The incidence of patients experiencing adverse events while hospitalized has proven a major problem [1] [2] [3] [4] . Reports and research on the occurrences of adverse events shows that a great number of patients experience adverse events while receiving health care all over the world [4] [5] [6] [7] . An international study conducted in England, Germany and the USA, with focus on medication errors, found that the prevalence of adverse event related to drugs alone ranged from 3.22% to 5.64% of the patients hospitalized in 2006 [8] . There have been a greater media and research attention to patient safety and the effect of staffing on the occurrence of adverse events [9] [10] [11] . International media have previously described patients being exposed to adverse events with severe consequences, of which many are caused by poor staffing [12] [13] . There are several studies on this topic, but with varying foci such as mortality, adverse events related to specific patient groups and specific patient diagnosis, medication errors, patient satisfaction, work environment and economic perspectives [14] [15] [16] [17] [18] . Evidence in this area has also previously been synthesized in literature reviews [19] [20] [21] , but no reviews have documented the direct and indirect relationship between understaffing of nurses and patient safety.

The US Department of Health & Human Services defines patient safety as “the freedom of accidental or preventable injuries produced by medical care”. Patient safety measures are defined as interventions or work intended to reduce the occurrence of preventable events [22] . Several different bodies ensure patient safety in hospitals. Health care providers have a responsibility to perform professionally appropriate health care. Each hospital must ensure that the health care providers work under conditions that promote safe practice, for example with appropriate equipment, adequate staffing and safe routines. Additionally, both health care workers and the health care system must comply with the law, professional standards and procedures to provide safe healthcare services [23] .

The requirement of professional responsibility is the most central demand in the legislation when it comes to safety practice in the healthcare service. The demand of professional responsibility is according to law, a shared responsibility, were the health workers are responsible for their own actions, and the healthcare system is responsible for the environment these actions are conducted in. Further on health workers, in this case nurses are obligated to perform safe healthcare through the nature of their public authorization, which is an arrangement built on common national and international demands in the nursing education and is a requirement to legally assess nursing tasks [23] .

The International Council of Nursing states that nurses carry a responsibility to perform safe practice and to obtain the knowledge to do so. They are obligated to provide holistic patient care, which include giving the patients and families accurate information and education. Nurses are expected to participate in maintaining safe working conditions and safe practice [24] .

The purpose of this study is to increase the knowledge of understaffing amongst nurses in hospitals, and the possible consequences of understaffing for patient safety. More specifically the study examines understaffing as a risk factor for hospitalized patients. The following research question has guided the study: How can understaffing amongst nurses in hospitals affect patient safety?

Understaffing is a term with numerous connotations and meanings. In the literature, understaffing is used, for instance, in reference to high patient-nurse ratios, heavy workload, large patient load, nursing hours per patient, and high bed occupancy. The common denominator of these definitions are that understaffing is a lack of personnel, in this case nurses, to conduct their required tasks. In this study, the term understaffing is defined as “a disparity between load of responsibilities/tasks and the possibility to conduct them in a professional manner”.

A nurse is in this study, defined as a person with a bachelor education in nursing, having regular contact with patients admitted to a hospital ward.

To be included in the review, articles had to be written in English, they had to have a clear qualitative or quantitative design, and they had to have been published between1997 and 2016. Eligible studies had to be concerned with nurses with patient contact, working in hospitals, some form of understaffing (excessive workload, high patient-to-nurse ratio, number of working hours per patient) and patient safety.

The search strategy was developed in accordance with Bettany-Saltikov [25] methodology for literature reviews in nursing. Literature searches were conducted in the Cinahl, Medline, Isi Web of Science, Cochrane Library and Academic Search Premiere databases between early February and the end of January 2016. All databases focus on health and social science. In addition, secondary searches were made based on the reference lists of included articles, and a manual search was conducted in selected journals of specific interests such as BMC Health Services Research [29] . The terms used when searching the databases included understaffing, nurse, patient safety, lack of nurses, staffing levels, manning levels, downsizing, short-staffed, short-handed, inadequate staffing, insufficient staffing, workload, patient outcome, patient security, patient mortality and adverse events. Terms which disclosed understaffing (lack of nurses, staffing levels, manning levels) were combined with terms disclosing nurses (healthcare worker, RN, trained nurse) and hospital (healthcare facility, Hospital ward, medical institution) in a systematic manner which included all possible combinations of these words. (A comprehensive list of the searches done in each database can be found in Attachment 1)

