By Daphne Lyn G. Rivera, RN

Roison Andro Narvaez, MSN RN

St. Paul University Philippines – Graduate School, Tuguegarao, Cagayan

Citation: Rivera, D. L., Narvaez, R. A.  (2023). Application of Machine Learning approaches in Cancer prediction. Canadian Journal of Nursing Informatics, 18(2).  https://cjni.net/journal/?p=11581

Abstract

Background: Cancer has long been a major concern in the healthcare industry. Scientists’ countless studies to find resolutions, conduct trials, correct errors have required enormous financial expenditures. Recently, this quest has led some to Machine Learning (ML) approaches to cancer prediction, susceptibility and prognosis.

Aim: The purpose of this paper is to investigate the various ways in which machine learning methods can be used to predict cancer.

Design: An integrative review

Methods: This study makes use of five search engines and databases namely: PubMed, the Science Citation Index, Sci-Hub, Google, and Google Scholar. Makes use of terms “cancer and machine learning”, “cancer machine learning for cancer prediction”, “machine learning in cancer prognosis”, and “cancer risk”. Inclusion criteria were assessed by carefully scanning abstracts to find ones that point to machine learning and cancer.

Results: Six studies met the inclusion criteria. Most of the scholarly articles showed ML effectiveness on individuals with carcinoma and described machine learning methods able to detect cancer in the future.

Conclusion: Machine Learning methods have advanced enough to serve oncological research and practice. It has already contributed to improving the quality of life for cancer patients. Allowing individuals to determine their cancer expectations and survivability and providing them with the most accurate choice of treatment according to their needs are other benefits of ML use in oncology.

Implications for Practice: Machine learning can improve success in cancer detection and treatment if scientists, engineers, and medical professionals incorporate it in their studies and practice. It is also advantageous to have an adequate number of health professionals engaged in oncology institutions to better provide appropriate care to achieve healthier outcomes.

Background

Cancer has a significant impact on the lives of those who have been diagnosed with it. Even though certain forms of cancer are curable, the mere mention of the disease has long been associated with death. According to the National Cancer Institute (2021), cancer is a condition wherein the body’s cells develop abnormally and metastasize to distant parts of the system caused by a deficiency in cell division, DNA damage, or genetic inheritance. It remains one of the most debilitating diseases despite technical developments in the medical industry around the world. A historical study conducted by Weir et al. (2015) demonstrated that the total number of new cancer cases would continue to rise by more than 20%, thus, to mitigate the impact of senescence and expanding population on cancer incidence and mortality, primary prevention and early detection must receive priority. Regardless of the large decline in the number of cases, this is still a significant amount regardless of the continuous promotion of healthy living and lifestyle changes to reduce cancer incidence. An important factor that impacts people at risk is being diagnosed at a higher, more dangerous stage making it less curable. As Kumaraswamy (2022) pointed out, deaths from cancer are often documented at a high rate because of inaccurate or delayed diagnosis. If diagnosed and treated early, mortality rates can be reduced and even eliminated. Determining the possible prognosis of an illness is among the most frustrating and complicated challenges for medical experts. Cancer diagnosis consists of three prognostic predictions pertaining to “cancer risk assessment”, “cancer recurrence”, and “cancer survivorship” (Kumar et al., 2022). However, fundamentally speaking there is a clear distinction between the objectives of cancer detection and diagnosis and those of cancer prediction and prognosis.

Machine Learning (ML) can be used for cancer susceptibility prediction (Ayer et al. 2010). Ayers et al. (2010) described how successful clinical decisions require the accurate differentiation of benign from malignant breast lesions as well as the prediction of the risk of breast cancer for each patient. This is supported by Osareh and Shadgar (2010) who determined the optimal accuracy rating for breast cancer detection to be 98.80% and 96.33 percent, respectively, utilizing support vector machines and a classifier model versus two frequently utilized breast cancer benchmark datasets. According to Stojadinovic et al., (2011) to build a model that can forecast overall risk survival in people who have peritoneal surface cancer, an application of a learning and practical artificial intelligence platform to the study population that uses pertinent clinical, pathological, treatment-related, and oncological variables and outcome data are still under consideration. This is supported by the study conducted by Naji et al. (2021) where the support vector machines did a better job in comparison to other classifiers with the highest accuracy of 97.2%. Identified machine learning techniques offering high precision in classification and efficient screening skills using cross validation demonstrated an error rate of only 2.49% to 3.81% (Amrane et al., 2018). Moreover, the result of the study conducted by Garg & Gupta (2020) concluded that deep learning has the most accuracy in breast cancer prediction with minimum loss at 98.24%.

