Tracing frequent users of regional care services using emergency medical services data: a networked approach

Objectives This study shows how a networked approach relying on ‘real-world’ emergency medical services (EMS) records might contribute to tracing frequent users of care services on a regional scale. Their tracing is considered of importance for policy-makers and clinicians, since they represent a considerable workload and use of scarce resources. While existing approaches for data collection on frequent users tend to limit scope to individual or associated care providers, the proposed approach exploits the role of EMS as the network’s ‘ferryman’ overseeing and recording patient calls made to an entire network of care providers. Design A retrospective study was performed analysing 2012–2017 EMS calls in the province of Drenthe, the Netherlands. Using EMS data, benefits of the networked approach versus existing approaches are assessed by quantifying the number of frequent users and their associated calls for various categories of care providers. Main categories considered are hospitals, nursing homes and EMS. Setting EMS in the province of Drenthe, the Netherlands, serving a population of 491 867. Participants Analyses are based on secondary patient data from EMS records, entailing 212 967 transports and 126 758 patients, over 6 years (2012–2017). Results Use of the networked approach for analysing calls made to hospitals in Drenthe resulted in a 20% average increase of frequent users traced. Extending the analysis by including hospitals outside Drenthe increased ascertainment by 28%. Extending to all categories of care providers, inside Drenthe, and subsequently, irrespective of their location, resulted in an average increase of 132% and 152% of frequent users identified, respectively. Conclusions Many frequent users of care services are network users relying on multiple regional care providers, possibly representing inefficient use of scarce resources. Network users are effectively and efficiently traced by using EMS records offering high coverage of calls made to regional care providers.


REVIEWER
Yohann Chiu Université de Sherbrooke, Canada REVIEW RETURNED 19-Feb-2020 GENERAL COMMENTS This is an interesting piece of work about tracing frequent users of care services through Emergency medical services, in the province of Drenthe. It is well written and the results are very encouraging. I have a few comments, though I suspect that they are mostly minor.
1) EMS should appear in full at first mention in the text. It is in full mention in the abstract, but I think that it also should be written in full at page 5 (line 45). In the same spirit, Netherlands should appear at first mention of the province of Drenthe (page 6 line 8).
2) How did the authors gain access to EMS data? Since there is no detail in the Data or Privacy sections, it is not clear to me as a non Deutch researcher how easy the access would be. As the authors state themselves, there are many hurdles in gaining access to those data, so I am curious as to how they .
3) This is related to my previous comment; more globally, has any research on frequent use been done using those same databases? The results suggest that using EMS data allows for tracing a lot more frequent users than any single care providers. This seems a very good improvement and, depending on the answer to my previous question, I was wondering what other studies have shown using those databases. 4) "We found how inclusion of ambulance transports to unknown, i.e., not recorded destinations in data analysis may result in higher numbers of frequent users" (page 12 line 28). What proportion of ambulance transports was related to unknown destinations? I did not see any mention of this in the text.

5)
In order to get a full grasp on the results scope, I would be interested in seeing which type of frequent users are "captured" using EMS databases (e.g. patients with drug abuse, geriatric patient with comorbidity, etc.). Are the reason for EMS calls included in the databases? If so, would it be possible to investigate those reasons, or at least to discuss it in the discussion? The study appears to be easy to interpret and reproduce. Methods are linear and appropriate.

VERSION 1 -AUTHOR RESPONSE
Answer: Thank you for giving us this opportunity to revise our paper. The constructive comments indeed guided us in making relevant changes to the manuscript.
1. The introduction should be complemented by better clarifying the appropriate way in which EDs manage frequent users with regard to current scientific recommendations. It would be recommended to add more bibliography references.
"Frequent users, i.e., patients that make repetitive calls for health care services, may be responsible for a relatively large share of regional care consumption. They represent a minority of Emergency Department (ED) patients (4.5-8%), yet, they may account for up to 21-28% of all ED visits.1 2" Many studies have shown that a customized Case Management approach helps Frequent Users in finding an appropriate answer to their needs reducing visits to the ED and, in some cases, healthcare costs through Individual Care Plans, telephone contact and facilitated contacts with healthcare providers.
"Due to their high impact on care providers' workload and associated costs they are a focal group for regional policy makers and clinicians aiming to make best use of scarce resources." Answer: We thank the reviewer for providing us with recommendations to include more references regarding the management of frequent users, especially by a case management approach, individual care plans, telephone contact and facilitated contacts with healthcare providers. While these solutions might indeed reduce unnecessary use of services it also remains clear that not all individuals can be identified prospectively, i.e., before such individuals become frail and frequent users, that would profit from some sort of advance care planning. We propose that exploiting routinely collected data as described would be complementary, and could allow factual identification and subsequently taking appropriate measures as suggested.
Accordingly, new sentences were added (see lines 74-80 in paper, and text below), including additional references: "Different solutions have been devised for frequent users once identified. Subsequently, the appropriate answers to their needs, and consequently reducing the visits to ED and ambulance transports may be achieved. These solutions range from case management4-6, to individual care plans7-9, and facilitated contacts with healthcare providers10. However, to be able to offer and consider such a form of advance care planning for apparently frail patients they first need to be identified. The latter in reality may escape attention or appear difficult with data scattered over various institutions." 4. The authors showed EMS dataset (ambulance call log data) can be used to identify heavy users of health care resources. The authors concluded this is a better approach compared to those depending on single facility dataset. I think this is so obvious and not need to be proved by any research. In addition, there are already many previous studies providing deeper insights on this subjects which were not mentioned in this paper.

