Table 2 Predictive models for knee arthroplasty management
From: Artificial intelligence in knee arthroplasty: current concept of the available clinical applications
Authors | Patients | Year | Type | Assessment | Factors | Conclusion/Algorithms |
|---|---|---|---|---|---|---|
Judge et al. [63] | 1991 | 2012 | PM | Satisfaction, OKS | Age, sex, BMI, Primary diagnosis, ASA score, Index of Multiple Deprivation, OKS, EQ. 5D | Strongest determinants of outcome: pain/function (less severe preop disease obtain best outcomes); diagnosis in relation to pain outcome (RA > OA); deprivation (poorer areas = worse outcomes); anxiety/depression (=worse pain) |
Lungu et al. [64] | 141 | 2014 | PM | WOMAC | 5 preoperative WOMAC questions: difficulty of taking off socks, getting on/off toilet, performing light domestic duties and rising from bed as well as degree of morning stiffness after the first wakening | Predictive rule, based on 5 preop WOMAC questions |
Dowsey et al. [65] | 615 | 2016 | PM | WOMAC (using OMERACT-OARSI responder criteria) | BMI, radiographic degree of OA (K.L. scale), WOMAC, SF-12, sex, age, ASA score, Charlson comorbidity, smoking status, etiology, SEIFA, rurality, contralateral TKA, constraint, patella, computer navigation, LOS, discharge destination, complication/adverse event | Better probability of clinical response with lower BMI, lower SF-12 MCS disability level, lower K.L., higher (worse) preoperative WOMAC |
Pua et al. [66] | 1096 | 2016 | PM | Walking limitations (time before severe difficulty) | Age, BMI, hypertension, fall history, walking aids, contralateral knee pain, reconstruction specialist, walking ability, fast gait speed and knee pain, sex | Lower risk of walking< 15 min with younger age, lower BMI, no HTA, less fall history, less preop walking aids, no contralateral knee pain, adult reconstruction specialist surgeon, better preop walking ability, faster 1-month gait speed, lower 1-month knee |
Van Onsem et al. [67] | 113 | 2016 | PM | KSS satisfaction score | Questions selections based on KOOS, OKS, PCS, EQ-5D, KSS, age and sex | Algorithm: Satisfaction at M3 = 26.10 + 2.3*sex+ 0.13*age + 1.58*Q3–1.40*Q4–1.08*Q5–0.75*Q6–1*Q7–1.12*Q8–0.88*Q9–1.10*Q10 |
To et al. [68] | 737 | 2017 | PM | Transfusion | Preop variables | Valid |
Garriga et al. [69] | 221 | 2018 | PM | Non-satisfaction | Demographic preop pain, function | Country dependent |
Shim et al. [70] | 721 | 2018 | PM | OKS (score less than 26 classified as poor). | OKS, chronic widespread pain, high expectations of knee pain after recovery, lack of active coping | Better (higher) postop OKS with better preop OKS, less chronic widespread pain, lower expectations of knee pain after recovery, better active coping strategies |
Kunze et al. [71] | 484 | 2018 | PM | Satisfaction after TKA | Â | 97.5% sensitivity, 95.7% VPN |
Navarro et al. [72] | 141,446 | 2018 | PM | LOS, Cost | Age, race, sex, comorbidity scores | Excellent validity |
Sanchez et al. [73] | 1649 | 2018 | PM | OKS | Age, sex, marital status, Index of Multiple Deprivation, BMI, anxiety/depression, OKS, ASA score, etiology, previous knee arthroscopy, flexion contracture, ACL status | Better (higher) postoperative OKS with better (higher) preoperative OKS, no anxiety/depression (E.Q. 5D-3L Q5), fit and healthy ASA grade, no other conditions affective mobility, no previous arthroscopy, lower IMD 2004 score, lower BMI, presence of fixed flexion deformity, damaged/absent ACL, females aged < 80 or males aged > 60. |
Van Onsem et al. [74] | 57 | 2018 | PM | KOOS, KSS, OKS | Preop ROM, quadriceps and hamstring force, sit-to-stand test, 6-min walk test | High postop PROMs showed higher postop functional outcomes. A model to predict the cluster allocation contained sex, ROM improvement and 6MWT improvement (sensitivity 91.1%, specificity 75%) |
Calkins et al. [75] | 145 | 2019 | PM | Satisfaction (KSS satisfaction subscale, score less than 20 classified as unsatisfied). | KOOS, OKS, PCS, EQ-5D, new KSS, age, sex, diagnosis, previous surgery on knee, BMI, radiographic degree of OA, coronal alignment | Higher KSS score with male sex, older age, higher pain (EQ-5D-5L Q4), less knee joint stiffness (KOOS Sy1), less grinding/clicking noise (KOOS Sy4), knee felt ‘normal’ (KSS: Symptoms Q3), less awareness of knee problem (KOOS Q1), less anxiety/depression (EQ-5D-5L Q5), pain not on mind (PCS Q9), less worried about serious problem occurring (PCS Q13) |