Park et al. (2014) found that the supervised ML approach is appropriate for prognosis prediction and analysis of the biological functions of genes associated with cancer recurrence because it makes use of numerous data points without class labels. Moreover, cancer survival prediction studies done by Park et al. (2013) and others, to assist medical healthcare experts and information specialists in the guideline election process when generating a conjecturing model for the medical field, have been recommended using a paradigm that overpass the information and medical domains. This contrasts with others who showed microarray data also highlights the importance of clinical data for classification. Three criteria should be met by prognostic markers: they must be reliable enough to distinguish between patients with slow-moving and fast-moving diseases, they must be able to predict a patient’s future response to a given therapy, and they must be able to recognize patients who are at risk of experiencing harmful side effects from a given therapy (Rosado et al., 2013).

As a result, a reliable, efficient, and speedy response can be provided by an automatic disease detection system, which not only reduces the risk of death but also assists medical professionals in the process of disease identification (Islam et al., 2020). Health-care systems are being supported by the ability of machines to direct cognitive activities to achieve a specific goal depending on available data (Goldenberg et al., 2019). A trained medical expert now has access to a wide range of diagnostic tools and treatment options for determining the best course of action for each individual patient as the potential for serious adverse effects on the patient’s life due to an inaccurate diagnosis is a risk that is also highly being addressed (Kumaraswamy, 2022).

As a result, a reliable, efficient, and speedy response can be provided by an automatic disease detection system, which not only reduces the risk of death but also assists medical professionals in the process of disease identification (Islam et al., 2020). Health-care systems are being supported by the ability of machines to direct cognitive activities to achieve a specific goal depending on available data (Goldenberg et al., 2019). A trained medical expert now has access to a wide range of diagnostic tools and treatment options for determining the best course of action for each individual patient as the potential for serious adverse effects on the patient’s life due to an inaccurate diagnosis is a risk that is also highly being addressed (Kumaraswamy, 2022).

Aim / Research Questions

The purpose of this study is to investigate the various ways in which machine learning methods can be used to predict cancer and the advantages of machine learning technologies in the context of cancer prediction. This includes the significance of the information obtained through machine learning. We looked at the question: How does the application of machine learning approaches aid in development of clinical decisions and support systems for cancer patients and their welfare? Also, how can we apply appropriate and applicable ML to a specific set of patients with specific oncological problems?

Method

Design

This research study applied Whittemore & Knafl ‘s (2005) integrative literature review process in its design. A scientific and strategic step-by-step procedural analysis of literature utilizing a combination of conventional and non-conventional approaches in research that can enhance evidence-based practice in nursing. Thus, the literature review clearly explains and focuses specifically on the presentation, analysis, and evaluation of gathered data.

Search Strategy

The review was conducted from December 2021 to January 2022. Several electronic databases were used in the preparation of this review including PubMed, Google Scholar, CINAHL, Web of Science and Scopus. Using search terms”Cancer and machine learning” “machine learning for cancer prediction,” “cancer machine learning and prognosis,” and “cancer risk”. A PRISMA Diagram (see Figure 1) illustrates the search process. Initial identification yielded 256 articles based on the title. Out of these, 25 abstracts were initially reviewed, and 12 items were comprehensively reviewed. A final six studies were deemed eligible to be included in this review based on the criteria.

Inclusion and Exclusion Criteria

The inclusion includes studies on multiple assessment and machine learning subqueries using machine learning techniques and algorithms on cancer prediction, risk and survival. Those that lacked details, duplicates and incomplete literature were excluded. The literature reviewed was authored and collected in a scientific database from 2003-2021 and were all published in English.

Figure 1. PRISMA Flow Diagram of Application of Machine Learning Approaches in Cancer Prediction

Data Evaluation/ Quality Appraisal

The keywords were used to determine articles and literature that resulted in extensive information comparative to the focus studies. Scholarly writings relating to Machine Learning approaches in the prediction of cancer, cancer prognosis and cancer survivability that linked together were included. Whilst those that were not related to the topic, had incomplete data information, and those lacking thorough data samples and accuracy were excluded. Therefore, the Critical Appraisal Skills Programme (CASP) (2018) and the hierarchy of Level of Evidence (LOE) (Melnyk and Fineout-Overholt, 2011) were used to classify the study design and methodology, credibility, relevance, and implications for the delivery of health services and decide which studies should be included for review and evaluation.

