Propensity Score Matching Use in Nursing: A Synthetic Near Empty Review and Search for New Opportunities

1. Introduction

With the rapid increase in the volume of real-world data and evidence [1], Propensity score matching (PMS) has become widely used in quasi-experimental studies like retrospective analyses of healthcare data sets, registries, observational studies, and electronic medical record analyses based on individual patient covariates, equalizing potential confounding factors when comparing different groups of patients. More precisely, the method attempts to adjust recognized unbalanced factors at baseline such that the data, once analyzed, will better approximate analyses performed on a prospective gold standard, namely randomized data [2,3,4,5,6,7]. The use of propensity score matching in nursing research is much sparer than in medicine [8]. However, interesting results have been recently reported in 107 studies, for example, analyzing the impact of workforce violence against nursing staff [9], analyzing the impact of home and community care services pilot programs on healthy aging [10], the effects of special nursing units in nursing homes on healthcare utilization and cost [11], understanding the impact of chronic diseases on COVID-19 vaccine hesitancy [12], and analyzing if the treatment for insomnia symptoms is associated with reduced depression among older adults [13]. On the other hand, the number of observational, retrospective, and similar nursing studies where propensity score matching could be used is increasing, reaching more than 10,000 studies in total in just the last five years.

This study aimed to synthetically analyze, first, the content of the current nursing proximity score matching studies; second, the content of nursing observational, retrospective or other quasi experimental studies; and finally, based on the above analyses, explore possibilities for further use of PMS in nursing.

2. Materials and Methods

Synthetic near empty review (SNER) approach and Synthetic thematic analysis (STA) [14,15] were used in our study. SNER is a triangulation of bibliometric mapping, near empty reviews (reviews with very small number of eligible studies) [16,17] and STA. The bibliometric analysis was performed with VOSViewer [18], which visualizes bibliometric maps in various ways. In the author's keywords map, closely associated keywords are grouped into clusters, denoted by different colors. The proximity and links represent relations between keywords, and the size of a node reflects its popularity. During the thematic analysis, nodes are used as codes, and a cluster represents codes belonging to one

Following steps were performed:

• 1. Using Synthetic near empty reviews, we identified study themes by analysing publications describing PSM use in nursing published in the period 2020-2024 harvested from Web of Science bibliographic data base, using the search string “propensity score matching”, limited to the Research area Nursing.
• 2. Using Synthetic thematic analysis, we identified the most popular themes for observational, retrospective and other quasi experimental studies in nursing published in the same period and using the same bibliographic database as in Step 1. The search string used was observational or retrospective or "quasi?experimental" limited to research area Nursing
• 3. Comparing the themes identified in Step 1 and Step 2, and using themes emerging just in in Step 2 as keywords, we searched for influential articles in Web of Science and Scopus databases, where PMS has already been successfully used. The cases presented in these articles were finally identified as a sample of new opportunities for PSM use in nursing.

3. Results

3.1. Near Empty Synthetic Review of Nursing PSM Studies

For the period 2020-2024, only 76 papers describing the use of PSM in nursing were found. Comparing that number to the number of more than 10000 PSM studies in health care, we can safely conclude that the PSM used in nursing is extremely sparse. The bibliometric mapping of those 76 studies performed with VOSViewer resulted in the authors keyword landscape presented in Figure 1. The content analysis of the landscape revealed following themes:

• Red cluster: Psychological health (anxiety and depression) of nursing staff after workplace violence. PSM and regression analysis were used to compare depression and anxiety symptoms in physicians and nurses who had or had not experienced workplace violence [19] or whether workplace violence affects psychological health in general [20].
• Light blue cluster: Nurse-led management [21]
• Yellow: Resilience and regret in cancer patients: PSM was used to asses measurement invariance in using resilience instruments in cancer patients [22] and regret in parents of children with cancer [23].
• Violet cluster: Nurse-led management. PMS was used to compare groups of patients who received nurse-led multidisciplinary psychological management and who did not [24] to compare anesthesia-related outcomes between patients monitored by newly recruited nurse anesthetists and those monitored by newly recruited anaesthesiologists [25].
• Dark blue cluster: Group antenatal care. PSM was used to assess the effects of pregnant women in participating in Centering Pregnancy on maternal, birth, and neonatal outcomes [26] and compare antenatal complications in obese and non-obese women [27].
• Green cluster: Successful aging. PSM was used to assess the effect of hospitalization on successful aging [28].

3.2. Syntetic Thematic Analysis of Nursing Observational, Retrospective and other Quasi Experimental Studies

For the period 2020-2024, 9463 publications reporting results from nursing observational, retrospective and other quasi experimental studies were harvested. The author's keyword landscape of these studies is shown in Figure 2. The STA performed on the landscape shown in Figure 2 resulted in the following themes:

• Nursing Education
• Emergency and Critical care nursing
• Primary care nursing
• Patient safety and quality of care
• Pandemics
• Midwifery
• CVD rehabilitation
• Quality of life and self-care/management in all ages
• Pain management
• Epidemiology from the nursing perspective

3.3. Identifying a Sample of New Opportunities

As the analyses above show, the range of current PSM use in nursing is limited, however on the other hand, the range of themes dealt with in observational, retrospective and other quasi experimental nursing studies is very broad and extensive. Comparing the themes from the nursing PMS studies and themes from nursing observational, retrospective and quasi excremental studies, it is interesting to note that they are not overlapping, opening a plethora of new opportunities for PSM use in nursing research.

In order to find the most suitable themes and thus new opportunities for the use of PSM in nursing, as already described above, we searched for influential studies where PSM has already been successfully used on comparable health topics. The results are presented in Table 1.

5. Conclusions

Unlike in general healthcare research, PSM is rarely used in nursing. The most likely reason is the lack of awareness among nursing researchers about its benefits and potential applications. However, successful and comparable examples of PSM use in health care show that PSM deserves to be more frequently employed in nursing research. In this way, observations, conclusions, and new knowledge gained from observational, retrospective, and other nursing quasi experimental studies might become more consistent, relevant, internally and externally valid and consequently more useful in clinical practice and research.