Our predefined search strategy identified 3361 titles; none were identified through searching reference lists of published systematic reviews, and review of included study reference lists identified 1 more. The screening process and reasons for exclusion at each step are presented in Figure 1.
We identified 6 studies describing 6 prediction models that that met our inclusion criteria, which included 7075 patients.[15-20] Four studies were from North America, one from Australia, and one from Europe. Four included general medicine patients and two included specific subpopulations that could be cared for on a general medical ward (stroke patients and patients with advanced cirrhosis) (Table 1).
Table 1 - Studies
Study
|
Participants and Setting
|
Design
|
Sample size derivation cohort
|
Validation type and sample size
|
Mean age of derivation cohort
|
Most common admitting diagnoses
|
Prognostic data collection timeframe*
|
Outcome
|
Fairchild et al, 1998
|
General medicine patients in an urban teaching hospital
|
Single center prospective cohort
|
387
|
Internal, n=327
|
55
|
Chest pain
|
<24hrs
|
Use of post-discharge medical services
|
|
|
|
|
|
|
Heart failure and shock
|
|
|
|
|
|
|
|
|
Bronchitis and asthma with complications
|
|
|
Simonet et al, 2008
|
General medicine patients in a teaching hospital
|
Single center prospective cohort
|
349
|
Internal, n=161
|
65
|
Not provided
|
<24hrs
|
Discharge to a post-acute care facility
|
Mehta et al, 2011
|
General medicine patients in a teaching hospital
|
Multicenter prospective cohort
|
885
|
External, n=753
|
78
|
Chronic lung disease
|
<24hrs
|
Need for ADL support
|
|
|
|
|
|
|
Peripheral vascular disease
|
|
|
|
|
|
|
|
|
Congestive heart failure
|
|
|
Stineman et al, 2013
|
Veterans hospitals stroke patients
|
Multicenter retrospective cohort
|
3909
|
Internal, n=2606
|
69
|
Stroke
|
Throughout hospitalization
|
Home discharge
|
Tapper et al, 2015
|
Liver transplant unit in an academic hospital
|
Single center retrospective cohort
|
490
|
Internal, n=244
|
57
|
Encephalopathy
|
<24hrs
|
Discharge to rehabilitation
|
|
|
|
|
|
|
Ascites
|
|
|
|
|
|
|
|
|
Gastrointestinal bleeding
|
|
|
Basic et al, 2015
|
Patients admitted under geriatricians with geriatric issues
|
Single Center Prospective Cohort
|
1055
|
Internal n=1070
|
83
|
Dementia
|
During the Hospitalization
|
New Admission to a Nursing Home
|
|
|
|
|
|
|
Delirium
|
|
|
|
|
|
|
|
|
Deconditioning
|
|
|
*Relative to time of admission
|
The mean age of patients in the derivation cohorts ranged from 55 to 83 years.
Risk of Bias
The risk of bias for included studies was moderate to high for most studies (Table 2).
Table 2 – Risk of Bias Assessment
Study
|
Study participation
|
Study attrition
|
Prognostic factor measurement
|
Outcome measurement
|
Study confounding
|
Statistical analysis and reporting
|
Overall
|
Tapper et al.
|
M
|
L
|
M
|
M
|
M
|
L
|
M
|
Stineman et al.
|
H
|
L
|
M
|
M
|
M
|
L
|
H
|
Fairchild et al.
|
H
|
H
|
M
|
M
|
M
|
M
|
H
|
Simonet et al.
|
M
|
M
|
L
|
L
|
L
|
L
|
M
|
Basic et al.
|
L
|
L
|
M
|
L
|
M
|
M
|
M
|
Mehtah et al.
|
H
|
L
|
L
|
L
|
M
|
L
|
M
|
|
|
|
Legend: L – Low, M – Moderate, H – High, For each category low risk of bias is defined as: Study participation: The study sample represents the population of interest on key characteristics, Study Attrition: Loss to follow-up is not associated with key characteristics, Prognostics Factor Measurement: Factors are adequately measured, Outcomes Measurement: Outcome of interest is adequately measured, Study Confounding: Important potential confounders are appropriately accounted for, Statistical Analysis and Confounding: The statistical analysis is appropriate for the design of the study, Overall: Majority of criteria met. Little or no risk of bias.
Inadequate description of the source population’s inclusion and exclusion criteria was found in 5 of 6 included studies.
Prediction Models
Four studies used univariable analysis to screen for predictors to be used for modeling while the other two studies used clinical reasoning to select predictor variables. To build the final model, five prediction models used various automated selection algorithms while one included all the variables that were hypothesized to be predictive (Table 3).
that were hypothesized to be predictive (Table 3).
