Retention of Nursing Staff

Methodology

Evaluation Methods and Tools

Pursuant to the aforementioned information, the primary means of data evaluation will be statistical analysis of the pencil-and-paper questionnaires gathered over the course of the research project. However, there is still more to be said about this questionnaire, as it is the main tool by which data will be collected. Though previously it may have seemed that one option was to leave at least some survey questions open-ended to maintain the possibility that participants would devise novel responses to general questions about nursing recruitment and retention issues, in fact, this will not be the final method chosen. Rather, a closed set of potential concerns regarding nursing staff retention shall be raised so that the statistical analysis can be as robust as possible. Yet in spite of this, it is worth a brief aside in order to justify this elimination of a potential approach.

Ultimately, as has been previously stated, one researcher will be doing the evaluation of the data, and despite the clustering techniques previously proposed, there will still be a veritable plethora of activities that must be performed in order to complete the research. Given this fact, it will be helpful to eliminate nonessential aspects of the study, and one of these components is the necessity of learning data-coding techniques. In addition, the ability of a researcher who has freshly learned such techniques to be consistent may be called into question, and thus this could prove confounding to data evaluation. However, such methods do exist, as is demonstrated by one study that “. . . illustrate[d] the use of a qualitative research technique, textual data analysis, in assessing the emotional content of open-ended survey responses” and found that “Dimension scores produced from the DAL and direct ratings of open-ended responses correlated with the quantitative measures in expected and consistent ways” (Mossholder, Settoon, Harris, & Armenakis, 1995, p. 335). Future studies may wish to employ such methods when approaching nursing retention issues from a more open-ended perspective. Alternatively, there are other methods that might be used, as given by Jackson and Trochim (2002), who “ . . . present[ed] concept mapping as an alternative method to existing code-based and word-based text analysis techniques for . . . open-ended survey questions . . . offer[ing] a unique blending of the strengths of these approaches while minimizing some of their weaknesses” (p. 307). Thus, it is quite apparent that there are a good number of viable alternative options to the strategy proposed here, but these have been reviewed in detail prior to their rejection in favor of the current tactics. In the scientific world, it is always at the very least educational to explore other options and eliminate them only after careful analysis; indeed, this is the very means by which hypotheses themselves are evaluated, and to apply similar logic to the selection of data evaluation methods is only natural and normal. Moving forward, then, the details of the data tools and evaluation strategies ultimately chosen must be presented.

The issue with retaining nursing staff is one of human decision-making in the end, as it is the nurses themselves who use their freedom to pick and choose among available employment options. Indeed, though part of the trouble with retention may be that there are simply too many choices available to nurses, who, thanks to their in-demand status, are free to hop around from position to position as they see fit, there is little hospitals and other medical establishments can do about this aspect. Instead, the focus must be on giving nurses a satisfying work environment, which is ultimately tied to the emotional needs of the nurses. As is well-known, some studies do seek to measure emotion quantitatively, such as with Likert-type scales, though of course human emotion is the most qualitative of factors of all. Using Likert-type scales allows the researcher greater access to the inner workings of participants’ minds while at the same time providing quantitative data. This helps combine the strengths of both qualitative and quantitative approaches, allowing statistical analyses to be performed with as close to their full potency as possible. With the Likert scale identified as the means of phrasing the questions, using the classic “strongly disagree” to “strongly agree” scale now so commonly known as to be even a part of layperson’s language, the next obvious question is how many points should be included on this scale.

In the course of performing preliminary research, it has been determined that this study will use a seven-point Likert-type scale, instead of any other number. This will be the means of administering a survey with subjective statements that seek to examine nurses’ beliefs about their own motivations for remaining in a given position or seeking employment elsewhere. Here, the choice of a seven-point scale for use in the survey is in reality rather arbitrary. However, as the literature on the topic seems to find no particular certain reason for which to use any certain number, in the balance the tack to take that makes the most sense is the one that is the most intuitive to the researcher in question. In addition, though it indeed may be a rather surprising piece of information, Matell and Jacoby (1971) found that, “Results show that (a) there were no specific patterns for reliability or validity coefficients using different number of alternatives; and (b) when the original scales were dichotomized or trichotomized, no significant differences were found in reliability or validity” (p. 657). This is obviously valuable for the support of choosing the “correct” number of points as being an endeavor with multiple right answers. Furthermore, Albaum (1997), too, discussed the issue, reaching an almost identical conclusion that no particular number need be favored over another. Therefore, as this researcher has most frequently seen seven-point scales used, that foundation on its own shall serve as sufficient means to direct the study toward the use of a seven-point scale. Though it may seem that much time has already been spent evaluating various components of the survey, no amount of investigation could truly be too much into this essential data tool and its evaluation in this research. Having established certain base premises upon which the foundation of the survey might rest, it is time to address the meat of the issue and provide examples with questions that might be used on the survey.