All included studies were of quantitative study design, and underwent a quality assessment according to the Cochrane Quality Assessment Tool for Quantitative Studies [30] . This assessment tool was developed for use in public health, and is suitable for quality assessment in most areas of public health [30] . The quality assessment involves grading six areas in each study (selection bias, study design, confounders, blinding, data collecting methods, withdrawals and drop-outs). Each domain is given points (1 - 3 points) based on relevant questions. Based on the total ranking score, each study was graded as having strong, moderate or weak quality. Studies graded “moderate” or “strong” were included in the review. (Attachment 2 gives a full overview of the quality assessment.)

We used a predesigned form for data extraction according to Dixon-Woods et al. [26] , describing the study design, quality, objectives, data material/partici- pants, main findings, and the location of the study. First author retracted data from all included studies under detailed supervision from the research group. Recognition of themes was made through “identification of prominent or recurrent results in the articles analyzed, to produce an account of evidence” [26] .

The database searches identified 2847 records. Six articles were found through secondary searches and recommendations from researchers in the field of patient safety. Of the 2609 articles screened, 2495 were excluded. The remaining 114 articles were read and evaluated in full text (see Figure 1, The PRISMA 2009 Flow diagram if article selection process). Following the full text evaluation, 43 articles were included based on the inclusion and exclusion criteria. 10 of these

studies did not pass the quality assessment, and were excluded, resulting in a total of 33 quantitative studies being included in the current review. First author performed the searches and undertook the screening of titles and abstracts against inclusion criteria, with supervision from the research group. First researcher then undertook the read-through of selected full-text articles. Where there was question of inclusion eligibility, the research group was consulted independently to assess full-text item suitability.

The results of the included studies were analyzed through Thematic analysis which involve finding prominent or recurrent themes in included articles, and gather the themes under suitable headings [26] . The predesigned form was used to organize the main results of the included studies, allowing the researcher to detect patterns or recurring subjects in the literature. The recurring subjects were then organized in to sub-themes and themes.

Data represented in this study emerged from already published peer reviewed articles. Data collection did not involve human subjects, and a written informed consent has therefore not been obtained.

All the data supporting the conclusions can be found in Table 1 with information on the included studies. Information on the search words, combinations and results, can be found in the supplementary file 2 Database search report.

Of the thirty-three studies included, there are nine cohort studies, thirteen cross-sectional studies, two correlation studies, one case control study, three retrospective observational studies, two retrospective longitudinal studies and one with a four-stage sampling design. It was not possible to categorize the methods used in two of the studies. (Table 1 lists the articles and study characteristics). All of the studies investigated the relationship between understaffing of nurses working in hospitals and patient safety.

The thematic analysis [26] [60] generated two main themes and two subthemes concerning understaffing and consequences for patient safety.

Theme 1: Direct consequences

Twenty-three studies reported that understaffing had direct and severe consequences for patients.

1A) Patient harm

Sixteen of the twenty-three studies examined several adverse events as a direct consequence of understaffing among nurses working in hospitals [16] [17] [18] [32] [35] [36] [37] [40] [42] [44] [46] [47] [48] [49] [50] [58] . Several studies found that understaffing increased the patient’s risk of contracting pneumonia, urinary tract infections or other hospital-related infections [16] [17] [35] [36] [40] [42] [44] [47] [48] [49] [50] . Cho and colleagues [16] found that one extra hour of work by registered nurses per patient per day was associated with a 8.9% decrease in the odds of the patients getting pneumonia. Cimiotti et al. [35] also found a negative association between number of hours of care provided by registered nurses and the risk of bloodstream infections in infants receiving neonatal care, but only in one of the two neonatal intensive care units (NICU) investigated. The authors attribute these differences in results to variation in the hospitals’ and infants’ characteristics.