Table 3 – Models
Study
|
Outcome
|
Number (%) with outcome
|
Variable screening method
|
Variable selection method
|
Variables included in final model
|
Discrimination (derivation)
|
Discrimination (validation)
|
Calibration
|
Other measures of model performance
|
Fairchild et al, 1998
|
Use of post-discharge medical services
|
134 (35)
|
Univariable analysis
|
Selection algorithm
|
Age>65, SF-36 Physical administered on admission <50, SF-36 Social <15
|
AUC= 0.75
|
AUC= 0.70
|
N/A
|
N/A
|
Simonet et al, 2008
|
Discharge to a post-acute care facility
|
104 (30)
|
Univariable analysis
|
Selection algorithm
|
Number of medically active conditions on admission, Inability of patient's partner to provide home help, Number of IADL and ADL disabilities, Age, Admitted via inter-hospital transfer
|
AUC= 0.82
|
AUC= 0.77
|
N/A
|
8-point cutoff: Sensitivity 0.87/Specificity 0.63, 16-point cutoff: Sensitivity 0.42/Specificity 0.91
|
Mehta et al, 2011
|
Need for IADL support
|
242(27)
|
Univariable analysis
|
Best subset algorithm
|
Age <80, 80-89, >90, Dependent in >3 IADLS prior to admission, Number of ADL dependencies at the time of admission 1, 2-3, 4-5, Metastatic cancer or stroke, Albumin <3g/dL, Mobility before admission
|
0.78
|
0.78
|
H-L P= 0.40
|
N/A
|
Stineman et al, 2013
|
Home discharge
|
3348(85)
|
Univariable analysis
|
Selection algorithm
|
Married, Location before admission extended care, hospital, home, Functional recovery grade at discharge, discharge cognitive grade, History of liver disease, no feeding tube required, No intensive care unit admission, Mechanical ventilation during admission
|
0.82
|
0.8
|
H-L P= 0.23
|
|
Tapper et al, 2015
|
Discharge to rehabilitation
|
199(15)
|
Clinical reasoning
|
None (all variables included)
|
Gender, Age, Ethnicity, Charlson Co-morbidity Index, Admission ADL score, Admission Morse fall risk score and Braden score, Admission MELD, Admission serum sodium, Infection, Cirrhotic decompensation, Hepato-cellular carcinoma, Admitting hepatologist
|
AUC= 0.85
|
AUC= 0.77
|
N/A
|
N/A
|
Basic et al, 2015
|
New Admission to a Nursing Home
|
62 (5.9%)
|
Clinical reasoning, literature review
|
Backwards selection Using LR test
|
Clinical frailty scale 7-point, Dementia, delirium, Age, Urinary retention, Deconditioning, Seizure disorder
|
N/A
|
N/A
|
N/A
|
N/A
|
Four models used predictors that were available with in the first 24 hours of hospital admission while two models used variables collected throughout the hospital stay. All prediction models had a binary outcome as the dependent variable; however, each study defined the outcome differently (Table 3). Four models defined the outcome as a place (home, care facility, or rehabilitation hospital) while two models defined the outcome as the need for support services after discharge. Model discrimination was generally good (range derivation C-statistics 0.75 – 0.85) and similar but slightly lower in the validation cohorts (range validation C-statistics 0.70 – 0.80) with one study not reporting discrimination statistics. Two models tested calibration with the Hosmer-Lemeshow goodness of fit test and found no evidence of poor fit. Other calibration metrics, such as calibration plots, were not reported.
Predictors
All variables associated with the need for post discharge supportive services in predictive models are presented in Table 3. Many variables were associated with the outcome in univariable analysis in a single study but not in multivariate analyses. Variables present in 2 or more studies were assessed for the strength of evidence using the GRADE tool (Table 4).
