Algorithm 1.

Train, tune, and evaluate a model to predict if a patient should be placed on CRRT.

1:	${\mathbb{D}}_{\text{train+val}}\Leftarrow$ choose from {UCLA: CRRT, UCLA: Control, Cedars: CRRT }
2:	${\mathbb{D}}_{\text{test}}\Leftarrow$ choose from {UCLA: CRRT, UCLA: Control, Cedars: CRRT }
3:	if ${\mathbb{D}}_{\text{train+val}}={\mathbb{D}}_{\text{test}}$ then
4:	${\mathbb{D}}_{\text{train}},{\mathbb{D}}_{\text{val}},{\mathbb{D}}_{\text{test}}\Leftarrow {\mathbb{D}}_{\text{train+val}}\text{*}\left[0.6,0.2,0.2\right]$	▷ Split dataset into 60/20/20% respectively.
5:	else
6:	${\mathbb{D}}_{\text{train}},{\mathbb{D}}_{\text{val}}\Leftarrow {\mathbb{D}}_{\text{train}\text{+val}}\text{*}\left[0.8,0.2\right]$	▷ Split dataset into 80/20% respectively.
7:	end if
8:	$w\Leftarrow \left\{1,2,3,4,5,6,7,10,14\right\}$ days	▷ look-back window size
9:	$m\Leftarrow \left\{\text{lgbm},\text{xgb},\text{rf}\right\}$	▷ model type
10:	$i\Leftarrow \left\{\text{mean}/\text{mode},k\text{nearest}\text{neighbors}\right\}$	▷ imputation method
11:	$f\Leftarrow \left\{k\text{best},\text{correlation}\text{threshold}\right\}$	▷ feature selection method
12:	$ℍ\Leftarrow \left\{w,m,i,f\right\}$	▷ Set hyperparameter grid
13:	function Tuning $\left(ℍ,{\mathbb{D}}_{\text{train}},{\mathbb{D}}_{\text{val}}\right)$
14:	metrics ⇐ {}
15:	for $w_j,m_j,i_j,f_j\in ℍ$ do
16:	Train $m_j\left(f_j\left(i_j\left({\mathbb{D}}_{\text{train}}\right)\right)\right)$
17:	$\text{metrics}\stackrel{+}{\Leftarrow }\text{Evaluate}{m}_j\left(f_j\left(i_j\left({\mathbb{D}}_{\text{val}}\right)\right)\right)$
18:	end for
19:	$m^{\text{*}}\Leftarrow m_j$ such that corresponding metric is max/min (best)
20:	return m*
21:	end function
22:	$m^{\text{*}}\Leftarrow \text{Tuning}\left(ℍ,{\mathbb{D}}_{\text{train}},{\mathbb{D}}_{\text{val}}\right)$
23:	metrics ⇐ Evaluate $m^{\text{*}}\left({\mathbb{D}}_{\text{test}}\right)$
24:	subpopulations ⇐ (heart, liver, infection) × (male, female, race)
25:	for subpopulation in subpopulations do
26:	metrics by subpopulation ⇐ Evaluate $m^{\text{*}}(\mathbb{D}\text{test}{}^{\text{*}}\text{subpopulation})$
27:	end for