Evaluating Patients' Perception of the Risk of Acute Care Visits During Systemic Therapy for Cancer

Implementation of machine learning in the clinic: challenges and lessons in prospective deployment from the System for High Intensity EvaLuation During Radiation Therapy (SHIELD-RT) randomized controlled study

Background

Artificial intelligence (AI) and machine learning (ML) have resulted in significant enthusiasm for their promise in healthcare. Despite this, prospective randomized controlled trials and successful clinical implementation remain limited. One clinical application of ML is mitigation of the increased risk for acute care during outpatient cancer therapy. We previously reported the results of the System for High Intensity EvaLuation During Radiation Therapy (SHIELD-RT) study (NCT04277650), which was a prospective, randomized quality improvement study demonstrating that ML based on electronic health record (EHR) data can direct supplemental clinical evaluations and reduce the rate of acute care during cancer radiotherapy with and without chemotherapy. The objective of this study is to report the workflow and operational challenges encountered during ML implementation on the SHIELD-RT study.

Results

Data extraction and manual review steps in the workflow represented significant time commitments for implementation of clinical ML on a prospective, randomized study. Barriers include limited data availability through the standard clinical workflow and commercial products, the need to aggregate data from multiple sources, and logistical challenges from altering the standard clinical workflow to deliver adaptive care.

Conclusions

The SHIELD-RT study was an early randomized controlled study which enabled assessment of barriers to clinical ML implementation, specifically those which leverage the EHR. These challenges build on a growing body of literature and may provide lessons for future healthcare ML adoption.

Trial registration: NCT04277650. Registered 20 February 2020. Retrospectively registered quality improvement study.

A Systematic Review of Discrete Choice Experiments in Oncology Treatments

BACKGROUND

As the number and type of cancer treatments available rises and patients live with the consequences of their disease and treatments for longer, understanding preferences for cancer care can help inform decisions about optimal treatment development, access, and care provision. Discrete choice experiments (DCEs) are commonly used as a tool to elicit stakeholder preferences; however, their implementation in oncology may be challenging if burdensome trade-offs (e.g. length of life versus quality of life) are involved and/or target populations are small.

OBJECTIVES

The aim of this review was to characterise DCEs relating to cancer treatments that were conducted between 1990 and March 2020.

DATA SOURCES

EMBASE, MEDLINE, and the Cochrane Database of Systematic Reviews were searched for relevant studies.

STUDY ELIGIBILITY CRITERIA

Studies were included if they implemented a DCE and reported outcomes of interest (i.e. quantitative outputs on participants' preferences for cancer treatments), but were excluded if they were not focused on pharmacological, radiological or surgical treatments (e.g. cancer screening or counselling services), were non-English, or were a secondary analysis of an included study.

ANALYSIS METHODS

Analysis followed a narrative synthesis, and quantitative data were summarised using descriptive statistics, including rankings of attribute importance.

RESULT

Seventy-nine studies were included in the review. The number of published DCEs relating to oncology grew over the review period. Studies were conducted in a range of indications (n = 19), most commonly breast (n =10, 13%) and prostate (n = 9, 11%) cancer, and most studies elicited preferences of patients (n = 59, 75%). Across reviewed studies, survival attributes were commonly ranked as most important, with overall survival (OS) and progression-free survival (PFS) ranked most important in 58% and 28% of models, respectively. Preferences varied between stakeholder groups, with patients and clinicians placing greater importance on survival outcomes, and general population samples valuing health-related quality of life (HRQoL). Despite the emphasis of guidelines on the importance of using qualitative research to inform attribute selection and DCE designs, reporting on instrument development was mixed.

LIMITATIONS

No formal assessment of bias was conducted, with the scope of the paper instead providing a descriptive characterisation. The review only included DCEs relating to cancer treatments, and no insight is provided into other health technologies such as cancer screening. Only DCEs were included.

CONCLUSIONS AND IMPLICATIONS

Although there was variation in attribute importance between responder types, survival attributes were consistently ranked as important by both patients and clinicians. Observed challenges included the risk of attribute dominance for survival outcomes, limited sample sizes in some indications, and a lack of reporting about instrument development processes.

PROTOCOL REGISTRATION

PROSPERO 2020 CRD42020184232.

Physician treatment of metastatic triple-negative breast cancer in the immuno-oncology era: a discrete choice experiment

Aim: Guidelines list atezolizumab with nab-paclitaxel (ANP) as the preferred first-line (1L) therapy for metastatic triple-negative breast cancer (mTNBC) with PD-L1 expression ≥1%, but which clinical attributes impact ANP prescribing? Materials & methods: Medical oncologists participated in a discrete choice experiment (DCE) with four hypothetical mTNBC clinical scenarios to assess influences of: PD-L1 expression, menopausal status, prior adjuvant therapy and bulky liver metastases. Results: A total of 47% chose ANP in 1L irrespective of menopausal status, prior adjuvant therapy or tumor bulk. PD-L1 expression was the only attribute with a significant impact on ANP preference, with 69% choosing ANP for those with ≥1% expression versus only 26% for those with <1% (p < 0.00001). Conclusion: ANP choice for 1L mTNBC deviated from guidelines.