Health professionals’ attitudes toward religiosity and spirituality: a NERSH Data Pool based on 23 surveys from six continents

Systematic searches

Eligible surveys were found using a combination of a network search (the NERSH collaboration), and both a citation search and a systematic literature review.

The network search utilized the global collaboration of researchers in NERSH. Past, on-going or planned surveys using either the Religion and Spirituality in Medicine, Perspectives of Physicians (RSMPP) questionnaire¹⁴ or the NERSH Questionnaire are the topic of frequent correspondence between collaborators. The NERSH Questionnaire being basically RSMPP with several additions including support for the DUREL index. Also, several research groups have joined the collaboration with a priority to share their survey data once collected and published locally. Local restrictions may apply so that survey data are not released to be included into the data pool until certain criteria are met, why included data are not necessarily added in chronological order.

Citation and literature searches were performed by the first author in the period of January to February 2020. For the citation search we looked up citations in Web of Science referencing eight articles on the original RSMPP-survey published by Curlin in the period 2005 to 2008.^14-21 The eight articles by Curlin used in the citation search were selected by Hvidt and Kørup in 2016 for the first systematic search. It was important for us to use the exact same approach in this study. Articles were chosen because they all rest on the RSMPP and were the first articles published by Curlin in the years after the RSMPP was published. Looking for publications using the RSMPP we judged it unlikely that at least one of these articles would not be referenced. All citing articles were reviewed on abstract level, and if the data source was the RSMPP, or the data source was unclear based on the title or abstract, the entire article was screened. To ensure we found all surveys based on the RSMPP, and to also find potential surveys based on the more recent NERSH Questionnaire, our search strategy also included a literature search in Google Scholar, Web of Science, Embase, Medline and PsychInfo using the search strings in Table 1. Survey data already included in the data pool were ignored. All searches were limited to English productions, and as the previous version of the data pool was based on the same search strategy in 2016, we only assessed publications from the year 2016 and forth.

Inclusion and exclusion criteria

We only included data on health professionals based on either of the two questionnaires: Religion and Spirituality in Medicine, Perspectives of Physicians (RSMPP)¹⁴ or NERSH Questionnaire,¹¹ the latter being a further development of the first. Customized versions were accepted if they were mainly the RSMPP or NERSH Questionnaire. Exclusion criteria were: Observations that were missing information on gender or only contained empty answers were excluded. Also, We also excluded observations who had entered we set a minimum an age of less than 18 years. 

Building the NERSH Data Pool

The data pool versions were created consecutively upon earlier releases. Thus, only new datasets not already included in the 2nd version were added to the 3rd version. All importations were based on original raw datasets sent to us by the local researchers. Data were sent to us in various data formats (Stata, RRID:SCR_012763; IBM SPSS, RRID:SCR_002865; or Microsoft Excel, RRID:SCR_016137), and all were converted into Stata datasets (.dta) before import. For an open-access alternative to the statistical software the R Project for Statistical Computing (RRID:SCR_001905) can perform equivalent analyses. In case interpretation of the raw data was not straightforward, or vital information could not be extracted from published articles, the relevant researchers were contacted by e-mail or phone until the issue was resolved.

The data pool was created to comprise a total of 98 variables, of which 76 stem from the RSMPP, two are part of the DUREL-index not included (in complete form) in the RSMPP, and the remaining 20 variables are calculated variables aiding study categorization and usage of the included scales.

A codebook was created documenting available original and calculated variables and scales.²²

Data security and ethical considerations

All observations in the data pool have been anonymized. The data pool is physically located in a secure server environment in Odense, Region of Southern Denmark, and containing only anonymized data upholds the latest security requirements of the General Data Protection Regulation (GDPR) of the European Union. The project was approved by the University of Southern Denmark Research & Innovation Organization (reg.nr: 10.312).

Systematic searches

For the network search, knowledge about past or on-going research projects within the NERSH group were assessed. We knew from an earlier query that Baumann and Lee were in possession of survey data from 138 German chaplains using the NERSH Questionnaire from 2012 to 2014, and that their data had now been approved to be included into the NERSH Data Pool.²³ Also, a Swiss data sample of 104 general practitioners from a survey by Hefti et al. in 2015 was now available.²⁴ The Swiss survey was published in German language and would thus not have been found by the citation or literature searches.