Results

Six studies met the eligibility criteria that considered substantial results of machine learning methods that considered several cancer risk factors while subsequently determining cancer predictability, cancer prognosis and cancer susceptibility.

Table 1 outlines the six studies’ pertinent characteristics that satisfied the eligibility requirements and preferences. There is no doubt that developed nations with highly standardized healthcare systems and technologically advanced researchers conducted most of the studies that were reviewed. The studies were mainly performed in well-established medical facilities that focus on cancer patients. Place of study research included the USA, Hungary, South Korea, Belgium, and Spain. In addition, six research study designs can be summarized as Retrospective (n=1), Prospective, Descriptive, Comparative, and Qualitative study which were presented as Levels of Evidence II (n=1), IV (n=4) and VI (n=1).

Table 1. Characteristics of the Studies

The experimental subjects in the sample all had oncological medical issues. Most of the study participants in the six selected studies had breast cancer (Ayer et al., 2011; Park et al., 2014; Park et al. 2014; Gevaert et al. 2006). Other oncological participants included oral carcinoma (Rosado et al. 2013) or colon carcinomatosis (Stojadinovic et al., 2011). Methods used to gather specific candidates for the research were mostly sought from their data based on previous history and current oncological problems, were referred or sought referral, and were randomly selected.

Considering the search queries’ enormous return of 256 articles, it was necessary to scrutinize the results to keep only the most pertinent articles in this study’s sample. Based on the keywords from the three predictive tasks that were found in the titles and abstracts of each publication, each study’s relevance was evaluated. In the end, we found only a few were relevant after reading their abstracts and titles. Publications that explored one of the three areas of “cancer susceptibility” (Ayer et al., 2010; Stojadinovic et al., 2011) “cancer recurrence” (Park et al., 2014) or “survivability prediction” (Park et al. 2013; Gevaert et al. 2006; Rosado P. et al. 2013) were included.

Table 2 summarizes the aim or goal of the six selected studies that used ML in the prognosis and prediction of cancer and the findings that resulted in each of the studies.

Table 2. Aim and Findings of the Studies

Discussion

In the current research on cancer prediction, pertinent machine learning methods have been introduced. The aim of the selected studies was to determine the effectiveness of machine learning methods. After a conscientious analysis, three identifying factors that can be used to facilitate efficacy in each cancer characteristic – cancer susceptibility, cancer recurrence and cancer survivability – rooted in machine learning emerged. There is sizable literature published that contributed precise outcomes decades ago. Nevertheless, there are still imperfections that need to be corrected currently and in the future. Future researchers, engineers and health professionals can supplement the existing clinical information in the prediction of cancer results to further develop these areas of study.

Factors affecting Machine Learning Methods

Over the past decades, scientists and researchers have explored many ways to identify, control and cease cancer progression. Histological, clinical, and population-based data with the integration of characteristics like family history, age, diet, weight, risky habits, and exposure to environmental carcinogens have played a crucial role in prognosticating the development of cancer in recent years. However, if this type of large-scale information was related to a small number of variables, then statistical significance would be low (Kourou, et al., 2015). Recent studies have shown that combinations or models of multiple molecular biomarkers are even more significant than the individual tests or readings (tumour type, aspects, risk factors), and that cancer prognoses and predictions are becoming more reliable as a result of the combination of these molecular patterns with extensive clinical data (Cruz & Wishart, 2006). However, given the number of parameters, measurement is becoming increasingly important and with it the challenge of trying to make sense of all this information.

It is undeniably true that cancer is one of the leading causes of mortality around the world. Its disease progression and underlying causes have been an important issue in medicine. Cancers are sometimes hard to detect unless sudden severe symptoms are experienced that lead people to seek medical consultation. According to Kourou, et. al., (2015), application of specific machine learning techniques could enhance the accuracy of cancer vulnerability, recurrence, and prognostication of survival, which could enhance detection before symptoms become severe.

