A Two-Part Framework for Estimating Individualized Treatment Rules From Semicontinuous Outcomes

Diagnostics for regression models with semicontinuous outcomes

Semicontinuous outcomes commonly arise in a wide variety of fields, such as insurance claims, healthcare expenditures, rainfall amounts, and alcohol consumption. Regression models, including Tobit, Tweedie, and two-part models, are widely employed to understand the relationship between semicontinuous outcomes and covariates. Given the potential detrimental consequences of model misspecification, after fitting a regression model, it is of prime importance to check the adequacy of the model. However, due to the point mass at zero, standard diagnostic tools for regression models (eg, deviance and Pearson residuals) are not informative for semicontinuous data. To bridge this gap, we propose a new type of residuals for semicontinuous outcomes that is applicable to general regression models. Under the correctly specified model, the proposed residuals converge to being uniformly distributed, and when the model is misspecified, they significantly depart from this pattern. In addition to in-sample validation, the proposed methodology can also be employed to evaluate predictive distributions. We demonstrate the effectiveness of the proposed tool using health expenditure data from the US Medical Expenditure Panel Survey.

Risk from delayed or missed care and non-COVID-19 outcomes for older patients with chronic conditions during the pandemic

Background

During the COVID‐19 pandemic, patients with chronic illnesses avoided regular medical care, raising concerns about long‐term complications. Our objective was to identify a population of older patients with chronic conditions who may be at risk from delayed or missed care (DMC) and follow their non‐COVID outcomes during the pandemic.

Methods

We used a retrospective matched cohort design using Medicare claims and electronic health records at a large health system with community and academic clinics. Participants included 14,406 patients over 65 years old with two or more chronic conditions who had 1 year of baseline data and up to 9 months of postpandemic follow‐up from March 1, 2019 to December 31, 2020; and 14,406 matched comparison patients from 1 year prior. Risk from DMC was defined by 13 indicators, including chronic conditions, frailty, disability affecting the use of telehealth, recent unplanned acute care, prior missed outpatient care, and social determinants of health. Outcomes included mortality, inpatient events, Medicare payments, and primary care and specialty care visits (in‐person and telehealth).

Results

A total of 25% of patients had four or more indicators for risk from DMC. Per 1000 patients annually, those with four or more indicators had increased mortality of 19 patients (95% confidence interval, 4 to 32) and decreased utilization, including unplanned events (−496 events, −611 to −381) and primary care visits (−1578 visits, −1793 to −1401).

Conclusions

Older patients who had four or more indicators for risk from DMC had higher mortality and steep declines in inpatient and outpatient utilization during the pandemic.