Early on in the survey, the questions asked will simply be about the nurses’ satisfaction with their current positions, taking care to ask about the same variable from several different angles for a sum of six questions per category, more or less. One example might be: (1) I feel satisfied with my job. From that point, the survey will journey into broader questions about nurses’ perceptions of their positions as compared to other potential positions, e.g., (2) My position is better than the average nursing position available to someone of my skill level and educational background, or (3) I am lucky to have this job. The survey will proceed apace from there, going into greater detail.

From that point, gradually narrowing the questions but still asking them in multiple ways in batteries of six apiece, give or take a few, the variable will change to touch on the issues identified in the problem statement for this research—i.e., the quality of team-building in the workplace and the possibility of upward mobility. Then, the questions become similar to: (4) I feel valued as a member of a team in my workplace; and, (5) I think it is likely that if I work hard and do my job well, I will be promoted as soon as there is an opening. Again, these two aspects will be addressed separately, and, as with the other components of the survey, will be asked several different times in several different ways to generate greater validity.

In addition, although it might seem strange in the beginning, the opposite statements must be presented for evaluation on the Likert scale. For example: (6) I am unlucky in the position I have compared to other nurses. Even though the responses may be facile “strongly disagree” answers, in truth, asking these converse alternatives to the previous questions gives this research greater validity and helps erase some of the possibility suggested previously that invalid assumptions may have been made. The pairing of the survey items will also serve as a form of validation. Thus, by querying respondents with the opposite questions, this study will confirm that the originally asked questions exist along valid axes for investigation in the course of this research.

As for data evaluation, though statistical methods have been discussed before, an appropriate p-value has not yet been chosen. Given that this study exists essentially in the realm of the “soft” sciences, dealing in part with the areas of sociology, education, and psychology, the appropriate p-value to choose for significance is p < 0.05. In all analyses, the questions on the survey shall be evaluated as groups depending on the banks of questions from which they were drawn, and the aforementioned clustering techniques shall be applied only after this action has been performed.

Since the topic has been broached, it may now behoove this work to justify the use of a multivariate approach in the data tools and evaluation strategies used. Though it may seem at first that this study is merely following the recent scientific trend of including as many variables as might be possible, given that the advent of the current technological times has made computers able to process enormous quantities of data on reasonable time scales, in this case, the choice is reasonable. It is true that there are quandaries, from a statistical point of view, that are involved in multivariate studies. Yet it is still worth the possibility of gaining additional data to adopt this method. In addition, though issues arise inherently from the question of whether the survey responses collected using a Likert scale can be really considered ratio data instead of ordinal data, there have been no substantial challenges to this method of approaching research. The major notable objection, in fact, comes out of the area of geographical mapping, and as such is not terribly related to the subject matter at hand—Fortheringham and Wong’s (1991) depressing results need not allay the researcher here, though the pair did find that, “ . . . multivariate statistical analysis . . . is therefore a much greater problem than . . . univariate or bivariate analysis. The results of this analysis are rather depressing in that they provide strong evidence of the unreliability of any multivariate analysis . . . ” (p. 1025). The survey nature of the research approach for this study is unavoidable here, and yet with only such paltry objections to going forward with a multivariate approach, surveys will prove to be an indispensable tool for use in gathering this type of multivariate data. Though it may not on the surface of the matter appear to be the cleanest scientific approach, ultimately, there is a compromise to be made between the breadth of data gathered and its depth.

Maintenance of Data Integrity

Given that this data will contain aspects related to nurses’ satisfaction with their current jobs, it is, without question, necessary to maintain the highest standards of data integrity in particular as it relates to data security. For an insightful overview of the topic of data integrity in general, it is helpful to review Finkenzeller (2003), who discussed the potential of researchers handling data that might prove sensitive. In addition, it will be helpful to keep in mind that the greater the participants’ sense of security, the more they will be likely to disclose even truths they feel to be uncomfortable. The security aspect of the data can be reasonably handled by taking care to ensure the slotted box into which participants will deposit their survey responses is closed by means of a padlock to which the research alone bears the key. Because the recipients of the emailed survey are responsible for providing the email address with which they feel most comfortable using for discussing work in ways that may be negative, the participants retain control over this aspect of data security. Likewise, the selection of a printer for use in depositing the hardcopy version of the survey is up to the participant, and is thus out of the hands of the researcher.

Data integrity issues unrelated to security in specific may also arise. As the survey will be distributed with instructions suggesting the use of pencil in filling out the information, so that corrections can be quickly and easily marked, there is a potential concern that repeated handling could lead to degradation of the data. Should marks become illegible, or should they start out that way due to participant or researcher error, the final method of recourse that remains always available is to email the specific respondent asking for clarification. However, the option of making copies in advance and securing them at a location other than that in which the originals are kept, ensuring that this location is also safeguarded, is always present. To counter the troubles inherent in using paper for this type of study, there is the fact that the information can be quickly “cleansed” at the end of the study, as Maletic and Marcus (2000) advocate for, by means of using a high-quality shredder to dispose of both the originals and any copies made. Thus, in some ways, the relatively more primitive approach of using actual paper written upon by hand, rather than, for example, printed out from a fillable pdf file, is preferable from the point of view of data integrity. This is but one tiny aspect of the overall research design, and yet no detail can be neglected in the pursuit of a broad overview.

