Keywords
Artificial Intelligence, Digital Health, Healthcare Information Systems, Automated Diagnoses.
This article is included in the Artificial Intelligence and Machine Learning gateway.
Artificial Intelligence, Digital Health, Healthcare Information Systems, Automated Diagnoses.
The development of Artificial Intelligence (AI) has been very fast in the past decade. AI and robotics are now utilized in almost all business sectors and this will impact the global labor market in the near future1. The following are AI services that have gained the most publicity: robotics, autonomous cars, computer vision applications, and natural language processing services. However, AI is already in use or is emerging on a wide scale in different business sectors2.
When considering new models for AI utilization in healthcare processes, it is generally believed that soon AI will automatically give accurate diagnoses, provide and invent new care instructions and new treatments for diseases, and will act as human healthcare professionals but in a much more accurate, precise, and effective way. In this scenario, AI could provide a remarkable advantage to enhance health outcomes and healthcare processes, as well as help caregivers in their demanding work. For example, AI can provide help regarding the shortage of the healthcare workforce and can also provide healthcare resources for aging populations3. In addition, in situations when the global healthcare system is burdened, such as the COVID-19 pandemic, healthcare resources have an increased workload and there is strong demand for AI solutions, which will provide governments and healthcare organizations with the necessary tools to perform actions faster and more efficiently, especially in pandemics4. These are the final goals of utilizing AI in healthcare, but all of these are currently not achievable at the time of writing this article.
To achieve these goals, we need to study healthcare professionals’ viewpoints concerning current utilization of AI in healthcare and ask professional end-users (clinicians and administrators) of healthcare services about the best use cases where AI could benefit different healthcare processes. Moreover, before we can take full advantage of AI in healthcare, we must tackle ethical and legal obstacles and comply with all regulatory requirements5. Furthermore, metrics to measure benefits and better care outcomes for patients, better care outcomes for healthcare service providers, effectiveness of service, and response time for service queries, must be constructed.
In this study, we provide the results of a survey conducted to healthcare professionals about use of AI in healthcare practices. Based on the results we provide some evaluation of AI features that could be widely accepted by healthcare professionals. Moreover, we present new ideas that will be used in our future research work and insights for a novel upcoming healthcare AI-service platform. We hope that this novel AI-based healthcare solutions platform will be able to support healthcare service providers, caregivers, and customers in a more effective way compared with traditional healthcare IT services.
The aim of the survey was to evaluate utilization and acceptance of different AI-based services of healthcare professionals working in the healthcare sector. We also provide a comparative analysis of our survey and an earlier survey6, where the adoption of AI services by healthcare professionals and public users were studied.
The main areas of AI-based healthcare are clinical decision making, healthcare interventions, automatic care or healthcare process recommendations, patient administration, and patient monitoring. For this study, we searched the literature for research and utilization of AI technology in the healthcare sector. Based on our search, we selected healthcare services that included AI technologies in natural language processing, neural networks and deep learning, computer vision or robotics, rule-based expert systems, and physical robots7,8. This selection was made on the basis that there are many studies concerning these AI technologies and they are commonly in use in the healthcare sector. Therefore, healthcare services utilizing AI that were used as examples in our survey are as follows: robot-assisted surgery, virtual assistants for nursing and consultation, assistant for administration and workflow, fraud detection, dosage error reduction, connected machines, clinical trial participation, preliminary diagnosis, image diagnosis, cybersecurity, medication management, health monitoring, and drug creation.
The survey was designed to analyze the most suitable services and features to support healthcare professionals and administrators in their work, as well as which features and services could provide best care outcomes for the patients. The survey contained three parts: part 1 asked for demographic characteristics of respondents; part 2 asked general questions concerning technology and AI; and part 3 asked about an individual’s acceptance of AI technology and its advantages for certain use cases. Question types were single selection, multi-selection, open number value, and 5-step Likert Scale9. The survey can be found in full in the Extended data.
The survey was created using Google Forms. A link to the survey was sent to healthcare professionals by informing healthcare professionals in our own network in Finland via Facebook and using SurveyCircle scientific survey service10. No individual requests to complete the survey were sent. The survey was accessible from December 19, 2019 to February 25, 2020.
The survey was designed to provide data in a formal structured manner to support our analysis in an optimal way and enable us to smoothly continue with our practical implementation efforts of future research work ideas.