Machine learning methods have been studied and developed to determine cancer progressions that could help prevent terminally ill cancer patients through future diagnostics, preventative measures, and cures. Therefore, it is highly beneficial to use machine learning methods for cancer prediction and prognosis if larger data are available and thoroughly undergo a series of tests to be able to set better outcomes.

Impact of Machine Learning

The capability of machines to perform cognitive tasks based on the data they are provided to accomplish a particular goal is restructuring and transforming healthcare systems. With access to “big data,” cognitive computers can competently recognize intricate patterns while scanning billions of unstructured data bits (Goldenberg et al., 2019). Decision support systems based on computational machine learning (ML) have the potential to completely transform medicine through the improvement of diagnostic accuracy, productivity efficiency, clinical workflow, reduced personnel costs, and improved treatment options.

Machine learning also classifies cancer according to its risk: high or low, which can help researchers from biomedical or bioinformatics fields in the application of appropriate machine learning methods (Kourou et al., 2015). These characteristics have numerous expanding applications in diagnostic imaging, surgery, skill training and assessment, digital pathology, and genomics, which may be particularly helpful in the treatment of prostate cancer. Collaborative partnerships in different settings such as oncology, radiology, and pathology need to understand this evolving science and acknowledge that cooperation and participation between data scientists, computer scientists, and engineers is necessary to create exceptionally accurate AI-based decision support applications to reap the benefits of ML.

The overwhelming development and improvement in the health care world through ML greatly impacts the recognition of diseases like cancer and their progression. However, with the immense machine learning methods evolution, the worth of procedure was left in question. One of the challenges will be its cost effectiveness. If cost is a concern, affordability may be restricted to wealthy and affluent patients.

Some patients may be deprived of better health care if they are unable to pay the cost of the procedure and treatment. Machine learning allows individuals with cancer to have an opportunity for alternative medical intervention that suits their cancer diagnosis, and provides the facts in accordance with their percentages of success. However, ways to support equity in access and choice must be supported to harness the benefits of ML for all patients, regardless of socioeconomic status.

Barriers to Machine Learning

One of the many reasons why cancer is the leading cause of mortality is due to setbacks in diagnosing and shortfalls in determining accuracy in the detection process. A proficient and professional radiologist may experience errors with the use of machine learning classifiers as they may face various issues during the training, testing and validation phases of detecting cancer (Kumaraswamy, 2022). It is undeniably possible to reduce and withstand cancer mortality if detected and treated in early phases. While the use of ML methods can improve our understanding of cancer progression, an adequate level of validation is required to consider these methods in daily clinical practice (Kourou, et al., 2015). Naturally, it depends on how well a disease prognosis turns out. But a prognostic prediction should be as accurate as a medical diagnosis, considering factors other than just a single diagnostic option when negotiating care. When relating to cancer prognosis and prediction, three things can be predicted: (i) predictive prognosis and risk assessment (ii) the cancer prognostic; and (iii) the cancer recurrence prognostic. There are two areas researchers are focusing on, first is the possibility of acquiring a certain type of cancer and second, the feasibility of cancer remission and exacerbation. The foremost purpose of the two areas is predicting the survivability rate after trials of treatments.

Therefore, a thorough assessment, history taking, and family tracing for cancer helps in the approach to cancer prediction. Patients play a crucial role in the success of cancer prediction by ensuring all personal health information given is accurate and unbiased. Missed or inaccurate information may affect the effectiveness of machine learning methods. Moreover, healthcare workers, engineers, and scientists must have the appropriate training and knowledge in diagnosing cancer as patients depend on them for its prognosis and prediction.

Implication for Practice

Machine learning is important to medical advancement as this helps classify cancer tumours and understand how they invade and spread in the body, determine whether they are benign or malignant, and examine cancer cells based on tumour size and patient age. In practice, machine learning methods can support nurses’ role in more positive ways by helping nurses identify at-risk patients by accessing more detailed patient information in ML enhanced databases. The application of ML in routine could be more successful if nurses engage with it and take part in its development and deployment. Clinical advocacy could complement nursing advancement and improvement, effective delivery of care, and better patient outcomes. ML boosts nurses’ roles to adequately provide more care to cancer patients through patient advocacy and recognizing challenges and differences experienced by each individual patient dealing with cancer.