Research Design

There is a good deal of research supporting the selected design. As is well-known, surveys have enjoyed a long history of use in scientific studies thanks to the fact that they are simple, easy to execute, and inexpensive. Babbie (1990) provided an excellent work touching on many aspects of this approach. It is upon the basis of such previous investigation that several of the precepts that have guided this study have been formed. In proceeding, such guidance will prove invaluable.

Research Methods and Justification

In particular, this study has benefitted from modeling upon previous work in the specific field of nursing staff retention strategies. For example, though Aiken, Clarke, Sloane, Sochalski, and Silber (2002) focused on the situation specifically in California at the time of their writing, the rousing language used therein is still evocative of the nursing staff shortage and retention problem today: “The worsening hospital nurse shortage and recent California legislation mandating minimum hospital patient-to-nurse ratios demand an understanding of how nurse staffing levels affect patient outcomes and nurse retention in hospital practice” (p. 1987). Given that the team for that study also chose a survey design in order to approach the topic of the retention of nursing staff, it would seem that there is strong support for proceeding along these lines.

It is worth noting, if momentarily, that the possibility of discussing the interaction of nursing staff retention with the education of new nurses, a prospect that was raised previously, has subsequently proven beyond the scope of this study. Still, in the unlikely event that resources should prove overly abundant, both in terms of time and in terms of budgetary considerations, this is a possible angle to explore in further depth, perhaps leading to an adjunct work of research. As one study has put it, “The newly developed nursing self-concept of the graduate nurse may provide a key indicator for predicting graduate retention” (Cowin & Hengstberger-Sims, 2006, p. 59). This is also an area rife with possibility for future study, and it would be interesting to take a new tack based upon Cowin and Hengstberger-Sims’s (2006) research that combines their exploration of newly graduated nurses’ self-concept with the ways in which that idea of the self might interact with career prospects. Taken altogether, however, this angle is not as immediately relevant as the one already described.

Participants’ Relationship to Target Demographic

In previous sections, the discussion has already centered on the topic of the demographics of the participants, and so repeating these details proves unnecessary. To recap the existing work, the fact that the nurses shall be randomly selected from an appropriate age range is a sufficient starting basis to satisfy concerns about demographics. Of course, comparisons of sex, race, and other proportions in the sample to those in the population may be needed should the sample seem inherently skewed, but this is a topic better discussed after the actual data are gathered, and if such difficulties arise, they are fair fodder for subsequent discussions of possible confounds to the study.

Obtaining Needed Permissions

Obtaining the needed permissions, too, is at this point a relatively simple matter, as all other aspects of the study have already been established. Informed consent forms will be signed by all the participants, and these will be distributed via email to everyone involved. With the surveys themselves, these forms will be returned via a slotted box in a central location for all to deliver. Any survey returned without this vital permission shall be promptly shredded and discounted should obtaining the needed permission from the participant prove to be a matter of any significant length of time, as it would not be proper to retain survey data without a participant’s full consent.

References

Aiken, L. H., Clarke, S. P., Sloane, D. M., Sochalski, J., & Silber, J. H. (2002). Hospital nurse staffing and patient mortality, nurse burnout, and job dissatisfaction. JAMA: The Journal of the American Medical Association, 288(16), 1987-1993.

Albaum, G. (1997). The Likert scale revisited. Journal of Market Research, 39, 331-348.

Babbie, E. R. (1990). Survey research methods (Vol. 2). Belmont, CA: Wadsworth Publishing Company.

Cowin, L. S., & Hengstberger-Sims, C. (2006). New graduate nurse self-concept and retention: A longitudinal survey. International Journal of Nursing Studies, 43(1), 59-70.

Finkenzeller, K. (2003). Data integrity. Hoboken, NJ: John Wiley & Sons, Ltd.

Fotheringham, A. S., & Wong, D. W. (1991). The modifiable areal unit problem in multivariate statistical analysis. Environment and Planning A, 23(7), 1025-1044.

Jackson, K. M., & Trochim, W. M. (2002). Concept mapping as an alternative approach for the analysis of open-ended survey responses. Organizational Research Methods, 5(4), 307-336.

Maletic, J. I., & Marcus, A. (2000, October). Data cleansing: Beyond integrity analysis. IQ, 200-209.

Matell, M. S., & Jacoby, J. (1971). Is there an optimal number of alternatives for Likert scale items? Study I: Reliability and validity. Educational and Psychological Measurement, 31(3), 657-674.

Mossholder, K. W., Settoon, R. P., Harris, S. G., & Armenakis, A. A. (1995). Measuring emotion in open-ended survey responses: An application of textual data analysis. Journal of Management, 21(2), 335-355.