We used basic statistical analysis tools from Excel Analysis ToolPak11, presenting mean values and standard deviation for each question, and analyzed respondents’ staff group, work experience, and age against AI technology adoption. We analyzed how respondents utilize IT technology in their work, as well as their willingness to adopt new healthcare services that use AI technology. Finally, we compared our survey results to the survey results conducted by YouGov Research in November 20166 to the general public in 12 countries across Europe, the Middle East, and Africa (n=12000). Our aim was to find out if our survey results correlate with the survey results from YouGov Research.
Ethical approval for the study protocol and survey was waived by the University of Eastern Finland Research Ethics Committee. This waiver of approval was provided since the study did not contain any of the following: research deviates from principle of informed consent; research involves intervening in the physical integrity of participants; research is on minors under the age of 15 years; research exposes participants to strong stimuli; research involves risk of causing mental harm to participants; and research involves threat of safety to participants.
Participants were informed about the aim of the questionnaire and that this information would be used in research, which was shown at the beginning of the questionnaire. Completion of the questionnaire was taken as consent to participate.
We created an Internet survey to examine the usage, usability, and end-user acceptance of AI-based healthcare services by healthcare professionals. We received a total of 125 responses. Of these, only 121 responses were 100% complete and therefore only these responses were analyzed.
The study population gender distribution was female 63.6% (77), male 35.6% (43), and other 0.8% (1). Mean age and mean work experience was not able to be calculated accurately because of the age and experience questions did not contain exact age or years of experience. Respondents staff group distribution was as follows: nursing staff, 48.8% (59); office and administration staff, 14.9% (18); research staff, 14.0% (17); physicians, 14% (17); social work, 5.0% (6); and maintenance staff, 3.3% (4). The demographics of the study population are shown in Figure 1.
The survey measured healthcare professionals’ opinions about how AI technology will be able to provide enhancement to care processes and functions, and if AI can replace some activities or tasks that are traditionally done by healthcare professionals or traditional healthcare ICT services. The survey results were analyzed by each question and with basic statistical analysis tools (mean values and standard deviation).
In the survey we asked about the current use of different digital services and features among healthcare professionals (question G1). When analyzing the results, we noticed that respondents stated that the most currently used ICT services listed in the survey were for drug discovery (31.4%), medical data security (28.1%), medical diagnosis (24%), signal analysis (23.1%), and finding participants to take part in clinical trials (23.1%). The least used services were robot surgery (6.6%) and fraud detection (9.1%). In total, 19% of professionals did not use any of the listed digital services in their work (Figure 2).
For the utilization of these ICT services, 19% of respondents did not use any of the services listed in G1. For the technology usage level (question G3), we found that 20% of the respondents used these services daily and 24% at least once per week (Figure 3).
We also asked health professionals which healthcare ICT services and features were considered the most beneficial in their work and/or care processes (question G2). There was the possibility to select multiple technologies. The respondents reported that the following were the most beneficial/used: health monitoring (43.8%), medication management (32.2%), connected machines (27.3%), dosage error reduction (23.1%), and signal analysis (18.2%). The least beneficial services for healthcare professionals were cybersecurity (8.3%) and fraud detection (9.1%). In total, 10.7% of respondents do not want to use any of the listed services in their work (Figure 4).
One of the questions in the survey was for healthcare professionals to consider the advantages of AI-based services (question G4; Figure 5). The response was a 4-step Likert scale from 0 to 4, where 0 is “Do not agree at all” and 4 is “I fully agree”. The results showed that three most considered advantages of AI by healthcare professionals were “Simple and repeating tasks can be performed by AI/robots” (mean score 3.0), “AI can provide help in healthcare professionals workload” (mean score 3.0), and “AI can provide easier access for more people/patients to healthcare services” (mean score 3.0). These services have also lowest standard deviation (0.86 – 0.95). Other lesser considered advantages were “AI can give better treatment recommendations” (mean score 2.3) and “AI can perform surgery more accurately than healthcare professionals” (mean score 2.3). When analyzing the average scores from all replies there is an indication that healthcare professionals mostly consider AI-based services and features in healthcare to have advantages.
Likert scale from 0 to 4, where 0 is “Do not agree at all” and 4 is “I fully agree”.
We also asked what disadvantages healthcare professionals consider AI to have when used in healthcare (question G5, Likert scale as for question G4; Figure 6). The highest considered disadvantage was given for “AI or robots cannot be in contact with patients. Patients need human interaction when providing healthcare services.” (mean score 2.8), “Too complicated for healthcare professionals to access AI technology solutions” (mean score 2.1) and “AI does not provide accurate results for decision making. Only healthcare professionals can make right decisions.” (mean score 2.1) The lowest value was given to “I can see only disadvantages when using AI or robotics in healthcare.” (mean score 1.2). The results of this question show that some healthcare professionals express their concern that AI should not have contact with patients without human interaction and the final decision in healthcare processes should be kept in human hands. We can also see that healthcare professionals consider that AI-based solutions bring advantages rather than disadvantages.