Limitations and Recommendations

One of the biggest limitations in this study was the small sample size (n=6) and the small samples used within these six studies. When modeling a disease with classification schemes, size matters. The training datasets must be sufficiently large. A relatively large dataset enables an adequate division into training and testing sets, resulting in the estimators’ validity being validated. A tiny, sizeable sample, when compared to data dimensionality, can lead to errors in the estimators’ classifications which could result in unstable and biased models. There’s no denying that they used a wealthier group of patients. The generalizability of the predictor can be improved by survival prediction model.

Furthermore, what was not discussed in this study was the current ML methods that are currently widely used. There have been few studies about this, and it may be beneficial if further research were conducted, published, and used to raise consciousness for future improvements in healthcare and the continuing education of medical and other health professionals.

Further ML research is important, if we want to get precise results through predictive models, we need to use independent features that might result in better validation. According to Narang, et al. (2023), ML studies involve both inside and outside validation that make it possible to extract predictions that are more accurate and trustworthy while minimizing bias. Except for data size and dataset quality, selection schemes are essential for effective machine learning, and as a result, for accurate cancer prediction. Better statistical processing of diversified datasets would yield more precise findings and would explain why certain diseases manifest themselves in a certain way.

An example is a substantial program that consolidates multiple sets of data on cancer. One of the sample programs is the DREAM project that uses the best crowd-source model, a decade of follow-up with thorough clinical information and genetic material expression, to assess consistency of the model. Although there are still limitations in the program, a need for further research and study is necessary.

Conclusion

In this review, machine learning (ML) concepts were described in the context of how they are used in the prognosis and prediction of cancer. Most of the studies that have been suggested in recent years have a development-focused approach, using supervised ML techniques and classification algorithms of predictive models with the goal of accurately forecasting the course of diseases. It is clear from an analysis of their findings that the combination of multidimensional heterogeneous data integration with the use of various feature selection and classification techniques can yield positive result tools for making inferences in cancer and improving care.

Author Biographies

Daphne Lyn Gigante Rivera, RN is a BSN Graduate of Our Lady of Fatima University, Valenzuela, and a Graduate student of St. Paul University Philippines. ORCHID: https://orcid.org/0000-0001-7265-7827

Roison Andro Narvaez, MSN RN is a BSN Graduate of Manila Doctors College, Clinical Case Manager of Ace Home Health and Hospice and a Graduate School Professor of St. Paul University Philippines. ORCHID: https://orcid.org/0000-0001-7555-542

References

Amrane, M., Oukid, S., Gagaoua, I., & Ensari, T. (2018). Breast cancer classification using machine learning. Electric Electronics, Computer Science, Biomedical Engineerings’ Meeting (EBBT), Istanbul, Turkey, 2018, pp. 1-4. https://doi.org/10.1109/EBBT.2018.8391453

Ayer, T., Alagoz, O., Chhatwal, J., Shavlik, J. W., Kahn Jr, C. E., & Burnside, E. S. (2010). Breast cancer risk estimation with artificial neural networks revisited: discrimination and calibration. Cancer, 116(14), 3310-3321. https://doi.org/10.1002/cncr.25081

Critical Appraisal Skills Programme (CASP) (2018) CASP Systematic Review Checklist. https://casp-uk.net/casp-tools-checklists/

Cruz, J. & Wishart, D. (2006). Applications of machine learning in cancer prediction and prognosis. Cancer Informatics, 2, 59–77.  https://doi.org/10.1177/117693510600200030

Garg, S. & Gupta, P. (2020). Breast cancer prediction using varying parameters of machine learning models. Procedia Computer Science, 171, 593-601. https://doi.org/10.1016/j.procs.2020.04.064

Gevaert, O., De Smet, F., Timmerman, D., Moreau, Y., & De Moor, B. (2006). Predicting the prognosis of breast cancer by integrating clinical and microarray data with Bayesian networks. Bioinformatics (Oxford, England), 22(14), e184–e190. https://doi.org/10.1093/bioinformatics/btl230

Goldenberg, S., Nir, G., & Salcudean, S. (2019). A New Era: Artificial intelligence and machine learning in prostate cancer. Nature Reviews Urology, 16(7), 391-403. https://doi.org/10.1038/s41585-019-0193-3

Islam, M., Haque, M., Iqbal, H., Hasan, M., Hasan, M., & Kabir, M. (2020). Breast cancer prediction: A comparative study using machine learning techniques. SN Computer Science, 1(5). pp. 1-14. https://doi.org/10.1007/s42979-020-00305-w

