and Hispanics 51–56 years old.
Resulting Action/ChangeResults led to change in TPMG policy to start screening African Americans at age 45, including mailed FIT. PROMPT updates are in process.
Additional RecommendationsOperational leaders can consider repeat evaluation to assess response rates with repeated invitations (similar approaches currently used to increase responses for older patients).
Implementation ToolsStudy outreach letter.
Implementation and
Follow-up Measures
Percentage of screening uptake among African Americans age 45–50 in comparison to previously unscreened African Americans, whites, Hispanics and Asian/Pacific-Islanders 51–56 years old (implementation); changes in cancers/cancer stage/ advanced polyps detected (effectiveness); changes in need for surgery/chemotherapy (utilization).
Fecal immunochemical test (FIT)  colorectal cancer screening among African Americans ages 45–50

doi: 10.1053/j.gastro.2020.07.011

AuthorsMatthew D. Solomon, MD, PhD; Grace Tabada, MPH; Amanda Allen; Sue Hee Sung, MPH; Alan S Go, MD
ChallengesValvular heart disease is common, but it is difficult to study the completion and effectiveness of guideline- consistent surveillance clinically, given widely-used diagnosis code-based approaches are inaccurate.
Existing EvidenceCurrent diagnoses codes for heart disease conditions may be unreliable and no systematic methods exist with KPNC for accurately identifying patients for research and population management programs. Regional cardiologists have requested a population management program for valvular heart disease to aid in clinical care and to evaluate the effectiveness of current surveillance guidelines.
Target PopulationAdult patients with at least one physician-read echocardiogram report from 2008–2018.
Intervention or ExposureIdentification of aortic stenosis and associated parameters using a novel natural language processing algorithm for large, unstructured echocardiogram reports.
Outcomes/Key Findings957,505 eligible echocardiograms were identified among 522,633 patients. The final NLP algorithm achieved positive and negative predictive values of >95% for identifying people with aortic stenosis; this was much more accurate than using codes alone. It classified 104,090 (10.9%) echocardiograms as having AS; only 36,070 (34.7%) of these patients had a diagnosis code for AS around the time of the echocardiogram and 35% of these unidentified patients had hemodynamically significant AS (i.e., moderate or severe disease).
Resulting Action/ChangeThe study created the first accurate database of KPNC patients — and one of the world’s largest — with aortic stenosis to allow for: 1) identification of center/provider variation; 2) improved understanding of the natural history of disease; 3) studying the effectiveness of surveillance intervals (ongoing); 4) creating a new regional effort for standardized reporting; and 5) informing a guideline consistent tracking/disease management program for surveillance.
Additional RecommendationsThe results will inform next-steps for standardized regional reporting and further development of operational tracking and centralized surveillance for high-risk patients.
Implementation ToolsNatural language processing algorithms, valvular heart disease database.
Implementation and
Follow-up Measures
Implementation of echocardiographic data base showing mild, moderate or severe aortic stenosis (implementation); variation in appropriate follow-up of patients for surveillance (clinical effectiveness); proportions of patients with follow-up echo, surgery, etc. (utilization).

Application of Validated NLP Algorithm vs. Diagnosis Codes to Identify Aortic Stenosis Among All Echocardiograms

Manuscript pending. 

AuthorsJoseph Presti Jr, MD; Stacey Alexeeff, PhD; Brandon Horton, MPH; Stephanie Prausnitz, MA; Andrew L Avins, MD, MPH
ChallengesIn 2012, the US Preventive Services Task Force (USPSTF) recommended against PSA-based screening for prostate cancer for all men. The impact of the resulting marked decrease in screening on clinical outcomes (including metastatic disease) are unknown and would inform whether more targeted screening may be advisable, such as for higher-risk groups (e.g. African-Americans).
Existing EvidencePSA-based prostate cancer screening was suggested to be minimally effective for prostate cancer decreasing mortality in randomized trials. Thus, the USPSTF’s downgrading of PSA-based screening to “recommended against”, theoretically, this should have little impact on morbidity and deaths from prostate cancer, but little community-based data exist.
Target PopulationScreen eligible men without a history of prostate cancer.
Intervention or Exposure2012 USPSTF Statement stating “Do not screen anyone for prostate cancer.”
Outcomes/Key FindingsAfter the USPSTF recommended against routine prostate cancer screening, screening rates declined 23.4% (95% CI 23.0–23.8%) and biopsy rates declined 64.3% (95% CI 62.9–65.6%). Subsequently, incident prostate cancer diagnoses declined 53.5% (95%
CI 50.1–56.7%), resulting in 1871 fewer incident cancers detected, but metastatic cancer rates increased 36.9% (95% CI 9.5–71.0%) resulting in 75 more stage IV cancers detected.
Resulting Action/ChangeThe finding of more advanced cancers informs next-step already-started analyses for identifying the impact on high-risk populations (e.g. African Americans) who may benefit from more targeted prostate cancer screening, to lower the rate of metastatic cancer while minimizing over-screening of populations not likely to benefit (or to have harm).
Additional RecommendationsThese findings inform potential next steps such as outreach to higher risk groups for informed decision- making regarding screening and informing primary care about the consequences of not screening.
Implementation ToolsN/A
Implementation and
Follow-up Measures
Development of risk-stratified screening tools (implementation); changes in stage IV cancer (effectiveness); PSA testing, surgery, chemotherapy (utilization).