We asked healthcare professionals about willingness to use healthcare services that utilize AI (questions A1-A15). Respondents could evaluate each service independently with a 4-step Likert scale between 0 to 4, with 0 as “Not willing” and 4 as “Very willing”. A mean utilization score of 2.8 was given to services used for automatic health monitoring, providing enhanced data security, providing fraud detection, providing medical risk prediction, and conducting health assessment questionnaires. These services also have the lowest standard deviation (0.80 – 0.90). Lowest utilization scores were given to services such as AI-assisted major surgery (mean score 1.9), AI giving diagnosis of patient (2.3), and AI providing instant healthcare advice for the patient (2.3). Overall results are shown in Figure 7.
Likert scale between 0 to 4, with 0 as “Not willing” and 4 as “Very willing”.
Finally, we asked healthcare professionals how they are to have AI make decisions or give recommendations in healthcare processes (question G6). The responses showed that most healthcare professionals are willing to use AI in healthcare in all possible cases with the final decision made by humans (65%) or AI to give decisions without human interaction (3%). Also 29% of respondents consider that AI can ease healthcare work and processes, but only by giving recommendations. Only 3% of respondents consider that AI is not suitable for healthcare.
We analyzed how respondents’ staff group and age group correlate to the willingness to use AI technology. Social work and maintenance staff had a total 10 respondents each and were therefore combined. We found that there was only minor deviation among different staff groups supporting AI utilization (Figure 8).
Blue, “AI can be used in all possible cases, but humans should always make final decision”; orange, “AI can ease clinical work and clinical processes but only with restricted matter (only by giving recommendations)”; black, “AI is not suitable for healthcare”; grey, “AI can be used in healthcare and can independently provide diagnoses and decisions for care”.
Secondly, we analyzed age groups which have over 15 respondents. Age groups (55–64 years and over 65 years) had total 9 respondents and those were combined. We found that there was only minor deviation among different age groups supporting AI utilization (Figure 9).
Blue, “AI can be used in all possible cases, but humans should always make final decision”; orange, “AI can ease clinical work and clinical processes but only with restricted matter (only by giving recommendations)”; black, “AI is not suitable for healthcare”; grey, “AI can be used in healthcare and can independently provide diagnoses and decisions for care”.
Thirdly, we analyzed work experience from the groups that have over 15 respondents. Work experience groups (55 – 64 years and over 65 years) had total 9 respondents and those are combined to last pie chart. We found that there was only minor deviation among different work experience groups supporting AI utilization (Figure 10).
Blue, “AI can be used in all possible cases, but humans should always make final decision”; orange, “AI can ease clinical work and clinical processes but only with restricted matter (only by giving recommendations)”; black, “AI is not suitable for healthcare”; grey, “AI can be used in healthcare and can independently provide diagnoses and decisions for care”.
From the survey results, we found out that the AI technology that respondents consider to be most useful for healthcare work processes (question G2) are services that provide health monitoring, medication management, and networked devices and databases which can exchange data between each other. In addition, the three most currently used healthcare IT services by respondents are drug discovery, medical data security, and medical diagnosis (question G1).
We compared our research results to an earlier survey conducted by YouGov Research6. We found similarities in perceived advantages of AI between the two surveys. In the YouGov survey, the top two advantages that respondents reported were “Healthcare would be easier and quicker for more people to access” and “Advanced computers/robots can make diagnosis faster and more accurately”. From our survey, the top three perceived advantages identified by healthcare professionals were “AI can provide easier access for more people/patients to healthcare services”, “AI can provide help in healthcare professionals workload”, and “Simple and repeating tasks can be performed by AI/robots”.
When we compared disadvantages between surveys, we found that the top three disadvantages reported in our survey were “AI or robots cannot be in contact with patients. Patients need human interaction when providing healthcare services”, “Too complicated for healthcare professionals to access AI technology solutions”, and “AI does not provide accurate results for decision making. Only healthcare professionals can make right decisions”. The disadvantages reported by the YouGov survey were “People need the human touch when it comes to their healthcare”, “If something unexpected is found during surgery or in a test, I do not trust robots to make decisions on what to do”, and “Only a human healthcare professional can make the right decisions”. Therefore, a correlation in perceived disadvantages by both groups of respondents can be seen.
