Table 3. Assessment of prediction models in identified studies using the CHARMS checklist (CHecklist for critical appraisal and data extraction for systematic reviews of prediction modelling studies.
Study | Source | Participants | Outcomes | Predictors | Development | Sample Size (EVR) | Missing Data | Performance | Evaluation |
---|---|---|---|---|---|---|---|---|---|
Nagamatsu [37] | Cohort (R) |
Individuals undergoing oesophagectomy 97% M | Cardiorespiratory | VO2 max AT VC FEV1, V50/V25 ratio, DLCO, PaO2 |
Multiple regression, unclear model building strategy | 91 (2.4) |
N/A | ROC No AUC |
Not done |
Hartley [29] | Cohort (R/P) |
84%M, 23% >80 undergoing open and EVAR AAA repair | Mortality(30D) | Repair type Sex, Age, DM, IHD, HTN, Antiplatelets, Statins, Anaemia, Urea, Creatinine AAA size & location, AT, Peak VO2, VE/VCO2 |
Univariable prior to Multi, backwards using AIC | 451 (3.5) |
Excl. if >15% missing Median imputation for continuous classed as absent if unclear for categorical | Not done | Not done |
Junejo [33] | Cohort (P) |
Undergoing hepatic resection >65 or with complex resection, average age 71, 2:1M:F | Cardiorespiratory | Age ASA BMI RCRI AT VE/CO2, |
Unclear. P threshold for progression to Multiple Regression analysis of <0.1 mentioned | 92 (19.5) |
N/A | Not done | Not done |
Ting [34] | Cohort (P) |
Kidney transplant, average age 42, 91% living kidney recipients, 60% Male | Unplanned ICU | BMI Desensitisation Age, Sex Months of Dialysis, HTN, Dyslipidemia, DM Carotid-Femoral PWV (m/s) Augmentation index AT |
Multivariable after univariable P of <0.10, FW and BW step. BMI and dialysis time incl. as known confounders of postop cardiovascular state. AT used as binary value (11.3) | 70 (4.7) |
N/A | AUC HL test |
Cross-validation (no details) |
Snowden [35] | Cohort (P) |
Consecutive for open hepatic resection, excluded those unable to obtain an AT | Mortality | AT Peak VO2 Peak Work Peak HR, Age, BMI, Sex IHD, CVA, DM, CCF, HTN, Asthma, COPD RCRI |
Univariable then included those with P<0.1, forward step regression with assessment of co-linearity. Only present final model containing AT alone. | 389 (18) |
N/A | AUC | Not done |
Barakat [27] | Cohort (P) |
Single centre study, Male preponderance(89%), Unclear percentage of eligible population sent for CPET | Cardiac Respiratory | Age, Sex Method of repair AT VO2 peak VE/VCO2 |
Included set variables, no evidence of progression from univariable to multivariable. No process of model simplification given. | 130 (R = 4) (C = 3) |
Not stated | Not done | Not done |
Lam [36] | Cohort (R) |
Single centre, aimed for all consecutive patients undergoing oesophagectomy for cancer | Cardiorespiratory | AT Peak VO2 Type of Surgery Gender Comorbiditty Smoking Stage and histology Adjuvant therapy |
Included CPET variables and all with univariable p of <0.10 | 80 (20) |
Not stated | Not done | Not done |