Table 4 – GRADE
|
|
|
|
Univariable
|
Multivariate
|
GRADE Factors
|
Prognostic Factor
|
# of studies
|
No. of participants
|
Phase of Investigation
|
+
|
0
|
-
|
+
|
0
|
-
|
Risk of Bias
|
Inconsistency
|
Indirectness
|
Imprecision
|
Moderate/Large effect size
|
Dose Effect
|
Overall Quality
|
Age
|
5
|
6585
|
3
|
4
|
0
|
0
|
4
|
0
|
0
|
Moderate
|
✘
|
✘
|
✘
|
✔
|
✔
|
++++
|
Impaired Physical Function*
|
4
|
5530
|
3
|
3
|
0
|
0
|
4
|
0
|
0
|
Moderate
|
✘
|
✘
|
✘
|
✔
|
✔
|
++++
|
ADL disabilities
|
3
|
1724
|
3
|
3
|
0
|
0
|
3
|
0
|
0
|
Moderate
|
✘
|
✘
|
✘
|
✔
|
✔
|
++++
|
Frailty **
|
2
|
1545
|
3
|
1
|
0
|
0
|
2
|
0
|
0
|
Moderate
|
✘
|
✘
|
✘
|
✔
|
✔
|
++++
|
Stroke
|
2
|
4794
|
3
|
3
|
0
|
0
|
1
|
0
|
0
|
Moderate
|
✘
|
✔
|
✘
|
✔
|
✔
|
+++
|
Supportive environment prior to admission
|
2
|
4258
|
3
|
2
|
0
|
0
|
2
|
0
|
0
|
Moderate
|
✘
|
✔
|
✘
|
✔
|
✘
|
+++
|
IADL disabilities
|
2
|
1234
|
3
|
2
|
0
|
0
|
2
|
0
|
0
|
Moderate
|
✘
|
✘
|
✘
|
✔
|
✘
|
+++
|
Cognitive impairment***
|
4
|
6198
|
3
|
3
|
0
|
0
|
1
|
0
|
0
|
Moderate
|
✘
|
✔
|
✘
|
✘
|
✔
|
++
|
Marital status (Married) vs Other
|
3
|
5181
|
3
|
0
|
0
|
3
|
0
|
0
|
2
|
Moderate
|
✘
|
✘
|
✘
|
✘
|
✔
|
++
|
Heart Failure
|
2
|
4794
|
2
|
2
|
0
|
0
|
0
|
0
|
0
|
Moderate
|
✘
|
✘
|
✘
|
✘
|
✘
|
++
|
Metastatic Cancer
|
2
|
4794
|
2
|
2
|
0
|
0
|
1
|
0
|
0
|
Moderate
|
✘
|
✘
|
✘
|
✔
|
✘
|
++
|
Lives with informal care giver
|
2
|
736
|
3
|
0
|
0
|
2
|
0
|
0
|
1
|
Moderate
|
✘
|
✘
|
✘
|
✘
|
✘
|
++
|
Sex (Female)
|
4
|
5633
|
3
|
1
|
3
|
0
|
1
|
1
|
0
|
Moderate
|
✔
|
✘
|
✘
|
✘
|
✘
|
+
|
Heart Valve Disease
|
2
|
4296
|
3
|
2
|
0
|
0
|
0
|
0
|
0
|
High
|
✘
|
✘
|
✘
|
✘
|
✘
|
+
|
Comorbidities Increasing number
|
4
|
2111
|
3
|
3
|
1
|
0
|
0
|
1
|
0
|
Moderate
|
✔
|
✘
|
✔
|
✘
|
✘
|
+
|
* Physical function as measured by mobility, gait, ability to transfer or physical function scores
|
|
** Measured using Rockwood frailty scale, Braden risk and Morse fall risk
|
|
*** Cognitive impairment defined as low mini-mental state score, severe cognitive impairment, dementia or cognitive stage
|
|
Strength of Evidence + very low; ++ low; +++ moderate; ++++ high. The overall quality of evidence for a factor is rated as high if it comes from explanatory research aimed at understanding causal pathways (phase of investigation 3) or moderate if it comes from prediction research aimed at identifying associations (phase of investigation 2). The quality of the evidence is then downgraded if there are study limitations, inconsistency, indirectness, imprecision or publication bias. The quality is upgraded if there is moderate or large effect size ore and exposure-response gradient identified.[12]
|
|
There is high quality evidence that age, impaired physical function, ADL disabilities, and frailty increase the probability of needing supportive services after discharge. There is moderate evidence that a diagnosis of stroke, IADL disabilities and receiving supportive services prior to hospital admission increase the probability of needing supportive services after discharge. Interestingly a greater number of comorbidities was a significant predictor in univariable analyses in 3 studies but not in any multivariate analyses. Indicators of absent support at home (marital status and absence of an informal care giver) had moderate-weak evidence of predicting discharge with supportive services. Lastly several specific diagnoses (heart failure, metastatic cancer and heart valve disease) had weak to very week evidence for predicting the need for supportive services.
Causal Pathway
We constructed a causal pathway to explain the relationships between the predictor variables and the outcome (Figure 2). Activities of daily living andcare giver at home directly affected the need for supportive services after discharge whereas age and comorbidities act by their effects on physical function and cognitive function to affect ADL disabilities. Stroke directly causes ADL and IADL disabilities while the causative impact of other specific diagnoses is less certain.