Along with hospital-related infections, the incidence of pressure wounds was also a problem that, according to the literature, increased in proportion to understaffing [17] [44] [47] [48] [52] . Twigg et al. [47] reported that surgical patients experienced significant lower pressure wounds rates in all three hospitals investigated, after those hospitals adopted a new staffing method called Nursing Hour Per Patient Day (NHPPD). This staffing method significantly increased the staffing levels. In other words, increased staffing levels led to decreased rates of pressure wounds. A lower proportion of licensed nurses working was associated with a higher incidence of pressure wounds in the hospital investigated by Unruh [48] . Van den Heede et al. [59] in contrast, did not find any significant relationship between acuity adjusted nursing hours per patient day and pressure wounds. Kovner et al. [60] also investigated the direct consequences of understaffing but did not find any significance between registered nurses (RNs) hour per patient per day and three types of adverse events: pulmonary compromise, urinary tract infection, and thrombosis. Other studies, however, found a significant relationship between understaffing and both thrombosis and shock/cardiac arrest [17] [40] [47] . Schreuders et al. [57] found an inconsistency in the relationship between understaffing and patient complications such as urinary tract infections and surgical wound infections related to patient characteristics such as age, gender and health status. The direction of the association between nurse staffing and inpatient complications was not consistent for different patient complications.

Lastly, one study found that higher night staffing reduced the prevalence of extended hospital stays [37] .

2B) Patient mortality

Nine studies found understaffing to affect mortality in hospitalized patients. Both surgical and medical patients were investigated [14] [31] [33] [34] [38] [39] [41] [43] [45] .

In one observational study from 2014 and one cross-sectional analysis from 2002 Aiken and colleagues found that after adjusting for patient and hospital characteristics, each additional patient per nurse was associated with a 7% increase in the likelihood of dying within 30 days after admission. Carthon et al. [33] reported similar results, finding that an increase of one patient in the nurse’s patient load increased the occurrence of deaths in hospitalized patients. They also found that older surgical patients were affected more by understaffing than other groups of patients. Older surgical patients had a poorer post-surgical course when their nurses had a heavier workload, including a higher chance of death and failed resuscitation [33] . One study investigated mortality in the form of failure to rescue (from an adverse health-related event), finding that understaffing had a negative effect also in this area [33] .

In two studies, no correlation between mortality and understaffing were found [38] [68] . Amarvadi and colleagues [32] investigated the understaffing-mortality problem by taking a closer look at the night shift. They used the night time nurse-to-patient ratio to calculate staffing, and found no increased risk of dying in hospital whether one night nurse cared for one or two patients or for three or more. Kiekkas et al. [38] found that an increased workload among the nurses did affect hospital mortality, but these numbers did not reach statistical significance.

Theme 2: Indirect consequences

Six studies found understaffing to affect patient safety, not by causing the patient direct harm but by increasing the risk for direct harm later on.

2A) Poor quality of basic care.

Four studies concluded that understaffing led to poor quality of care [15] [17] [54] [69] . Several nursing tasks, such as patient communication, skin care, oral hygiene, documentation, responding to patient alarms and mobilization were left undone, postponed or not prioritized due to lack of time, as a consequence of understaffing of nurses [17] [55] . Zhu and colleagues [55] found that 30.39% of nurses reported that the care that they delivered was of moderate or poor quality. Duffield et al. [17] found in their longitudinal study that patient communication was not prioritized in as many as 39.5% of the shifts. Skin care and back rubs (24.0%) and oral hygiene (19.3%) were also frequently reported undone. Patients’ ability for self-care and patient satisfaction was also negatively associated with understaffing [15] [54] .

2B) Errors in administration of medication.

Two studies focused on understaffing and medication errors [53] [56] , reporting that heavy nursing workload, interruptions and being rushed increased the risk of adverse events in conjunction with the administration of medication. Several other studies found an association between understaffing and medication errors, even when this was not their focus [15] [17] [52] . These findings showed that understaffing led to missed or late doses of medication, poor pain management resulting from difficulties in administering needed pain medication, and medication errors in general [15] [17] .