Incident prostate cancer rate by month
​​Incident Stage IV prostate cancer rate by  month

doi: 10.1007/s11606-019-05561-y

AuthorsDan Li, MD; Liyan Liu, MSc; Helene B Fevrier, MPH; Stacey E Alexeeff, PhD; Amanda R Doherty, MD; Menaka Raju, MD; Laura B Amsden, MSW, MPH; Jeffrey K Lee, MD, MPH; Theodore R Levin, MD; Douglas A Corley MD, PhD; and Lisa J Herrinton PhD
ChallengesSerrated colon polyps (SPs) are precursors to 20% to 30% of cases of colorectal cancer (CRC), but patients’ long- term risk after polyp removal is poorly understood, which may lead to inappropriate follow-up colonoscopy intervals. This study investigated the risk of CRC in individuals with a history of SPs.
Existing EvidenceEvidence around SPs and CRC risk is limited. Current surveillance guidelines suggest relatively frequent need for follow-up colonoscopy but the appropriateness relative to risk (and other polyp types) is unknown.
Target PopulationPatients undergoing colonoscopy.
Intervention or ExposurePresence of serrated colon polyps.
Outcomes/Key FindingsAmong 233,393 individuals undergoing colonoscopy, 445 developed a subsequent CRC. 173,257 had no polyp on first colonoscopy; 11,505 had proximal SPs, 12,080 proximal SPs and synchronous adenomas, 19,410 distal SPs, and 17,141 distal SPs and synchronous adenomas. Among patients with SPs, risk of CRC was not increased until 3 years or more after the first colonoscopy (HR for small proximal SPs 2.6; 95% CI, 1.7–3.9 and HR for large proximal SPs 8.0; 95% CI, 3.6–16.1). The risk was higher if an adenoma was also diagnosed (HR for proximal SPs with synchronous adenomas 4.0; 95% CI, 3.0–5.5; and HR for distal SPs with synchronous adenomas 2.4; 95% CI, 1.7–3.4).
Resulting Action/ChangeThe study provided some of the first community-based evidence for post-colonoscopy risk stratification; this is influencing national and KP guidelines for follow-up colonoscopy surveillance after SP diagnosis (found on 1 of 10 colonoscopies).
Additional RecommendationsDissemination of findings to practitioner and follow-up analysis with compliance of these data into KPNC practices can inform evidence-based follow-up intervals.
Implementation ToolsColonoscopy surveillance guidelines and risk measures.
Implementation and
Follow-up Measures
Serrated polyp diagnosis and guideline-concordant follow-up (implementation); cancer risk after colonoscopy (effectiveness); changes in colonoscopy and cancer care (utilization).
Relative risk of colorectal cancer in patients with a history of serrated polyps vs. no polyp
​Cumulative incidence of colorectal cancer in patients with a history of serrated polyps

doi: 10.1053/j.gastro.2020.04.004. PMID: 32277950 

AuthorsDavid R Vinson, MD Dustin G Mark, MD et. al. and the eSPEED Investigators of the KP CREST Network
ChallengesMany low-risk patients with acute pulmonary embolism (PE) in the emergency department (ED) are eligible for outpatient care, but are hospitalized nonetheless. One impediment to home discharge is the difficulty of identifying which patients can safely have care at home.
Existing EvidenceHome discharge for pulmonary embolism in most medical centers globally is low, ranging from 1% to 8%, despite evidence that patients with acute PE are eligible for safe outpatient management. Specific methods for safely guiding decision-making are limited, such as proven, evidence-based decision-support systems.
Target PopulationADult patients with acute PE presenting to the emergency department.
Intervention or ExposureTen intervention sites received a multidimensional technology and education intervention — including a clinical decision support system at month 9 of a 16-month study period (1/2014 to 4/2015); the remaining 11 sites were controls.
Outcomes/Key FindingsA clinical decision support system significantly increased safe home-discharge rates for patients presenting with pulmonary embolism to the emergency room.  Among 881 eligible PE patients at intervention sites vs. 822 at control sites, adjusted home discharge increased (17.4% to 28.0% pre/post vs. 15.1% to 14.5% pre/post at intervention vs. control, respectively, an absolute increase of 11.3 percentage points (95% CI, 3.0 to 19.5 percentage points; P = 0.007), without increases in relevant 5-day return visits or 30-day major adverse outcomes.
Resulting Action/ChangeThe project provided both the evidence and the specific
electronic tool, accessible from the electronic medical
record, for broader implementation: structured
promotion of computer decision support for physicians
in site-of-care decision making for ED patients with
acute PE safely increased outpatient management.
Additional RecommendationsDissemination across ED and other providers and folloe-up measures of spread can assess and inform uptake
Implementation ToolsClinical decision support system.
Implementation and
Follow-up Measures
Utilization of decision support and appropriate discharge to home from either the ED or a short-term (<24-hour) ED observation unit for patients with PE (implementation); adverse outcomes (e.g. return visits for PE-related symptoms within 5 days, recurrent thromboembolism, hemorrhage, and all-cause mortality within 30 days) (effectiveness); hospital admissions from ED for PE (utilization)

Effect of intervention (from pre- to post-intervention periods) on home discharge of emergency department patients with acute PE.

