Reported Versus Actual Assault Rates

TO THE EDITOR: In the November issue, Dr. Staggs reports on the results of a survey that collected data on injurious assault rates on 461 adult inpatient psychiatric units in 327 U.S. hospitals (1). The outcomes were mostly counterintuitive, such as higher hours per patient day (HPPD) for both registered nurses (RNs) and non-RNs were associated with higher rates of assault against staff. A clever hospital administrator would jump on these data and conclude that the hospital could employ fewer staff, achieve a lower rate of assaults against staff, and save money. Perfect.

Does the article report the data necessary to support the study findings? I think not. Dr. Staggs indicates throughout the article, including in the title, that he is analyzing “injurious assault rates.” But he is not. The data are on reported assault rates, not actual rates. In order to conduct the study he has made the assumption that the validity and reliability of reported assault rates are constants, unaffected by staffing or HPPD.

In my experience of evaluating psychiatric facilities in half the states, Dr. Staggs’ assumption is unfounded. In general, the better the staffing, meaning the higher the HPPD, the more accurate is the reporting of assaults. Poorly staffed units or hospitals with low HPPD underreport most reportable events because the staff cannot get around to doing the paperwork. Understaffed units and hospitals are more likely to cut corners and slide on their adherence to policies and protocols, and reporting of some injurious assaults would expose this. I am not critical of the personnel unfortunate enough to work in inadequately staffed facilities, where just surviving one’s shift can be a challenge.

Dr. Staggs has no way of knowing whether the association between HPPD and injurious assaults is a function of the rate of assaults or of the accuracy of reporting of assaults. The percentage of assaults that are actually reported is an uncontrolled variable in the study, which basically invalidates the findings.