Answer:
Part of the remark put forward by the reviewer is true, i.e., a single comprehensive record of some sort capturing all healthcare use of individuals indeed would allow simple and efficient identification of heavy users. However, only those countries or regions that have a single payer or single service providing organization indeed have those possibilities. In most settings, funding as well as service provision is dispersed and impossible to completely trace. This study presents an exemplar of an inner service provider, i.e., EMS that links the majority of acute and chronic health services and thus is in a unique position that other service providers within the network will never achieve. We maintain that for many regions and settings this insight is novel and worthwhile to consider. Indeed, we respectfully disagree with the reviewer. Importantly, also the third reviewer working in a setting more comparable to ours also recognizes the relevance and novelty.
We also added the following sentences in section "Discussion", see lines 239-243: "The opportunity we identified and seized might seem trivial in settings where individuals are easily traced, i.e., single payer or service provider systems. In these systems the necessity to take appropriate action is no less urgent, yet the effort to obtain a listing and pattern of use might be simpler. Nevertheless, we provide a worked out exemplary approach that may be applied in many settings like the Netherlands." Reviewer: 3 Reviewer Name: Yohann Chiu Université de Sherbrooke, Canada This is an interesting piece of work about tracing frequent users of care services through Emergency medical services, in the province of Drenthe. It is well written and the results are very encouraging. I have a few comments, though I suspect that they are mostly minor.
Answer: Thank you for giving us this opportunity to revise our paper. The constructive comments indeed guided us throughout all the changes we made in this revised version.
1. EMS should appear in full at first mention in the text. It is in full mention in the abstract, but I think that it also should be written in full at page 5 (line 45). In the same spirit, Netherlands should appear at first mention of the province of Drenthe (page 6 line 8).
Answer: Thank you for helping us improving our manuscript, we incorporated the suggestions on EMS (see lines 57 and 82-83). Also we have inserted "the Netherlands", following reviewer's suggestion (see line 109).
2. How did the authors gain access to EMS data? Since there is no detail in the Data or Privacy sections, it is not clear to me as a non Dutch researcher how easy the access would be. As the authors state themselves, there are many hurdles in gaining access to those data, so I am curious as to how they .
Answer: Thank you for this question. We were able to overcome the mentioned hurdles by getting a strong commitment from the EMS provider serving the Province of Drenthe, the Netherlands. The EMS services play a key role in our regional emergency care networks, and they were at times experiencing congestion. They thus were very much interested in analyzing their data and looking for causes and potential solutions. Indeed, they may be considered problem owners looking for expert partners such as our team to collaborate with. We obviously also obtained a full waiver for using anonymized data from the EMS services from our institutional ethical review board as stated in lines 132-134 (see also below): "We obtained a full waiver for using anonymized data from the EMS services from our institutional ethical review board." Furthermore, to clarify the interest of the EMS in identifying frequent users and cooperating with us we add the following lines to the Introduction, see lines 82-86.
"In the Netherlands and possibly other settings the Emergency Medical Services (EMS) are increasingly overburdened, and at times encounter backlogs at the EDs of hospitals11 12 Indeed, the role of EMS in triage and adequate and timely referral is increasingly recognised in acute care networks. Accordingly, identifying opportunities to relieve an overburdened acute care system from frequent and inappropriate may be considered an impending responsibility of EMS." 3. This is related to my previous comment; more globally, has any research on frequent use been done using those same databases? The results suggest that using EMS data allows for tracing a lot more frequent users than any single care providers. This seems a very good improvement and, depending on the answer to my previous question, I was wondering what other studies have shown using those databases. Answer: To the best of our knowledge, this is the first manuscript addressing the use of EMS data as key and unique data source for tracing frequent users.
4. "We found how inclusion of ambulance transports to unknown, i.e., not recorded destinations in data analysis may result in higher numbers of frequent users" (page 12 line 28). What proportion of ambulance transports was related to unknown destinations? I did not see any mention of this in the text. Answer: In section "Results" (lines 165-167), we explain which data is used in analysis: "EMS records for 2012-2017, refer to 212,967 calls for services, involving 126,758 patients. Data cleaning resulted in 2,494 calls being removed. In addition, 13,156 calls were discarded due to their lack of information on transport destination, i.e., care providers. The latter 13,156 calls refer to unknown destinations." To clarify matters, we adapted the above text using same terminology as being used on the first paragraph of section "Results", see lines 165-167.
"EMS records for 2012-2017, refer to 212,967 calls for services, involving 126,758 patients. Data cleaning resulted in 2,494 calls being removed. In addition, 13,156 calls (6%) were discarded due to unknown, not recorded destinations (i.e. care providers)." 5. In order to get a full grasp on the results scope, I would be interested in seeing which type of frequent users are "captured" using EMS databases (e.g. patients with drug abuse, geriatric patient with comorbidity, etc.). Are the reason for EMS calls included in the databases? If so, would it be possible to investigate those reasons, or at least to discuss it in the discussion? Answer: We thank the reviewer for this excellent and very relevant remark.
EMS data include reasons for calls building on diagnostics provided by the ambulance nurse and 911 (Europe 112). We added the following text in "Discussion", see lines 250-252 to clarify how reasons for EMS calls may be determined using novel techniques for diagnostics data analysis: "Whereas process mining may be helpful in capturing patients' routing along care providers, data mining may assist in analysing patients' care needs further using text analysis of diagnostic data, thereby unravelling their reasons for calls."