DOI: 10.7326/M18-1206

AuthorsElizabeth Suh-Burgmann, MD; Tracy Flanagan, MD; Todd Osinski, MD; Mubarika Alavi, MS; Lisa Herrinton, PhD
ChallengesNo established evidence-based, integrated decision systems exist for evaluating ovarian or adnexal masses/ cysts.  If achieved, it would identify high risk women for prompt surgical evaluation and avoid unnecessary surgery and morbidity for women at low risk.
Existing EvidenceAdnexal masses/cysts are common, present in 7–12% of asymptomatic women. The high prevalence of incidentally discovered benign masses on ultrasound, low cancer prevalence, and overlap between benign and malignant ultrasound characteristics explains the lack of benefit of ovarian cancer screening. However, ultrasound detection leads to concerns regarding ovarian cancer, subsequent surgical removal, or serial monitoring with ultrasound. Standardized risk assessment methods have been adopted for other abnormal imaging findings, such as the Breast Imaging Reporting and Data System (BIRADS) for mammography, the Fleischner system for lung nodules. Algorithms have been proposed for adnexal masses, but none have been widely adopted.
Target PopulationAverage-risk women undergoing ultrasonography.
Intervention or ExposureRisk stratification system for adnexal masses based on standardized ultrasound characteristics.
Outcomes/Key FindingsA new evidence-based risk stratification system for ovarian cysts/masses, with follow-up recommendations, was developed, validated, and integrated into radiology reports. Reporting categories 1, 2, 3, and X allowed risk stratification (table) relative to women with normal examinations (category 0). Categories 1, 2, 3, and X were associated with increasing risks of ovarian cancer diagnosis: 0.2% (95% CI 0.05–0.3%) for category 1, 1.3% (95% CI 0.7–1.9%) for category 2, 6.0% (95% CI 3.0–8.9%), for category 3, and 13.0% (95% CI 9.5–16.4%) for category X while Category 0 studies were associated with a risk of 0.1% (95% CI 0.07–0.14%).
Resulting Action/ChangeThis category system provides the first standardized risk stratification system for adnexal masses integrated into routine care through radiology reporting in a community-based setting. This is changing current care and will further inform ongoing data-driven care.
Additional RecommendationsDevelopment, validation, and implementation of similar risk estimating methods for other conditions requiring surveillance can inform evidence-based follow-up and decrease patient and provider uncertainty for care intervals.
Implementation ToolsRadiology reporting templates with evidence-based classification system and a Practice Resource that provides clinical recommendations.
Implementation and
Follow-up Measures
Utilization of risk stratification and appropriate follow-up (implementation); cancer detection (effectiveness); use of surgery and ultrasound (utilization).

Risk of ovarian cancer by ultrasound reporting category by baseline ultrasound findings

doi: 10.1097/AOG.0000000000002939 

AuthorsRobert W Chang, MD; Lue-Yen Tucker, BA; Kara A Rothenberg, MD; Rishad M Faruqi, MD; Hui C Kuang, NP; Alexander C Flint, MD; Andrew L Avins, MD; Mai N Nguyen-Huynh, MD
ChallengesContemporary outcomes and long-term stroke risk for asymptomatic carotid stenosis management in patients who receive primary medical vs. surgical therapy are lacking. Such data would inform whether surgery is likely to be of benefit beyond medical therapy.
Existing EvidenceStroke is a leading cause of death in US. Carotid disease accounts for 12–20% of all strokes (historically) and 7–22% of elderly patients have carotid disease. Carotid stroke-related preventive intervention cost $21B annually. Both medical and surgical care decrease carotid stroke rates, but no current data exist for modern management of asymptomatic carotid stenosis. Randomized trial on medical vs. surgical therapies indicated a benefit for surgery, though these used treatments available in the 1990s.
Target PopulationPatients with severe carotid stenosis and without prior intervention, prior ipsilateral stroke or transient ischemic attack.
Intervention or ExposureMedical vs. surgical treatment (carotid endarterectomy or carotid artery stenting).
Outcomes/Key FindingsAmong patients with severe asymptomatic carotid stenoses, stroke rates for medical treatment were lower than historical estimates and comparable to surgery. 95.3% of patients did not have a stroke on the same side as their stenosis after five years (95%CI 94.3%–96.1%). 1572 (42.1%) patients underwent 1676 carotid interventions (mean months diagnosis to intervention 6.2±12.5). In the cohort (n=4230) prior to any intervention, 129 strokes were attributable to same-side stenosis (annual rate 1.0%; 95% CI 0.7–1.3%). Among 2327 severe but not ‘high-grade’ stenoses without intervention, 385 (16.5%) progressed to high-grade and 89 (3.8%) to occlusion.
Resulting Action/ChangeThe results suggest comparable contemporary likely risk/benefit between medical and surgical therapy; pending modern trials, these data inform clinicians and their patients with stenosis for shared decision-making regarding treatment choice.
Additional RecommendationsThe development of information for decision-making and broader dissemination of these results would inform provider-patient informed decision-making.
Implementation ToolsModern risk estimates for medical treatment for carotid endarterectomy (CEA) or carotid artery stenting (CAS) treatment.
Implementation and
Follow-up Measures
Proportions of patients for med vs. surgery (implementation); stroke risk (effectiveness); use of surgery, stenting and other interventions (utilization).

Kaplan-Meier estimate of 5-year freedom from same-sided carotid-related stroke.

Manuscript pending. 

AuthorsThomas H Urbania, MD; Jennifer R Dusendang, MPH; Lisa J Herrinton, PhD; Stacey Alexeeff, PhD; Douglas A Corley, MD, PhD, MPH; Sora Ely, MD; Ashish Patel, MD; Todd Osinski, MD; Lori C Sakoda, PhD, MPH
ChallengesLung cancer diagnoses require accurate, standardized CT nodule reporting and follow-up methods to optimize timely, appropriate care but none have been validated within KPNC.
Existing EvidenceAlthough lung cancer is usually diagnosed at a late stage, when diagnosed early, 5-year survival is >50%. Standardized reporting and follow-up may reduce time to diagnosis and provide more accurate diagnoses and more rapid stage-specific care for lung cancer. Methods have been proposed (Fleischner guidelines), but they need integration into KPNC workflows, testing for local accuracy, and potential modification to optimize performance.
Target PopulationKP Northern California members undergoing nonscreening chest CT imaging.
Intervention or ExposureStandardized tagging and classification of chest CT pulmonary findings, auto-generated recommendations embedded in CT reports, and coordinated patient followup/referral for patients with findings tagged high risk (suggesting lung cancer) by a multidisciplinary care team.
Outcomes/Key FindingsAmong 2,856 patients (2.9%) diagnoses with lung cancer, 28% had early-stage disease. 40% percent of all patients received the intervention.  The intervention was associated with 9% greater odds of diagnosing any lung cancer (OR 1.09; 95% CI 1.00–1.18); 24% greater odds of early-stage diagnosis (OR 1.24; 95% CI 1.09–1.41); no change in the odds of late-stage diagnosis (OR 1.04; 95% CI 0.95–1.14); and no change in surgical treatment within 120 days.
Resulting Action/ChangeThese findings supported increased use of standardized tagging, classification, and multi-disciplinary care navigation for identifying early stage lung cancer patients. The intervention did not decrease time to diagnosis; this can inform efforts to decrease time to therapy.
Additional RecommendationsEvaluation of steps for time to follow-up and misclassification can further optimize accuracy and expedite next-steps in patient care.  Similar imaging tagging/standardized recommendation can be considered more broadly for other conditions.
Implementation ToolsPlaybook/workflow for care navigation, reporting system integrated into radiology reports.
Implementation and
Follow-up Measures
Proportions of cancers using risk stratification (implementation), early-stage lung cancer diagnosis and time to follow-up following implementation of standardized reporting system (effectiveness); appropriate use of biopsy and surgery (utilization).

​Tags, Descriptions, and Recommendations Used to Code Lung-Specific Findings on Diagnostic Chest CT Imaging

doi: 10.1016/j.chest.2020.05.595

AuthorsFernando S Velayos, MD, MPH; Jennifer R Dusendang, MPH; Julie A Schmittdiel, PhD, MA
ChallengesA growing number of individuals require biologics, prednisone, or oral immunomodulators to suppress a dysregulated immune system. It is unknown whether these therapies modify the risk for severe illness from SARS-CoV-2.
Existing EvidenceEvidence of immunosuppressant treatment and risk of severe outcomes from SARS-CoV-2 is lacking.
Target PopulationAdults who have tested positive for SARS-CoV-2.
Intervention or ExposureUse of biologics, prednisone, or oral immunomodulators.
Outcomes/Key FindingsUsing immunosuppressants prior to a SARS-CoV-2 diagnosis was not associated with a higher (or lower) risk for the composite risk of severe illness, with the exception of prednisone. Out of 39,686 adults who tested positive for SARS-CoV-2, 2.4% (n=958) had a prior prednisone prescription, 0.9% (n=366) animmunomodulator, and 0.3% (n=130) for a biologic (proportions similar to the background population). A total of 10.0% (n=3,977) had at least one outcome of interest (hospitalization, ICU admission or death). Oral prednisone prior to SARS-CoV-2 diagnosis was associated with hospitalization (OR= 1.40, 95% CI 1.15– 1.70), ICU admission (OR 1.96, CI 1.47–2.63), and death (OR 2.01, CI 1.37–2.93). A prescription for biologics or oral immunomodulators did not increase the risk for the composite outcome, although there was an association between oral immunomodulator therapy and mortality (OR 2.39; 95%CI 1.18–4.84).
Resulting Action/ChangeThese findings support the ongoing use of immunosuppressive medications in patients who need them and the ability of such patients to continue routine work, medical care, and other activities with appropriate (average risk) caution.
Additional RecommendationsClinical leaders can consider broader dissemination of this information to relevant providers and patient populations to address concern and patient management
Implementation ToolsN/A
Implementation and
Follow-up Measures
Use of immunosuppressants (implementation); risk of illness from inappropriate discontinuation (effectiveness); hospitalizations or other adverse events from inappropriate discontinuation (utilization)

