Correction to “Predicting preventable hospital readmissions with causal machine learning”

In Marafino et al., ¹ the caption and the third covariate “DCO 4” in Table 1 should have been amended. The correct table is as follows:

TABLE 1.

The covariates used in this study and their distributions

Covariate	Description	Mean (median; IQR)
AGE	Patient age in years, recorded at admission	65.0 (67; 54‐78)
MALE	Male gender indicator	47.5% (–)
DCO_4	Code status at discharge (4 categories)	84.3% (–)
HOSP PRIOR7 CT	Count of hospitalizations in the last 7 d prior to the current admission	0.05 (0; 0‐0)
HOSP PRIOR8 30 CT	Count of hospitalizations in the last 8 to 30 d prior to the current admission	0.11 (0; 0‐0)
LOS 30	Length of stay, in days (with stays above 30 d truncated at 30 d)	4.6 (3; 2‐5)
MEDICARE	Indicator for Medicare Advantage status	58.8% (–)
DISCHDISP	Discharge disposition (home, skilled nursing, home health)	72.7% (–)
LAPS2	Laboratory‐based acuity of illness score, recorded at admission	55.7 (49; 16‐84)
LAPS2DC	Laboratory‐based acuity of illness score, recorded at discharge	44.5 (40; 24‐60)
COPS2	Comorbidity and chronic condition score, updated monthly	44.7 (24; 10‐66)
HCUPSGDC	Diagnosis supergroup classification	–
W (or Wi )	Treatment: Transitions Program intervention	5.2% (–)
Y (or Yi )	Outcome: Nonelective readmission within 30 d postdischarge	12.4% (–)

A more comprehensive listing of characteristics, stratified respective to the implementation of the Transitions Program, as well as definitions of the HCUPSGDC variables, can be found in Appendix S1. For binary variables, only means are presented; for DCO_4 and DISCHDISP, the quantities presented correspond to the proportion of discharges who were full code, and those discharged to home, respectively.

Abbreviations: COPS2, COmorbidity Point Score, version 2; HCUPSGDC, Health Care and Utilization Project Super Group at discharge; IQR, interquartile range; LAPS2, Laboratory‐based Acute Physiology Score, version 2.