How Good Is Machine Learning in Predicting All-Cause 30-Day Hospital Readmission? Evidence From Administrative Data.

OBJECTIVES: Hospital readmission is a main cost driver for healthcare systems, but existing works often had poor or moderate predictive results. Although the available information differs in different studies, improving prediction is different from the search for important explanatory variables. With large sample size and abundant information, this study explores state-of-the-art machine-learning algorithms and shows their performance in prediction. METHODS: Using administrative data on 1 631 611 hospital stays from Quebec between 1995 and 2012, we predict the probability of 30-day readmission at hospital admission and discharge. We compare the performance between traditional logistic regression, logistic regression with penalization, and more recent machine-learning algorithms such as random forest, deep learning, and extreme gradient boosting. RESULTS: After a 10-fold cross-validation on the training set (80% of the data), machine learning produced very good results on a separate hold-out test set (20% of the data). The importance of explanatory variables is not the same for different algorithms. The area under receiver operating characteristic curve (AUC) reached above 0.79 at hospital admission and above 0.88 at hospital discharge. Diagnostic codes, which include many different categories, are among the most predictive variables. Logistic regression with penalization also produced good results, but a standard logistic regression failed without penalization. The good results are confirmed by calibration curves. CONCLUSION: Although the identification of those at highest risk of readmission is just 1 step to preventing hospital readmissions, 30-day readmission is highly predictable with machine learning.

End-Of-Life Medical Spending In Last Twelve Months Of Life Is Lower Than Previously Reported.

Although end-of-life medical spending is often viewed as a major component of aggregate medical expenditure, accurate measures of this type of medical spending are scarce. We used detailed health care data for the period 2009-11 from Denmark, England, France, Germany, Japan, the Netherlands, Taiwan, the United States, and the Canadian province of Quebec to measure the composition and magnitude of medical spending in the three years before death. In all nine countries, medical spending at the end of life was high relative to spending at other ages. Spending during the last twelve months of life made up a modest share of aggregate spending, ranging from 8.5 percent in the United States to 11.2 percent in Taiwan, but spending in the last three calendar years of life reached 24.5 percent in Taiwan. This suggests that high aggregate medical spending is due not to last-ditch efforts to save lives but to spending on people with chronic conditions, which are associated with shorter life expectancies.

Physician payment mechanisms, hospital length of stay and risk of readmission: evidence from a natural experiment.

We provide an analysis of the effect of physician payment methods on their hospital patients' length of stay and risk of readmission. To do so, we exploit a major reform implemented in Quebec (Canada) in 1999. The Quebec Government introduced an optional mixed compensation (MC) scheme for specialist physicians working in hospital. This scheme combines a fixed per diem with a reduced fee for services provided, as an alternative to the traditional fee-for-service system. We develop a model of a physician's decision to choose the MC scheme. We show that a physician who adopts this system will have incentives to increase his time per clinical service provided. We demonstrate that as long as this effect does not improve his patients' health by more than a critical level, they will stay more days in hospital over the period. At the empirical level, we estimate a model of transition between spells in and out of hospital analog to a difference-in-differences approach. We find that the hospital length of stay of patients treated in departments that opted for the MC system increased on average by 4.2% (0.28 days). However, the risk of readmission to the same department with the same diagnosis does not appear to be overall affected by the reform.

Health and income: a robust comparison of Canada and the US.

This paper uses sequential stochastic dominance procedures to compare the joint distribution of health and income across space and time. It is the first application of which we are aware of methods to compare multidimensional distributions of income and health using procedures that are robust to aggregation techniques. The paper's approach is more general than comparisons of health gradients and does not require the estimation of health equivalent incomes. We illustrate the approach by contrasting Canada and the US using comparable data. Canada dominates the US over the bottom part of the bi-dimensional distribution of health and income, though not generally over the uni-dimensional distributions of health or income. The paper also finds that welfare for both Canadians and Americans has not unambiguously improved during the last decade over the joint distribution of income and health, in spite of the fact that the uni-dimensional distributions of income have clearly improved during that period.