​Research Letter submitted for publication.

45-day severe outcomes of 39,686 patients for SARS-CoV-2 by prior medication

AuthorsLisa Gilliam, MD, PhD; Julie Schmittdiel, PhD, MA; Rick Dlott, MD; Bharathi Ramachandran, MPH; Wendy Dyer, MS
ChallengesPatients with controlled diabetes may “flip” from good to poor control (A1c>8%), no helpful prediction tools exist. Accurate prediction of which patients are likely to flip could inform workflows (i.e. closer monitoring and earlier interventions), decreasing likelihood of deterioration of glycemic control.
Existing EvidenceAccountable Population Managers (APMs) are empaneled with ~1300 patients with DM, of whom about 68% have “good” glycemic control, as assessed by an A1c <8%. Because the group of patients with higher A1c’s include many patients with compliance issues and social issues, this group is time consuming and receives the bulk of the APM’s focus. Unfortunately, among those in the much larger “controlled” group (A1c<8%), we are not good at predicting which patients will subsequently “flip” out of control (A1c>8%) at follow up. A single medical center analysis determined that among patients with A1c 7.6 to 7.9%, 40% of these patients had an A1c>8% at the subsequent follow up a1c test. Only high performing population managers were focusing on this controlled, but at-risk group, most were focusing on the patients currently out of control.
Target PopulationKP members, ages 18–75, in the DOR Diabetes registry with an A1c <8%.
Intervention or ExposureSignificant predictors of glycemic deterioration.
Outcomes/Key FindingsA prediction tool was created with predictors of glycemic deterioration, including gender, age, race, number of oral DM meds, insulin use, index a1c value, smoking, BMI, AbLAPS, COPS2, EVS, and NDI. The training data validation demonstrated good correlation between observed vs. predicted risk generating a C-statistic of 0.691. Based on this model, if 5,000 members are targeted, the model identified 2,500 who would have a >1% increase in a1c in the next 2 years.
Resulting Action/ChangeThe creation of a prediction tool is informing a pilot intervention study looking at efficacy of incorporating the prediction algorithm into clinical practice.
Additional RecommendationsOnce efficacy has been demonstrated, the prediction algorithm can be incorporated into the PROMPT care pathway to facilitate broader utilization.
Implementation ToolsN/A
Implementation and
Follow-up Measures
Completion of pilot study and number of population care managers subsequently using the prediction tool in clinical practice (implementation): changes in deterioration rates following identification and more intensive management of at-risk patients using the prediction algorithm in a feasibility/intervention study and subsequently in practice (effectiveness). Risk of deterioration resulting in hospitalization, medication use, other complication (utilization).

​Internal report.

Validation data: Demonstrating good correlation between observed vs.  predicted risk,  C-statistic = 0.691

AuthorsMara Greenberg, MD; Monique Hedderson, PhD; Fei Xu, MS
ChallengesPatients frequently have pregnancies post-bariatric surgery within KPNC, but the effectiveness of existing nutritional management efforts to improve pregnancy outcomes is unknown.
Existing EvidencePrior studies women post-bariatric surgery had small numbers inconclusive results regarding impact adverse pregnancy outcomes. Approximately 73% of post-bariatric surgery pregnancies are referred to the Regional Perinatal Service Center (RPSC), which requires multiple laboratory evaluations and adjustment of nutritional supplements. It is unknown whether this service improves pregnancy outcomes.
Target PopulationKPNC enrolled women with pregnancy post-bariatric surgery.
Intervention or ExposureEnrollment in Regional Perinatal Service Center for nutritional management.
Outcomes/Key FindingsAmong women post-bariatric surgery, pregnancies enrolled in the Regional Center were less likely to have a preterm birth, experience hypertensive disorders (including pre-existing HTN, gestational HTN and preeclampsia), or to be admitted to the NICU. There were no differences for cesarean deliveries and gestational or pre-existing diabetes. Among all women post-bariatric surgery, >20% had HTN, >40% had impaired glucose tolerance or diabetes, and many (39%) were delivered by cesarean.
Resulting Action/ChangeThe results support the effective use of the Regional Perinatal Service Center for nutritional management and monitoring of post-bariatric surgery patients. These findings support improving rates of referral and uptake and efforts to identify which program components are most associated with better outcomes.
Additional RecommendationsCOnsideration of additional measures for identifying and referring appropriate patients for nutritional evaluation.
Implementation ToolsNone; tools could be developed around increasing awareness and referral.
Implementation and
Follow-up Measures
RPSC referrals among women post-bariatric surgery (implementation); ongoing perinatal outcomes following referral (effectiveness); complications (utilization).

