The MPSMS is a nationwide surveillance system designed to identify rates of specific adverse events within the hospitalized fee-for-service Medicare population. An adverse event is an unintended harm, injury, or loss that is more likely associated with an individual's interaction with the healthcare delivery system than with diseases the individual may have. The MPSMS determines the national rates for Medicare beneficiaries in the following adverse event categories: central venous catheter (CVC)-associated blood stream infections; CVC-associated mechanical adverse events; CVC-associated blood stream infection (BSI) adverse drug events; HAPU- and ventilator-associated pneumonia; hip and knee replacements; and postoperative rates of venous thrombolic event, cardiac events, and pneumonia.
The MPSMS PU study uses secondary analyses of the Hospital Payment Monitoring Program (HPMP) sample. The HPMP medical record sample is an existing database, selected randomly each month from the Medicare National Claims History File by CMS from a pool of approximately 1 million fee-for-service Medicare beneficiary hospital discharges. For this study, the records of hospital discharges between January 1, 2006, and December 31, 2007, totaling 51,842 Medicare fee-for-service inpatient discharges across the 50 states, Washington, DC, Puerto Rico, and the U.S. Virgin Islands, were selected for use.
Trained medical record abstractors collected documentation describing individuals with HAPUs that developed during their index hospitalizations. The medical record abstractors were trained on model charts. Interrater reliability (between principal investigator and medical abstractors) was established at 90% before the medical abstractors were allowed to abstract the medical records.
Charts of individuals with PUs present on admission (prevalence) were reviewed to determine whether new ulcers (incidence) developed during the hospital stay. The National PU Advisory Panels Stage I (2001) and Stage II to IV (1989) definitions of PUs were used to distinguish a PU from another potential skin injury.[6,7] The comorbidities for PUs were defined as congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), cerebrovascular disease (CVD), diabetes mellitus, usage of corticosteroids, obesity, and smoking. These were selected because of their association with PUs in the literature. The PU diagnosis was determined based on physician and nurse documentation in the medical record during admission (prevalence) or at any time during hospitalization. Newly developed PUs (incidence) were also monitored during hospitalization.
The abstractors determined whether an individual developed a HAPU based on documentation of a HAPU in the medical record or on the description of ulceration. The locations of PUs, newly acquired and those documented on admission, were also recorded. To further distinguish HAPUs from PUs already present on admission, new PUs found in the same body region as PUs present on admission were not counted as HAPUs. Any individual who developed at least one new PU during the hospitalization was included in the analyses. Participant demographic information (e.g., age, sex, race), International Classification of Diseases, Ninth Revision (ICD-9) diagnoses (707), and other characteristics (e.g., obesity, CHF, COPD as determined by ICD coding) were obtained from the Medicare Enrollment Database.
The outcomes of interest were in-hospital mortality rates, readmission or mortality during the 30 days after discharge, and hospital length of stay. The CMS National Claim History database was the source of in-hospital mortality and readmission information. The Medicare Enrollment Database was the source of 30-day mortality information.
Descriptive and bivariate analyses were conducted to identify participants' baseline demographics and medical diagnoses and to compare the observed differences in characteristics and outcomes (mortality, readmission, and length of stay) of participants who developed HAPUs and those who did not. The chi-square test was used to compare dichotomous and categorical variables, and the t-test was used to compare continuous variables. The hierarchal generalized linear modeling (HGLM) approach[8–10] was applied to assess the association between participant characteristics and development of HAPUs by modeling the log-odds of HAPUs as a function of participant demographic and clinical variables. This approach was used to determine whether there was a relationship between the outcomes and development of HAPUs after adjustment for participant characteristics.
Two modeling steps were constructed for each outcome. The first step was fitted to a model without adjusting for participant characteristics, and the second step was fitted to a model with adjustments for participant demographics and medical conditions. To examine differences in HAPU rates across states and regions, all HGLMs were fitted with a random state-specific effect to account for within-state correlation of the observed adverse events and outcomes to distinguish between within-state variations from between-state variation. The CMS regions were used to cluster states. A 95% confidence interval (CI) was calculated for each estimate for the models. To take into account differences in Medicare fee-for-service volume between states, all HGLMs were weighted by a total number of discharges from each state. All statistical analyses were conducted using Stata version 8.0 (StataCorp., College Station, TX) and SAS version 8.12 (SAS Institute, Inc., Cary, NC). Hierarchical models were estimated using the GLIMMIX macro in SAS.
J Am Geriatr Soc. 2012;60(9):1603-1608. © 2012 Blackwell Publishing