Kourou, K., Exarchos, T., Exarchos, K., Karamouzis, M. & Fotiadis, D. (2015). Machine learning applications in cancer prognosis and prediction. Computational and Structural Biotechnology Journal, 13, 8-17. https://doi.org/10.1016/j.csbj.2014.11.005

Kumar, Y., Gupta, S., Singla, R., & Hu, Y. (2022). A systematic review of artificial intelligence techniques in cancer prediction and diagnosis. Archives of Computational Methods in Engineering 29(4), 2043–2070. https://doi.org/10.1007/s11831-021-09648-w

Kumaraswamy, E. (2022). Key challenges in the diagnosis of cancer using artificial intelligence methods. AIP Conference Proceedings, 2418(1), 1–6. https://doi.org/10.1063/5.0081712

Melnyk, B.M. & Fineout-Overholt, E. (2011) Making the Case for Evidence-Based Practice. In B.M. Melnyk and E. Fineout-Overholt (Eds). Evidence-Based Practice in Nursing & Healthcare: A Guide to Best Practice. 2nd Edition, Lippincott, Williams & Wilkins, pp. 3-24.

Montazeri, M., Montazeri, M., Montazeri, M., & Beigzadeh, A. (2016). Machine learning models in breast cancer survival prediction. Technology and Health Care, 24(1), 31-42. https://doi.org/10.3233/THC-151071

Naji, M., El Filali, S., Aarika, K., Benlahmar, E., Abdelouhahid, R. & Debauche, O. (2021). Machine learning algorithms for breast cancer prediction and diagnosis. Procedia Computer Science, 191, 487-492. https://doi.org/10.1016/j.procs.2021.07.062

Narang, A., Kumar, R. and Dhiman, A. (2023). Machine learning applications to predict the axial compression capacity of concrete filled steel tubular columns: a systematic review. Multidiscipline Modeling in Materials and Structures, 19(2), 197-225. https://doi.org/10.1108/MMMS-09-2022-0195

National Cancer Institute (2021). What is Cancer?  https://www.cancer.gov/about-cancer/understanding/what-is-cancer

Osareh, A., & Shadgar, B. (2010). Machine learning techniques to diagnose breast cancer.  2010 5th International Symposium on Health Informatics and Bioinformatics, Ankara, Turkey, 2010, pp. 114-120. https://doi.org/10.1109/HIBIT.2010.5478895

Park, C., Ahn, J., Kim, H., & Park, S. (2014). Integrative Gene Network construction to analyze cancer recurrence using semi-supervised learning. PLoS ONE, 9(1), 1–9. https://doi.org/10.1371/journal.pone.0086309

Park, K., Ali, A., Kim, D., An, Y., Kim, M., & Shin, H. (2013). Robust predictive model for evaluating            breast cancer survivability. Engineering Applications of Artificial Intelligence, 26(9), 2194-2205. https://doi.org/10.1016/j.engappai.2013.06.013

Rosado, P., Lequerica-Fernández, P., Villallaín, L., Peña, I., Sanchez-Lasheras, F., & De Vicente, J. C. (2013). Survival model in oral squamous cell carcinoma based on clinicopathological parameters, molecular markers, and support vector machines. Expert Systems with Applications, 40(12), 4770-4776. https://doi.org/10.1016/j.eswa.2013.02.032

Siegel, R., Miller, K., Fuchs, H., & Jemal, A. (2021). Cancer Statistics, 2021. CA: A Cancer Journal for Clinicians, 71(1). 7-33. https://doi.org/10.3322/caac.21654

Stojadinovic, A., Nissan, A., Eberhardt, J., Chua, T. C., Pelz, J. O. W., & Esquivel, J. (2011). Development of a Bayesian Belief Network Model for personalized prognostic risk assessment in colon carcinomatosis. The American Surgeon, 77(2), 221–230. doi:10.1177/000313481107700225

Weir, H., Thompson, T., Soman, A., Møller, B. & Leadbetter, S. (2015). The Past, Present, and Future of Cancer Incidence in the United States: 1975 through 2020. Cancer (0008543X), 121(11), 1827–1837. https://doi.org/10.1002/cncr.29258

Whittemore, R., & Knafl, K. (2005). The integrative review: Updated methodology. Journal of Advanced Nursing, 52(5), 546–553. https://doi.org/10.1111/j.1365-2648.2005.03621.