​Internal report.

Risk for perinatal outcomes among women not referred to RPSC


DARE Supported Projects 2019-2020

Delivery Science Grants Program
Adult Hospital MedicineSomalee Banerjee,  Alyce AdamsImproving Outcomes and Care Experience Among Dual Eligible Members: the Role of Health System Factors
AFMJeff East, Mark Moeller, Tracy LieuDeveloping New Strategies for Managing High Volume Patient-Physician Electronic Communication
AFMJoan Lo, Kendal Hamann,  Mehreen KhanAssessment of Fracture Prevention Quality Measures
AFMTR LevinOutreach with Fit Testing increases Detection of Polyps/CRC Among Younger African Americans (45-50 Years)
AFMJonathan Volk,  Michael SilverbergLeveraging EHR Data To increase Uptake of HIC Preexposure Prophylaxis
CardiologyAmir Axelrod,  Andrew AmbrosyManagement Optimization Via Telemonitoring and Resource Utilization and Outcomes in Heart Failure (Monitor-Hf)
CardiologyEd McNulty,  Alan GoPersonalizing Risk of Transcatheter Aortic Valve Replacement: Quality of Life, Complications, Mortality, and Utilization (TAVR - Predict)
CardiologyJamal Rana,  Alan GoIdentifying Variation and Barriers to Use of Non-Invasive Cardiac Imaging Tests for Suspected Coronary Heart Disease
DermatologySangeeta Marwaha,  Lisa HerrintonDermoscope Use Improves Cancer Detection While Decreasing Biopsy and in Person Visits
Emergency MedicineDana Sax,  Mary ReedDevelopment of A Machine-Learning Tool to Risk Stratify Emergency Department Patients with Acute Heart Failure
Emergency MedicineDustin Mark,  Mary ReedElectronic Decision Support Safely Reduces Objective Cardiac Testing Among Emergency Department Patients with Chest Pain At Low Risk of Major Adverse Cardiac Events
GastroenterologyVarun Saxena,  Julie SchmittdielOptimization of Hepatocellular Carcinoma Surveillance Protocols
Infectious DiseasesJulia Marcus, Michael SilverbergShort-Course Treatment (8 Weeks) as Effective as 12 Weeks Treatment for Black Patients with Hepatitis C Virus (HCV) infection
Mental HealthKathryn Erickson-Ridout,  Connie WeisnerCollaborative Care for Depression and Anxiety Requires Active Outreach, Accurate Diagnosis, and Regular Symptom Tracking
NephrologySijie Zheng,  Alan GoUpstream Management of Patients with Chronic Kidney Disease to Delay and Avoid Renal and Major Cardiovascular Events
NeurologyAlex Flint,  Jeff KlingmanOutcomes of Door-To-Needle Times in Stroke Patients
Ob/GynYvonee Crites,  Anne RegensteinSimilar Neonatal Developmental and Delivery Outcomes, Fewer NICU Admissions When Gestational Diabetes Treated with Glyburide (Vs. insulin)
Ob/GynTracy Flanagan,  Lyndsay AvalosThe Impact of Patch “Prenatal Care and Maternal & Child Health Outcomes” On Health Outcomes and Health Care Utilization.
OncologyTatjana Kolevska, Yan Li, Ai KuboMobile Health Mindfulness in Cancer Palliative Care
OncologyRaymond Liu,  Gabriel EscobarAutomating Risk Stratification for Hospital-Acquired Thromboembolism Guides Provider Decision Making and Improves Patient Outcomes
OncologyJed Katzel,  Stephen Van Den EedenElectronic Collection of Patient-Reported Outcomes in Head and Neck Cancer Patients to Improve Survival and Minimize Hospitalization
OncologyDavid Baer, Tracy LieuPrognostic information System (Prism): Refinement and Testing in Actual Practice
OncologyPiyush Srivastava,  Stephen Van Den EedenPatient Reported Outcomes in Pancreatic Cancer
OncologyAndrea Harzstark,  Lisa HerrintonRegionalization of Testicular Cancer Diagnosis and Treatment Planning Effective and increased Satisfaction Among Oncologists
OncologyBethan Powell,  Larry KushiStreamlining Genetic Counseling increases Genetic Testing Among Women with Ovarian Cancer
OncologyCharles Meltzer,  Lori SakodaConsolidated Multidisciplinary Care Improves Survival for Head and Neck Cancer
OphthalmologyDariusz Tarasewicz,  Oleg SofryginDevelopment and Implementation of an Evidence-Based Risk Calculator for Diabetic Retinopathy Screening and Population Management
PediatricsLisa Chyi,  Michael KuzniewiczEat, Sleep, Console (ESC) Assessment Tool to Escape Postnatal Opioid Exposure in infants with Neonatal Opioid withdrawal Syndrome
PediatricsMustafa Bseikri,  Elizabeth FelicianoPediatric Obstructive Sleep Apnea: Predictive Modeling to Streamline Care
