data sets. Staffing data were obtained from Medicaid cost reports for the 10 states, which were found to have a higher correlation with payroll data than the Medicaid On-line Survey and Certification Report (OSCAR) data that are provided to state survey agencies and the federal government by facilities.

Analysis involved the generation of resident-level risk models for each outcome, which were then used to estimate resident-level risk scores, calculate a facility average risk score, and assess the association between staffing levels and rate of adverse events, adjusting for the facility average risk score. Facilities in the worst 10th percentile were considered to have an inappropriately high level of untoward events, which generally reflected a rate that was three or more times the mean rate for the outcome (e.g. overall UTI hospitalization mean = 0.03; 10th percentile mean = 0.09). Consistently, associations were found between different staffing levels and whether facilities were in the worst 10th percentile. These significant associations persisted until a staffing threshold was reached, above which there was no further detectable benefit from additional staffing. These findings occurred for all three types of nursing staff separately (nursing assistant [NA], licensed [LPN/LVN and RN combined], and RN). The thresholds occurred at staffing levels that exceeded the current levels of 75–90 percent of facilities, depending on the type of staff and the measure. Thus, most facilities fell considerably below the staffing thresholds. These thresholds were between 2.4 and 2.8 hours per resident day for NAs, between 1.1 and 1.3 hours per resident day for licensed staff, and between 0.55 and 0.75 hours per resident day for RNs. However, incremental improvements in quality occurred at all levels until these staffing thresholds were reached.

This study also found (based on an analysis of 631 facilities in California for which information on staff turnover and retention was available) a strong relationship between staff retention and outcomes related to patient safety. For example, improved annual retention of nursing staff up to a threshold of about 51 percent (i.e., half the staff stay for a full year) was associated with a substantially higher likelihood (odds ratio 3.66) that a nursing home would not be in the worst 10 percent of facilities. However, retention of less than 51 percent was associated with a high risk of adverse events, such as hospitalizations for UTIs and pressure ulcers.

Explanations for the Causal Relationship Between Staffing Levels and Patient Outcomes

Several studies have attempted to explain the relationship between higher levels of nurse staffing and improved patient outcomes. The results of these studies support the position that as the numbers of nursing staff increase, the staff are proportionately able to provide increasing amounts of



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement