Question

Multidimensional Severity Index (MSI) validity in knee/hip osteoarthritis (OA): Evaluate the construct and criterion validity, and responsiveness, of a composite MSI combining pain and function PROs, physical-performance tests, and imaging among adults ≥70 years with knee/hip OA.

Answer (protocol-grade plan + expected benchmarks)

1) What to measure (domains & instruments)

Domain	Instrument	Metric	Interpretation anchor
Pain	WOMAC Pain (0–20)	Higher = worse	MCID ≈ 2
Function	WOMAC Function (0–68) or KOOS/HOOS ADL	Higher = worse	MCID ≈ 6–9
Performance	4-m gait speed; 30-s chair-stand; 6MWD	Speed m/s; reps; meters	MCID: 0.05 m/s; 2 reps; 20–30 m
Strength	Handgrip dynamometry	kg (best of 3)	Frailty cutpoints (sex-specific)
Structure	Radiograph KL (0–4) and/or MRI WORMS	Ordinal / summed	Severe = KL≥3 or WORMS high tertile
Global	Pain NRS (0–10); Patient Global (0–10)	Higher = worse	MCID ≈ 1–2

2) MSI construction

MSI_raw = Σ_i (w_i * z_i) ;   MSI_0_100 = 100 * (MSI_raw - min_train) / (max_train - min_train)

3) Study design

• Population: n = 320 adults ≥70 with symptomatic knee/hip OA (community + clinic), excluding recent arthroplasty (<12 months).
• Visits: Baseline, 6 months, 12 months.
• Imaging: Fixed-flexion radiographs at baseline; MRI subset (n≈120) for WORMS/cartilage thickness.
• Data splits: 60% train (CFA/weights), 40% test (validation).

4) Statistical analysis plan (SAP)

• Construct validity: CFA fit indices (CFI≥0.95, TLI≥0.95, RMSEA≤0.06) for hypothesized factor structure; measurement invariance by sex (ΔCFI < 0.01).
• Criterion validity: AUC of MSI for severe structural disease (KL≥3 or MRI high tertile) and for high symptom burden (WOMAC pain ≥75th percentile). Target AUC ≥0.80.
• Convergent validity: Spearman r with KOOS/HOOS, gait speed, 6MWD (expect |r| 0.5–0.7).
• Known-groups validity: Mean MSI differences ≥0.5 SD across KL strata, assistive-device use, obesity, and multimorbidity tertiles.
• Responsiveness: Standardized Response Mean (SRM) for those reporting meaningful change at 6–12 months (anchor-based NRS/Global). Target SRM ≥0.6.
• Minimal clinically important difference (MCID): Anchor-based (Global change ≥“somewhat better/worse”) + distribution-based (0.5 SD). Anticipate ~6–8 MSI points.
• Calibration: Regress outcomes (pain flare, mobility loss) on MSI; check calibration slope ~1.0 and Brier score improvement vs PRO-only model.
• External validation (optional): Refit weights in an independent cohort; compare C-statistics and calibration.

5) Power & sample size (sketch)

• CFA: ~20 indicators → N≥10 per parameter ⇒ N≈300 adequate for stable loadings.
• AUC test: To detect AUC 0.80 vs 0.70 (null) with 35% severe cases, α=0.05, power=0.90 ⇒ N≈300.
• Responsiveness: Detect SRM 0.5 with 80% power ⇒ n≈130 with change data (allowing 20% attrition).

6) Handling practicalities

• Missingness: Multiple imputation by chained equations (MICE); sensitivity complete-case.
• Ceiling/floor effects: Examine proportion at bounds in PROs; consider item-response theory (IRT) refinement if >15% at extremes.
• Measurement error: Test–retest in a stable subset within 2 weeks; compute ICC(2,1) for each metric and for MSI composite.
• Equity: Assess differential item functioning by sex, race/ethnicity, language; adjust weights if bias detected.

7) Benchmarks to claim validity (pre-specified)

Property	Threshold to meet
CFA fit	CFI/TLI ≥ 0.95; RMSEA ≤ 0.06
Criterion (KL≥3)	AUC ≥ 0.80 (95% CI lower bound ≥ 0.75)
Convergent	|r| with gait speed/6MWD ≥ 0.5
Known-groups	ΔMSI ≥ 0.5 SD across severity strata
Responsiveness	SRM ≥ 0.6; MCID ~ 6–8 points
Reliability	ICC ≥ 0.85 for MSI

8) Cut-points (provisional; to validate)

• Mild: MSI < 33 — conservative care; activity progression.
• Moderate: 33 ≤ MSI < 67 — structured PT, weight loss, bracing, analgesic optimization.
• Severe: MSI ≥ 67 — consider injections, surgical evaluation (context-specific).

9) Pseudocode (clinic-ready computation)

# Inputs (all oriented so higher = worse):
# womac_pain, womac_function, gait_speed, chair_stand_reps, sixmwd, grip_kg, kl_grade
# sex ('F'/'M'), age

# Standardize (examples; use cohort means/SDs or external norms):
z_pain      = (womac_pain     - mean_pain)  / sd_pain
z_function  = (womac_function - mean_func)  / sd_func
z_gait      = -(gait_speed    - mean_gait)  / sd_gait      # negative: slower = worse
z_chair     = -(chair_stand_reps - mean_chair) / sd_chair  # fewer reps = worse
z_6mwd      = -(sixmwd        - mean_6mwd)  / sd_6mwd
z_grip      = -(grip_kg_adj(age, sex) - mean_grip) / sd_grip
z_kl        = (kl_grade       - mean_kl)    / sd_kl

# Weights (from CFA on training set; example placeholders):
w = {'pain':0.28, 'function':0.26, 'gait':0.14, 'chair':0.10, 'sixmwd':0.10, 'grip':0.06, 'kl':0.06}

MSI_raw = (w['pain']*z_pain + w['function']*z_function + w['gait']*z_gait +
           w['chair']*z_chair + w['sixmwd']*z_6mwd + w['grip']*z_grip + w['kl']*z_kl)

MSI_0_100 = 100 * (MSI_raw - min_train) / (max_train - min_train)

10) Anticipated findings & interpretation

• MSI shows stronger linkage to mobility outcomes than any single component (e.g., MSI r≈0.65 with 6MWD vs KL r≈0.30).
• Patients with MSI ≥67 exhibit ~2–3× odds of assistive-device use and ~0.15–0.20 m/s slower gait speed vs MSI <33.
• MSI is responsive to PT/bracing/weight loss with SRM ≈0.6–0.8; change ≥7 points aligns with patient-reported meaningful improvement.

Abbreviations — WOMAC: Western Ontario & McMaster Universities Osteoarthritis Index; KOOS/HOOS: Knee/Hip injury and Osteoarthritis Outcome Score; KL: Kellgren–Lawrence; WORMS: Whole-Organ MRI Score; 6MWD: 6‑minute walk distance; ICC: intraclass correlation coefficient; MCID: minimal clinically important difference; CFA: confirmatory factor analysis.