PediatricsPaul Espinas,  Stacy SterlingScreening for Aces in Pediatric Clinics Are Feasible and Acceptable
PediatricsMeghan Davignon,  Lisa CroenSetting the Stage to Measure the Long-Term Benefits of the More Than Words Program on KPNC’S Parents, Children, and Providers
Population ManagementGabriel Escobar,  Laura MyersTools for Outpatient and Population Management of Sars-Cov-2 infections (Tops2)
Surgery OncologyVeronica Shim,  Laurie HabelDevelopment of A Clinical Pathway for Selective Oncotype Testing in Early Breast Cancer
Surgery OrthopedicAdrian Hinman,  Andy AvinsA Noninferiority Trial of the Adductor Canal Catheter in Total Knee Replacement
Surgery Plastic SurgeryAmanda Graff-Baker,  Marilyn KwanLYMPHA – Collaboration in the Operating Room Results in Improved Outcomes for Breast Cancer Patients
UrologyJoseph Presti,  Andrew AvinsEasy-To-Implement intervention Reduced inappropriate Prostate Cancer Screening in Men 70 and Over
Women's HealthMara Greenberg,  Assiamira FerraraPandemic Associated Obstetric Care Delivery and COVID infection in Pregnancy: Impact on Outcomes
Rapid Analytics Unit
Anesthesia, Surgery, CardiologyEdward Yap,  Lisa HerrintonAdhering to Cardiovascular Risk Reduction Guidelines Decreases Myocardial Injury After Noncardiac Surgery
CardiologyRichard Birnbaum,  Andy AvinsSystemic Identification and Management of Familial Hypercholesterolemia Optimizes Patient Recognition and Treatment
DermatologyPatrick McCleskey,  Lisa HerrintonChilblains Is Not the Canary in the Covid19 Coal Mine
Emergency MedicineDale Cotton,  Mary ReedCovid-19 in the Ed Encounter: Characteristics and Predicting Outcomes
Emergency MedicineMamata Kene,  Mary ReedOpioid Safety Education Associated with Decreased Opioid Prescribing by Emergency Physicians
GastroenterologyKrisna Chai & Joanna Ready,  Andy AvinsAn Organized Hepatitis B Surveillance Program increases Patient Identification and Optimizes Management
GastroenterologyFernando Velayos,  Julie SchmittdielCovid-19 Complications Are Not More Common Among Immunosuppressed Populations in KPNC
GastroenterologyCraig Munroe,  Doug CorleyTelemedicine (Compared to in-Person Gastroenterology Visits) Had High Patient Satisfaction and Comparable Physician Decision-Making
Gastroenterology/HepatologyVarun Saxena,  Julie SchmittdielBenefits of Hepatitis C Virus Cure
Hematology/OncologyAshok Pai & Gwendolyn Ho,  Julie SchmittdielChronic Anticoagulant and Antiplatelet Use Is Not Associated with Decreased Disease Severity in Sars-Cov-2 infection
Interventional RadiologyMaud Morshedi,  Lisa HerrintonImprovements in Coordinated Hepatocellular Screening, Care Coordination, and Treatment Associated with 50% Decrease in Mortality
Mental HealthKathryn Ridout,  Esti IturraldeMental Health Service Demand in the Face of Covid-19
NeurologyAlex Flint,  Andy AvinsKPNC Stroke Express Program Markedly Shortens Time-To-Thrombolysis for Patients with Ischemic Stroke
Ob/GynEve Zaritsky,  Andy AvinsCovid-19 and Shelter-In-Place Results in Delayed Presentations for Emergency Gynecologic Care
OncologyAndrea Harzstark,  Julie SchmittdielConcomitant Cancer Treatment and SARS-Cov-Infection increased Risk of Noninvasive Ventilation Compared to Those without Cancer
OncologyAndrea Harzstark,  Liyan LiuRapid Ascertainment of New Bladder Cancer Diagnoses informs Regional Multi-Disciplinary Case Management
OncologyTilak Sundaresan,  Liyan LiuRapid Case Ascertainment Using NLP Is Effective and Feasible in Pancreatic Cancer Management
OncologyLisa Law,  Lisa HerrintonRegionalizing Subspecialized Acute Myeloid Leukemia Care increased induction Therapy and Bone Marrow Transplantation and Decreased Mortality.
Oncology and PulmonologyNareg Roubinian,  Julie SchmittdielThe incidence of Venous Thromboembolism Is Similar in Outpatients with and without Sars-Cov-2 infection
Pediatric surgeryAlbert Chong,  Lisa HerrintonLow Recurrence and Complication Rates with Minimally invasive Repair of Pediatric inguinal Hernia
Pediatrics/Adolescent MedicineJosephine Lau,  Julie SchmittdielA Systematic Evaluation of Eating Disorders in Children and Adolescents Identifies Patient Populations Under Care and Potential Needs
Radiology, pulmonary, thoracic surgerytom Urbania,  Lisa HerrintonStructured Reporting of Lung Nodules Detected on Chest Ct Was Associated with Greater Chance of Detecting Early Stage Lung Cancer
