DISTANCE EDUCATION – PROGRAM 2: A NEW START IN MEASUREMENT FOR HEALTH TECHNOLOGY ASSESSMENT MODULES 1 – 5
A NEW START IN MEASUREMENT FOR HEALTH TECHNOLOGY ASSESSMENT
For fifty years, HTA has confused numbers with measures. Scientific practice requires adherence
to representational measurement theory, from Stevens’ linkage of arithmetic to scale type to the
representation and uniqueness theorems of Krantz, Luce, Suppes, and Tversky. Rasch provided
the operational bridge for latent traits, transforming ordered responses into an invariant logit
ruler. HTA instead fixated on QALYs and multiattribute valuations, guaranteeing persistent
failure. The remedy is non-negotiable: only linear ratio scales for manifest claims and Rasch
logit ratio scales for latent trait possession are admissible.
MODULE 1: THE DENIAL OF FALSIFICATION IN HEALTH TECHNOLOGY ASSESSMENT
Falsification requires stable units, replicability, and clear disconfirmation conditions. RMT
supplies these prerequisites. HTA denies falsification because its core quantities are not
measures: ordinal utilities are treated as interval and multiplied by time into QALYs. Reference-
case models embed these non-measures, producing outputs that cannot be empirically refuted.
Without validated units, HTA abandons normal science.
MODULE 2: THE RASCH MODEL – LATENT TRAITS AND ITEM SELECTION
Latent traits are measurable only if they yield invariant scales. Rasch specifies a single trait, tests
items against it, and maps responses via a logistic function of person location minus item
difficulty. Items must fit the model to ensure constant relative differences and invariance. Item selection is critical: well-targeted items span the trait and cluster near 50% endorsement. Misfit
signals failure and functions as falsification.
MODULE 3: THE RASCH MODEL – THE UNIQUE RASCH LOGIT RATIO SCALE
Rasch measurement operationalizes conjecture and refutation through fit statistics, local
independence, DIF, and invariance testing. Surviving these tests earns the status of “measure.”
The result is the unique Rasch logit ratio scale: additive logits defining trait possession and
multiplicative odds with a true zero, where differences translate to invariant odds ratios.
MODULE 4: THE RASCH MODEL – POSSESSION AND FALSIFICATION
The quantity of interest is possession of a single latent trait, measured in logits. When Rasch
assumptions hold, equal logit differences have equal meaning across persons and items.
Estimation yields person and item locations with standard errors governed by targeting. Group
comparisons and change analyses are conducted on the logit scale, with odds-ratio
interpretations. For subjective responses, Rasch is the only framework consistent with RMT.
MODULE 5: THE RASCH MODEL – THE EXISTENTIAL CRISIS FOR DISEASE SPECIFIC INSTRUMENTS
Disease-specific instruments used in HTA rarely meet Rasch standards. Summed scores lack
unidimensionality, invariance, and additivity, so totals are not measures. Apparent change often
reflects instrument behavior rather than trait change. The Rasch model uniquely enforces
representational measurement for latent traits; instruments that fail its tests produce numerical
storytelling, not evidence.