The eight citation searches performed found a total of 763 hits including many duplicates. References were screened by the first author who identified three eligible surveys that were based on parts of the original RSMPP. A survey by Cordero et al. from Seville, Spain, examined 75 graduate students (nurses, podiatrists and physiotherapists) using the RSMPP and the DUREL index in 2017,²⁵ and the same research group also examined Portuguese nursing students in 2016 using the same questionnaires.²⁶ In addition, Menegatti-Chequini et al. in performed two surveys of psychiatrists in 2013–2014, a local facility sample in São Paulo (N = 84), and a nation-wide sample among members of the Brazilian Psychiatric Association (N=508) using a questionnaire based on the RSMPP.^27,28 Details of the citation search hits are found in Table 1.

The literature search resulted in a total of 4,929 hits. Restricting the search to articles published 2016 or later reduced the count to 1,133 hits. 20 articles were retrieved in full length but did not prove eligible. In summary, the literature search did not find any eligible surveys not already found by the network and citation searches.

In total, six new surveys were eligible to import into the NERSH Data Pool. The researchers were contacted and invited to submit their original data, and all agreed with written confirmation. The surveys varied from surveys sampled across an entire nation to surveys done at a single facility or hospital. The study with Spanish nursing students (Cordero, 2019) was sampled among graduate students from The Faculty of Nursing, Podiatry and Physiotherapy in Seville.²⁵ By the same research group, the Portuguese nursing students (Cordero, 2018) were 3^rd- and 4^th-year students from the School of Health of University of Algarve and School of Health of Polytechnic Institute of Santar.²⁶ The data on German hospital chaplains were collected in a nation-wide survey by Lee et al.,²³ while the Swiss sample of General practitioners was a random sample from the region of Bale and Aarau by Hefti et al.²⁴ From Menegatti-Chequini et al. we received two samples, the first based on a survey sent to members of the Brazilian Psychiatry Association (ABP) and an additional sample from a single psychiatric department Instituto de Psiquiatria, Hospital das Clínicas, Faculdade de Medicina, Universidade de São Paulo (IPqHC-FMUSP). Duplicates from the ABP and IPqHC-FMUSP samples were removed.

Demographics

After applying exclusion criteria, a total of 1,068 unique observations were added to the NERSH Data Pool. The entire 3^rd version of the data pool ended up comprising 7,323 observations, including 4,070 females and 3,253 males. Mean age (SD) was 41.4 (12.5). Looking at the entire data pool response rates ranged from 18% (116 responses out of 642 questionnaires sent in New Zealand with no possibility for follow-up on non-responders) to 95% (Brazil) and 99% (Indonesia), the latter two secured due to tight follow-up including personal meetings and encouragements to complete the forms. Crude response rate was 54.7% for all currently included studies.

After inclusion of the found datasets the complete data pool comprised 4,872 physicians, 1,319 nurses, 286 midwives and other HP occupations in smaller numbers (see Table 2 for details). Medical specialties were categorized by major specialty groups (Table 3). The largest group of specialties were gynecology/obstetrics with 1,788 participants mainly from the German sample from perinatal care professionals, followed by 1,591 working in psychiatry, 953 from Internal Medicine, 842 from general practice, 447 from surgical specialties, 236 within pediatric medicine, and 143 from paraclinical specialties (i.e. laboratory sciences and supportive branches of medicine not directly involved patient care).

Almost all health professionals supplied information about their religious affiliation (N = 7,158; 97.7%). If the responder did not want to supply this information the answer was treated as missing. All questionnaires had a “No affiliation” option, and some also included the options “None”, “Atheism” and/or “Agnosticism”. These options were all grouped together based on their common denominator of not being affiliated with a religion. Answers of religious affiliation were categorized in groups of the major faiths: Buddhism, Hinduism, Judaism, Mormonism, Islam and Christianity. Christian denomination (i.e. Orthodoxy, Catholicism or Protestantism) was registered for samples providing this information. Answers that did not fit any of these groups were placed in the “Other” group.

Religious affiliations from predominantly Islamic cultures were almost entirely Muslim (Saudi Arabia 100% and Indonesia 86%), whereas Hinduism was predominant in the Indian sample, N = 195; 71%.

Looking at the entire data pool the largest group was Christian denominations (N = 4,189; 59%), represented in all samples apart from the sample from Saudi Arabia. The second largest was the group of responders that declared themselves not affiliated with a religion, atheist or agnostic (N = 1,529; 21%). Muslims comprised the third largest group in the data pool (N = 504; 7%) (Table 4).

Influence of R/S on patient health

Responders were asked about the potential influence of R/S on patient health in general. Across all samples 3,700 physicians answered this question, of which 1,767 (48%, CI95% 46.4%–49.6%) replied that R/S has “Much” or “Very much” influence on patient health (Figure 1). The number of physicians believing that R/S has at least “Some” influence on patient health was 3,078 (83%, 81.8%–84.2%). In comparison, for 1,020 nurses these proportions were 658 (65%, 62.1%–67.9%) and 955 (94%, 92.5%–95.5%) respectively. Due to the large between-group heterogeneity, statistical significance test of differences was not feasible.