SurgeryVinnie Liu,  Andy AvinsEnhanced Recovery After Surgery (Eras) intervention Was Associated with Reduced Opioid Prescriptions After Surgery
SurgeryGillian Kuehner,  Mary ReedImplementation of Telemedicine within Surgical Specialties Before and After Covid-19: Adjusting to A Changing Landscape
SurgeryBrooke Vuong,  Julie SchmittdielOutpatient Mastectomy: Factors influencing Patient Selection and Predictors of Return to Care
SurgeryRobert Li,  Lisa HerrintonRegionalization of Sub-Specialized Gastric Cancer Care increased Use of Laparoscopic Approaches, Recommended Staging, and increased Survival
SurgeryReza Rahbari,  Lisa HerrintonRegionalizing Sub-Specialized Adrenal Surgery Decreases Operative Time, Hospital Stay, and Major Complications
SurgerySimon Ashiku,  Andy AvinsStreamlined Surgical and Perioperative-Care Benefit Esophageal Cancer Patients Undergoing Esophagectomy
Surgery Head and NeckKevin Wang & Janet Lai,  Andy AvinsIdentifying Optimal Strategies for Improving Human Papilloma Virus Immunization Rates in Young-Adult KPNC Members
Surgery OrthopedicDavid Ding,  Andy AvinsRisk of Total Hip Replacement After Hip Arthroscopy increases with Age
Surgery ThoracicJeffrey Velotta,  Lisa HerrintonOutcomes Following interventions To Sustain Body Weight in Esophageal Cancer Patients Starting Preoperative Therapy
Physician Researcher Program
CardiologyMatt Solomon,  Alan GoNatural Language Processing Tool Accurately Identifies Aortic Stenosis and Severity To inform New Clinical Tracking and Surveillance Programs
Emergency MedicineDana Sax,  Mary ReedBuild and Pilot Testing of a Machine Learning Acute Heart Failure Risk Prediction Tool
Emergency MedicineDavid Vinson,  Mary ReedNew Clinical Decision Support intervention increased Safe Outpatient Management of Emergency Department Patients with Pulmonary Embolism
GastroenterologyDan Li,  Lisa HerrintonNew Risk Estimates for Colon Cancer Among Persons with Serrated Colon Polyps inform Guidelines for Timing of Repeat Colonoscopy
GastroenterologyT.R. Levin,  Jeffrey LeePredicting Serious Colonic Growths to Risk Stratify People Coming Due for Surveillance Colonoscopy
Gynecologic OncologyBetty Suh-Bergmann,  Lisa HerrintonIntegration of Standardized Ovarian Cyst Risk Stratification System into Radiology Reports Estimates Risk and informs Follow-Up
Infectious DiseasesJacek Skarbinski,  Larry KushiSars-Cov-2 Serological Antibody Testing for Disease Surveillance and Clinical Use
Mental HealthKathryn Erickson-Ridout,  Constance WeisnerHigh Ability to Deliver New and Ongoing Mental Health Care Visits During the Covid-19 Pandemic
Mental HealthMatthew Hirschtritt,  Stacy SterlingService Use Following Initial Mental Health Evaluation and Referral Differs By Patient Characteristics
NeurologyMai Nguyen-Huynh,  Alan GoThrombectomy for Stroke Patients with Large Vessel Occlusion and Delayed Presentation:  Community-Based Results Are Comparable to Trials
RadiologyVignesh Arasu,  Laurel HabelImproved Selection of BRCA-Negative High-Risk Women for Breast MRI Screening Through Validation of Ibis Risk Model Variants
Surgery VascularBobby Chang,  Mai Nguyen-HuynhPatients with Severe Asymptomatic Carotid Stenosis Are at Low Risk of Stroke with Contemporary Medical Management
UrologyJoe Presti,  Stacey AlexeefDecreased Prostate Cancer Screening Following the 2012 USPTF Statement Resulted in A Significant increase in Metastatic Cancer
Targeted Analysis Program
EndocrinologyLisa Gilliam,  Julie SchmittdielAn Electronic Algorithm Predicts Deteriorating Glycemic Control in Patients with Diabetes
Hospital MedicineVincent LiuInpatient Outcomes Associated with Regional Implementation of a Benzodiazepine-Sparing Alcohol withdrawal Orderset
Mental HealthMatthew Hirschtritt,  Stacy SterlingTelepsychiatry Provides Rapid Mental Health Evaluation and Referral for Treatment Among Adults with Mild-to-Moderate Symptoms
Ob/GynMara Greenberg,  Monique HeddersonPerinatal Nutritional Management Associated with Improved Pregnancy Outcomes Among Women Post-Bariatric Surgery
UrologyMark Gasparini,  Stephen VanDenEedenSignificance of Borderline High and High Serum Calcium Levels in High Risk Kidney Stone