Zabawa et al. [76] | 203 | 2019 | PM | Patient dissatisfaction following TKA | KOOS, OKS, PCS, EQ-5D, new KSS, age, sex, diagnosis, previous surgery on knee, BMI, radiographic degree of OA, coronal alignment, payment method, education, income, diabetes mellitus, HTA, hyperlipidemia, insurance provider, comorbidities | External validation of a new prediction model; Less pain prior to surgery (Q3), lesser anxiety/depression prior to surgery (Q9) and better ability to control pain symptoms (Q9); Also found lower BMI and past medical history of hypertension through additional analysis |
Twiggs et al. [77] | 330 | 2019 | PM | Knee pain | Age, sex, KOOS items, back pain, occurrence of hip pain, occurrence of falls in past year | Predictive model with a web application KOOS: activities of daily living, pain and symptom subscores, pain when pivoting on knee, pain when standing, difficulty bending the knee fully, frequency of back pain, severity of back pain, occurrence of hip pain, occurrence of falls in preceding year, age, sex |
Tolk et al. [78] | 7071 | 2019 | PM | Residual symptoms (pain at rest and activity, sit-to-stand movement, stair negotiation, walking, performance of activities of daily living, kneeling and squatting) | Age, sex, ASA score, BMI, smoking, previous knee surgery, Charnley score, KOOS-PS, OKS, EuroQoL 5D-3L, NRS | Predictive model for residual symptoms |
Kunze et al. [71] | 484 | 2019 | PM | Patient-reported health state, KSS, ROM, satisfaction = > Knee survey score | BMI, drug allergies, osteophytes, soft tissue thickness, flexion contracture, diabetes, opioid use, comorbidities, previous knee surgery, surgical indication, smoking | Knee survey score on 110 pts; 4 risks of experiencing postoperative dissatisfaction: Score 96.5–110 = low risk Score 75–96.4 = mild risk Score 60–74.9 = medium risk Score < 60 = high risk |
Huber et al. [79] | 34,110 | 2019 | PM | EQ-VAS (MID), OKS (MID). | All 81 variables in NHS dataset (April 2015 – March 2016); including sociodemographic information such as living status, age groups, sex, disease affliction, EQ-5D-3L, EQ-VAS, OKS scores | Preop OKS score, often limping (OKS Q6), preop EQ-VAS, revision surgery, no disability, not interfering with work (OKS Q9), no previous knee surgery, no diabetes, extreme difficulty doing shopping (OKS Q11), age 50–59 |
Gronbeck et al. [80] | 61,284 | 2019 | PM | Inpatient admission after TKA | Demographic, comorbidity, perioperative variables | Reliable identification of candidates for inpatient admission |
Bini et al. [5] | 22 | 2019 | PM | PROMs | 35 variables (PROMS, demographic …) | Valid |
Jo et al. [81] | 1686 | 2019 | PM | Transfusion after TKA | 43 preop variables | Validated – good performance |
Pua et al. [82] | 4026 | 2019 | PM | Walking limitation | Socio-demographic data outcomes | Better (higher) postop score with lower preop knee pain levels, lower preop depression levels, lower preop knee flexion range and Chinese race |
Itou et al. [83] | 50 | 2020 | PM | satisfaction | KSS FJS12 | Low utility |
Li et al. [84] | 1826 | 2020 | PM | LOS | ASA, diabetes, comorbidities, anesthesia, operation time | LOS prediction model for TKA |
Kunze et al. [85] | 430 | 2020 | PM | Dissatisfaction after TKA | Demographics, medical history, flexion contracture, knee flexion, outcome scores | Good discriminative capacity |
Turcotte et al. [86] | 2266 | 2020 | PM | Ambulatory surgery for TKA | Demographics, comorbidities | Good validity |
Harris et al. [87] | 587 | 2020 | PM | PROMs Improvement | PROMs health data | Improve decision support and decision making |
Goltz et al. [88] | 10,155 | 2020 | PM | Risk prediction of TKA for discharge location | 45 variables (sociodemographic data, postop labs, comorbidity) | Excellent accuracy to predict discharge location |
Farooq et al. [89] | 897 | 2020 | PM | Satisfaction | 15 variables (sociodemographic – surgery) | Valid - multifactorial |
El Galaly et al. [90] | 25,104 | 2020 | PM | Revision TKA | Patient’s characteristics and surgical information | Inable to predict revision |
Anis et al. [91] | 5958–2391 | 2020 | PM | LOS, 90 days readmission, PROMs | Age, sex, BMI, race, educational level, smoking, comorbidities, KOOS items, 12PCS, 12MCS | Scalable predictive tools Can accurately estimate the likelihood of improved pain, function, and quality of life 1 year after TKA as well as LOS and 90 day readmission. |
Ko et al. [92] | 5757 | 2020 | PM | Acute kidney injury | 18 variables | 6 major variables – valid |
Andersen et al. [93] | 538 | 2021 | PM | Revision TKA | Age, EQ-5D, comorbidities | Partially validated |
Han et al. [94] | 1298 | 2021 | PM | LOS | 36 variables | Valid |