NurseMatcher had many nurses working in part-time contracts who were always looking into the shifts they would like to work. Searching and finding the best suitable shift match was tedious, hectic, and monotonous when done manually.
Hence, developing a perfect shift finder without going through painstaking searches was a needed solution for them.
The healthcare industry continues to struggle with a shortage of nurse staffing. According to an article in the Nursing Times, “According to the American Nurses Association, the US is projecting 1.1 million RNs are needed to avoid a critical shortage of nurses by 2022.” One of the primary reasons for the high turnover in healthcare staff is burnout. There aren’t enough resources in the facilities, hence forcing nurses to work extra in inflexible hours.
Many organizations and hospitals are coping with the shortage by opting for per diem health staff as per the patient’s outflow; however, acquiring skilled nurses in desired shifts in real-time is still a challenge.
NurseMatcher (a proprietary solution that we have built) allows nurses to find suitable shifts in the facilities they chose. The AI-powered shift matching connects nurses to shifts as per their need in real-time. Caboom helped NurseMatcher by incorporating AI in healthcare. We streamlined the processes to build a smart AI-based recommendation engine that allows us to intelligently match nurses to shifts available without hiring in-house a data science team at the client’s end.
Caboom team with NurseMatcher partnered to provide a solution to automate the task involved in matching nurses' shifts. We came up with an AI-based approach to matching the nurses to the desired best recommendations. By filtering the ineligible candidate, we recommended shifts to nurses based on the application, acceptance, and rating into their account.
Check below for details. :
In the candidate selection process, we filtered out the nurses who are not eligible to apply for the shift. This reduces the number of nurses. Hence it also reduces the computation time to generate a score for every nurse. The filtration criteria are:
- Negative feedback from the hospital to the nurses
- Blacklisted from the hospital
- If a shift has already been applied by the nurses
- It also checks whether the nurse is permissible to work on the state.
From the pool of selected candidates, we further analyzed the temporal behavior of the nurses with respect to their shifts. For this, we took the past 10 latest highly rated shifts or latest accepted or latest applied shifts by the organization to the user to create the profile of the nurses. Three types of engine deployed to analyze the nurse’s behavior with respect to shifting are as below:
Rating engine: Engine that recommends the nurses shift based on their rating.
Application engine: Engine that recommends the nurses based on the application.
Acceptance engine: Engine that recommends the nurses based on the shift acceptance by the nurses.
In case users do not have a past history, to solve this problem, we have taken the average feature created based on their qualification and preferable location as the default user profile.
Ultimately after creating the nurse profile, the three recommendation models mentioned above generate the top 2 recommendations from each model based on rating, application, and acceptance.
The AI-based recommendation engine implemented in NurseMatcher helped provide easy means for nurses to find shifts as per their preference. At a time when healthcare facilities struggle with the nursing shortage, NurseMatcher provides them with choice and transparency by the use of AI-based recommenders.
As seen in our recommendation engine, we used three types of a content-based recommendation engine, taking ratings into account. It recommends the best shift to nurses based on applied history, acceptance history, and ratings. Hence for a single shift, we use three different ranking engines to compute the ranking score due to which platform recommends a diverse set of nurses.
The applied model keeps nurses at the forefront. It recommends shifts that consider the nurse's choice and their preferences for facilities to find very similar shifts. The acceptance model considers those shifts where the nurse is highly likely to be accepted and recommends according to it. Recommendation based on ratings takes the nurse experience in a particular shift, finds similar shifts, and recommends it. The information based on the nurses' certification could help track the similar nurses and help us recommend the shift according to it even though we don't have any historical data. Hence it tackles the cold start problem or fewer data problems as well.
Some nurses chose shifts based on personal relationships. This kind of favoritism would interfere with the shift's profile and end up affecting the nurse matching algorithm.
Nursing is a very specialized job. Hence there were few cases where recommending shifts could be done just by heuristic filtering rather than applying machine learning algorithms. For example, a nurse working in eye care usually specializes in eyes, and hence the shift gets narrowed down to 5 or sometimes fewer than that. Therefore in scenarios like this, the recommendation process becomes somewhat redundant.