Figure 1. Physicians' and nurses' opinion about the influence of R/S on patient health.

Physicians, N = 3,700. Nurses, N = 1,020.

R/S Training of HPs

Across all samples 649 (%, CI95%) (16%, 14.9%–17.1%) physicians reported to have undergone any formal training regarding R/S in medicine with the lowest proportion was 4% (Lucchetti, 2016) and the highest 28% (Ramakrishnan, India, 2014). For nurses this amounted to 264 (23%, 20.6%–25.4%) ranging from 18% (Schouten-Wermuth, 2016) to 35% (Hvidt-Frick, 2014) (Figure 2).

Figure 2. Physicians' and nurses' formal training in R/S.

Physicians, N = 3,987. Nurses, N = 1,158.

Data pool characteristics

Using both network, citation and systematic literature searches we found six new survey samples based on the RSMPP or NERSH Questionnaire, which were added to the preexisting second version of the data pool, now released as the NERSH Data Pool 3.0. Here we have presented the demographics of the complete data pool and select variables regarding attitudes toward influence of R/S on patient health, and whether responders had any formal training in R/S in medicine.

The included samples were collected in very distinct locations, some were collected from national organizations, and others were collected at single facilities. The cultures and religious landscapes of the represented countries differ, and participants' attitudes are likely affected thereof. Some samples include only HPs of a specific profession (i.e. only physicians, nurses or chaplains), where others included a variety of health professions.

The heterogeneity between studies was expected as part of the design and is the reason why it did not make sense to test between-sample differences of descriptive variables. As expected, when looking at attitudes toward potential influence of R/S on patient health, we see large variations between studies, and based on the 95% confidence intervals nurses more often reported this influence on patient health than the physicians did. This is in line with earlier research on differences in nurses' and physicians' attitudes toward R/S,^29,30 although we must underline that the crude descriptive findings reported here are at study-level and not controlled for within-study confounders. This was beyond the scope of this descriptive analysis of the data pool.

Also, the reported R/S training of the physicians and nurses in the data pool varied largely. Again, we read two important points out of this. 1) Physician and nurses are largely under-educated when it comes to handling R/S in the clinical setting (i.e. 16% and 23% respectively had some education, leaving approximately four out of five physicians and nurses as having no formal education in this matter). We have no information about the degree or length of educations that were reported. 2) Large differences between sample may signify how the focus on R/S education local is influenced by local culture.

MAs and IPDMAs

Traditional meta-analyses (MAs) are the recommended approach when comparing results of several studies. MAs examining HPs’ attitudes in clinical practice are still scarce, and IPDMAs even more so.

In this study we have reported how we have built a pool of individual HP survey responses that enables us to perform IPDMA. Already a single recent IPDMA study compared the religiosity of physicians from seven countries, and their self-reported influence of their religiosity on their clinical practice. Religiosity and influence of religious beliefs were most pronounced in India, Indonesia, and a European faith-based hospital. Half (50%) of the physicians examined reported being influenced that their work as a physician was influenced by their own religiosity.³¹ Using individual participant data, the authors were able to conduct a sensitivity analysis of potential confounders at sample level, and thus demonstrated the potential of this data pool.

Koenig et al. recently expressed detailed concerns about the biases introduced when using meta-analyses within R/S research.³ They describe R/S and mental health research as a social science where meta-analyses should only be used to describe heterogeneity and not as much searching for consistency and generalizability of study findings across populations. Koenig et al. highlights several common critique points of MAs including when studies measure different variables, incomplete or unstandardized results, inability to account for inter-study variation, heterogeneity due to broad inclusion criteria without the possibility to limit the analysis to sub-groups within the samples, and susceptibility to the ecological fallacy. All these critique points are mitigated by the improved IPDMA-design, which we enable when collecting individual participant data.³²

Meta-analyses are considered one of the gold standards behind evidence-based health care, and the number of published meta-analyses have increased markedly over the last decades although average study quality has been questioned.³³ One of the caveats is that when collections of samples are very heterogeneous, comparisons of variables using a traditional meta-analysis design are likely to lead to biased results if within-sample confounders are not controlled for at the individual-level (i.e. ecological fallacy).³⁴ This fallacy is suspected to bias interpretations of meta-analyses where individual participant data were not available to the researchers, who thus had to rely on simple aggregated effect measures of the included studies (i.e. at study-level). Sometimes effect measures even differ between the included studies, hence they must be converted to a common effect measure by meta-analysis researchers. This introduces yet another step in the meta-analysis and thus a risk for bias/error.