When analyzing the demographic data of respondents and survey responses, we concluded that age and staff group among healthcare professionals does not have significant effect on adoption of AI features or willingness to use healthcare services utilizing AI methods. There was also an indication from the survey results that respondents of a younger age and shorter work experience in the healthcare sector correlated negatively with willingness to adopt AI technology in healthcare.
In addition, we found that willingness to use novel features and services utilizing AI by healthcare professionals is high when novel features are used to help healthcare professionals in their decision making or work processes. Perceived disadvantages when using AI services mainly concerned loss of human interaction when robotics is used in healthcare and unwillingness to let AI make clinical decisions.
Overall, we conclude that using AI in healthcare is accepted by healthcare professionals and there are many AI enabled features and services where there could be of high potential in healthcare. However, developers and researchers of these novel AI services must keep in mind that the perceived disadvantages should be solved and human decision making and human interaction in these novel services is necessary, until these new services gain user acceptance and adoption in the healthcare sector. From the YouGov survey, these same acceptance criteria also apply when considering healthcare AI service adoption within the general public.
In our future research work we will evaluate more results of this survey to find the most effective combination of AI-services and technologies that could provide the best benefits for healthcare work processes, health outcomes, and preventive health processes. We will focus on providing services that could be easily adopted by healthcare professionals and the general public, and that should provide imminent enhancement in care processes. Moreover, we will provide a comparative study of healthcare technologies where AI is used as a key element and evaluate the effectiveness of these technologies in the field of healthcare. Finally, we will propose a technology platform where these studied AI technologies could be easily adopted by healthcare professionals and their customers or patients.
Open Scientific Framework: Survey to healthcare professionals on the practicality of AI services for healthcare, https://doi.org/10.17605/OSF.IO/674ZM12.
Open Scientific Framework: Survey to healthcare professionals on the practicality of AI services for healthcare, https://doi.org/10.17605/OSF.IO/674ZM12.
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).
Views | Downloads | |
---|---|---|
F1000Research | - | - |
PubMed Central
Data from PMC are received and updated monthly.
|
- | - |
Is the work clearly and accurately presented and does it cite the current literature?
Partly
Is the study design appropriate and is the work technically sound?
Partly
Are sufficient details of methods and analysis provided to allow replication by others?
Partly
If applicable, is the statistical analysis and its interpretation appropriate?
Not applicable
Are all the source data underlying the results available to ensure full reproducibility?
Yes
Are the conclusions drawn adequately supported by the results?
Partly
References
1. Yu KH, Beam AL, Kohane IS: Artificial intelligence in healthcare.Nat Biomed Eng. 2 (10): 719-731 PubMed Abstract | Publisher Full TextCompeting Interests: No competing interests were disclosed.
Reviewer Expertise: Artificial intelligence, Medical imaging, Radiology, Imaging informatics
Is the work clearly and accurately presented and does it cite the current literature?
Partly
Is the study design appropriate and is the work technically sound?
No
Are sufficient details of methods and analysis provided to allow replication by others?
No
If applicable, is the statistical analysis and its interpretation appropriate?
I cannot comment. A qualified statistician is required.
Are all the source data underlying the results available to ensure full reproducibility?
Partly
Are the conclusions drawn adequately supported by the results?
No
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Philosophy of medicine, informatics, future of the health professions
Is the work clearly and accurately presented and does it cite the current literature?
Yes
Is the study design appropriate and is the work technically sound?
Yes
Are sufficient details of methods and analysis provided to allow replication by others?
Yes
If applicable, is the statistical analysis and its interpretation appropriate?
Partly
Are all the source data underlying the results available to ensure full reproducibility?
Yes
Are the conclusions drawn adequately supported by the results?
Partly
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Artificial Intelligence, eHealth
Alongside their report, reviewers assign a status to the article:
Invited Reviewers | |||
---|---|---|---|
1 | 2 | 3 | |
Version 1 22 Jul 20 |
read | read | read |
Provide sufficient details of any financial or non-financial competing interests to enable users to assess whether your comments might lead a reasonable person to question your impartiality. Consider the following examples, but note that this is not an exhaustive list:
Sign up for content alerts and receive a weekly or monthly email with all newly published articles
Already registered? Sign in
The email address should be the one you originally registered with F1000.
You registered with F1000 via Google, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Google account password, please click here.
You registered with F1000 via Facebook, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Facebook account password, please click here.
If your email address is registered with us, we will email you instructions to reset your password.
If you think you should have received this email but it has not arrived, please check your spam filters and/or contact for further assistance.
Comments on this article Comments (0)