A COMPREHENSIVE NATURAL HISTORY MODEL OF PRESSURE ULCER TO ESTIMATE THE CLINICAL IMPACT OF PREVENTION AND TREATMENT OPTIONS

Purpose:   Pressure ulcers (PU) are common in a variety of care settings and associated with adverse health outcomes, litigation potential and high societal aggregate cost. Existing PU models are limited in design and scope for evaluating the cost-effectiveness of available prevention and treatment options. We described the development and calibration of a natural history model of PU with linked population-based data from the Ontario Resident Assessment Instrument - Minimum Dataset (RAI-MDS).

Methods: The model was developed with content advice from an expert panel.

Model population: Typical cohort of long-term care (LTC) residents in Ontario Canada (mean 80 years of age, 70% females, and average 3 life-years post-institution).  

Model structure: A Markov model was used to simulate the natural history disease of PUs among LTC residents across 52 mutually exclusive health states, stratified by: (A) underlying risk for PU (high/low risk), (B) PU stage (I-IV), (C) PU prognosis (healable, chronic or healing wound), (D) related complications (local/systemic infection) and (E) care setting, LTC or hospitals.

Input Data: RAI-MDS data pertaining to demographics, death in LTC and hospitalization, PU incidence and prevalence, healing, chronicity and infection was from a cohort of 18,325 residents in Ontario (91 of 631 LTCs, average follow-up 12 months). RAI-MDS is a well validated assessment instrument for a resident’s health status, care needs, and preferences. RAI-MDS records were linked to hospital discharge abstracts to obtain in-hospital mortality data. LTC residents was stratified into high and low risk for developing PU stage 2-4 according to validated risk-adjustment scale based upon 17 RAI-MDS items.

Calibration: A first-order calibration was performed: projected stage-specific prevalence estimates from the average cohort at one year after admission were matched to the observed prevalence in the age group 80-84 from the RAI-MDS cross-sectional data. Other calibration included average life-years post-admission, hospitalization and in-LTC mortality.

Results: The model projected lifetime probability of PU (45%), chronic PU (19%), PU-related local infection (10%) and systemic infection (1.78%), and PU-related death (0.7%) for an approximate cohort of 90,166 residents.

Conclusions:   The model provides a platform for policy analysis of PU prevention and treatment options, especially those target high risk populations. Refinement of the model is on-going with plans to adapt the model to other care settings, including acute care and community care.