Koenig et al. also problematized the risks introduced when study results are reduced to a single value.³ For the social and health sciences this is however a key concept that has driven scientific research to where it is today. This is not saying that this reductionism does not include biases and caveats that researchers must understand and respect in their interpretation of their results. Especially within psychometrics (i.e. regarding R/S and mental health) measurements may lack the validity and reliability compared to measurements from more objective sciences. Still limited by these known biases, psychometrics have not only made quantifiable scientific comparisons possible within mental health research, but have also generated a wealth of crucial clinical instruments positively affecting lives and disease courses of patients, some examples are scales for measuring severity of mental illnesses during treatment like Hamilton Depression Rating Scale,³⁵ Global Assessment of Functioning,³⁶ IQ tests like WAIS-VI³⁷ and more. Measuring aspects of R/S is notoriously difficult, mostly due to the personal and subjective dimensions that are natural parts of R/S experiences. Challenges caused by imprecise measurements, and/or attempts to compare results from studies using different or custom instruments (i.e. comparing apples and oranges) does not implicitly negate those instruments, but rather demonstrate a lack of collaboration amongst researchers of this field.

We thus believe, that the use of an instrument, or scientific method like meta-analyses, that over time has demonstrated its ability to advance health care and/or enrich the research thereof, should not be discontinued because of its imprecisions before another instrument or scientific method with improved characteristics are suggested to take its place.

Assimilating research designs and measurements

It has been argued that MAs are so complex that mistakes are inevitable.³³ We truly acknowledge the difficulty in conducting MAs, and most likely not a single MA is 100% perfect. Still, this is not an argument against using this method, because the argument of embedded error is applicable to practically most research methods including both quantitative and qualitative methods. Planning and deciding on research designs, deciding how to collect and filter the data, how strictly to enforce an interview guide, the handling of missing answers/observations, the weighing of pros and cons of different statistical strategies for analysis, choosing which statistics to report and how to interpret them. These are just some examples where research projects are susceptible to subjective decisions by human researchers that are prone to make mistakes and bad judgements. Rather than giving up on these methods we believe this calls for our continuing focus on our own human biases, and also for the need to systematize research methods using internationally recommended gold standards (like PRISMA) that precisely aims to document and limit these errors.³⁸ Also, instead of limiting research on R/S and health to theological and qualitative methods, we should welcome diverse and mixed research strategies, all adding valuable perspectives on a theoretical common consensus among researchers. This way we will continually equip ourselves with the latest and best instruments, with which we will attempt to prove or disprove the current hypotheses about how R/S is related to patient health and health care in general. Only this way can we make reason for decision makers to level the importance of R/S in health care with other central health topics crucial for patient health and well-being.

Like any other scientific research design, IPDMA has weaknesses, the largest being heterogeneity, which is largely caused by differences in sampling (location, culture, profession and sampling method), and differences in subjective judgements in the local samples.

Still, pooling survey data in the NERSH Data Pool will enable us to perform meta-analyses using the individual data of survey participants. In summary IPDMA carries both statistical and clinical advantages over regular traditional MA.^32,39 Some advantages include: 1) the ability to utilize data from yet unpublished studies and or outcome measures thus reducing publication bias; 2) standardized statistical analysis across studies; 3) ability to perform analysis on sub-groups participants (i.e. certain religious affiliations, occupations or medical specialties); 4) ensure consistent inclusion and exclusion criteria; 5) standardized handling of missing values across samples; and 6) overcoming the ecological fallacy of traditional meta-analysis by enabling analysis at the individual level.³⁴

Future work and clinical implications

In the near future, we will use the described data pool to test and retest hypotheses about R/S in medicine using IPDMA designs. Some planned analyses are: 1) the association of HPs’ attitudes and self-reported behavior regarding R/S in the clinical encounter; 2) R/S characteristics and attitudes of physicians from different medical specialties; and 3) Attitudes of HPs considering controversial ethical situations in healthcare.

We believe this work will better our understanding of how HPs values work in clinical practice, and aiding the development of R/S curricula that will help HPs learn how to incorporate spiritual care into their treatment of patients from any culture, and despite potentially differing world-views.

Extended data

Open Science Framework: Extended data for ‘Health professionals’ attitudes toward religiosity and spirituality: a NERSH Data Pool based on 23 surveys from six continents’. http://doi.org/10.17605/OSF.IO/J79PT.¹⁹

This project contains the following extended data:

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).